Rotary Wing Structural Dynamics and Aeroelasticity

Rotary Wing
Structural Dynamics
and Aeroelasticity
Second Edition
“prelims” — 2005/11/22 — page i — #1

Rotary Wing
Structural Dynamics
and Aeroelasticity
Second Edition
Richard L. Bielawa
EDUCATION SERIES
Joseph A. Schetz
Series Editor-in-Chief
Virginia Polytechnic Institute and State University
Blacksburg, Virginia
Published by
American Institute of Aeronautics and Astronautics, Inc.
1801 Alexander Bell Drive, Reston, VA 20191
“prelims” — 2005/11/22 — page iii — #3

American Institute of Aeronautics and Astronautics, Inc., Reston, Virginia
1 2 3 4 5
Library of Congress Cataloging-in-Publication Data on File
Copyright © 2006 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
Printed in the United States. No part of this publication may be reproduced, distributed, or transmitted,
in any form or by any means, or stored in a database or retrieval system, without the prior written
permission of the publisher.
1-56347-698-3
Data and information appearing in this book are for information purposes only. AIAA is not responsible
for any injury or damage resulting from use or reliance, nor does AIAA warrant that use or reliance
will be free from privately owned rights.
“prelims” — 2005/11/22 — page iv — #4

Dedication
To my wife, Joan, ever supportive, ever encouraging, and ever patient!
“prelims” — 2005/11/22 — page v — #5

AIAA Education Series
Editor-in-Chief
Joseph A. Schetz
Virginia Polytechnic Institute and State University
Editorial Board
Takahira Aoki David K. Holger
University of Tokyo Iowa State University
Robert H. Bishop Rakesh K. Kapania

University of Texas at Austin Virginia Polytechnic Institute
and State University
Claudio Bruno
University of Rome Brian Landrum
University of Alabama,
Aaron R. Byerley Huntsville
U. S. Air Force Academy
Achille Messac
Richard Colgren Rensselaer Polytechnic Institute
University of Kansas
Michael Mohaghegh
Kajal K. Gupta The Boeing Company
NASA Dryden Flight Research
Center Todd J. Mosher
Utah State University
Albert D. Helfrick
Embry-Riddle Aeronautical Conrad F. Newberry
University Naval Postgraduate School
Rickard B. Heslehurst David K. Schmidt

Australian Defence Force University of Colorado
Academy Colorado Springs
David M. Van Wie

Johns Hopkins University
“prelims” — 2005/11/22 — page vii — #7

Foreword
It is a privilege and pleasure to help bring this second edition to the attention of
the widest possible readership. It follows a very well received first edition and, as
did its predecessor, provides a comprehensive yet accessible account of many of
the important concepts and developments in the field by a master of the subject.
Dr. Bielawa is superbly qualified by his experience in industry and academe to
contribute an authoritative yet clear account of this fascinating set of topics.
Those of us who have on occasion worked in rotorcraft structural dynamics and
aeroelasticity, but whose primary work has been in fixed wing aircrafl including
turbomachinery, have only the highest respect for those engineers who have taken
up the high challenges that rotorcraft provide. These challenges emerge across
a spectrum of activities from the intellectual, computational, and experimental
to perhaps the greatest challenge of all in design. They arise from a wide range
of disciplines including fluid dynamics, structural dynamics, dynamic response,
control, and stability. To bring coherence and structure to this wide range of relevant
activities and disciplines in the context of rotorcraft is a major challenge in itself
and one that is well met in this volume.
This book will be of great value to all those who participate in one or more
of these activities and disciplines. The scope of the treatment is broad yet deep.
Experienced rotorcraft engineers will benefit from an account of new ideas as well
as from a ready reference for now classical material. The beginning student, with
the help of an experienced guide such as Dr. Bielawa or one of his peers, will find
this a reliable and engaging introduction to this fascinating topic. Self-study is also
possible and frequently this book will be one of the best places to begin the study
of a topic in rotorcraft structural dynamics and aeroelasticity.
Dr. Bielawa has again performed a great service by bringing forth this new
edition. I am confident that it will find a well deserved and highly valued place in
the libraries of rotorcraft engineers, both present and future.
Earl Dowell
Duke University
ix
“prelims” — 2005/11/22 — page ix — #9

Contents
Preface to the Second Edition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Preface to the First Edition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii
Chapter 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Rotary Wing vs Fixed Wing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Chapter 2 Basic Analytical Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.1 Linear Single-Degree-of-Freedom System . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Fourier Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Linear Two-Degree-of-Freedom System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4 Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.5 Structural Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.6 Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.7 Vector Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.8 Theorem of Coriolis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Chapter 3 Rotating Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3.1 Basic Equations for Bending . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.2 Reference Uniform Blade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.3 Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.4 Approximate Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.5 Two-Bladed Rotor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.6 Blade Torsion Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.7 Coupling Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Chapter 4 Gyroscopics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

4.1 Rotational Motion of a Solid Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.2 Simplified Gyroscope Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4.3 Precession and Nutation Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.4 Gyroscopic Characteristics of Rotor Blades. . . . . . . . . . . . . . . . . . . . . . . . . 138
Chapter 5 Drive System Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

5.1 Shaft Critical Speeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
5.2 Torsional Natural Frequencies of Shafting Systems . . . . . . . . . . . . . . . . 162
5.3 Special Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
xi
“prelims” — 2005/11/22 — page xi — #11

xii CONTENTS
Chapter 6 Fuselage Vibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189

6.1 Dynamic Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
6.2 Harmonic Rotor Hub Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6.3 Nonrotor Sources of Fuselage Excitation . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
6.4 Rotor-Fuselage Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Chapter 7 Methods for Vibration Control. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

7.1 Basic Modification Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
7.2 Modification of Blade Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
7.3 Modification of Fuselage Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
7.4 Vibration-Suppression Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
Chapter 8 Vibration Test Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

8.1 Basic Shake Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
8.2 Other Test Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
Chapter 9 Stability Analysis Methods: Linear Systems . . . . . . . . . . . . . . . 297

9.1 Basic Concepts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
9.2 Basic Tools for Linear Systems: Constant Coefficients . . . . . . . . . . . . . 299
9.3 Linear Multiple-Degree-of-Freedom Systems:
Constant Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
Periodic Coefficients (Floquet Theory) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
9.5 Nyquist Criterion for Multiple-Degree-of-Freedom Systems . . . . . . 325
Chapter 10 Mechanical and Aeromechanical

Instabilities of Rotors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
10.1 Unsymmetrical Rotor Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
10.2 Quasi-Steady Aerodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
10.3 Rotor Weaving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
10.4 Blade Pitch-Flap-Lag Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
10.5 Rotordynamic Instabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
Chapter 11 Mechanical and Aeromechanical Instabilities of

Rotor-Pylon Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
11.1 Multiblade Coordinates and Rotor Modes. . . . . . . . . . . . . . . . . . . . . . . . . . . 387
11.2 Rotor-Nacelle Whirl Flutter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
11.3 Ground Resonance Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404
11.4 Air Resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424
11.5 Air Mass Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433
Chapter 12 Unsteady Aerodynamics and Flutter of Rotors . . . . . . . . . . . . 447

12.1 Introduction and Classification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
12.2 Two-Dimensional Frequency-Domain Theories . . . . . . . . . . . . . . . . . . . . 451
12.3 Two-Dimensional Arbitrary Motion Theories . . . . . . . . . . . . . . . . . . . . . . . 469
12.4 Bending-Torsion Flutter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
12.5 Three-Dimensional Aerodynamic Theories . . . . . . . . . . . . . . . . . . . . . . . . . 496
12.6 Dynamic Stall and Stall Flutter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508
“prelims” — 2005/11/22 — page xii — #12

CONTENTS xiii
Chapter 13 Analysis of Nonlinear Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529

13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529
13.2 Simple Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 530
13.3 Transient Solutions Using Numerical Integration . . . . . . . . . . . . . . . . . . . 535
13.4 Quasi-Linearization for Explicit Nonlinearities . . . . . . . . . . . . . . . . . . . . . 546
13.5 Numerical Methods for Stability Estimation . . . . . . . . . . . . . . . . . . . . . . . . 548
13.6 Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 553
Chapter 14 Model Rotor Testing for Aeroelastic Stability . . . . . . . . . . . . . . 557

14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557
14.2 Scaling Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557
14.3 Model Construction Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562
14.4 Instrumentation and Test Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564
14.5 Aeroelastic Considerations for Nonaeroelastic Testing . . . . . . . . . . . . . 574
Chapter 15 Elastomeric Devices for Rotorcraft . . . . . . . . . . . . . . . . . . . . . . . . . . 579

15.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 579
15.2 Examples of Elastomeric Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 579
15.3 Basic Characteristics of Elastomeric Materials . . . . . . . . . . . . . . . . . . . . . 581
15.4 Elastomeric Lead-Lag Dampers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 584
15.5 Other Applications of Elastomerics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
Chapter 16 Blade-Section Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 591

16.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 591
16.2 Generalized Blade Elastic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 591
16.3 Beams with Thin-Shell Closed-Cell Construction . . . . . . . . . . . . . . . . . . 594
16.4 Multicell Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 605
16.5 Total Section Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 616
16.6 Hygrothermal Effects in Composites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 620
16.7 Prismatic Bars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 622
Chapter 17 Cross-over Topics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625

17.1 Interactions of the Rotor Drive and Engine/Fuel Control Systems 625
17.2 Aeroelastic Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
Chapter 18 Concluding Thoughts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651

18.1 Key Historical Milestones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
18.2 What’s on the Horizon? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654
Appendix A Glossary of Rotorcraft-Related Terms . . . . . . . . . . . . . . . . . . . . . . 657
Appendix B Charts for Blade Frequency Estimation . . . . . . . . . . . . . . . . . . . . 665
Appendix C Generalized Frequency-Domain

Substructure Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 677
“prelims” — 2005/11/22 — page xiii — #13

xiv CONTENTS
Appendix D Basic Equations of Motion for Ground Resonance and

Air Resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 685
Appendix E Composite Materials—Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 707
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 719
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 729
Supporting Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 751
“prelims” — 2005/11/22 — page xiv — #14

Preface to the Second Edition
While the first edition was being used, both the technology base and I have
matured. I am still learning my craft! Most importantly, perhaps, is the ubiquitous
availability of first-rate computer resources, both in hardware and software. A
prime driver for upgrading this text has been the challenge of providing up-to-
date material for the course I have been teaching within the University of Kansas
Aerospace Short Course Program. Also, to my chagrin, is the opportunity to correct
the several errors that invariably crop up in any new text. (Hopefully this revision
will be free of such errors!) Because much of the original material was geared
to non-computer solution techniques using at most only hand calculators, that
material was subsequently discarded in favor of more up-to-date techniques. Also,
where possible, new material that I and others have since developed has been
included. I have endeavored to maintain the text as a tool principally for the
practicing engineer and have attempted to reformulate some material that originally
was rather academic into more practical forms. New chapters have been included
dealing with elastomeric devices, airfoil sections with an emphasis on composites,
“cross-over” topics, and a historical perspective on the subject material. A new
appendix has been provided presenting basic material on composites.
As with the first edition, I have received help from a number of sources both
in supplying me with excellent graphic material and with reviewing the new text
material. The Lord Corporation’s help in providing me with their photo and graph-
ical contributions is greatly appreciated. Steve Winckler’s help in reviewing the
material on composites and Dean Bellinger’s help in providing the historical data
on NASTRAN are both gratefully acknowledged. All other persons who have
taken time out of their busy professional lives to contribute requested photos are
also gratefully acknowledged.
Richard L. Bielawa
October 2005
xv
“prelims” — 2005/11/22 — page xv — #15

Preface to the First Edition
This text evolved in spurts over the course of several years. Each evolutionary
growth spurt was, however, in response to a somewhat different objective. Like so
many other texts, this one had its inception in a specific need to rapidly provide an
engineering team with the specialized knowledge they needed to do their jobs. In
this case, the team consisted of fixed wing aeronautical engineers and designers
at the Lockheed California Company who were developing innovative helicopter
concepts during the 1960s. To this end Lockheed organized an in-house education
program in rotary wing technology, which included material on structural dynamics
and aeroelasticity. Thus, the original objective of the embryonic lecture notes that
resulted was to provide rapid augmentation of working fixed wing capabilities with
appropriate new rotary wing capabilities. The lecture notes were later expanded to
meet the needs of a similar short course given by the University of California at
Los Angeles, again for full-time practicing engineers. In each of these instances
the accent was on the rapid dissemination of fundamental practical information
with minimum attention to rigorous derivation.
Further evolution of the text occurred after several years’hiatus with the author’s
entry into an academic career at the Rensselaer Polytechnic Institute, and his
participation in the Rotorcraft Technology Center of Excellence Program residing
at that institute. The original lecture notes were greatly expanded to provide suitable
material for graduate courses in the subject. Consequently, the objectives of the
evolving text became modified to the extent that, although the practical aspects
of the text were still to be retained, additional material was required to provide
in-depth specialized education to individuals who were preparing for careers in
the rotary wing industry. Inasmuch as a large portion of this material inherently
consists of specialized applications of fundamental concepts in basic (non-rotary-
wing) structural dynamics and aeroelasticity, several of these fundamental concepts
needed to be addressed. Finally, with the passage of time the emphasis of rotary
wing aircraft technology has changed from that of a purely helicopter orientation
to a more general nature that now includes the development of operational tilt-rotor
and tilt-wing aircraft.
Thus, the final objectives that have driven the evolution of the text are an amal-
gamation of both the prior and later objectives: 1) to provide an exposition of
some of the fundamental concepts of structural dynamics and aeroelasticity (by
way of review), 2) to provide the practicing engineer with the basic, fundamental
knowledge in a practical format which can be used advantageously in the practice
of his/her art, and 3) to provide a vehicle for the teaching of the material in a uni-
versity environment at the graduate level. An additional fourth objective is that the
first three objectives are to be met while concurrently serving the special require-
ments of such newly emerging rotor concepts as tilt-wing and tilt-rotor aircraft
configurations.
A complete exposition of all aspects of rotary wing structural dynamics and
aeroelasticity would be an impossibility due not only to the sheer accumulated
volume of pertinent material, but because of the constant growth of the technology.
xvii
“prelims” — 2005/11/22 — page xvii — #17

xviii PREFACE TO THE FIRST EDITION
The material presented herein thus deals with fundamental issues that form a
matrix into which all of the more detailed and newly emerging engineering studies
can be placed. To be certain, much of the text deals with diverse methodologies
relating to the rotational dynamics of rotors. At an even more fundamental level
basic mathematical tools are presented which relate to the analysis requirements
of multiple-degree-of-freedom dynamic systems. For many practicing engineers
this mathematical material might be of limited novelty, but for others its inclusion
should prove to be quite useful.
Because the subject technology has two main drivers, the minimization of vibra-
tion and the assurance of aeromechanical and aeroelastic stability, this text has been
organized sequentially as much as possible into these two basic areas. As required
for both of these areas, Chapters 1–4 present various basic analysis tools which lay
the necessary groundwork for the remainder of the text. Beyond the basic tools pre-
sented in these first chapters, additional methodologies are presented throughout
the text on an as-needed basis. Chapters 5–8 are focused on the airframe vibration
problem with an intended emphasis on tools for designing vibration out of a new
design and/or fixing unaccounted-for vibration problems once they have surfaced.
The rotor dynamic environment is rich in possibilities for a variety of aeronauti-
cal and aeroelastic instabilities. Indeed, those instability phenomena presented in
Chapters 10–13 are not intended to constitute an exhaustive exposition. Instead,
the phenomena presented were selected on the bases that they either constitute the
more pertinent phenomena driving the rotorcraft design, or they form the bases for
other more complex phenomena. Finally, because the available analytical tools are
as yet not completely reliable (much less “user-proof”), successful integration of
the technology into real-world applications must have experimental verification.
Consequently, Chapters 8 and 14 are respectively directed at experimental tech-
niques found to be useful in each of the two basic technology areas. The text has
been organized in this manner so that, hopefully, it will be a practical and conve-
nient teaching resource for either a two-semester or three-quarter course sequence
at the graduate level. The objective of providing a vehicle for the formal teaching
of the material in a university environment consequently required the composition
and incorporation of suitable exercise problems. The problems compiled include
a variety of relatively simple ones as well as more difficult ones; in all of the prob-
lems presented the intent is to teach and thereby to help readers achieve mastery
of the material.
The objective of addressing the emerging, more advanced rotary wing concepts
required the inclusion of material relating to propeller-nacelle whirl flutter, as well
as some analysis material relating to multiple rotor dynamic systems. The extensive
analytical development presented in Appendix D was primarily formulated as
a rudimentary analysis tool for helicopter ground resonance and air resonance.
However, its inclusion in this text was justified on the grounds that it also serves
the useful purpose of defining a basic, practical aeroelastic rotor module, to be
used as a building block for formulating larger analyses of multiple rotor (tilt-rotor
and/or tilt-wing) aircraft aeroelastic stability.
This textbook was not written in a vacuum and in many instances the text is
a selective compilation of the findings of many other researchers. Furthermore,
without the encouragement of friends and colleagues it would not have been writ-
ten. Special acknowledgments in this category go to William Twomey, William
“prelims” — 2005/11/22 — page xviii — #18

PREFACE TO THE FIRST EDITION xix
Flannelly, and Robert Loewy. The Rensselaer Polytechnic Institute Rotorcraft

Technology Center provided the nurturing environment for the teaching of the
material and the requirement for a coherent textbook. Although students are noto-
rious for ferreting out errors in their professors’ lecture notes, special thanks go to
Nicolas Theron and Walter Hassenpflug, whose efforts in this regard were meticu-
lous. Besides providing productive working environments for the many researchers
who contributed to the textbook, Kaman Aerospace Corporation and the Sikorsky
Division of United Technologies Corporation donated the computational resources
without which the timely wordprocessing, computations, and figure generation
would not have been possible. This generous donation is gratefully acknowledged.
Much of the pictorial material was enthusiastically donated by numerous people
in the rotary wing industry, both in America and abroad, too numerous to name
individually; this help was much appreciated and my thanks go to all of these
people. Finally, the indefatigable help provided by Judy O’Connor in transfor-
ming rough wordprocessing into a finished manuscript and in the composing of
some of the figures is gratefully acknowledged.
Richard L. Bielawa
June 1992
“prelims” — 2005/11/22 — page xix — #19

1
Introduction
At its basic roots, the subject matter of rotary wing structural dynamics
and aeroelasticity is concerned with the following problem: How can structural
integrity and passenger comfort be attained in the vibratory environment pecu-
liar to rotary wing aircraft? From this broad problem definition has evolved the
two principal areas of concern in rotary wing structural dynamics: vibrations and
aeroelastic stability. The emphasis is currently on vibration reduction, but with
close attention being paid, nonetheless, to the fact that rotors are subject to a
variety of potentially unstable response phenomena. The advent of new rotor con-
cepts increases the probability of an otherwise benign characteristic rotor response
becoming a major aeroelastic instability issue. In any practical design the insta-
bilities must be well understood, and ways must be found to suppress them in all
conceivable flight conditions. Last, it should be noted that one area of concern
that closely relates to structural dynamics and aeroelasticity is that of noise. The
high degree of noise characteristic of most rotorcraft has spawned the growing
technical areas of rotor far-field noise acoustics and structure-borne interior noise
acoustoelasticity. Although both these subjects are important and timely, this text
is limited to considerations of only vibrations and aeromechanical and aeroelastic
instabilities.
1.1 Rotary Wing vs Fixed Wing

The agenda for structural dynamics and aeroelasticity of rotorcraft is distin-
guished from that of fixed wings and/or airborne spacecraft in a variety of ways.
First, vibration is a major problem with rotorcraft in all forward-flight conditions,
both steady and maneuvering flight. However, this vibration is not broadband;
rather, it occurs at discrete frequencies related to the main and tail rotor rotational
frequencies. Indeed, the principal driver of the rotorcraft vibration problem is the
essentially unsteady aerodynamics of the main rotor even in steady forward flight.
Second, not only is the aerodynamic environment of the rotor unsteady, but at the
same time it is also nonlinear, governed by non-infinitesimal motion and periodicity
in conditions traditionally held constant with fixed wing applications. Presently,
the understanding of rotorcraft aerodynamics is a slowly growing technology, and
despite its key role in analyzing rotorcraft vibration and aeroelastic stability, it still
requires much more work for practical analysis.
In addition to the problems directly arising from unsteady aerodynamics are
those related to the fact that the rotor blades of contemporary helicopter rotors
1
“chapter_1(new)” — 2005/10/18 — page 1 — #1

2 ROTARY WING STRUCTURAL DYNAMICS
are “structural lightweights” compared with their fixed wing counterparts. They
achieve a large measure of their stiffening from the tension induced by the rota-
tional centrifugal force field. The rotational environment of the rotor blades also
gives rise to a host of rotation-related phenomena: gyroscopic characteristics,
Coriolis forces, and a variety of nonlinear inertial loadings. While still governed
principally by linear operators, the resulting aeroelastic description of rotor blade
elastic responses is consequently fraught with nonlinearities that modify the results
obtained using only linear analysis.
The relatively high degree of rotor flexibility also drives the significant interac-
tion occurring between the rotor and the also flexible rotorcraft airframe. At present,
the stubbornness of typical modern-day helicopter airframes to yield to accurate
dynamic analysis continues to pose a very important challenge to the rotorcraft
structural dynamist. Furthermore, contrary to the case of fixed wing aircraft, where
aeroelastic stability characteristics are the priority issue, the dynamic analysis of
rotorcraft airframes must be relatively much more accurate to enable a reasonably
accurate calculation of the vibration characteristics.
1.2 Methodology
Stated in the context of an engineering solution, the structural dynamics problem
of rotorcraft entails three general concurrent avenues of approach: 1) Knowledge
of the vibratory environment of the rotorcraft must be acquired. Principally, this
means knowing the essentially unsteady aerodynamic characteristics of rotorcraft.
2) The extent to which the structure responds to the environment must be calcu-
lated. 3) The resulting responses must be judged for acceptability; if they are not
acceptable, ways must be found to make them so.
1.2.1 Knowledge of the Environment

To know the vibratory environment requires that many of the concepts relating
to frequency-domain analysis be brought to bear. An understanding of unsteady
aerodynamic loadings is required, as are ways of formulating these loadings in both
the frequency domain (for both vibration and instability problems) and the time
domain (for transient responses involving nonlinearity problems). Also, knowledge
of the “real-world” characteristics of the dynamic components used in rotorcraft
is a continuing challenge. Elastomeric devices and composites structures are see-
ing increased usage, and their characteristics are often unique and either deviate
to a degree from completely linear descriptions or present new linear coupling
relationships.
1.2.2 Calculating the Responses

Calculating the dynamic responses of structures to their environment requires
the idealizations that linear differential equations of motion provide. In some cases
linear systems are insufficient for an accurate formulation, and non-linear differ-
ential equations must be used. In general, two types of responses command our
attention almost exclusively: resonant responses and unstable responses. Exam-
ples of resonant responses are the once per rotor rev accelerations imparted to a
“chapter_1(new)” — 2005/10/18 — page 2 — #2

INTRODUCTION 3
helicopter fuselage by an unbalanced rotor, and the (number of blades) per rev
accelerations resulting in the fuselage from a periodic main rotor wake impinge-
ment on the horizontal stabilizer. An example of an unstable response is the flutter
of a wing or rotor blade. Customarily, a resonant response is ideally thought of
as the excitation of a structure at one or more of its natural frequencies by an
external energy source, whereas an unstable response is thought of as a structure
driving itself from an internal energy source. It is a complication with rotorcraft
that, in practice, it is often difficult to ascertain which of these responses is in fact
occurring. This is because both types of response can exhibit measured behavior
that “blossoms” for periods of time, and furthermore, both types can exhibit limit
amplitudes of sinusoidal motion.
1.2.3 Evaluation and Modification

To pass engineering judgment on any given structural configuration, acceptable
limits of response must be defined in an appropriate quantitative sense. Such
limits are typically defined by the primary considerations of structural integrity
and passenger comfort. If the structure is stressed too high, static failure of the
material will occur. If the structure is dynamically stressed too high, too often, it
fails in a fatigue mode. Furthermore, failure by instability may be precipitated by
a partial failure or weakening in a subsidiary structure. Ultimately the structural
integrity problem is one of achieving a configuration that is both fail-safe and has
an acceptable component life. Whereas the structural problem has an objectively
measurable solution (structure either is or is not fail-safe, or has the required
x number of hours of minimum useful life), the problem of achieving passenger
comfort is measured only relatively. Certainly a reasonable goal is to achieve cabin
acceleration levels that are less than those of the last ship built, or, better yet, that
are “jet smooth.” Present goals are to achieve vibratory levels that are generally
less than 0.05 g in all components.
Perhaps of most importance to the engineering judgment aspect of the broad
problem is the ability to change the structure to achieve a desired improvement
in the dynamic responses of the structure. As yet, this ability is still somewhat
of a craft involving educated trial and error. However, this situation is rapidly
changing, due to the emergence of a variety of powerful analysis tools and their
increasingly more efficient and cost-effective implementations on computers (both
large mainframe machines and personal computers).
1.2.4 Organization of the Text

After an initial three chapters covering basic concepts needed for all aspects of
rotorcraft structural dynamics, this text is divided into three main sections. The
first two sections reflect the emphasis on the two main problem areas: vibration
and aeroelastic stability. Chapters 5 through 8 are specifically devoted to vibra-
tion issues, with an emphasis on methods for structural modification for achieving
vibration reduction. Chapters 9 through 14 are devoted to mechanical, aerome-
chanical, and aeroelastic stability issues. In addition to presenting expositions
of well-established instability phenomena of rotors, techniques are presented for
analyzing the stability of general multiple degree-of-freedom systems. For both of
“chapter_1(new)” — 2005/10/18 — page 3 — #3

these problem areas, i.e., vibration and aeroelastic stability, chapters are presented
dealing with appropriate experimental procedures.
The third section of the text consists of Chapters 15 through 18, which deal with
material that has special applications to both of the two main areas. Chapter 15
presents material relating to elastomeric devices, as these have found a permanent
niche in rotorcraft structural dynamics applications. Chapter 16 presents specifics
for characterizing the blade section with an emphasis on the application of compos-
ite materials. Chapter 17 describes instances of cross-over interactions of “outside”
dynamics-related material with the more standard material of rotorcraft structural
dynamics covered in the previous chapters. Finally, Chapter 18 presents some
concluding thoughts together with an abbreviated description of key milestones in
the development of rotary wing structural dynamics and aeroelasticity.
It is to be hoped that this text will provide an in-depth introduction to the subject
material as well as a useful reference resource both for actual formulations and
bibliographies. The theory of linear differential equations is the primary requisite
mathematical discipline employed in this text; most of the analytical formulations
involve only the integrating of a handful of basic, relatively simple concepts.
“chapter_1(new)” — 2005/10/18 — page 4 — #4

2
Basic Analytical Techniques
The essential tools needed for analyzing the structural dynamics of rotorcraft
must address a variety of issues. The principal problem areas are characterized by
1) small motions, usually sinusoidal in nature; 2) several degrees of freedom of
motion; and 3) a variety of loadings arising from rotational effects. Much attention
is therefore paid to identifying and reviewing pertinent methods for analyzing
linear dynamic systems, starting from the simplest configuration (single degree
of freedom) and progressing to general configurations involving several degrees
of freedom. Basic techniques dealing with “frequency-domain” problems are also
identified and reviewed.
2.1 Linear Single-Degree-of-Freedom System

The archetypical dynamic system to be understood and used repeatedly
as an analog for understanding more complicated systems is the single-degree-of-
freedom (spring-mass-damper) system shown in Fig. 2.1, where the force elements
can be assigned engineering units [U( ), based on lbf, ft, and s], given as
U (mass, m) = lbf-s2 /ft (2.1a)
U (damper, c) = lbf-s/ft (2.1b)
U (spring, k) = lbf/ft (2.1c)
U [applied force, F(t)] = lbf (2.1d)
The appropriate dynamic equation is obtained by taking the mass as a free body
and examining and equilibrating the forces acting on it. One indispensable tool for
analyzing dynamic systems is D’Alembert’s principle, which allows the conversion
of dynamic problems to static ones. Typical usage of this principle consists of
representing the statement of Newton’s second law (that F = ma) by an inertial
load equal to ma, but directed opposite to the acceleration a. With this approach
the forces acting on the mass can be equilibrated, as shown in Fig. 2.2, where the
5
“chapter2(new)” — 2005/11/22 — page 5 — #1

Fig. 2.1 Simple spring-mass-damper dynamic system.
individual force elements can be expressed mathematically as
Fspring = −kx (2.2a)
F(viscous) damper = −cẋ (2.2b)
Finertia = −mẍ (2.2c)
where ( ˙ ) ≡ d( )/dt, and Fapplied is an arbitrary, time-dependent applied force

F(t).
Upon summing up the forces and equating the sum to zero, the following basic
differential equation of motion is obtained:
−mẍ − cẋ − kx + F(t) = 0 (2.3a)
or, more usefully, as
mẍ + cẋ + kx = F(t) (2.3b)
Fig. 2.2 Equilibration of loads acting on mass element.
“chapter2(new)” — 2005/11/22 — page 6 — #2

BASIC ANALYTICAL TECHNIQUES 7
The principal properties of this equation are as follows:

1) The equation is a linear, ordinary differential equation.
2) The total solution to the equation x(t) consists of a sum of two separate
solutions:
x(t) = xh (t) + xp (t) (2.4)
where xh (t) is the homogeneous or “transient” solution obtained by setting F(t) to

zero and xp (t) is the particular or “steady-state” solution obtained by considering
the particular (nonzero) excitation afforded by F(t).
3) For a complete solution x(t), initial conditions relating the state of x at t = 0,
x(0) = α, and ẋ(0) = β are required.
4) Initial conditions are not required to find gross “characteristics” of the
dynamic system.
2.1.1 Homogeneous Solution

Here the appropriate differential equation of motion is
mẍ + cẋ + kx = 0 (2.5)
Because the equation is linear, an exponential solution can be assumed, so that
xh (t) = x̄eλt
where λ ≡ a scalar quantity (a number). Substitution of this solution in the

differential equation yields an algebraic equation of the following form:
(mλ2 + cλ + k)
x̄eλt = 0
(not, in general, = 0)
Therefore,
mλ2 + cλ + k = 0 (2.6)
This is the basic characteristic equation whose solution (the two values of λ) is
given by

λ1 , λ2 = −c/2m ± i k/m − (c/2m)2 (2.7)
Thus,
xh (t) = x̄1 exp(λ1 t) + x̄2 exp(λ2 t) (2.8)
“chapter2(new)” — 2005/11/22 — page 7 — #3

We can obtain a somewhat more useful form by noting first that
exp(iωt) = eiωt ≡ cos ωt + i sin ωt (2.9)
Thus, the homogeneous solution for x(t) can then be written in the following form:
x(t) = eσ t [C1 cos ωD t + C2 sin ωD t] (2.10)
where
σ ≡ exponential decay coefficient = −c/2m (2.11a)

ωD ≡ damped natural frequency = k/m − (c/2m)2 (2.11b)
and C1 and C2 are arbitrary constants determined from the initial conditions. For
the undamped case c = 0, the homogeneous solution becomes a simple sinusoid:

x(t) = C1 cos k/m t + C2 sin k/m t (2.12)
Note: The frequency increases with spring stiffness k and decreases with inertia m.
Effect of damping. The value of the damping coefficient c determines the

attenuation characteristics of the transient responses, as described by the homo-
geneous solution. Examples of transient responses for varying values of the
damping coefficient are shown in Fig. 2.3.
Critical damping. When the damping coefficient c is increased to the point

just where ωD = 0, the solution is no longer oscillatory, and a condition of critical
damping exists. This value of damping is denoted the critical damping value cc .
Fig. 2.3 Attenuation characteristics of damping for homogeneous solutions.
“chapter2(new)” — 2005/11/22 — page 8 — #4

At this condition both of the characteristic values λ1 and λ2 become equal, and the
appropriate solution then becomes
x(t) = (C1 + C2 t) exp(−cc t/2m) (2.13)
The condition of zero damped natural frequency (ωD = 0) then yields

√
k/m − (c/2m)2 = 0 −→ cc = 2 km (2.14)
For systems with damping levels less than the critical value, the responses will
always be oscillatory. Those systems with damping values equal to or in excess of
the critical value will always be aperiodic, and the response with the critical value
can be shown to approach quiescence aperiodically most quickly. These points are
depicted in Fig. 2.4.
Although it is unlikely that a new class of structures will exhibit damping char-
acteristics anywhere approaching critical damping in the near future, the concept of
critical damping is nonetheless quite useful because it separates periodic responses
from aperiodic ones. This feature provides a clear-cut yardstick for measuring
damping. Because the dynamic response is seen to depend on the interrelationship
of inertia, damping, and stiffness, some meaningful measure is needed to describe
how much damping is present in any given system, irrespective of the scale of the
system (as defined by the mass and stiffness).
One method for achieving this measure of damping is afforded by first recasting
the form of the basic differential equation of motion by dividing it by the mass:
ẍ + (c/m)ẋ + (k/m)x = F(t)/m (2.15)
Fig. 2.4 Attenuation characteristics of critical damping.
“chapter2(new)” — 2005/11/22 — page 9 — #5

The term now multiplying the velocity can then be rewritten as

c/m = (c/cc )(cc /m) = 2(c/cc ) k/m = 2ζ ωn (2.16)
where ζ is denoted the critical damping ratio for the system and ωn is the undamped
natural frequency. With these two concepts the basic equation of motion can then
be put into the following useful form:
ẍ + 2ζ ωn ẋ + ωn2 x = F(t)/m (2.17)
To summarize, ζ is a nondimensional damping parameter and, hence, is a more

useful way of quantifying the damping present; and ωn is a frequency parameter
that is independent of damping and thus, when taken with the mass, becomes a
useful way of describing stiffness. With the use of these two parameters, the general
solution to the homogeneous equation can then be written as before, except that
the σ and ωD parameters can be written as
σ = −ζ ωn (2.18a)

ωD = 1 − ζ 2 ωn (2.18b)
Relationship between period and frequency. The sinusoidal content of

a response is measured alternatively in terms of its frequency and/or period of
oscillation. The frequency is typically denoted by ω when it has the dimensions
of radians per second and by f when it has the dimensions of cycles per sec or hertz
(Hz). These two ways of dimensioning frequency are related to the period of the
oscillation T as follows:
(one cycle) 1 2π (rad)

= (Hz) = (2.19)
T (s) T T (s)
To summarize,
2π/T = ω; f = 1/T ; 2π f = ω (2.20)
where
U( f ) = cycles/s = (Hz); U(ω) = rad/s (2.21)
2.1.2 Particular Solutions

Although the applied force F(t) can be virtually any continuous, single-valued
function imaginable, we will concentrate here on only a relatively few basic types.
When the particular solution is considered, it is assumed that a sufficiently long
period of time has elapsed so that all transients have died out. The imposition of
“chapter2(new)” — 2005/11/22 — page 10 — #6

initial conditions for a complete time-history solution requires that both homo-
geneous and particular solutions be taken together before the values of the
undetermined coefficients in the homogeneous equation can be ascertained.
A lot of practical engineering use can be made of both the homogeneous and
particular solutions separately without resorting to constructions of complete solu-
tions at this point. Problems requiring the detailed construction of such complete
time-history solutions do arise, however, and will be treated in later chapters as
needed.
Steady load. For this case let

0, t<0
F(t) = (2.22)
F0 , t≥0
The particular solution in this case is obtained by neglecting the differentiated

terms in the equation of motion (setting ẍ and ẋ equal to zero):
m(0) + c(0) + kx = F(t) −→ xp = F0 /k (2.23)
which is merely another way of stating a problem in statics: displacement =

force/stiffness.
If it were necessary to construct the entire solution (using appropriate initial
conditions), the approach would be to combine the homogeneous and particular
solutions to give

x(t) = exp(−ζ ωn t) [C1 cos 1 − ζ 2 ωn t

+ C2 sin 1 − ζ 2 ωn t] + F0 /k (2.24)
where, again, C1 and C2 are determined from the initial conditions.
Sinusoidal loads. The particular solution corresponding to sinusoidally

applied loads forms the basis of the very important study of vibration analysis
and is therefore essential to our understanding of vibrational structural dynamics
of rotorcraft. The sinusoidal applied load problem defines what we typically refer
to as a frequency-domain solution or approach. Questions of initial conditions
are disregarded because the motion is assumed to have been going on for an
(mathematically speaking) infinitely long time in the past.
Let
F(t) = P0 cos t = Re[P0 eit ] (2.25)
Because we have assumed that the forcing function has been operating on the
linear spring-mass-damper system for a sufficiently long time, the homogeneous
solution can therefore be assumed to have died out. The resulting response
produced will then also be sinusoidal and at only one frequency: the same frequency
“chapter2(new)” — 2005/11/22 — page 11 — #7

as the applied load frequency . Therefore, the particular solution for this case
will have the following form:
xp = A cos t + B sin t (2.26)
Consequently,
ẋp = (−A sin t + B cos t) (2.27a)
ẍp = −2 (A cos t + B sin t) = −2 xp (2.27b)
Substitution into the basic differential equation and collecting terms yields
[(ωn2 − 2 )A + 2ζ ωn B] cos t

P0
+ [−2ζ ωn A + (ωn2 − 2 )B] sin t = cos t (2.28)
m
Then, upon equating coefficients of like trigonometric functions,
P0
(ωn2 − 2 )A + 2ζ ωn B = (2.29a)
m
−2ζ ωn A + (ωn2 − 2 )B = 0 (2.29b)
These equations then constitute a set of algebraic equations in the coefficients

A and B, the solution of which can be easily accomplished using Cramer’s rule.
The final form for the sinusoidal particular solution can be conveniently expressed
in terms of amplitude and phase:
xp (t) = X cos(t − φ) = X[cos φ cos t + sin φ sin t] (2.30)
Therefore,
A = X cos φ (2.31a)
B = X sin φ (2.31b)
so that
tan φ = B/A = 2ζ ωn /(ωn2 − 2 ) (2.31c)
and

X= A2 + B2 (2.31d)
“chapter2(new)” — 2005/11/22 — page 12 — #8

or, in final useful form:
P0 /m
X= ≡ response amplitude (2.32)
(ωn2 − )2 + (2ζ ω
2
n )
2
2ζ ωn
φ = tan−1 ≡ phase angle (2.33)
ωn2 − 2
Plots of these two basic functions with variations in forcing frequency and
damping are given in Fig. 2.5.
Fig. 2.5 Frequency-response functions for linear single-degree-of-freedom system.
“chapter2(new)” — 2005/11/22 — page 13 — #9

Significant points are as follows:
1) The sharp rise in amplitude response occurs when ≈ ωn is denoted the

resonance condition.
2) At = 0, the response amplitudes for all values of damping are the same
(=1/ωn2 ). This condition is merely the static deflection result:
X 1 P0 P0
= 2 −→ X= = (2.34a)
( P0 /m) ωn mωn2 k
3) As → ∞, all responses attenuate.

4) The range of the phase angle variation is 0 ≤ φ ≤ π .
a) For ζ = 0, φ = 0 or π, depending on whether or not the excitation frequency
is respectively below or above the natural frequency.
b) For = ωn , φ = π/2 for all finite levels of damping (the zero damping case
has the same value in the limit).
c) For all damping levels, the following is true:
For = 0, φ=0
For → ∞, φ=π (2.34b)
The preceding material summarizes the basic material used for simple harmonic
(sinusoidal) motion. The next section generalizes these concepts to the case
wherein the simple single-degree-of-freedom system is excited by multiple
frequencies, which are also integral multiples of a basic frequency. As we shall
see in later sections, this situation is of prime importance to rotorcraft dynamic
vibration problems.
2.2 Fourier Methods

2.2.1 Multiharmonic Responses
In this instance we are concerned with the general (particular) solution to the
single-degree-of-freedom system problem wherein the excitation is an integer
multiple harmonic of the already considered frequency :
F(t) = Fnc cos nt + Fns sin nt n = 1, 2, 3, . . . (2.35)
Again, because the dynamic system is linear, the response will also be at the same
frequency:
xp (t) = An cos nt + Bn sin nt (2.36)
“chapter2(new)” — 2005/11/22 — page 14 — #10

where
Fnc (ωn2 − n2 2 ) − Fns 2ζ ωn n

An = (2.37a)
m[(ωn2 − n2 2 )2 + (2nζ ωn )2 ]
Fns (ωn2 − n2 2 ) + Fnc 2ζ ωn n

Bn = (2.37b)
m[(ωn2 − n2 2 )2 + (2nζ ωn )2 ]
If the excitation were now to consist of the steady load as well as a summation
of such harmonic excitations, wherein n is varied with increasing integer values
(n = 1, 2, 3, . . .), the excitation is then expressible as a Fourier series:
N
F(t) = F0 + [Fnc cos nt + Fns sin nt] (2.38)
n=1
Because of the linearity of the dynamic system, superposition can be invoked so

that the response is also expressible as a Fourier series:
N
xp (t) = A0 + [An cos nt + Bn sin nt] (2.39)
n=1
where the An and Bn coefficients derive from the Fnc and Fns excitation Fourier
coefficients as given earlier.
2.2.2 Discrete Fourier Analysis

The applicability of the preceding development is contingent on the excitation
being periodic and on the ability to obtain the Fourier coefficients of the excitation.
In some cases the periodic function F(t) (with period T ) is known analytically,
in which case the Fourier coefficients can be determined by use of the following
expressions:

2 T 2π t 
Fnc = F(t) cos n dt 


T 0 T 
n = 1, 2, 3, . . . (2.40a)
T 

2 2π t 
Fns = F(t) sin n dt 

T 0 T
1 T
F0 = F(t) dt (2.40b)
T 0
(Note that T = 2π/ .)

For cases where the excitation function is not known analytically, numerical
methods must be employed. One such instance is when the function represents an
“chapter2(new)” — 2005/11/22 — page 15 — #11

experimentally measured value; in such a case the value is available only as an

analog or discretely sampled voltage. Eventually the function becomes available
as a set of discrete numerical (ordinate) values corresponding to a set of equally
spaced points in time within the period T . Assume further that the period is divided
into 2N + 1 points (so that the f0 and f2N values are equal). Furthermore, let the
corresponding (abscissa) values of time be given by
tj = jT /2N(= jπ/N) j = 0, 1, 2, 3, . . . , (2N − 1), 2N (2.41)
where is the characteristic “fundamental” frequency.

The periodic function to be (discretely) Fourier analyzed can then be represented
over one period as shown by Fig. 2.6. Note that the function is really known only
at the f0 , f1 , etc., values.
One straightforward method of obtaining the Fourier coefficients would be to
approximate the integrals defining these coefficients as finite summations. This
procedure could involve any number of approximation techniques (various poly-
nomial fitting functions, etc.) with varying degrees of accuracy. One convenient
method that can be used manually (with a hand calculator) or can be programmed
for use within some computer-driven application is presented herein. The method
is based on the recognition of the existence of even and odd functions, so that the
general function to be Fourier analyzed f (t) can be used to define two auxiliary
functions F(t) and G(t):
F(t) = 21 [ f (t) + f (−t)] (2.42)
G(t) = 21 [ f (t) − f (−t)] (2.43)
Fig. 2.6 Graphical representation of a discretely known periodic function.
“chapter2(new)” — 2005/11/22 — page 16 — #12

so that
f (t) = F(t) + G(t) (2.44)
Thus, for the discretization shown in Fig. 2.6 the following expressions can be
written:
F0 = f0 (2.45a)
F1 = 21 ( f1 + f2N−1 ) G1 = 21 ( f1 − f2N−1 (2.45b,c)
F2 = 21 ( f2 + f2N−2 ) G2 = 21 ( f2 − f2N−2 ) (2.45d,e)
.. ..
. .
FN−1 = 21 ( fN−1 + fN+1 ) GN−1 = 21 ( fN−1 − fN+1 ) (2.45f,g)
FN = fN (2.45h)
Note that, by virtue of Eq. (2.43), G0 and GN are identically zero. The approximate
resulting Fourier series is then expressed as
N−1
f (t) = A0 + [An cos n(2π t/T ) + Bn sin n(2π t/T )]
n=1
+ AN cos N(2π t/T ) (2.46)
The Fourier coefficients in this representation are obtainable from the F and G
functions as follows:

1 1 1
A0 = F0 + F1 + F2 + · · · + FN−1 + FN (2.47)
N 2 2

2 1
An = F0 + F1 cos nx1 + F2 cos nx2 + · · ·
N 2
1
+ FN−1 cos nxN−1 + (−1)n FN (2.48)
2

1 1 1
AN = F0 − F1 + F2 − · · · + (−1)N−1 FN−1 + (−1)N FN (2.49)
N 2 2
2
Bn = [G1 sin nx1 + G2 sin nx2 + · · · + GN−1 sin nxN−1 ] (2.50)
N
BN = 0 (2.51)
“chapter2(new)” — 2005/11/22 — page 17 — #13

where
xi = 2π ti /T (2.52)
An example of the use of approximate formulas is shown here: six harmonics,

that is, f (tm ) values every 30 deg:
A0 = [1/2(F0 + F6 ) + F1 + F2 + F3 + F4 + F5 ]/6 (2.53a)

√
A1 = [F0 + F2 − F4 − F6 + 3(F1 − F5 )]/6 (2.53b)
A2 = [F0 + F1 − F2 − 2F3 − F4 + F5 + F6 ]/6 (2.53c)
A3 = [F0 − 2F2 + 2F4 − F6 ]/6 (2.53d)
A4 = [F0 − F1 − F2 + 2F3 − F4 − F5 + F6 ]/6 (2.53e)

√
A5 = [F0 + F2 − F4 − F6 − 3(F1 − F5 )]/6 (2.53f)
A6 = [1/2(F0 + F6 ) − F1 + F2 − F3 + F4 − F5 ]/6 (2.53g)

√
B1 = [G1 + 2G3 + G5 + 3(G2 + G4 )]/6 (2.54a)
√
B2 = 3[G1 + G2 − G4 − G5 ]/6 (2.54b)
B3 = [G1 − G3 + G5 ]/3 (2.54c)

√
B4 = 3[G1 − G2 + G4 − G5 ]/6 (2.54d)
√
B5 = [G1 + 2G3 + G5 − 3(G2 + G4 )]/6 (2.54e)
2.2.3 Fourier Transform

The development of the previous section can be extended and generalized by
taking the limit as the period T → ∞. In effect this provides a way of trans-
forming arbitrary aperiodic functions into the frequency domain. This procedure
is mathematically accomplished using the Fourier transform, which is defined for
the arbitrary function f (t):
1 +∞
F(ω) = f (t) e−iωt dt (2.55)
2π −∞
The primary usefulness of the Fourier transform is that it provides a very general
method of determining particular solutions for a wide range of (aperiodic) exci-
tations. Some problems for which the excitation cannot be expressed in the time
domain, but rather only in the frequency domain (and, even then, sometimes only
using statistical concepts), are intractable with any solution method other than the
Fourier transform. In particular, assume for the moment that the applied excitation
“chapter2(new)” — 2005/11/22 — page 18 — #14

to the dynamic system F(t) is expressible only in the form of its frequency content
[i.e., F(ω), a Fourier transform of the excitation]. This is often the case with exper-
imentally acquired data wherein the Fourier transforms are produced in software
by the test equipment. The frequency response function (FRF) of the dynamic
system H(ω) is a frequency-domain representation of the sinusoidal output of the
dynamic system caused by a sinusoidal input. For the single-degree-of-freedom
system, H(ω) would be expressed as
P0 /m
H(ω) = (2.56)
ωn2 − ω2 + i2ζ ωn ω
Then, the Fourier-transform method could be used to obtain the Fourier transform
of the product response X(ω):
X(ω) = H(ω)F(ω) (2.57)
Finally, the actual time domain solution x(t) is then obtained using the inverse
Fourier transform:
+∞
x(t) = H(ω)F(ω)eiωt dω (2.58)
−∞
One significant development that makes Fourier methods increasingly more

applicable for solving vibration-related problems is that Fourier transforms can
now be calculated numerically (using discretely sampled data) very quickly using
appropriate software. These software (algorithms) are typically known as fast
Fourier transforms (FFT) and are readily available. Furthermore, the wide avail-
ability of such software makes the use of Fourier transform methods practical in
some applications where other more traditional techniques might appear outwardly
more direct. This tradeoff arises on the basis that the more traditional methodology
might often become less efficient from a CPU resource standpoint. For instance,
the FFT could also be used to obtain discrete Fourier analyses (DFA), as well.
However, care must be taken in the use of the FFT because it operates only on
discretized data and is therefore subject to specific problems such as aliasing and
leakage. These problems, although having a battery of solution techniques devel-
oped to compensate for the inaccuracies introduced, are beyond the intent of the
present chapter and will only be addressed later, as appropriate.
2.3 Linear Two-Degree-of-Freedom System

The next step that must be taken to bring ourselves closer to the realities of
rotorcraft structural dynamics is to increase the number of degrees of freedom
that we can analyze. Initially, we begin with only two degrees of freedom and
then generalize to an arbitrary number. This higher level of complexity is required
because most vibrational and aeroelastic stability issues of rotorcraft involve two or
more degrees of freedom. Virtually all of the additional concepts we must acquire
in order to investigate the multiple-degrees-of-freedom system can be exemplified
with the use of a relatively simple elastomechanical model, as shown in Fig. 2.7.
“chapter2(new)” — 2005/11/22 — page 19 — #15

Fig. 2.7 Basic configuration of a two-degrees-of-freedom dynamic system.
2.3.1 Obtaining the Equations of Motion

Two basic methods can be used to obtain the equations of motion; these are
briefly described and contrasted in the material to follow.
Newtonian approach. The Newtonian approach is a direct extension of the

method used for obtaining the single-degree-of-freedom equations given in the
previous section: Represent the accelerations of the masses as equivalent inertia
forces, and then use static force and moment equilibrium to define the equa-
tions of motion. For the multiple-degrees-of-freedom problem such statements of
equilibrium must be made for each mass and rotational inertia.
Using the spring-mass-damper configuration shown in Fig. 2.7, we then proceed
by summing separately the forces acting on each of the masses. Thus, the forces
on m1 can be written as
k2 (x2 − x1 ) − k1 x1 + c2 (ẋ2 − ẋ1 ) − c1 ẋ1 − m1 ẍ1 = 0 (2.59)
or, more compactly, after rearranging,
m1 ẍ1 + (c1 + c2 )ẋ1 + (k1 + k2 )x1 − c2 ẋ2 − k2 x2 = 0 (2.60)
Similarly, the forces acting on m2 can be summed and equilibrated:
−k2 (x2 − x1 ) − c2 (ẋ2 − ẋ1 ) − m2 ẍ2 + F(t) = 0 (2.61)
or, again, in more compact form:
−c2 ẋ1 − k2 x1 + m2 ẍ2 + c2 ẋ2 + k2 x2 = F(t) (2.62)
Lagrangian approach. In contrast to the Newtonian approach, the approach

using Lagrange’s equation is more elegant and finds application in those cases
wherein the inertial loads cannot be easily formulated by inspection. The
Lagrangian approach has the distinct advantage that the terms in the differential
equations relating to inertia loads can be more easily and accurately identified.
The disadvantage with the method is that, being a variational approach, it is
“chapter2(new)” — 2005/11/22 — page 20 — #16

a step removed from the details of the actual configuration and hence from an
understanding of the physics of the resulting dynamics.
Use of the Lagrangian approach hinges on the concepts of kinetic and potential
energy, T and U, respectively, as defined for the total system. The Lagrangian
approach utilizes a single differential equation that can be used for any dynamic
system, linear or nonlinear, and is applicable to every one of the (M) degrees of
freedom:

d ∂T ∂T ∂U ∂D
− + + = Fi i = 1, 2, . . . , M (2.63)
dt ∂ q̇i ∂qi ∂qi ∂ q̇i
where
T ≡ kinetic energy,
U ≡ potential (strain) energy,
D ≡ dissipation function,
qi ≡ ith generalized coordinate, and
Fi ≡ generalized force for ith generalized coordinate
Although the concepts of kinetic and potential energy should be familiar to

the reader, the concepts of dissipative function and generalized forces need ampli-
fication. The dissipative function is basically constructed using the potential energy
as an analog, wherein the velocities are substituted for displacements and damping
rates are substituted for stiffness rates. The generalized forces must be obtained
from a statement of the total system incremental work δW produced by variations
in the generalized coordinates that is,
M
δW = Fi δqi (2.64)
i=1
For the example shown in Fig. 2.7,
T = 21 (m1 ẋ12 + m2 ẋ22 ) (2.65a)
U = 21 [k1 x12 + k2 (x1 − x2 )2 ] (2.65b)
D = 21 [c1 ẋ12 + c2 (ẋ1 − ẋ2 )2 ] (2.65c)
where the generalized coordinates and generalized forces are simply given as
q1 ≡ x1 ; q2 ≡ x2 ; F1 ≡ 0; F2 ≡ F(t)
It can be readily verified that application of Lagrange’s equation yields equa-

tions of motion identical to those obtained by the Newtonian approach. Note that
a characteristic of these derivations is that the forces resulting from simple iner-
tia, stiffness, and dissipative contributions show up as symmetrical terms in the
equations. This feature will become more evident when we discuss matrices in a
subsequent section.
“chapter2(new)” — 2005/11/22 — page 21 — #17

2.3.2 Characteristic Frequencies and Modes

As in the case with the single-degree-of-freedom system, the solution to a set of
differential equations involves a homogeneous solution and a particular one. For
the type of linear systems that we are treating here, the homogeneous problem is
defined by setting the excitation force F(t) to zero. The solution to the resulting
problem set has an exponential form. Therefore, we assume exponential solutions
for each of the degrees of freedom of the form
x1 = x 1 eλt (2.66a)
x2 = x 2 eλt (2.66b)
With these substitutions in the equations of motion, the equation set then becomes
a set of algebraic equations and, in essence, an eigenvalue problem:
[m1 λ2 + (c1 + c2 )λ + (k1 + k2 )]x 1 − (c2 λ + k2 )x 2 = 0 (2.67a)
− (c2 λ + k2 )x 1 + (m2 λ2 + c2 λ + k2 )x 2 = 0 (2.67b)
We can make the following observations of this set of equations:

1) The eigenvalue problem so defined entails finding a value of the complex
frequency or eigenvalue λ, which leads to nonzero values of the unknown response
amplitudes (eigenfunction or eigenvector), [x̄1 , x̄2 ].
2) For linear equations of this type, the eigenvalue is to be found by setting the
determinant to zero:

[m1 λ2 + (c1 + c2 )λ + (k1 + k2 )] − (c2 λ + k2 )
≡ (λ) = 0 (2.68)
− (c2 λ + k2 ) + (m2 λ2 + c2 λ + k2 )
and solving the resulting polynomial for λ. This polynomial is called the
characteristic equation.
3) Once the various values of λ are known, for each λ a mode shape x̄1 /x̄2 can
be found. That is, only the relation between x̄1 and x̄2 is obtainable; either x̄1 or x̄2
can have any arbitrary value we might select for it. Typically, the selected value is
unity.
Undamped case. For the general damped case the resulting characteristic
equation is a quartic and is typically not solved using simple analytical formulas.
Also, both the eigenvalues and eigenvectors are usually complex valued. However,
the undamped case leads to a quartic polynomial characteristic equation devoid
of odd powers that can then be readily solved using the binomial theorem. Thus,
the characteristic equation (λ) = 0 has two imaginary root pairs given by the
“chapter2(new)” — 2005/11/22 — page 22 — #18

following solutions:
 1/2
2 1/2
 m2 (k1 + k 2 ) + m 1 k2 k1 k2 m2 (k1 + k2 ) + m1 k2 
λ = ±i ± − +
2m1 m2 m1 m2 2m1 m2
(2.69)
The two quantities multiplying (±i) in the preceding equation constitute the
resulting natural frequencies for the dynamic system ω1 and ω2 . The mode shapes
can then be found from either of the system differential equations: If we arbitrarily
set x̄1 = 1, then
[m1 λ2 + (k1 + k2 )] − k2 x 2 = 0 (2.70)
or
m1 λ2 + (k1 + k2 )
x2 = (2.71)
k2
Example: Let us take a simplified version of the aforementioned dynamic system

wherein the masses and stiffnesses are equal:
m1 = m2 = m (2.72a)
k1 = k2 = k (2.72b)
The roots are then given by

√
k/m √
λ = ± iω = ± i √ (3 ± 5)1/2 (2.73)
2
Thus,

ω1 = 0.618 k/m (2.74a)

ω2 = 1.618 k/m (2.74b)
These are then the two (undamped) natural frequencies for this two-degrees-
of-freedom spring-mass system. If we alternatively let x̄2 = 1, then for each of
the two natural frequencies (m = 1, 2),
λ2m + (k/m) √
(m)
x1 = =1− 1
2 3± 5 (2.75)
(k/m)
“chapter2(new)” — 2005/11/22 — page 23 — #19

Fig. 2.8 Natural modes for example two-degrees-of-freedom system.
Thus,
(1)
x 1 = 1 − 0.382 = 0.618 (mode 1) (2.76a)
(2)
x 1 = 1 − 2.618 = −1.618 (mode 2) (2.76b)
These results are given pictorially in Fig. 2.8. This figure depicts the relative
positions the masses have with respect to each other when they are oscillating
freely at the respective natural frequencies. The positions of x̄1 and x̄2 (relative to
each other) are the natural mode shapes.
2.3.3 Orthogonality (Undamped Case)

In this section the very important property of orthogonality is formulated.
This property, a feature of the characteristic (or natural) frequencies and
modes described earlier, provides a means of effectively reducing multiple-
degrees-of-freedom systems to an equivalent system comprised of a multiplicity
of single-degree-of-freedom systems. Furthermore, and as a corollary, it enables
reductions to be made in the number of degrees of freedom initially used to model
any given dynamic system. This is especially important, as we shall find, for reduc-
ing descriptions of continuous systems (which have an infinite number of degrees
of freedom) to a handful of normal modal variables.
Basic proof of orthogonality. The proof offered for the orthogonality of

natural modes is based on the two-degrees-of-freedom configuration defined earlier
in Fig. 2.7. However, the veracity of the orthogonality principle is not limited to
“chapter2(new)” — 2005/11/22 — page 24 — #20

this simple problem, but can easily be generalized to other more complicated, but
linear, dynamic systems. We begin by rewriting the two homogeneous equations
as follows:
−m1 ẍ1 = (k1 + k2 )x1 − k2 x2 (2.77a)
−m2 ẍ2 = −k2 x1 + k2 x2 (2.77b)
For the natural modes these equations then become

(for ω = ω1 )

m1 ω12 x 1 = (k1 + k2 )x 1 − k2 x 2 
(1) (1) (1)
mode 1 (2.78)
(1) (1) (1) 
m2 ω12 x 2 = −k2 x 1 + k2 x 2
(for ω = ω2 )

m1 ω22 x 1 = (k1 + k2 )x 1 − k2 x 2 
(2) (2) (2)
mode 2 (2.79)
(2) (2) (2) 
m2 ω22 x 2 = −k2 x 1 + k2 x 2
Each of the two equations of motions can then be written symbolically (for the ith
equation and pth mode and for n = 2) as
n
( p) ( p)
mi ωp2 x i = Kij x j (2.80)
j=1
and similarly for the rth mode the equations become
n
(r) (r)
mi ωr2 x i = Kij x j (2.81)
j=1
[Note that Kij is a shorthand representation for combinations of k1 and k2 (the j

index) as they appear in each of the two equations (the i index).] If now the first of
(r) ( p)
these equations is multiplied by x̄i and the second by x̄i , and if then summations
are taken over (i) with each equation (separately), the results are as follows:
n n n
(r) ( p) (r) ( p)
ωp2 mi x i x i = Kij x i x j (2.82a)
i=1 i=1 j=1
n n n
( p) (r) ( p) (r)
ωr2 mi x i x i = Kij x i x j (2.82b)
i=1 i=1 j=1
“chapter2(new)” — 2005/11/22 — page 25 — #21

Because for conservative systems Kij = Kji , the sums on the right-hand sides
of these two equations are equal. Therefore, if the two equations are differenced,
the following result is obtained:
n
( p) (r)
(ωr2 − ωp2 ) mi x i x i =0 (2.83)
i=1
Because, for p = r, the first term is clearly nonzero, the second term must be zero.
This result is summarized as follows:
n
( p) (r) 0, r = p
mi x i x i = (2.84)
Mp , r=p
i=1
Examples (using results from the preceding example) follow:

r = 1, p = 1:
= m[(0.618)2 + (1.0)2 ] = 1.382m (2.85a)
r = 1, p = 2 :
= m[(0.618)(−1.618) + 1.0] = 0 (2.85b)
r = 2, p = 1 :
= m[(−1.618)(0.618) + 1.0] = 0 (2.85c)
r = 2, p = 2 :
= m[(−1.618) + (1.0)] = 3.618m (2.85d)
Thus, these results indeed demonstrate that the orthogonality principle is real.
The property is no less real for larger systems and actually proves to be a good test
for the accuracy of modes, as calculated by any given eigenvalue solution scheme.
Furthermore, although the orthogonality principle is explicitly formulated earlier
for undamped systems, the principle is still valid for damped systems (both sym-
metric and unsymmetric terms), but requires a different formulation. The reader
is referred to Hurty and Rubenstein for a discussion of both proportional and non-
proportional damping cases. (The term proportional damping refers to damping
terms in the equations of motion that are proportional to linear combinations of
already existing inertia and/or stiffness terms.)
“chapter2(new)” — 2005/11/22 — page 26 — #22

2.3.4 Decoupling with the Use of Normal Modes

The orthogonality principle proves to be a useful tool for simplifying systems
of multiple degrees of freedom. This can be demonstrated, again, using the two-
degree-of-freedom example given earlier using the following procedure:
1) First, a linear coordinate transformation is made wherein the original degrees
of freedom x1 and x2 are replaced by combinations of two new variables q1 and
q2 , which we denote the generalized coordinates or, more specifically, modal
variables:
(1) (2)
x1 (t) = x 1 q1 (t) + x 1 q2 (t) (2.86a)
x2 (t) = x (1) (2)

2 q1 (t) + x 2 q2 (t) (2.86b)
( p)
where x̄m represents the mth component (degree of freedom) of the pth mode
( p)
shape. [Note that here we have assigned the value of 1 to x̄2 .]
2) This coordinate transformation is then substituted into the differential
equations of motion.
3) For each of the two values of p (1 and 2), the first equation is multiplied by
( p) ( p)
x̄1 and the second equation by x̄2 .
4) For each value of p, this multiplication operation is then followed by a
summation of the two equations. By virtue of the orthogonality principle, each
of the two resulting equations ( p = 1, 2) is then uncoupled from each other to
produce equations of the following form:
Mp q̈p + ωp2 Mp qp = p (t) (2.87)
where
qp (t) ≡ pth modal variable, a scalar quantity denoting how much the pth natural
mode is being excited
Mp ≡ pth generalized mass corresponding to the pth natural mode
ωp ≡ pth natural frequency
p ≡ pth generalized forcing function or excitation, the measure of how
much the pth mode is being excited
Note that a complete solution of the equations is then afforded by solving for the
generalized (modal) variables first and then combining them as per the coordinate
transformation just given. The aforementioned development, although formulated
in terms of only two degrees of freedom, can be easily generalized to any number of
degrees of freedom. Thus, by means of natural modes multiple-degree-of-freedom
systems can be reduced to an ensemble of single-degree-of-freedom problems. For
systems with (viscous) damping similar (but not identical in detail) techniques can
be used to uncouple the modes one from the other.
The procedure just given is an example of a more generalized procedure known
as the Galerkin method, which will be covered in more detail in a later section.
The Galerkin method is a powerful solution technique that can be used not only
for devising modal solution approaches to linear systems, but for treating modal
“chapter2(new)” — 2005/11/22 — page 27 — #23

formulations to nonlinear systems of equations and some classes of nonmodal

solution formulations as well.
2.3.5 Frequency Response

Undamped case. A further important element of the two- (or more) degree-
of-freedom problem is the response of the system to sinusoidal excitations. Similar
to the tack taken with the single-degree-of-freedom system, let the excitation to
the example two-degrees-of-freedom system just defined be given by a sinusoidal
variation:
F(t) = P0 cos t (2.88)
Then, in the absence of damping the two degrees of freedom x1 and x2 can be
similarly defined:
x1 (t) = X1 cos t (2.89a)
x2 (t) = X2 cos t (2.89b)
Following the solution technique given earlier for the single-degree-of-freedom

system, it can be shown that the solutions for X1 and X2 can be written as
P0 k2
X1 = (2.90a)
m1 m2 (ω12 − 2 )(ω22 − 2 )
P0 (k1 + k2 − m1 2 )
X2 = (2.90b)
m1 m2 (ω12 − 2 )(ω22 − 2 )
Plots of the amplitude variations of these two responses are given in Fig. 2.9.
Points to be noted are as follows:
1) Similar to the single-degree-of-freedom results, the system exhibits
resonance conditions (i.e., the responses become unbounded) when the forcing
frequency approaches either one of the two natural frequencies ω1 and ω2 .
2) At a certain frequency ω̂3 , situated between ω1 and ω2 , the response of X2
becomes zero. Such a condition is typically denoted an antiresonance condition.
Physically (at this frequency) the first mass m1 is acting as a “vibration absorber”
to the second mass. The reason that the response of the second mass can be zero
(despite the fact that it is receiving the full brunt of the application of the external
load) is that the first mass, together with the springs, develops an equal and oppo-
site load “in tune” with the forcing frequency . From the equation for X2 , this
frequency is seen to be

ω̂3 = (k1 + k2 )/m1 (2.91)
“chapter2(new)” — 2005/11/22 — page 28 — #24

Fig. 2.9 Variation of amplitude responses of the two-degrees-of-freedom system with

excitation frequency.
From a consideration of the physics of the configuration, this is also the natural
frequency of the first mass operating as a single-degree-of-freedom system with
the second mass constrained to have zero motion, x2 ≡ 0.
3) For excitation frequencies between ω1 and ω̂3 and from ω2 to ∞, the X2
response curve shows negative values. Similarly, the X1 response curve is negative
between ω1 and ω2 . Generally, a negative value is to be interpreted to mean that
the response is 180 deg out of phase with the excitation F(t).
4) Apart from the antiresonance issue, each of the response curves can be
interpreted to be comprised of an ensemble of single-degree-of-freedom-system
response curves overlapping each other. These single-degree-of-freedom systems
are clearly the modal response degrees of freedom treated in a preceding section.
Damped case. The reinstatement of the dampers shown in Fig. 2.7 to
nonzero values would result in a system of four algebraic equations in the four
(real) coefficients needed to describe the response of two degrees of freedom. The
solution cannot be stated easily in a way so as to glean out the physics. However,
resorting to a modal interpretation produces the result that the inclusion of damp-
ing modifies the results given earlier for the undamped case in ways suggested by
the single-degree-of-freedom results:
1) The singular or unbounded results at the resonance conditions disappear to be
replaced by high (but bounded) responses that attenuate with increased values
of damping.
2) The antiresonance condition develops a finite (but still small) response level,
depending on the damping level. At the minimum response level the phase
angle of the second mass would be near 90 deg because the response would be
resisted principally by the damping.
Thus, it can be inferred that the optimal operation of vibration absorbers based
on an antiresonance principle requires a minimization of damping in the system.
“chapter2(new)” — 2005/11/22 — page 29 — #25

2.3.6 General Coupling Issues

One hallmark of the structural dynamic systems encountered in rotorcraft appli-
cations is an abundance of distinct coupling terms. Some examples are presented
in this section to show typical ways in which coupling can be encountered. Con-
sider first a very general example of coupling, as afforded by a rotational mass
supported by unsymmetrical support stiffnesses (see Fig. 2.10).
Let the pertinent dynamic response quantities be the plunge displacement and
angular rotation of the point P, z, and θ, respectively. The equations of this example
can be easily obtained using Lagrange’s equations:
T = 21 m(ż − 1 θ̇)2 + 21 I θ̇ 2 (2.92a)
U = 21 k1 z2 + 21 k2 [z − θ (1 + 2 )]2 (2.92b)
δW = F(t)(δz − 1 δθ) (2.92c)
Therefore,
Fz = F(t) (2.92d)
Fθ = −1 F(t) (2.92e)
The resulting equations of motion then become
mz̈ + (k1 + k2 )z − m1 θ̈ − k2 (1 + 2 )θ = F(t) (2.93a)
−m1 z̈ − k2 (1 + 2 )z + (I + m21 )θ̈ + k2 (1 + 2 )2 θ
= −1 F(t) (2.93b)
Fig. 2.10 Simple example of coupling in a structure.
“chapter2(new)” — 2005/11/22 — page 30 — #26

where the ( ) and ( ) underlined terms represent, respectively, the inertia and
stiffness coupling terms between the two degrees of freedom.
Free-free systems. Often the dynamic system is not rigidly fixed to a sta-
tionary point. In such a case a rigid-body mode appears that has zero frequency and
involves zero spring deflections. Furthermore, for free vibrations the system then
has a center of gravity that does not move in space and hence becomes a system
node point. Consider the example given in Fig. 2.11 of two flywheels connected
dumbbell-like to each other by a flexible shaft:
Let us find the natural frequency and node point for this system. The equations
of motion can be easily determined to be
J1 θ̈1 + Kθ1 − Kθ2 = 0 (2.94a)
−Kθ1 + J2 θ̈2 + Kθ2 = 0 (2.94b)
Assuming a sinusoidal (undamped) homogeneous solution,
θ1 = θ̄1 eiωt (2.95a)
θ2 = θ̄2 eiωt (2.95b)
yields the following simultaneous equations for the modal response values:
(K − ω2 J1 )θ̄1 − K θ̄2 = 0 (2.96a)
−K θ̄1 + (K − ω2 J2 )θ̄2 = 0 (2.96b)
which results in the following characteristic equation:
ω4 J1 J2 − ω2 K(J1 + J2 ) = 0 (2.97)
Fig. 2.11 Two flexible shaft connected flywheels.
“chapter2(new)” — 2005/11/22 — page 31 — #27

whose solutions are

ω1 , ω2 = 0, K(J1 + J2 )/J1 J2 (2.98)
The zero frequency corresponds to a rigid-body mode, whereas the nonzero one
corresponds to the only elastic mode possible. Solving for this elastic mode gives
θ̄2 /θ̄1 = (K − ω2 J1 )/K = −J1 /J2 (2.99)
Because the twist of the shaft is linear,
1 /2 = −θ̄1 /θ̄2 = J2 /J1 (2.100)
then
1 = J2 2 /J1 = J2 ( − 1 )/J1 (2.101)
or
1 / = J2 /(J1 + J2 ) (2.102)
where
= 1 + 2 (2.103)
The same result can be obtained by considering the node point to be built in, as
shown in the following sketch:
Here the natural frequency squared calculation can be made on the basis of a
single-degree-of-freedom system:
ω2 = K1 /J1 = K2 /J2 (2.104)
“chapter2(new)” — 2005/11/22 — page 32 — #28

where
K1 = GJ/1 (2.105a)
K2 = GJ/2 (2.105b)
Then substitution of these spring values yields
GJ/1 J1 = GJ/2 J2 −→ 1 = (J2 /J1 )2 (2.106)
As a final example of a system coupling, consider the more complicated dynamic

system of a spring-mass-damper idealization of a hovering helicopter approxi-
mated as a transmission-fuselage combination (see Fig. 2.12). Three degrees of
freedom are needed to describe this system. One possible combination is lateral
displacement of the gimbal yg and the roll angles of the transmission and fuselage
φt and φf , respectively. To keep the total center of gravity of the system from
translating (in the absence of lateral forces),
mf ( yg + 2 φf ) + mt ( yg − 1 φt ) = 0 (2.107)
which results in a dependence of the lateral translation of the gimbal on the fuselage
and transmission roll angles and, as a consequence, a reduction in the number of
independent degrees of freedom needed to describe the dynamics:
mt 1 φt − mf 2 φf
yg = (2.108)
mf + mt
Fig. 2.12 Idealization of a hovering helicopter with a flexibly coupled transmission

and fuselage.
“chapter2(new)” — 2005/11/22 — page 33 — #29

Using Lagrange’s equation, one can then write the appropriate kinetic and potential
energy relationships:
T = 21 mt ( ẏg − 1 φ̇t )2 + 21 It φ̇t2 (2.109)
U = 21 Kr φr2 + 21 Kg (φt − φf )2 (2.110)
The generalized “forces” appropriate to the transmission and the fuselage Fφt and
Fφf , respectively, are to be found by considering the incremental work accom-
plished by the forces acting on the two masses. In this case these loads are the
gravity loads Wt and Wf , which are equilibrated by the total rotor lift L. Note that,
in this case, because work is accomplished only by virtue of the two rotations the
generalized forces will actually be moments.
δW = (Wt 1 φt − Lt φt ) δφt + (−Wf 2 φf )δφf (2.111)

= Mφt = Mφf
Therefore,
Fφt (= Mφt ) = −(Lt − Wt 1 )φt (2.112a)

Fφf (= Mφf ) = −(Wf 2 )φf (2.112b)
“chapter2(new)” — 2005/11/22 — page 34 — #30

[Note that L (rotor lift) = Wt + Wf .] Substitution of the constraint relationship

into the expression for the kinetic energy yields the following amended form:
mt mf
T= (1 φ̇t + 2 φ̇f )2 + 21 (It φ̇t + If φ̇f )2 (2.113)
(mt + mf )
Application of Lagrange’s equation then yields the following preferred form of the
equations of motion:

mt mf 21
+ It φ̈t + (Kr + Kg + Lt − Wt 1 )φt
mt + mf
mt mf 1 2
+ φ̈f − Kg φf = 0 (2.114)
mt + mf

mt mf 1 2 mt mf 21
φ̈t − Kg φt + + If φ̈f + (Kg + Wf 2 )φf = 0 (2.115)
mt + mf mt + mf
These equations are then similar in form to those of the other two-degree-of-
freedom systems considered earlier in this chapter, wherein both inertia and
stiffness couplings are present.
2.4 Laplace Transform

In Section 2.1 we distinguish between homogeneous and particular solutions.
This distinction is made to serve the dual objectives of exploring stability (the
homogeneous solution) and vibration (the particular solution). However, for many
situations this distinction gets blurred because of the need to define tools that do not
exactly fit in either category. The Laplace transform provides a way of generalizing
solutions to linear differential equations and provides us with additional vocabulary
for describing our solutions. Indeed, we can couch much of the material of the
previous sections in terms of the results of applying the Laplace transform.
2.4.1 Definition
The Laplace transform is similar to the Fourier transform in that it transforms
functions of time into a new variable. In this case the new variable p is complex-
valued. The complete definition of the Laplace transform of the function f (t) is
given by
∞
L [f (t)] ≡ F( p) = f (t)e−pt dt (2.116)
0
“chapter2(new)” — 2005/11/22 — page 35 — #31

where, generally, p can be expressed as = σ + iω. Thus, we will find it useful to

refer to the σ and ω variables that are discussed in Sec. 2.1.1 as variables defined
in the Laplace transform domain. This will become more useful in the study of
instability in subsequent chapters.
Perhaps the most useful characteristic of the Laplace transform is that it trans-
forms differential equations into algebraic ones, thereby rendering them easier
to solve, especially in the case of multiple-degree-of-freedom problems. This is
caused by the transforming of the differentials of the function into factors of the
Laplace transform variable and the transform function:

df
L = pF − f (0) (2.117a)
dt

d2 f df (0)
L = p2 F − pf (0) − (2.117b)
dt 2 dt
where f (0) and df (0)/dt are the initial conditions relating to f (t).
In practice the completion of the solution is to reduce the transformed variable
F( p) into a linear combination of specific forms that have known transform pairs
and then use the pairs to get back into the time domain. Although many such pairs
are available in the literature (e.g., Churchill), four transforms that find a lot of
application in structural dynamics and aeroelasticity are
1
L[1] = (2.118a)
p
1
L[eat ] = (2.118b)
p−a
a
L[sin at] = 2 (2.118c)
p + a2
p
L[cos at] = 2 (2.118d)
p + a2
2.4.2 Reduction to Elementary Forms and

Inverse Transformation
It is typical that when a transformed variable is finally isolated (after using
elementary algebraic techniques) it is in the form of a rational function and most
commonly in the form of a quotient of polynomials:
N( p) Bm pm + Bm−1 .pm−1 + · · · B0
L [f (t)] = F( p) = = (2.119)
D( p) pn + An−1 pn−1 + · · · A0
This form can be reduced to elementary forms by first factoring the denominator
function D( p). This chore amounts to finding the roots of the denominator function
(using any of a variety of algorithms available, see Press et al.). All of the roots of
“chapter2(new)” — 2005/11/22 — page 36 — #32

the denominator polynomial can be placed in the complex plane, which we now
refer to as the Laplace transform domain. Equation (2.119) then becomes
N( p) Bm pm + Bm−1 · pm−1 + · · · B0
F( p) = = (2.120)
D( p) ( p − a1 )( p − a2 ) · · · [( p − σj )2 + ωj2 ] · · ·
A variety of systemized techniques comprise the use of partial fractions

whereby Eq. (2.120) can then be reduced to a summation of terms each having
one of the denominator factors of Eq. (2.120) in the denominator. Each of these
terms can then be inverse transformed using available elementary transform pairs.
A related technique that again requires knowing the roots of the denominator poly-
nomial D( p) (i.e., an ) is the Heaviside expansion technique. With this technique
and the condition that none of the roots of D( p) are repeated, one can inverse
transform Eq. (2.120) directly as follows:
n
−1 −1 N( p) N(aj )
L [F( p)] = L = exp(aj t) (2.121)
D( p) D (aj )
j=1
It is to be understood that if any of the roots are complex, Eq. (2.121) must be
evaluated in complex arithmetic. Also, if roots are repeated, the use of continued
fractions can still be used with appropriate modifications. The complete menu of
techniques available for using partial fractions to inverse transform expressions is
beyond the scope of this text, and the reader is referred to Churchill or an equivalent
for such usage.
2.5 Structural Damping

A mathematical description of structural damping that is both general and accu-
rate does not yet exist. Through experimental studies structural damping has been
found to be both small and a function of a wide range of parameters: material,
method of construction, response level, frequency, temperature, etc. Addition-
ally, although a number of mathematical approaches have been devised to model
structural damping, the more successful of these modelings all too often become
excessively difficult to apply to any practical multiple-degree-of-freedom structural
dynamics problem.
2.5.1 Approximations to Structural Damping

By far the most practical approach found for modeling this kind of damping
has been to represent it either by some form of equivalent viscous damping or by a
complex form of stiffness wherein the imaginary part of the stiffness is proportional
to the energy dissipative structural damping. These approaches are represented
mathematically for a simple single-degree-of-freedom system, as described in the
following.
“chapter2(new)” — 2005/11/22 — page 37 — #33

Equivalent viscous damping. Here we express the damping in a straight-

forward manner as a rate-proportional force:
ẍ + cẋ + ωn2 x = 0 (2.122)
where c̄ = gωn = 2ζeq ωn , g is the structural damping coefficient, and ζeq is the
viscous equivalent critical damping ratio.
Complex stiffness. Here the structural damping coefficient g is used to

approximate the physically observable result that for sinusoidal motion the work
energy loss per cycle is, to first approximation, invariant with frequency. This
characteristic can be modeled as a “complex stiffness”:
ẍ + (1 + ig)ωn2 x = 0 (2.123)
In each case the structural damping is represented by a linear term involving

some constant g or, alternatively, by some structural damping equivalent critical
damping ratio ζeq . This constant must then be assigned a value commensurate
with the energy dissipative properties of the structure at hand. Techniques for
determining this constant experimentally are given in a later subsection.
As stated earlier, an important experimentally observed aspect of structural
damping is that the energy loss per cycle of oscillation is relatively invariant
with respect to frequency. It is for this fact that the complex stiffness form of
structural damping (which does not have any frequency dependence) was for-
mulated. However, such a form of damping presupposes that the motion of the
structure is that of sustained simple harmonic motion: thus, its use in a more gen-
eral transient type of assumed motion is invalidated. It is for this reason that the
equivalent viscous damping form of structural damping is so often used, despite
the fact that it gives an energy loss per cycle of oscillation that is proportional to
frequency. However, despite their shortcomings the two approximations for struc-
tural damping given earlier continue to be used extensively in a wide spectrum
of applications for the reason that the terms are both simple and linear and thus
are easily implemented. Furthermore, within those narrow frequency ranges of
interest these approximations do a reasonably good job.
2.5.2 Application to Multiple-Degree-of-Freedom Systems

A significant advantage of formulating dynamic problems in terms of natu-
ral modes is that each such mode can be treated in large measure as if it were a
single-degree-of-freedom system. The formulations already given for approximate
structural damping, which are presented for single-degree-of-freedom systems,
are thus easily implemented to describe the structural damping of each mode.
However, one problem encountered with such a relatively simple implement-
ation relates to large built-up analyses, involving two or more such natural modes,
which are coupled by explicit mass and/or stiffness forces. However, this approxi-
mate damping resides in the multi-degrees-of-freedom system description in the
form of autogenous damping terms. Such a resultant formulation generally leads
“chapter2(new)” — 2005/11/22 — page 38 — #34

to (coupled) modes that are complex valued rather than real valued (which are
observed) because these damping terms do not generally obey the same ortho-
gonality relationship that the mass and stiffness terms do. However, again because
of the ease of implementation, these formulations are still extensively used in the
analyses of multiple-degrees-of-freedom systems despite such anomalies.
2.5.3 Techniques for Determining Structural

Damping Coefficients
Although there are a number of ways of deducing from experimental data the
values of the coefficients used in the damping approximations, only three are
described herein.
Logarithmic decrement method. This method presupposes that the struc-

ture has been excited in one of its natural modes so that only responses in that
mode are present. Then the excitation is stopped so that the structure undergoes a
damped oscillatory response, again only in that mode. The method requires that
the maximum amplitudes of the damped sinusoidal response at two points in the
response A0 and AN each respectively separated from each other in time by an
integral number N of periods, are recorded and then ratioed. From this ratio the
so-called logarithmic decrement is formed:
δ = (1/N) n(A0 /AN ) (2.124)
Using the results given earlier for the homogeneous solutions for the single degree
of freedom, this logarithmic decrement can be related to the critical damping ratio
ζ , as well as to the structural damping coefficient g:
ζeq = g/2 = δ/2π (2.125)
Resonance curve width method. Contrary to the previous method, this

method presupposes that the structure has been sinusoidally excited at a sufficient
number of frequencies near a resonance point so that a reasonably well-defined
amplitude response curve is available (see Fig. 2.13). From this response curve the
following items are recorded: 1) maximum response amplitude Xmax ; 2) frequency
for maximum response amplitude ω0 ; and 3) frequency difference for which the
response amplitudes on adjacent sides of the maximum response point are both
equal to 0.707 of the maximum response amplitude, ω.
Then, assuming that the frequency difference is small relative to the maximum
response frequency the following relationship can be written:
g= ω/ω0 (2.126)
Radius of curvature method. This method again presupposes that the

structure has been sinusoidally excited near a resonance point. For this method
the complex form of the structural response velocity data (mobility or mechanical
“chapter2(new)” — 2005/11/22 — page 39 — #35

Fig. 2.13 Typical amplitude response curve.
admittance) is used. For a viscously damped single-degree-of-freedom system the

locus of the complex-valued response velocity begins with zero value on the pos-
itive imaginary axis (exactly 90-deg phase angle) and grows in polar amplitude
with increasing (negative) phase angle. At some frequency the velocity amplitude
reaches a maximum value and then diminishes to zero as the frequency is increased
to infinity. This behavior is schematically shown in Fig. 2.14.
It can be shown that for a linear, viscously damped single-degree-of-freedom
system this locus is exactly a circle passing through the origin, with center
located a positive distance on the real axis, and with radius equal to this distance.
Because mechanical admittance has units inverse of those for viscous damping
Fig. 2.14 Locus of complex-valued response velocities.
“chapter2(new)” — 2005/11/22 — page 40 — #36

(i.e., in./lbf-s), the circle, as plotted from the admittance data, has these same units.
At the point of maximum velocity magnitude, the following conditions hold:
1) The frequency for maximum polar magnitude ω̂ is given by
ω̂ = ωn (2.127)
2) The radius (and center location) of the resulting circle R is related to the
equivalent viscous damping coefficient ceq and, hence, to the equivalent critical
damping ratio as follows:
1 1
R= = (2.128a)
2ceq 4ωn ζeq Mn
or
1
ζeq = (2.128b)
4ωn Mn R
where ωn and Mn are, respectively, the natural frequency and generalized mass
for the single-degree-of-freedom system being considered. Note that whereas the
natural frequency is directly available using Eq. (2.127), the generalized mass must
be obtained from some other method.
3) Near the resonance condition wherein the polar amplitude is maximum,
the rate of change of the argument of the locus vector (i.e., phase angle) is also at
a maximum condition. The point of maximum rate of change of phase angle occurs
at frequency ω̃ as given by
1/2
ω̃ = ωn 2 1 − ζeq
2 −1 (2.129)
Application of methods to multiple-degree-of-freedom systems. Suc-

cessful use of each of the three methods already described depends in some way
on the frequency separation of each of the natural frequencies. In the case of
the logarithmic decrement method, it is important to ensure that the structure is
responding mainly in the mode in question. In the case of the resonance curve width
method, compensation must be made to account for the fact that the response of the
structure near resonance of one mode will still include some small contributions
from all of the other off-resonance modes that will affect determining the frequency
width for reduction down to 0.707 of maximum amplitude (see Fig. 2.15).
In the case of the radius of curvature method, each of the various resonance
points (local maximum radii of curvature) will produce a locally circular locus
whose radius should be distinct and relatively unaffected by the other off-resonance
modes (see Fig. 2.16). Furthermore, for each of the circles respectively defining a
resonance mode, if the two excitation frequencies at opposite ends of a diameter
(taken perpendicular to the resonance point radius) ωD and ωC , respectively, are
available they can be used to define equivalent resonance curve widths. Thus
Eq. (2.126) can be used for each resonance condition if ω is taken to be equal to
the difference between ωD and ωC and ω0 is taken to be the ω̂ value (for that circle).
“chapter2(new)” — 2005/11/22 — page 41 — #37

Fig. 2.15 Response amplitude curve for multiple resonances.
Fig. 2.16 Locus of complex-valued response amplitudes for multiple resonances.
2.6 Matrices
2.6.1 Basic Definitions
A matrix is a rectangular array of elements of any size for which specific math-
ematical operations have been prescribed. These elements are usually numbers,
real or complex, whose established mathematical operations define a set of unique
characteristics. Because of these characteristics, matrices prove to be highly useful
in formulating and solving structural dynamics problems. The elements compris-
ing matrices can be quite general and can be functions of one or several variables. A
matrix is generally written as a bracketed quantity [A] or as a boldfaced variable A.
“chapter2(new)” — 2005/11/22 — page 42 — #38

Basic definitions of matrices, both general and specific are given as follows:
 
a11 a12 ... a1n
 a21 ... a2n 
[A] = A = [aij ] = 
 ... ..  (2.130)
. 
am1 ... amn
where aij represents an element in the ith row and jth column. The matrix A in the
preceding equation is said to be of order m × n.
The most frequently encountered matrices are square matrices. Square matrices
are matrices in which the number of rows equals the number of columns. If, for
example, A, the preceding matrix, were square, then m would be equal to n. In that
case the matrix A would be a matrix of the nth order. A few special types of square
matrices are listed in the following.
Diagonal matrix

0, i = j
aij = (2.131)
aii , i = j
Example:
$ %
a11 0 0
[A] = 0 a22 0
0 0 a33
Unit matrix (usually represented by [I ])

0, i = j
aij = (2.132)
1, i = j
Example:
$ %
1 0 0
[A] = 0 1 0 = [I]
0 0 1
Scalar matrix

0, i = j
aij = (2.133)
α, i = j
Example:
$ %
α 0 0
[A] = 0 α 0 = α[I]
0 0 α
“chapter2(new)” — 2005/11/22 — page 43 — #39

Symmetrical matrix
aij = aji (2.134)
Example:
$ %
3 7 1
[A] = 7 2 6
1 6 5
Skew-symmetrical matrix
aij = −aji (2.135)
Example:
$ %
0 1 −3
[A] = −1 0 2
3 −2 0
Note that the diagonal elements of the skew-symmetric matrix must necessarily
be zero because aij = −aji only for a value of 0.
Null matrix
aij = 0 (2.136)
Example:
$ %
0 0 0
[A] = 0 0 0
0 0 0
If a matrix consists only of a single row, it is termed a row vector, has order of
1 × n, and is denoted as
[A] = A = a1 a2 a3 · · · an (order = 1 × n) (2.137)
If a matrix consists only of a single column, it is termed a column vector and

denoted as
 
 a1 

a2 
[A] = {A} = .. (order = n × 1) (2.138)

 
. 
an
“chapter2(new)” — 2005/11/22 — page 44 — #40

Transpose of a matrix. The transpose of a matrix is a matrix whose rows

and columns have been interchanged:
$ %
a11 a12 a13
[A] = [aij ] = a21 a22 a23 (2.139a)
a31 a32 a33
$ %
a11 a21 a31
[A] = transpose of [A] = [aji ] = a12
T
a22 a32 (2.139b)
a13 a23 a33
Note that, if A is symmetric, A = AT .
Equivalence of matrices. Two matrices are equivalent if 1) they are of equal

order, and 2) aij = bij (all corresponding elements are equal).
2.6.2 Mathematical Operations

Matrix addition. If
[A] = [aij ]m,n (2.140a)
and
[B] = [bij ]m,n (2.140b)
then
[A] + [B] = [C] = [Cij ]m,n (2.140c)
where
cij = aij + bij (2.140d)
The commutative and associative laws of algebra hold true.
Matrix subtraction
[A] − [B] = [D] = [dij ]m,n (2.141a)
where
dij = aij − bij (2.141b)
“chapter2(new)” — 2005/11/22 — page 45 — #41

Matrix multiplication
Scalar multiplication
α[A] = α[aij ] = [αaij ] (2.142)
where α is a scalar.
To multiply a matrix by a scalar, all elements of the matrix must be multiplied
by the same scalar.
The basic properties are as follows:
1) (α + β)[A] = α[A] + β[A] (2.143a)

2) ([A] + [B])α = α[A] + α[B] (2.143b)
3) α(β[A]) = (αβ)[A] (2.143c)
where α and β are scalars.
Hadamard multiplication. This form of matrix multiplication is essentially

an element-by-element multiplication:
Given
[A] = [aij ] and [B] = [bij ] (2.144a)
then the Hadamard matrix product, denoted by ⊗, is given by
[A] ⊗ [B] = [C] (2.144b)
where
cij = aij bij (2.144c)
Conventional multiplication of two matrices. Given
[A] = [aij ]m,n and [B] = [bij ]p,q (2.145a)
then, for the product AB to exist, p = n (number of columns in A = number of

rows in B); that is, A and B must be conformable:
[A][B] = [aij ]m,n [bij ]n,q = [cij ]m,q = [C] (2.145b)

n
Cij = aik bkj (2.145c)
k=1
“chapter2(new)” — 2005/11/22 — page 46 — #42

that is,
n n
c11 = a1k bk1 , c12 = a1k bk2 , etc. (2.145d)
k=1 k=1
  b11 b12 ···

a11 a12 a13 · · · a1n
a21  b21 ··· 
 a22 a23 ···   = [C]
..  ... 
.
bn1
 
(a11 b11 + a12 b21 + · · · + a1n bn1 ) · · ·
 
= (a21 b11 + a22 b21 + · · · + a2n bn1 ) · · · (2.145e)
..
.
Example:
(2 × 3) (3 × 2) (2 × 2)
$ %
1 0
2 1 −1 7 −1
3 −1 =
0 −3 1 −11 3
−2 0
Linear transformations using matrix multiplications. Given the quantities

(x1 , y1 ) and (x2 , y2 ), related by
x1 = a11 x2 + a12 y2 (2.146a)

y1 = a21 x2 + a22 y2 (2.146b)
In matrix representation this relationship can be expressed as

x1 a a12 x2
= 11 (2.147)
y1 a21 a22 y2
Then let the quantities (x2 , y2 ) and (x3 , y3 ) be related by
x2 = b11 x3 + b12 y3 (2.148a)

y2 = b21 x3 + b22 y3 (2.148b)
Or, again in matrix representation,

x2 b b12 x3
= 11 (2.149)
y2 b21 b22 y3
“chapter2(new)” — 2005/11/22 — page 47 — #43

Using simple substitutions of the matrix representations yields

x1 a11 a12 b11 b12 x3
= (2.150)
y1 a21 a22 b21 b22 y3
In algebra, if ab = 0 then either or both a, b = 0, but in matrix algebra

AB = 0 does not imply that either A or B = 0.
Example:
$ % $ %
3 2 0 0 0 0
[A] = 0 9 0 = [0], [B] = 0 0 0 = [0]
0 1 0 7 14 100
but for this case [A][B] = 0.
Furthermore, if [A][B] = [C], then it is not generally true that [B][A] = [C].
Stated another way, [A][B] = [B][A], unless one of them is a unit, diagonal, or
inverse matrix.
Example:

2 −1 1 4 3 7
=
−1 2 −1 1 −3 −2

1 4 2 −1 −2 7
=
−1 1 −1 2 −3 3
Pre- and postmultiplication. Given that [A][B] = [C], then [B] is said to be
premultiplied by [A], and [A] is said to be postmultiplied by [B].
Continued product. Provided that they are conformable, matrix products can
be cascaded:
[A][B][C][D] . . . [N] = [R] (2.151a)
[A] [B] [C] [D] . . . [N] = [R]
(2.151b)
m×n n×p p×q q × ··· ··· × y m×y
Also, matrix multiplication is associative:
([A][B])[C] = [A]([B][C]) (2.152)
In algebra,

ab = c
b=d (2.153a)
ad = c
With matrices,

[A][B] = [C]
[B] = [D] (2.153b)
[A][D] = [C]
“chapter2(new)” — 2005/11/22 — page 48 — #44

Vector products. Given two vectors X and Y,

 


y1 


 
 2
y  n
[x1 x2 x3 . . . xn ] 3y

 .  = [A] = xk yk = scalar (2.154)
 .. 
 
 
1×n
 k=1
yn
n×1
 
 y1   

 
 x1 y1 x2 y1 ··· x n y1
y
  
2
x1 y2 x2 y2 ··· 
y3 [x1 x2 x3 . . . xn ]  
 . = x1 y3
 .
 (2.155)

 ..   .. 
  . 
1×n ..
 
yn
x 1 yn · · · x n yn
n×1
Transposition of a matrix product. Given
[D] = [A]m×n [B]n×p [C]p×q (2.156a)
then
[D]T = [C]Tq×p [B]Tp×n [A]Tn×m (2.156b)
Determinants. If [A] is any matrix, then the determinant of [A] exists only if
[A] is square. Det[A] = |A| implies a specific operation on the elements of [A]. A
minor of any element aij is the determinant of the matrix [A] wherein the ith row
and the jth column have been deleted:
a
11 a12 a13 · · · a1j · · · a1n
a
21 a22 a23 · · · a2j · · · a2n
· · · · ·

(minor), Mij ≡ · · · · · (2.157)
ai1 ai2 ai3 aij ain

· · · · ·
a
n1 an2 an3 · · · anj · · · ann
For any element aij whose minor is Mij , there exists a cofactor defined by
(cofactor of aij ), Cij ≡ (−1)i+j Mij (2.158)
According to Laplace’s expansion for determinant evaluation, the value of the

determinant can be found from the following:
n n
Det[A] = aij Cij = aki Cki (2.159)
j=1 k=1
where i can be any value.
“chapter2(new)” — 2005/11/22 — page 49 — #45

2.6.3 Systems of Simultaneous Equations and Matrix Inverses

Cramer’s rule: solution of simultaneous equations. Given a set of (linear)
simultaneous equations,
a11 x1 + a12 x2 + · · · · +a1n xn = y1

.. ..
. . (2.160)
an1 x1 + an2 x2 + · · · · +ann xn = yn
The solution is
|ai |
xi = (2.161)
|a|
where |ai | is the determinant of the coefficients with the ith column replaced by
the column of constants yj and |a| is the determinant of the coefficients.
Adjoint of a matrix. The adjoint of a matrix is defined as the transpose of the

cofactors:
adjoint [A] = [Cij ]T (2.162)
where Cij are the cofactors of aij . Note that every square matrix has an adjoint.
Properties of the adjoint matrix are as follows:
1) [A] (adj[A]) = (adj[A])[A] = |A|[I] (commutative)
2) |adj[A]| = |A|n−1
[Proof: From property a, (adj[A])[A] = |A|[I]; therefore, |adj[A]| |A| = ||A| I| =
|A|n . Thence, a division by |A| produces the desired result.]
Inverse of a matrix. The inverse of the matrix A, written as A−1 , is the

result of operations analogous to algebraic division in that the product of matrices
A and A−1 is the unit matrix I:
[A][A]−1 = [I] (2.163)
where
[A]−1 = adj[A]/|A| for |A| = 0 (2.164)
“chapter2(new)” — 2005/11/22 — page 50 — #46

Note that 1) A is singular if |A| = 0, 2) A is nonsingular if |A| = 0, and

3) therefore, a singular matrix has no inverse.
Inverse of a matrix product. Similar to the transposition of a matrix product,

the inverse of a matrix product manifests a transposition of the matrix ordering.
Given
[D] = [A][B][C] (2.165)
then
[D]−1 = [C]−1 [B]−1 [A]−1 (2.166)
Solution of simultaneous equations using the matrix inverse. The set of

simultaneous equations already considered can be written as
    
a11 a12 · · · a1n x1   y1 

 
 
 

a21 a22 ···  x  
 2
y
a  x 
2 
 31  3 = [A]{X} = y3 = {Y } (2.167)
 .  .   .. 
 ..  .  
.
 



.


an1 · · · ann xn yn
The solution can then be conceptually expressed as
{X} = [A]−1 {Y } (2.168)
Alternate methodology for efficient solutions. Except in the case of very

simple systems of equations (i.e., 2 × 2 equation sets or 2 × 2 matrix inversions),
Cramer’s rule or the formal definition given by Eq. (2.164) for matrix inversion
is inefficient or impractical. Press et al. present an exhaustive collection of alter-
nate methods together with tested FORTRAN coding of the more efficient of the
methods. Also, except for those cases wherein simultaneous equation solution
vectors are to be obtained (corresponding to multiple right-hand vectors) all for
the same coefficient matrix A, the use of the matrix inverse is a highly inefficient
method for obtaining solutions for simultaneous equation sets. For single equation
solutions, the algorithm to be used should always be one expressly for simultaneous
equations.
2.6.4 Matrix Partitioning

In many matrix applications certain regularities are obtained wherein portions
of a given matrix M can be treated as separate submatrices within the confines of
M. Examples of this situation would be where the submatrices are symmetric, or
better yet, diagonal, or where an inverse can be readily defined. In such instances
matrix algebra can be used to formulate a more efficient manipulation of the given
“chapter2(new)” — 2005/11/22 — page 51 — #47

matrix. To this end, let us consider the given matrix M to be of dimension n + m

and to be partitioned into four submatrices, A, B, C, and D:

[A]nn [B]nm
[M] = (2.169)
[C]mn [D]mm
where [A]nn is an n-dimensional square matrix, [B]nm is a rectangular matrix

having n rows and m columns, etc. Then the M matrix can be expanded to a series
of matrix multiplications:

[A] [0] [I] [0] [I] [A]−1 [B]
[M] = (2.170)
[0] [I] [C] [I] [0] [E]
where
[E] = [D] − [C][A]−1 [B] (2.171)
It can then be verified that this expansion allows for the following inverse for the
M matrix to be written:

−1 [I] −[A]−1 [B] [I] [0] [I] [0] [A]−1 [0]
[M] =
[0] [I] [0] [E]−1 −[C] [I] [0] [I]
(2.172)
Thus, the total inverse requires actual inverses to be calculated only of the A and
E matrices, each of which is of smaller dimension than M, thereby resulting in
savings in CPU time. Furthermore, if several inverses of M are required, wherein
only one or the other of the A and E matrices changes with each such inverse, then
the inverse of that submatrix need not be repeatedly calculated, resulting in still
further CPU reductions.
2.7 Vector Calculus

2.7.1 Definitions of a Vector
A vector can be defined in two ways: 1) as the entity embedded within some
physically identifiable concept that has a directional sense and 2) as a general
mathematical identity that obeys certain mathematical rules. It is the beauty of
vector calculus that both of these ideas can be so consistently combined. In the
material to follow, vector quantities are denoted either in matrix notation as given
earlier, or as boldface, italicized quantities.
From the physical standpoint, a vector is a mathematical quantity that possesses
both magnitude and direction. The concept of a physical vector can be visualized
as being embedded within some identifiable physical quantity such as position or
velocity. For example, the position vector R (= ROP , from point O to point P) can
be sketched using components in three directions.
“chapter2(new)” — 2005/11/22 — page 52 — #48

From the mathematical standpoint a vector can be considered to be an

n-tuple ensemble of coordinate numbers satisfying some transformation law in an
n-dimensional space. For example, using matrix notation,
 


r1 


 
 2
r 
R= 3 r (1 × n vector) (2.173)

 .. 

.
 
 
rn
These two concepts can be integrated to represent vectors in three-dimensional
space. Let i, j, and k be mutually orthogonal vectors of unit length in the x, y, and
z directions, respectively. Then

rx
R = rx i + ry j + rz k = [i j k] ry (2.174)
rz
The unit vectors in the x, y, and z directions, i, j, and k, respectively, can then be
denoted as

1 0 0
{i}, { j}, {k} = 0 , 1 , 0 (2.175)
0 0 1
2.7.2 Mathematical Operations

Addition (and subtraction). Given two vectors x and y, which are generally
not codirectional, then
z = Rac = Rab + Rbc = x + y (2.176)
or

z1 x1 y1 x1 + y1
z2 = x2 + y2 = x2 + y2 (2.177)
z3 x3 y3 x3 + y3
“chapter2(new)” — 2005/11/22 — page 53 — #49

Unit vectors. Any vector can be represented as a product of its (scalar)

magnitude and a unit vector defining the vector’s direction. Thus,
R = R1R (2.178)
where the magnitudes of the R and 1R vectors are respectively given by

R = |R| = r12 + r22 + r32 (2.179a)
|1R | = 1 (2.179b)
Thus,
1R = R/R (2.180)
Scalar multiplication (dot product). Two vectors can be multiplied in the

sense that the scalar (scalar) projection of one vector on the other can be multiplied
by the second vector’s scalar value. This is illustrated as follows between vectors
B and C:
1) Such a mathematical operation is denoted the dot product and is mathe-

matically defined as follows:
B · C = BC cos θ (2.181)
The result is a scalar quantity.

2) The dot products of the various unit vectors are as follows:
i·i =j·j =k·k =1 (2.182a)

i·j =j·k =k·i =0 (2.182b)
3) Vector algebra: In addition to the mathematical definition of the dot product

(which can also be used to evaluate the product when the magnitudes of the vectors
“chapter2(new)” — 2005/11/22 — page 54 — #50

and the angle between them θ are known), the following definition forms the basis
for evaluating the product when the components of the respective vectors are
known:
B · C = (bx i + by j + bz k) (cx i + cy j + cz k)
= bx cx + by cy + bz cz (2.183)
4) Matrix algebra representation: The dot product can be presented as a matrix

multiplication:

cx
BC = B · C = [bx by bz ] cy
cz
= bx cx + by cy + bz cz (2.184)
Vector multiplication (cross product). Two vectors can be multiplied in the

sense that the result is itself a vector that follows the “right-hand rule” between
vectors B, C thus defining the cross product vector product (B × C). This is
illustrated as follows between vectors B and C:
1) The cross product is mathematically defined as follows:
B × C = −C × B = 1n BC sin θ (2.185)
The result is a vector quantity with the unit vector 1n being perpendicular to both
the B and C vectors.
2) The cross products of the various unit vectors are as follows:
i×i =j×j =k×k =0 (2.186a)

i × j = k; j × k = i; k×i =j (2.186b)
“chapter2(new)” — 2005/11/22 — page 55 — #51

3) Vector algebra: In addition to the mathematical definition of the cross product

(which as in the case of the dot product can also be used to evaluate the product when
the magnitudes of the vectors and the angle between them θ are known), the fol-
lowing definition forms the basis for evaluating the product when the components
of the respective vectors are known:

i j k

B × C = bx by bz (2.187)
c cy cz
x
4) Matrix algebra representation: From the preceding definition it can then

be verified that the cross product can be represented as a matrix multiplication:
$ %
0 −bz by cx
{B × C} = bz 0 −bx cy (2.188)
−by bx 0 cz
Triple products. Because the vector cross product is itself a vector, it can
then be used in vector multiplications with other vectors. Two such products are
then defined: a scalar (dot) triple product and a vector (cross) triple product.
1) The scalar triple product is the dot product of one vector with the result of
a vector cross product of two other vectors:
A · (B × C) = (A × B) · C = B · (C × A)

ax ay az

= [A B C] = bx by bz (2.189)
c c c
x y z
2) Similarly, the vector triple product is the cross product of one vector with
the result of a vector cross product of two other vectors:
A × (B × C) = (A · C)B − (A · B)C (2.190)
Vector derivatives. Because vectors inherently represent the product of two

mathematical quantities (i.e., magnitude and direction), differentiation of vectors
must account for both of these factors:
R = R1R (2.191)
Therefore,
dR dR d1R
Ṙ = = 1R + R (2.192)
dt dt
dt

magnitude direction
change change
The differentiation problem then becomes one of determining an expression for

the time derivative of the unit direction vector. Consider the following geometric
“chapter2(new)” — 2005/11/22 — page 56 — #52

construction wherein this vector is allowed to change direction (i.e., undergo a

rotation dθ in the infinitesimal time interval dt):
By vector addition we have

(1R )t+dt = (1R )t + d1R (2.193)
Furthermore, all three of these vectors must be coplanar. As shown earlier, let
us define a vector perpendicular to this plane that has as its magnitude dθ . For
infinitesimal motion the magnitude of the differential vector d1R is therefore equal
to (1) dθ .
Thus,

dθ d1R
dθ = dt = ωR dt = |d1R | = dt (2.194)
dt dt
The rotational speed ωR can then be made into a vector by giving it an appropriate
unit vector normal to both the unit vector in the R direction (i.e., 1R ) and the d1R
vector. This is conveniently accomplished using the cross product:
d1R
ωR = 1R × (2.195)
dt
Then, crossing this vector with the unit R vector 1R produces
d1R d1R d1R

ωR × 1R = 1R × × 1R = −1R × 1R × = (2.196)
dt dt dt
With this expression for the derivative of the unit R vector, the derivation for the
time derivative of the R vector itself can be written:
dR dR
= 1R + ω R × R (2.197)
dt dt
This equation is the basic tool that we can now use to bridge the gap between vector
algebra and vector calculus. It shows that an intrinsic relationship exists between
the differential of a vector and its orientation. A more complete understanding of
the vector differentiation process is afforded by considering rotations of vectors,
generally, as developed in the following section.
“chapter2(new)” — 2005/11/22 — page 57 — #53

2.7.3 Coordinate Transformations

A fundamental concept for formulating the mathematics of coordinate system
transformations is that of Euler angles. Such angles are defined as the angles
required to orientate one orthogonal coordinate frame relative to another by making
three sequential rotations of one frame relative to the other about axes in the rotated
frames. Note that the order in which these rotations are made is crucial when the
angles are no longer subject to the small-angle assumption. For example, consider
the components of a vector, as measured in the { }1 coordinate system, upon
successive rotations being performed about the x1 , y2 , and z3 axes, so that the
vector is then measured in the final { }4 coordinate system (see Fig. 2.17).
Using matrix techniques developed earlier, the various coordinate system
transformations sequentially defined in the figure can be written as
1) rotation about the coincidental x1 , x2 axis by the angle α:
$ %
x2 1 0 0 x1
y2 = 0 cos α sin α y1 (2.198)
z2 0 − sin α cos α z1
2) rotation about the coincidental y2 , y3 axis by the angle β:
$ %
x3 cos β 0 − sin β x2
y3 = 0 1 0 y2 (2.199)
z3 sin β 0 cos β z2
3) rotation about the coincidental z3 , z4 axis by the angle γ :
$ %
x4 cos γ sin γ 0 x3
y4 = − sin γ cos γ 0 y3 (2.200)
z4 0 0 1 z3
Fig. 2.17 Euler angles defining one set of three sequential coordinate transformations.
“chapter2(new)” — 2005/11/22 — page 58 — #54

Cascade multiplication then yields the following combined transformation matrix

equation:
 
cos γ cos β [sin α sin β cos γ [− cos α sin β cos γ

x4  + cos α sin γ ] + sin α sin γ ]  x1
 
y4 = − sin γ cos β [− sin α sin β sin γ [cos α sin β sin γ  y1
z4  + cos α cos γ ] + sin α cos γ ]  z1
sin β − sin α cos β cos α cos β
(2.201)
The sequence of angle rotations selected is not unique; there are 12 sets of Euler
angles depending on the order in which they are taken: three axes available for
the first rotation, two axes available for the second rotation, and two available for
the third rotation. Now let the angles α, β, and γ be “small,” and, further, let us
redefine these angles to be explicit rotations about each of the three principal axes:
α = θx (2.202a)
β = θy (2.202b)
γ = θz (2.202c)
The aforementioned coordinate system matrix then becomes

$ %
x4 1 θz −θy x1
y4 = −θz 1 θx y1 (2.203)
z4 θy −θx 1 z1
[Note that if we had selected any one of the other five Euler angle sets involving
x, y, and t axes, the same coordinate system matrix equation for small angles, that
is, Eq. (2.203), would still have resulted.]
Then, after making the following shifts to vector notation,

x4
R4 = y4 (2.204a)
z4

x1
R1 = y1 (2.204b)
z1

θx
θ = θy (2.204c)
θz
the following vector equation can be written:
R4 = R1 − θ × R1 (2.205)
This important vector equation forms the basis of the following subsection.
“chapter2(new)” — 2005/11/22 — page 59 — #55

2.8 Theorem of Coriolis

This theorem provides us with a powerful tool for obtaining a derivative of
a vector whose components are measured in one coordinate system (such as a
moving and/or rotating one) with respect to another coordinate system (such as
a nonmoving and nonrotating inertial one). Drawing on the development of the
preceding section, let us define the four and one coordinate systems as follows: Let
the vector [x1 , y1 , z1 ] denote the basis of a stationary (inertial) coordinate system
(subscript S), and [x4 , y4 , z4 ] denote the basis of a variable (moving and rotating)
coordinate system (subscript V ). With these ideas in mind, we can then let R1 be
the vector R, as measured in the stationary coordinate frame, and R4 be the vector
R, as measured in the variable coordinate frame.
The vector equation derived in the preceding subsection can then be drawn upon
wherein at t = t1 , θ = 0, and at t = t1 + δt, θ = δθ . Then the equation evaluated
at t = t1 is subtracted from the equation evaluated at t = t1 + δt and divided by δt:
R4 (t1 + δt) − R4 (t1 ) R1 (t1 + δt) − R1 (t1 ) δθ × R
= − (2.206)
δt δt δt
Upon taking the limit as δt → 0,

d d
R= R − ωSV × R (2.207)
dt V dt S

pV pS
where pv is the derivative with respect to time, as measured in the variable or

rotating coordinate frame; and ps is the derivative with respect to time, as measured
in the stationary or nonmoving coordinate frame.
The preceding equation can then be rearranged into the following compact form:
pS R = pV R + ωSV × R (2.208)
where ωSV is the rotational velocity vector of the variable coordinate frame rela-
tive to the stationary one. This equation is the statement of Coriolis’ theorem for
coincident origins. As such, it defines the inertial velocity in terms of velocities
measured in the variable coordinate frame.
A similar equation can be derived for the inertial acceleration of the velocities
and acceleration measured in the variable coordinate frame:
pS ( pS R) = p2S R = p2V R + 2ωSV × pV R + pV ωSV × R

+ ωSV × (ωSV × R) (2.209)
An example of the use of Coriolis’ theorem is the derivation of Newton’s law

for an Earth-fixed reference frame. Newton’s law for an inertial frame is stated in
vector notation as
mp21 R = mG
+ F
(2.210)

gravitational external force
field intensity applied to mass, m
“chapter2(new)” — 2005/11/22 — page 60 — #56

Let us now define gravity to include the gravitational field intensity and the
centrifugal force caused by Earth’s rotation:
g = G − ωie × (ωie × R) (2.211)
where ωie is the Earth’s rotational velocity vector relative to inertial space and
measured in the Earth-fixed coordinate system. Using the equation of Coriolis
for inertial coordinate system accelerations, we can formulate an adjusted version
of Newton’s law wherein the accelerations are now measured in the Earth-fixed
coordinate system:
mp2e R = mg + F − 2mωie × pe R (2.212)
This equation thus relates the forces acting upon a mass (gravity mg and some
other external applied force F ) to the acceleration and velocity measured relative
to the Earth.
References
Section
2.1–2.3 Ewins, D. J., Modal Testing: Theory and Practice, Research Studies
Press, Herts, England, U.K. 1984.
Gaukroger, D. R., Skingle, C. W., and Heron, K. H., “Numerical
Analysis of Vector Response Loci,” Journal of Sound and Vibration,
Vol. 29, No. 3, 1973, pp. 341–353.
Hurty, W. C., and Rubinstein, M. F., Dynamics of Structures, Prentice–
Hall, Englewood Cliffs, NJ, 1964.
Kennedy, C. C., and Pancu, C. D. P., “Use of Vectors in Vibration
Measurement and Analysis,” Journal of the Aeronautical Sciences,
Vol. 14, No. 11, 1947, pp. 603–625.
Scanlan, R. H., and Rosenbaum, R., Introduction to the Study of
Aircraft Vibration and Flutter, Macmillan, New York, 1951.
Tong, K. N., Theory of Mechanical Vibrations, Wiley, NewYork, 1960.
2.4 Churchill, R. V., Operational Mathematics, McGraw-Hill, New York,

1958.
2.6 Frazer, R. A., Duncan, W. T., and Collar, A. R., Elementary Matrices,
Cambridge Univ. Press, Cambridge, England, 1950.
Press, W. H., Flannery, B. P., Teukolsky, S. A., and Vetterling, W. T.,
Numerical Recipies in FORTRAN, the Art of Scientific Computing,
2nd ed., Cambridge Univ. Press, New York, 1992.
2.7, 2.8 Wrigley, W., Hollister, W. M., and Denhard, W. G., Gyroscopic Theory,
Design and Instrumentation, MIT Press, Cambridge, MA, 1969.
“chapter2(new)” — 2005/11/22 — page 61 — #57

Problems
2.1 Consider the following differential equation for a single-degree-of-freedom
response variable X(t):
Ẍ + 25X = F(t)
where F(t) is alternatively taken to be the following:
a) F(t) = cos 5t
b) F(t) = 0.1Ẋ
c) F(t) = −0.1Ẋ
d) F(t) = cos 5t − 0.1Ẋ
and all cases have the same initial conditions on X and Ẋ as follows:
X(0) = 0.0; Ẋ(0) = 1.0
For each of the alternate excitations, sketch the solution as a function of

time, noting the qualitative differences between each solution and what each
represents.
2.2 A helicopter tail boom is subjected to periodic pressure fluctuations caused

by passage of the overhead main rotor blades. Consider the following
idealization of this interactional system:
“chapter2(new)” — 2005/11/22 — page 62 — #58

Find
(a) the undamped natural frequency, damping ratio, and fundamental
forcing frequency;
(b) the zeroth and first harmonic Fourier coefficients of F(t) using exact
and approximate numerical methods; and
(c) the zeroth and first harmonic amplitude responses of x(t) using the
exact Fourier coefficients for F(t).
2.3 The following sketch depicts a typical two-dimensional airfoil section:
The degrees of freedom are the plunge displacement z and the rotation angle θ
about the elastic center of the airfoil section. Also, let the section be quantified
by the following numerical values for the parameters:
m = 0.1 lb-s2 /in.; I = 10 lb-s2 -in.

xi = 2 in.; Kz = 400 lb/in.; Kθ = 5000 lb-in./rad
“chapter2(new)” — 2005/11/22 — page 63 — #59

Find
(a) the equations of motion for free oscillations;
(b) the natural frequencies and mode shapes; and
(c) the nodes (points with zero displacement) for each mode.
(Note: These points need not be located within the airfoil section.)
2.4 Derive the expressions used for each of the three methods of determining the
structural damping coefficients from experimental data.
2.5 For the radius of curvature method, derive the appropriate damping coeffi-
cient for the case of a complex stiffness representation for structural damping.
2.6 Derive the Euler angle transformation matrix relating an aircraft’s body axis
system {x, y, z}4 to the Earth-fixed coordinate system {x, y, z}1 (horizontal
plane with a normal vertical direction). The aircraft body-axis system is
defined by first, a heading angle ψ, as measured by compass heading; second,
a pitch attitude angle θ, that is, the angle the aircraft centerline makes with
the horizontal; and third, the aircraft roll angle φ, as measured about the
aircraft centerline from a wings-level position.
2.7 Consider a point P on a rotor blade with built-in coning β, rotating at a

constant rotational speed :
where X, Y , Z denote the inertial frame and x, y, z denote the rotating frame.
The components of R are

rx cos (β0 + β) 1 − (β0 β)
ry = r sin ≈r
rz sin (β0 + β) β0 + β
“chapter2(new)” — 2005/11/22 — page 64 — #60

Find the components of (linearized) inertial acceleration of point P as

measured in the rotating {x, y, z} coordinate system:

rx
p2I ry
rz
where
d
pI = pS ; pV =
dt
β0 = const; = const
2.8 Part 1. Show that the equations of motion (for small oscillations) of a freely
swinging pendulum can be written as
ẍ + (g/r)x + 2ωie sin Lẏ = 0

−2ωie sin L ẋ + ÿ + (g/r)y = 0
where
x, y ≡ north and east displacements of the pendulum bob

ωie ≡ magnitude of the Earth’s daily angular velocity
L ≡ Earth latitude of the pendulum location
g ≡ gravitational acceleration at pendulum location
r ≡ length of pendulum (distance from pivot to bob)
Part 2. Solve for the ( finite) frequencies of oscillation that exist. This is the
Foucault pendulum effect. [Remember that ωie is generally small compared
with the square root of (g/r).]
ωie = 0.0000728 rad/s; L = 34◦ N; r = 50 ft
“chapter2(new)” — 2005/11/22 — page 65 — #61

3
Rotating Beams
Because of the aerodynamic performance gains achievable, current helicopter

rotor blades have evolved into relatively high-aspect-ratio structures. From a struc-
tural dynamics standpoint such a structure is a simplification in that rotor blades
can then be treated as one dimensional, that is, their elastic characteristics can
be specified as functions of only the radius. A further consequence of this one
dimensionality is that we need define the elastic deflections of rotor blades only
in terms of bending deformations transverse to the blade, torsion, and, in some
special cases, axial elongation and cross-sectional warping.
In this chapter the basic concepts relating to blade vibration characteristics are
developed. These concepts include theory for and practical schemes for obtaining
natural frequencies and mode shapes in both in-plane and out-of-plane bending
and in torsion for relatively simple blades, that is, blades that have no pitch,
twist, or preconing. Such simplified blade configurations produce what we will
refer to as uncoupled mode characteristics. The uncoupled modal responses are
characterized as having, for any one mode, all of the motion in either the in-plane,
out-of-plane, or torsional degrees of freedom. In reality, rotor blades do have
all of these complicating characteristics (pitch, twist, and precone), and material
is therefore presented explaining how the basic uncoupled characteristics for the
simplified blade are modified or coupled by these additional considerations.
3.1 Basic Equations for Bending

The principal characteristics of rotating beams that distinguish them from con-
ventional beams are that they function in a tension field that is variable with span
and that they have additional dynamic loadings accruing from the rotational field.
Because of the one dimensionality of the beam, the appropriate starting point is to
take an infinitesimal spanwise element of the beam and then form the equilibrium
conditions governing it.
3.1.1 Equilibrium of a Spanwise Element

Figure 3.1 shows a (rotating) cantilevered beam with a distributed axial loading
fx (x) [arising from centrifugal force and resulting in distributed tension T (x)] and
a distributed transverse loading fz (x, t).
67
“chapter3(new)” — 2005/11/16 — page 67 — #1

Fig. 3.1 Loading distributions of a rotating cantilevered beam.
Let us then consider an arbitrarily located spanwise element and construct a

free-body diagram (see Fig. 3.2); note that we include the internal elastic loads
and the external loadings, as defined in Fig. 3.1. The element is then equilibrated
with respect to the axial and transverse forces and the moments:
Fx = 0; Fz = 0; M = 0 (3.1)
3.1.2 Basic Differential Equation for Transverse Bending

The aforementioned conditions of equilibrium, when taken together with the
standard beam bending equation
∂2
M(x, t) = EI z(x, t) (3.2)
∂x 2
yield the following basic differential equation for the beam in tension:

∂2 ∂ 2z ∂ ∂z
EI − T = fz (x, t) (3.3)
∂x 2 ∂x 2 ∂x ∂x
Fig. 3.2 Free-body diagram for a spanwise element of a rotating beam.
“chapter3(new)” — 2005/11/16 — page 68 — #2

ROTATING BEAMS 69
Boundary conditions. The appropriate boundary conditions for all rotating

beams are that they have zero deflection at the root and zero bending moment and
shear at the tip. These conditions are stated mathematically as
∂ 2z ∂ 3z
z(0, t) = (L, t) = (L, t) = 0 (3.4)
∂x 2 ∂x 3
Additionally, a fourth boundary condition must be imposed at the beam root based
on the type of fixity at that point. If the beam represents an articulated rotor blade,
then the beam has the boundary condition appropriate to a hinge, and the bending
moment is taken to be zero:
∂ 2z
(0, t) = 0 (3.5)
∂x 2
If the beam represents a hingeless rotor blade, then the beam root has the canti-
levered boundary condition, and the root slope is taken to be a constant equal to
some built-in coning value βB :
∂z
(0, t) = βB (3.6)
∂x
Distributed tension. The tension in the beam arising from the centrifugal
force field can be obtained by integrating the axial load distribution fx (=m2 x):
L
T (x, t) = 2
mx1 dx1 (3.7)
x
3.1.3 Transverse Bending Motion

At this point none of the development given earlier takes account of the fact
that the beam is in (transverse) motion. The dynamics of this motion can then be
simulated using D’Alembert’s principle; the transverse inertial load distribution
resulting from beam transverse motion is included explicitly in fz .
Transverse load distribution. Regarding the out-of-plane z motion, the

transverse load distribution is also in the z direction:
fz (x, t) = fz (x, t) applied + fz (x, t)inertial (3.8)

external
Correspondingly, the transverse load distribution for the in-plane motion case is
in the y direction and is similarly interpreted. Thus, for out-of-plane and in-plane
motions z and y, respectively, the transverse loadings are given as follows:
Out-of-plane bending:
∂ 2z
fz (x, t)inertial = −m (3.9a)
∂t 2
“chapter3(new)” — 2005/11/16 — page 69 — #3

In-plane bending:
∂ 2y
fy (x, t)inertial = −m + m2 y (3.9b)
∂t 2
The second term in the expression for the in-plane inertial load distribution arises
from the fact that the centrifugal force field is radial; hence, a component of this
force field will be in the direction of the in-plane deformation y.
3.1.4 Differential Equation for Out-of-Plane Bending

The reader can verify that all of the preceding formulations can then be com-
bined to yield the following differential equation for out-of-plane (transverse)
bending:
L
mz̈ + [EIz ] − z
2
mx1 dx1 = fapp (x, t) (3.10)
x
This equation can be seen to have the following properties:

1) The equation is linear, of fourth order in the spatial variable x, second order in
time t, and with spanwise variable coefficients.
2) The linearity of the equation allows a separation of variables solution scheme
combined with superposition:
∞

z(x, t) = γj (x)qj (t) (3.11)
j=1
where γj (x) is the jth natural mode shape (which satisfies the boundary
conditions) and qj (t) is the jth generalized coordinate or modal response
variable.
One mathematical significance of the γj (x) function is that it satisfies the
following orthogonality condition:
L
Mj , for k = j
mγj γk dx = (3.12)
0 0, for k = j
Calculation of the normal mode shapes γj (eigenvectors) and the corresponding

natural frequencies ωj (eigenvalues) defines the major computational task in analy-
zing rotating beams. Unfortunately, although the basic differential equation is
linear, general solutions do not exist even for spanwise constant properties, and
some form of numerical method must be employed. Two basic numerical methods
are presented in this chapter for obtaining these calculations, but first the basic
properties of the resulting solution for the free-vibration (eigenvalue) problem
must be addressed.
“chapter3(new)” — 2005/11/16 — page 70 — #4

ROTATING BEAMS 71
3.1.5 Equation for Modal Response Variable: Galerkin’s Method

The orthogonality property defined in Eq. (3.12) can be used to effect a sepa-
ration of variables solution for the basic partial differential equation for bending.
If the equation is first multiplied by the jth mode shape function γj (x) and the
equation is then integrated over the span of the blade, the orthogonality condition
then produces uncoupled response equations in the various modes:
L L
mγj (x) dx q̈j +
2
[γj (x)]2 EI dx
0 0

Mj
L L L
+ 2
[γj (x)]2 mx1 dx1 dx qj = γj (x) fapp dx (3.13)
0 x 0
This uncoupled equation for qj (t) then has the following general form:
Mj q̈j + [KSj + 2 KCFj ]qj = j (t) (3.14)

mass stiffness external force
where the generalized structure stiffness KSj and generalized centrifugal stiffness
KCFj follow directly from Eq. (3.13). Note that neither of these generalized stiff-
nesses, taken separately, demonstrates orthogonality, but when taken together with
the appropriate value of 2 they do, in fact, demonstrate orthogonality.
Galerkin’s method. The procedure defined by Eq. (3.13) was primarily

undertaken to effect an uncoupling of the modal degrees of freedom by availing
ourselves of the orthogonality principle. The technique can actually be general-
ized to include cases wherein the modes, although close to being orthogonal to
each other, are not strictly so. In this case a system of coupled equations will
result wherein both mass and stiffness coupling terms will be present. In a matrix
equation format these equations then have the following general form:
     
 q̈1  
 q1  1 

       
q̈2  q2 
   2 
[M] .. + [K] . = . (3.15)

 .  
 .. 
 
 .. 


   
   
q̈n       
qn n
Although in this case the procedure does not completely decouple the various
modes, a very useful result has still been achieved in that the equations are again
ordinary differential equations in time and a solution in the spatial variable has been
effected. The general technique defined by Eq. (3.13) is alternatively referred to as
the Galerkin technique or Galerkin’s method. The form of solution so defined can
be applied to a great variety of problems wherein the system is mostly governed by
linear processes with some nonlinearities present. It is an especially useful solution
technique for problems based on modal formulations. Generally, the Galerkin
technique can be used according to the following procedure:
“chapter3(new)” — 2005/11/16 — page 71 — #5

1) Isolate the linear portion of the equation as a left-hand side of the equation,
and (optionally) solve for the normal modes and natural frequencies.
2) Using the obtained normal modes or other approximate ones (which at least
satisfy the boundary conditions), premultiply the equation, in turn, by each of
these mode shapes.
3) Integrate the resulting equation over the length of the structure (beam), and
use the orthogonality condition to remove terms that are zero.
4) Solve the resulting ordinary equations of motion as either an assemblage of
single-degree-of-freedom equations or as a system of coupled equations.
The key to the use of Galerkin’s method is that the equations be “reasonably
linear.” Mathematically, this means that the nonlinear terms, if present, can be
expressed as a linear combination of the mode shapes (and modal variables) and that
any one of the resulting modal contributions is small relative to the corresponding
retained linear term. Thus, from a practical standpoint the nonlinear terms are
generally grouped together and put on the right-hand side of the equation as part
of the excitation. The remaining linear terms can either be made to yield a modal
solution or can be approximated by one.
3.1.6 Free-Vibration Characteristics

To determine the free-vibration characteristics of the rotating beam for out-of-
plane oscillations, we set the right-hand side of Eq. (3.15) to zero and assume
sinusoidal motion:
j (t) = 0; qj = q̄j eiωj t (3.16)
After dividing by the generalized mass Mj , we obtain the following equation for
the rotating natural frequency results:

KSj K
2 CFj
ωj =
2
+ = ωNR
2
j
+ 2 Kwj (3.17)
Mj Mj

2
ωNR Kwj
j
where ωNRj is the jth mode nonrotating blade natural frequency and Kwj is an
additional term to account for centrifugal stiffening and is alternatively denoted
herein as the rise factor or Southwell coefficient for the jth mode.
3.1.7 In-Plane Bending Natural Frequencies

Extension of the formulations given in Sec. 3.1.4–3.1.6 to the in-plane vibration
case is easily accomplished by substituting the inertial load distribution appropriate
for in-plane motion (see Section 3.1.3). For this case the resulting differential
equation looks identical in form to that for out-of-plane bending, except for the
addition of a centrifugal force like term involving in-plane deflection:
L

mÿ + [EIy ] − y 2
mx1 dx1 − 2 my = fapp (x, t) (3.18)
x
additional
term
“chapter3(new)” — 2005/11/16 — page 72 — #6

ROTATING BEAMS 73
Then, after the orthogonality principle has been applied a similar modal equation
for in-plane bending can be obtained:
ωk2 = ωNR
2
k
+ 2 Kvk (3.19)
where
Kvk = Kwk − 1 (3.20)
Note that in this case Kwk represents the rise factor of the same beam but rotated
so that the motion would be out of plane. Thus, the approximate expression for
the natural (rotating) modal frequency for in-plane bending is the same as the
corresponding one for out-of-plane bending except that the factor multiplying the
rotor speed squared, the rise factor, is reduced in value and can be taken to be that
for the corresponding kth out-of-plane bending mode, but reduced by unity.
3.1.8 Basic Rotor Speed Characteristics and Fan Plots

Generally, within the normal operating rotor speed ranges of rotorcraft the
rotating beam mode shapes (and their derivatives) are weak functions of rotor
speed. Therefore, the integrated quantities Mj , KSj , and KCFj are also weak
functions of rotor speed. Consequently, if these quantities are approximated as
constants, then Eq. (3.19), which defines the rotating natural frequency, essen-
tially defines a hyperbolic functional relationship between natural frequency and
rotor speed. The usefulness of this approximation is twofold: 1) It provides a
convenient tool for examining the qualitative characteristics of the modal frequen-
cies with variations in rotor speed; and 2) it is a reasonably accurate curve-fitting
function for estimating variations of the rotating natural frequencies over narrow
frequency ranges.
The Kwj constant appearing in the preceding equations is alternatively referred
to as the jth mode Southwell coefficient or rise factor. It is a measure of how
fast the rotating frequency for that mode increases with rotor speed. Figure 3.3,
which is alternatively called a Southwell diagram or a fan chart, shows the vari-
ability of the natural frequencies with rotor speed, and, by plotting the fan lines
for each harmonic of rotor speed it shows the proximities of the frequencies to
integral harmonics of rotor speed.
3.1.9 Detailed Characteristics of Uniform Blades

In reality, the principal assumption made in the preceding section, that is, that
the mode shapes do not change with rotor speed, is not true. This approximation is
quite useful for purposes of providing a basic understanding of the effects of rotor
speed and for calculating approximate trends within a small rotor speed band. In
this section this assumption is relaxed so that the effects of rotor speed can be
investigated in more detail. To this end, we can, in fact, obtain some analytical
results, but we will need to make the further assumption that the blade has uniform
mass and stiffness distributions along the blade. Additionally, it is assumed that
there is zero offset of the root of the blade with the rotation axis e. The objective
“chapter3(new)” — 2005/11/16 — page 73 — #7

Fig. 3.3 Typical Southwell coefficient plot for a rotating beam in bending.
in this section is to look at the limiting cases of the effective Southwell coefficient
at the two extremes of rotor speed variation: = 0 and → ∞. We will actually
look at these cases in reverse order, for simplicity.
Limiting case 1, → ∞. The case of rotor speed approaching infinity

implies that the centrifugal force stiffening term becomes significantly larger than
the elastic stiffening term. Thus, the appropriate differential equation for free
vibration in flapwise bending becomes
L
mz̈ − z 2
mx1 dx1 =0 (3.21)
x
If the assumption of uniform mass distribution is invoked, if the motion is nondi-

mensionalized by the rotor radius (z̄ = z/R), and if sinusoidal motion at the jth
natural frequency ωj is assumed, then this equation can be written as
ω 2
z̄j + [(1 − x 2 )z̄j ] = 0
j
2 (3.22)

Equation (3.22) can be recognized as a form of the Legendre differential equation
whose solution is given by any of the Legendre polynomials Pn (x). If we restrict
our consideration to the Legendre polynomials that are zero at the origin (n odd),
then the solution, in terms of natural frequency and mode shape, can be directly
written as
ω 2
j
= j(2j − 1) (3.23a)

γwj (x) = P2j−1 (x) (3.23b)
“chapter3(new)” — 2005/11/16 — page 74 — #8

ROTATING BEAMS 75
These results define the vibration characteristics of what amounts to a flexible

cable with vanishingly small elastic stiffness. As such, they are independent of
whether the blade has an articulated (hinged) or cantilevered (hingeless) root
boundary condition. However, P1 (x) is equal to x, which defines the mode shape
of the blade in “rigid” flapping. The rigid flapping mode is typically referred to
as the zeroth mode because (for zero offset) it is not an elastic mode. This is the
correct limiting case for the first elastic mode for a blade with a cantilevered root
boundary condition. However, P3 (x)(= −3/2x + 5/2x 3 ) is the correct limiting
case for the first elastic mode for a blade with the hinged root boundary condition.
Thus, by use of Eqs. (3.23a) and (3.23b) for hinged root blades, the j index on the
right-hand sides must be increased by one.
Limiting case of the Southwell coefficient for out-of-plane modes. Equa-

tion (3.23a) forms the basis of the limiting case for flapping modes, wherein
the ratio of nonrotating natural frequency to rotor speed (ωNRj / ) becomes
negligible; hence, the limiting case of the Southwell coefficient is given as

j(2j − 1); cantilevered root condition

Kwj →∞ = (3.24)
(2j + 1)( j + 1); hinged root condition
Limiting case of the Southwell coefficient for in-plane modes. For the free
vibration in the in-plane modes, the appropriate differential equation, subject to
the same assumptions of spanwise uniformity and very high rotor speed, becomes

ωk 2
2 + 1 ȳk + [(1 − x 2 )ȳk ] = 0 (3.25)

From this equation it can then be inferred that the solution is again expressible
in terms of Legendre polynomials. Thus, the limiting values of the Southwell
coefficient for in-plane motion are given as

k(2k − 1) − 1; cantilevered root condition
Kvk →∞ = (3.26)
(2k + 1)(k + 1) − 1; hinged root condition
The preceding results are summarized in Table 3.1.
Limiting case 2, → 0. The behavior of the blade characteristics when

the rotor speed becomes vanishingly low can be addressed by solving for the
derivative of natural frequency (squared) with respect to rotor speed (squared).
Although a straightforward numerical differentiation could be used to obtain this
result, it is instructive to formulate an analytical result. For this purpose we again
use Eq. (3.10), for out-of-plane bending, as a starting point. After uniformity in
the mass and stiffness distributions and free vibration in the out-of-plane direction
are assumed, the differential equation for the jth normal mode becomes
2{−ωj2 mγwj + [EIγwj ] } − 2 m[(1 − x 2 )γw j ] = 0 (3.27)
“chapter3(new)” — 2005/11/16 — page 75 — #9

Table 3.1 Limiting case ( → ∞) Southwell coefficients for

uniform beams
Hinged blades Hingeless blades

Mode no.
m KwmI KvmI KwmI KvmI
1 6 5 1 0
2 15 14 6 5
3 28 27 15 14
4 45 44 28 27
where ωj and γwj are, respectively, the natural frequency and normal mode shape
for the blade at zero rotor speed. The next step is to differentiate Eq. (3.27) with
respect to 2 and then set 2 to zero:

∂γwj ∂γwj ∂ωj2
2 −ωj m 2 + EI
2
−2 mγwj
∂ ∂2 ∂2
→0
− m[(1 − x 2 )γw j ] = 0 (3.28)
Equation (3.28) can then be solved using the Galerkin technique wherein the
equation is first multiplied by the jth mode shape and then integrated over the span
of the blade. The derivative of the mode shape with respect to 2 , ∂γ wj /∂2 can
be written as a linear combination of all of the mode shapes:
∞

∂γwj
= bjm γwm (x) (3.29)
∂2
m=1
Then, when Eq. (3.28) is multiplied by γwj and integrated over the blade span, the
orthogonality principle will leave the first term in braces equal to zero because it
defines the eigenvalue problem for ωj and γwj . The rest of the equation, after the
Galerkin technique has been invoked, becomes
1
γwj (x) 2(Kwj0 )γwj (x) + γw j (x) 1 − x 2 dx = 0 (3.30a)
0
where

∂ωj2
Kwj0 ≡ (3.30b)
∂2
→0
which, after an integration on parts with the second term in Eq. (3.30a), yields the
following simpler form in terms of modal integrals:
1 1
2(Kwj0 ) γwj (x) dx −
2
γw2j (x)(1 − x 2 ) dx = 0 (3.31)
0 0
“chapter3(new)” — 2005/11/16 — page 76 — #10

ROTATING BEAMS 77
This equation can then be reduced to the final desired result:

1 1 −1
Kwj0 = γw2j (x)(1 − x 2 ) dx 2 γw2j (x) dx (3.32)
0 0
Again, it can be readily verified that, for both hinged and hingeless rotor blades,
the similar limiting case for in-plane modes is
Kvk0 = Kwk0 − 1 (3.33)
Evaluation of the modal integrals. The modal integrals appearing in

Eq. (3.32) can be more readily evaluated by considering the actual solution for
the uniform beam with zero rotation. Consider the differential equation for the
amplitude of the jth mode of free vibration (with zero rotor speed) for out-of-plane
motion γwj :
−mωNR
2
γ + [EIγwj ] = 0
j wj
(3.34)
This equation is linear with constant coefficients and possesses an exact general
solution:

γwj (x) = Aj cos ω̄j x + Bj sin ω̄j x + Cj cosh ω̄j x + Dj sinh ω̄j x (3.35)
where the natural frequencies ωNRj are nondimensionalized by ω0 (a reference

frequency resulting from a nondimensionalization of the equation) to form the
solution eigenvalues:
ω̄j = ωNRj /ω0 (3.36a)
where

ω0 = EI/mR4 (3.36b)
and x is the spanwise variable nondimensionalized by R. Invoking the boundary

conditions appropriate for each type of beam leads to eigenvalue solutions for the
natural frequencies. Then, upon setting the deflections at the end of the beam to
unity, together with the boundary conditions, the arbitrary constants Aj , Bj , Cj ,
and Dj can be evaluated. These solution results are presented for each type of rotor
blade as follows:
1) Cantilever boundary condition (hingeless blades):
Frequency equation:

cos ω̄j cosh ω̄j = −1 (3.37)
“chapter3(new)” — 2005/11/16 — page 77 — #11

Solutions values:
ω̄1 = 3.516015
ω̄2 = 22.034490
ω̄3 = 61.697208
ω̄4 = 120.901922
Constants defining the mode shapes:

ω̄j + sinh ω̄j
sin
Aj = −Cj = − (3.38a)
2(sin ω̄j cosh ω̄j − cos ω̄j sinh ω̄j )

cos ω̄j + cosh ω̄j
Bj = −Dj = (3.38b)
2(sin ω̄j cosh ω̄j − cos ω̄j sinh ω̄j )
2) Hinged root boundary conditions (hinged blades):

Frequency equation:

tan ω̄j − tanh ω̄j = 0 (3.39)
Solution values:
ω̄1 = 15.418203
ω̄2 = 49.964866
ω̄3 = 104.247694
ω̄4 = 178.269709
Constants defining the mode shapes:
Aj = Cj = 0 (3.40a)
1
Bj = (3.40b)
2 sin ω̄j
1
Dj = (3.40c)
2 sinh ω̄j
Other methods for determining the nonrotating natural frequencies ωNRj and
mode shapes especially for nonuniform blades are presented in the succeeding
sections. Although closed forms for the modal integrals defined in Eq. (3.32) are not
directly available, evaluations of the integrals are feasible from Eqs. (3.35–3.39)
using either integral tables or numerical integration. Variations of the mode shapes
with rotor speed give rise to corresponding variations of the Southwell coefficients
with rotor speed. Examination of these variations, including the preceding ideas of
the limiting cases for the coefficients, is discussed in greater detail in the succeeding
section.
“chapter3(new)” — 2005/11/16 — page 78 — #12

ROTATING BEAMS 79
3.2 Reference Uniform Blade

Insofar as the rotor blade is reasonably uniform (constant mass and stiffness
distributions), a completely accurate solution can be devised using rigorous numer-
ical methods (to be described in a succeeding section). Furthermore, such a solution
can be scaled to yield accurate predictions of natural frequency for any combina-
tion of mass and stiffness distributions and rotor speed. Thus, because most rotor
blades tend to have properties that are generally uniform, with “small” spanwise
deviations, the reference uniform blade results can provide useful approximations
before rigorous frequency calculations are performed.
3.2.1 Scaled Differential Equation for Bending

As a starting point, we again use Eq. (3.10). This equation is first nondimension-
alized by rotor blade radius R and the reference frequency ω0 defined by Eq. (3.35).
After assuming sinusoidal motion, we obtain the following resulting equation for
free vibration in the out-of-plane direction:
1
2 ¯
−ω̄ z̄ + z̄ − z̄
iv 2
x̄1 dx̄1 =0 (3.41)
x
where the differentiations are with respect to x̄ (=x/R) and ω̄ and ¯ are,
respectively, the natural frequency and rotor speed nondimensionalized by ω0 .
3.2.2 Solution for Natural Frequencies

By use of one of the numerical methods described in succeeding sections,
Eq. (3.41) was rigorously solved for its natural frequencies with wide variation on
rotor speed. No attempt was made here to impose constraints on the mode shapes,
and, as a result, the frequency variation of the frequencies is exact. The concept
of the rise factor formulated in the preceding subsections is a very useful tool and
can still be retained by making the rise factor rotor-speed dependent.
Zero rotational speed frequencies. Table 3.2 presents the results for zero
rotational speed for hinged and hingeless blades in terms of the nonrotating natural
frequencies and Southwell coefficients for zero rotor speed for in-plane and out-
of-plane motion.
Nonzero rotational speed frequencies. The exact natural frequencies for

the uniform reference blade at any rotor speed can then be formulated using the
concept of a rotor speed variable rise factor. Thus, the variation of the (non-
dimensionalized) natural frequency with rotor speed can still be written in the
basic form given earlier, but now with the rise factor Kαm (where α = v or w, as
appropriate), which is itself variable with rotor speed:
ω̄α2 m = ω̄NR
2 ¯ 2 Kαm (
+ ¯ 2) (3.42)
m
“chapter3(new)” — 2005/11/16 — page 79 — #13

Table 3.2 Natural frequency characteristics for uniform reference blade, = 0
Hinged blades Hingeless blades

Mode no.
m ω̄NRm Kvm0 Kwm0 ω̄NRm Kvm0 Kwm0
1 15.4203 5.3969 6.3969 3.5160 0.1933 1.1933

2 49.9649 16.9030 17.9030 22.0344 5.4782 6.4782
3 104.2477 34.9928 35.9928 61.6972 16.8594 17.8594
4 178.2697 59.6559 60.6559 120.9019 35.0554 36.0554
where Kαm can be expressed in terms of the limiting values given in Tables 3.1 and
3.2 and a rise factor variation function ξm , defined as
¯ 2 ) − Kαm
Kαm (
¯ )=
ξm (
2 0
(3.43)
KαmI − Kαm0
The variations of the rise factor variation function with nondimensional rotor speed
squared are given in Figs. 3.4 and 3.5 for hinged and cantilevered root boundary
conditions, respectively. Equation (3.43) can then be rearranged to give the desired
variation of rise factor with rotor speed:
¯ 2 ) = Kαm + ξm (
Kαm ( ¯ 2 )(Kαm − Kαm ) (3.44)
0 I 0
Note that, again, the rise factors for in-plane modes (at any rotor speed) are the
corresponding out-of-plane rise factors minus one. The results of this section are
“exact” in that no simplifying mathematical assumptions were made. The results
of Figs. 3.4 and 3.5 required the use of an appropriate numerical method for
solving Eq. (3.41). Two basic numerical methods for accomplishing this solution
are presented in a subsequent section.
3.2.3 Approximations to the Mode Shapes

The results presented in the previous subsection can then be used to obtain
approximations to the mode shapes. The rise factor variation functions ξm , graphi-
cally depicted in Figs. 3.4 and 3.5, can be interpreted to represent deviations of the
dynamic properties from the zero rotor speed values. Using this interpretation, the
mode shapes can be approximated by adding to the zero rotor speed mode shapes,
as given by Eq. (3.35), appropriate adjustments that are functions of the rise factor
deviation functions. To this end, we first define a mode shape adjustment factor
χm , which is taken to be a direct function of ξm :
¯ 2 )]
χm = f [ξm ( (3.45)
“chapter3(new)” — 2005/11/16 — page 80 — #14

ROTATING BEAMS 81
Fig. 3.4 Rise factor variation functions for first four elastic bending modes, hinged
blades.
The functional variations of χm with the ξm values given in Figs. 3.4 and 3.5
are presented in Figs. 3.6 and 3.7, respectively, for the hinged and hingeless blade
configurations:
The mode shape adjustment factors are then used to “adjust” the nonrotating
mode shapes [i.e., Eq. (3.35)] to produce approximate mode shapes at rotor speed.
Fig. 3.5 Rise factor variation functions for first four elastic bending modes, hingeless
blades.
“chapter3(new)” — 2005/11/16 — page 81 — #15

Fig. 3.6 Mode shape adjustment factors for first four bending modes, hinged blades.
This is accomplished using appropriate mode shape adjustment factors together

with normalized mode shape adjustment functions γ̄m :
¯ 2) ∼
γm ( ¯ 2 )] γ̄m
= γm0 + χm [ξ( (3.46)
The normalized mode shape adjustment functions are presented in Figs. 3.8 and
3.9 again, respectively, for the hinged and hingeless blade configurations. Note
that in each of these figures the functions all achieve a maximum value of either
Fig. 3.7 Mode shape adjustment factors for first four bending modes, hingeless
blades.
“chapter3(new)” — 2005/11/16 — page 82 — #16

ROTATING BEAMS 83
Fig. 3.8 Normalized mode shape adjustment functions for first four bending modes,
hinged blades.
Fig. 3.9 Normalized mode shape adjustment functions for first four bending modes,
hingeless blades.
“chapter3(new)” — 2005/11/16 — page 83 — #17

plus or minus unit value somewhere along the span and are then scaled by the
mode shape adjustment factors. Also, in each of the figures the spanwise vari-
ations of the maximum points over the nondimensional rotor speed range are
indicated.
3.3 Numerical Methods

In the preceding discussions much has been said about the characteristics of
the solution of the basic differential equation of motion for the rotating beam.
Although the linearity of the equation allows modal techniques to be used to great
advantage, little has been offered as to how, in fact, these modal properties can
be calculated with reasonable accuracy. The one basic difficulty that arises with
respect to obtaining the required modal solution is the wide spanwise variabil-
ity of the coefficients in the equation of motion even with constant structural
properties.
Conventional rotorcraft blades are typically constructed with some spanwise
variation in the mass and stiffness distributions. Over and apart from this variability,
however, the tension in the rotating beam varies significantly along the span of the
beam. This variability generally denies us the ability to obtain closed-form exact
analytical solutions for the modal solution even for beams with constant structural
properties. It is thus mandatory for numerical methods to be used in obtaining the
modal properties.
3.3.1 Natural Frequencies and Mode Shapes by the

Holzer–Myklestad Technique
The practical vibration analysis of actual structures for the various natural frequ-
encies and their associated mode shapes requires idealizations that account for the
various details of the structure to a greater extent than do “back-of-the-envelope”
approximations. The method presented herein is an orderly procedure for finding
these quantities to any degree of precision desired. The method has the advantage
of being amenable even to handheld calculator operation as well as to use on a
digital computer. The basis of the method is the (assumed) ability to idealize the
structure into a discrete number of rigid masses with interconnecting elements.
The various masses are mathematically tied together in an orderly fashion that
thereby facilitates computation.
Example 3.1
Consider a longitudinal string of masses such as that given in Fig. 3.10. Let us
examine in Fig. 3.11 the forces acting on a single spring-mass element. Assuming
oscillating motion (based on a trial value of frequency), we can relate the forces
and displacements across this element:
ẍn = −ω2 xn (3.47)
“chapter3(new)” — 2005/11/16 — page 84 — #18

ROTATING BEAMS 85
Fig. 3.10 Multiple-degrees-of-freedom spring-mass system.
Then the following two equations can be written:

Fn
xn+1 = xn − = xn − un Fn (3.48)
kn
Fn+1 = Fn + mn+1 ω2 xn+1 (3.49)
or, expressed as a matrix equation,

xn+1 1 −un xn
= (3.50)
Fn+1 mn+1 ω2 (1 − un mn+1 ω2 ) Fn
where, starting at the right-hand (i.e., free) end,
F1 = m1 ω2 , x1 = 1 (3.51)
The preceding matrix equation defines a transfer matrix between the state con-
ditions at either end of the spring-mass element. This basic equation can then be
used repeatedly, starting with n = 1 and an assumed trial value for ω2 , until the
left-hand side of the structure is reached. The correct value of ω2 will produce
the correct boundary conditions at this end of the structure. In general, any given
trial frequency will be incorrect, and an iteration must be used to reach a suitably
accurate value. Iteration in this case consists of plotting or otherwise “tracking”
the value of the left-hand end boundary condition vs trial values of ω2 , as shown
in Fig. 3.12. Iteration consists of varying the frequency squared value and testing
for a crossing of the frequency squared axis, thereby indicating a zero value.
Fig. 3.11 Free-body diagram for a typical spring-mass element (Example 3.1).
“chapter3(new)” — 2005/11/16 — page 85 — #19

Fig. 3.12 Variation of left-hand end boundary condition with trial frequency
(Example 3.1).
For noncrossings of the frequency axis, a variety of appropriate search algo-

rithms can be employed to accelerate and/or improve the accuracy of locating
the frequency at which the curve crosses the axis.
Application to a rotating beam. The analysis of the free-vibration charac-

teristics of a rotating beam using the Holzer–Myklestad method requires a more
extensive element state vector. Although four state vector elements are a mini-
mum, any number of state variables can be accommodated. Consider in Fig. 3.13
the flapping motion of a rotating beam that has been discretized into a series of
spanwise elements, each of which consists of a massless elastic beam segment and
a concentrated point mass.
Fig. 3.13 Elastomechanical discretization of a rotating beam.
“chapter3(new)” — 2005/11/16 — page 86 — #20

ROTATING BEAMS 87
We wish to describe the properties at the endpoints of an arbitrary nth beam

segment:
Again, the solution technique is to form a transfer matrix statement relating the
state variables at the node point defined on one beam element to those at the
similarly defined point on the next beam element. Note that the sketch shows
the beam element to be lopsided, with the concentrated mass located at the left-
hand end; this will be discussed in more detail later. For the time being let the
appropriately defined point on the beam element be denoted { }L . Then the transfer
matrix statement we require, with n increasing from right to left, can be written as
{ }Ln+1 = [Tn ]{ }Ln (3.52)
The various transfer matrices can then be cascade multiplied to relate the boundary
conditions at the root to those at the tip:
{ }LN(root) = [TN ][TN−1 ] · · · [T1 ]{ }L1(tip) (3.53)
where the required boundary conditions are incorporated in { }LN and { }L1 , as
discussed earlier.
Consider now the makeup of a typical transfer matrix as defined over an arbi-
trary spanwise element. Let us divide the element into two distinct parts: a rigid
concentrated mass-inertia part (associated with the coincident mass mn+1 and iner-
tia In+1 ) and an elastic massless part, each attached to each other at the spanwise
point just to the right of the concentrated mass (denoted { }Rn+1 ), as shown in the
following sketch:
“chapter3(new)” — 2005/11/16 — page 87 — #21

Fig. 3.14 Free-body diagram of arbitrary spanwise concentrated mass element.
Concentrated mass/inertia part: (point transfer matrix). Let us first

look at the concentrated mass element and define equilibrium load conditions
(see Fig. 3.14), where the tension loads Fn and Fn+1 in Fig. 3.14 are respectively
given by

n
Fn = mi xi 2 (3.54a)
i=1
Fn+1 = Fn + mn+1 xn+1 2 (3.54b)
Then, upon balancing the transverse forces and moments on the mass element
and equating the displacements and rotations across the point mass, we obtain the
following recursive relationships:
L
Sn+1 = Sn+1
R
+ mn+1 ω2 zn+1
R
(3.55a)
L
Mn+1 = R
Mn+1 + In+1 ω 2 R
ψn+1 (3.55b)
L
zn+1 = zn+1
R
(3.55c)
L
ψn+1 = R
ψn+1 (3.55d)
which can then be expressed in matrix form

 L    R

 z 1 0 0 0   z


ψ   0 
ψ 
 1 0 0

=  (3.56)

 S 
 mn+1 ω2 0 1 0  S


   
 
M n+1 0 In+1 ω2 0 1 M n+1

= point transfer matrix, [Pn+1 ]
“chapter3(new)” — 2005/11/16 — page 88 — #22

ROTATING BEAMS 89
Elastic massless part: field transfer matrix. Let us now turn our attention
to the massless elastic portion between { }Rn+1 and { }Ln (see Fig. 3.15).
The equation for elastic deformation of this arbitrary elastic element can be
obtained by considering the elastic deformation characteristics of a simple canti-
levered beam in tension F, of length L, and with constant section properties m
and EI. This beam element is supported at the left-hand end so that it has inboard
boundary conditions of z = z0 and z = z0 , and is loaded at the right-hand end
by the vertical shear and axial loads ST and FT , respectively, and by a bending
moment MT . Upon taking the spanwise variable x to start at the left-hand end,
it can be verified that the differential equation for transverse deformation is then
given by
EIz = ST (xT − x) − FT (zT − z) + MT (3.57)
where zT is the transverse deflection at the tip (right-hand end) of the elastic
element. Solution of this equation is facilitated by a simple change of variable. Let
ζ = x − x0 and η = z − zT , so that
EIη − FT η = ST (L − ζ ) + MT (3.58)
where ζ varies from 0 to L and where the solution η(ζ ) is subject to the following
boundary conditions:
η(0) = z0 − zT and η (0) = z0 (3.59)
The condition that η(L) = 0 follows automatically from the definition of η. It can
then be verified as an exercise that the solution of this equation is given by the
Fig. 3.15 Free-body diagram of arbitrary spanwise massless elastic element.
“chapter3(new)” — 2005/11/16 — page 89 — #23

following two equations:

tanh λL ST L tanh λL
zT = z0 + z0 L + 1−
λL FT λL

MT 1
+ 1− (3.60)
FT cosh λL

1 ST L 1 MT
zT = z0 + 1− + [λL tanh λL] (3.61)
cosh λL FT cosh λL FT
√
where λ = F/EI, and for λL → 0 (zero-tension case), these expressions
simplify to the more elementary expressions for beam deflections:
zT = z0 + z0 L + ST L 3 /3EI + MT L 2 /2EI (3.62)

zT = z0 + ST L 2 /2EI + MT L/EI (3.63)
The elastic transfer function description is then completed by relating the shears
and moments at the two ends to each other:
M0 = MT + ST L − FT (zT − z0 ) (3.64)
S0 = ST (3.65)
Equations (3.60) through (3.65) can then be rearranged and combined into a matrix
format. In doing so, let us rewrite the left-hand and right-hand quantities in terms
of the variables defined in Fig. 3.15. The resulting matrix equation is then given by
 R    L

 z 1 −Ln fn (Ln /Fn )( fn − 1) ( gn − 1)/Fn  z


ψ  0  
 gn −( gn − 1)/Fn 
−fn hn  ψ 

=  (3.66)

 S 
 0 0 1 0  S


   
 
M n+1 0 −Fn Ln fn Ln fn gn M n

=[Fn ]
where

sinh λL
fn = (3.67a)
λL n
gn = cosh λLn (3.67b)
hn = Ln /EIn (3.67c)
“chapter3(new)” — 2005/11/16 — page 90 — #24

ROTATING BEAMS 91
and

λn = Fn /EIn (3.68)
When the preceding results, that is, for both the point and field transfer matrices
[Pn+1 ] and [Fn ], respectively, are combined, the required transfer matrix for the
total (nth) element [Tn ] is obtained:
 
1 −Ln fn (Ln /Fn )( fn − 1) ( gn − 1)/Fn
 0 gn −( gn − 1)/Fn −fn hn 
 
 
 −mn+1 [1 + mn+1 ω2 [mn+1 ω 2 
[Tn ] = mn+1 ω2 
 ×ω2 Ln fn ×(Ln /Fn )( fn − 1)] ×(gn − 1)/Fn ] 
 
 
[In+1 ω2 gn [−In+1 ω2 ( gn − 1)/Fn [−In+1 ω2 fn hn
0 − Fn Ln fn ] + Ln fn ] + gn ]
(3.69)
As described earlier, all of the spanwise transfer matrices are then used in a
cascade multiplication to relate the state conditions at the root of the beam to
those at the tip. Because each of the transfer matrices is frequency dependent, an
eigenvalue problem is thereby defined:
{ }N(root) = [TN ][TN−1 ] · · · [T2 ][T1 ]{ }1(tip) (3.70)

= [T (ω2 )]
The rest of the eigenvalue problem consists of rewriting the transfer matrix
equation [Eq. (3.70)] utilizing a statement of the boundary conditions that will
result in a standard (matrix) form of an eigenvalue problem. To this end, let us first
review appropriate statements of the boundary conditions:
At the tip:
 

 z1 


ψ  
1
{ }1 = (3.71a)
0
 

  
0
At the root:
   

 0  
 
0 

ψ   
 0 
N
{ }N = (3.71b)

 SN  
 SN 


   
  
0 MN

articulated hingeless
rotor rotor
“chapter3(new)” — 2005/11/16 — page 91 — #25

Then, if these vectors are inserted into the transfer equation and only those
nonzero terms are retained, the following equations are formed:
0 = t11 z1 + t12 ψ1 (3.72a)
(ψN ) = t21 z1 + t22 ψ1 (3.72b)
SN = t31 z1 + t32 ψ1 (3.72c)
(MN ) = t41 z1 + t42 ψ1 (3.72d)
where one or the other (but not both) of the terms in parentheses is set to zero.
Thus, for each type of rotor, articulated as well as hingeless, the explicitly written
matrix equation will contain two scalar equations involving z1 and ψ1 [Eq. (3.72a),
and either Eq. (3.72b) or (3.72d)], each of which is equal to zero.

t11 (ω2 ) t12 (ω2 ) z1 0
2 2
= (3.73)
ti1 (ω ) ti2 (ω ) ψ1 0
where i = 4 for articulated rotors and i = 2 for hingeless rotors.

The necessary characteristic equation is formed by equating the determinant
of this equation to zero. As before, an eigensolution is obtained when the iterated
value of frequency squared drives this determinant to zero (determinant iteration
solution):

t (ω2 ) t (ω2 )
11 12
(ω ) =
2
i = 4 or 2 (3.74)
ti1 (ω2 ) ti2 (ω2 )
A root, and hence natural frequency, is obtained when
(ω2 ) = 0 (3.75)
Once the various roots are known, the mode shape calculations follow in a
convenient manner. Because the various n transfer matrices will have already
been evaluated for the solution root (in the process of obtaining the elements of the
characteristic determinant), they are available to relate the inner element responses
to those at the tip. The steps required to find the mode shape for each root are as
follows:
1) Establish the normalizing condition, that is, set z1 or ψ1 equal to some
reference value (usually 1.0).
2) Solve for the other unknown variable at the tip using t12 ψ1 = −t11 z1 . (Note
that the shear and moment are zero at the tip.) This then completely defines the
state vector at the tip.
3) Use the individual transfer matrices for the interior elements to find the state
vectors at the other spanwise stations.
Clearly, knowledge of the state vectors at the several spanwise stations inher-
ently constitutes a solution for the natural mode shape in terms of displacement and
slope. However, even more importantly, as we shall come to appreciate later, the
state vector also provides knowledge of the shears and moments at the spanwise
stations for unit deflection of the mode.
“chapter3(new)” — 2005/11/16 — page 92 — #26

ROTATING BEAMS 93
3.3.2 Finite Element Methods

An alternate generic numerical method for calculating natural frequencies and
mode shapes is a variant of the finite element method. This method can be used not
only for the normal mode calculation task as described herein, but also as a basis
for formulating more general, fully coupled rotor-blade aeroelastic problems. The
principal elements of this methodology are as follows:
1) The equations of motion are obtained using Hamilton’s principle as a starting
point:
t2
(δU − δT − δW ) dt ≡ =0 (3.76)
t1
where δU, δT , and δW are, respectively, the variation of strain energy, the vari-
ation of kinetic energy, and the virtual work done. Generally, these expressions
are subject to considerable diversity among analysts because of the variety of
assumed coordinate systems used for defining the deformed state of the blade and
of assumed coupling effects to be included.
2) For the rotor-blade problem the deformed state of the blade typically consists
of three or four components: a) an out-of-plane displacement component w; b) an
in-plane displacement component v; c) a torsion component φ; and d) occasionally
a radial displacement component u. All of these components are assumed to be
continuous functions of span.
3) The blade structure is discretized into spanwise finite elements, wherein the
boundaries of these elements or spanwise segments are referred to as the nodes.
The values of the deformed state components (i.e., w, v, φ, etc., and their respective
spanwise derivatives) at these nodes are then taken to be the degrees of freedom
of the system (see Fig. 3.16).
4) The distributions of the deflections (w, v, φ, etc.) over each element are then
represented in terms of the nodal displacements using appropriate shape functions.
One common approach to this is to use Hermite interpolation polynomials for the
shape functions. Thus, for the ith spanwise element the deflection components
would be expressed as
 
w 
v = [H]{qi } (3.77)
 
φ i
where the shape function matrix [H] is given by

 
H1 H2 H3 H4 0 0 0 0 0 0
[H] =  0 0 0 0 H1 H2 H3 H4 0 0  (3.78)
0 0 0 0 0 0 0 0 Hφ1 Hφ2
and the vector of element degrees of freedom {qi } is defined as
{qi }T = w1 , w1 , w2 , w2 , v1 , v1 , v2 , v2 , φ1 , φ2 i (3.79)
“chapter3(new)” — 2005/11/16 — page 93 — #27

Fig. 3.16 Typical identification of nodal variables for a blade finite element modeling.
The shape functions are the following Hermite polynomials:
H1 (xi ) = 2(xi /i )3 − 3(xi /i )2 + 1 (3.80a)

H2 (xi ) = i [(xi /i ) − 2(xi /i ) + xi /i ]
3 2
(3.80b)
H3 (xi ) = −2(xi /i )3 + 3(xi /i )2 (3.80c)
H4 (xi ) = i [(xi /i )3 − (xi /i )2 ] (3.80d)
Hφ1 (xi ) = 1 − xi /i (3.80e)
Hφ2 (xi ) = xi /i (3.80f)
where i is the length of the ith element and xi is the local axial coordinate for that
element, as measured from the left end of the element.
5) By use of the preceding formulation, the virtual displacements δw, δv, and
δφ over the ith element can then be defined:
 
δw
δv = [H]{δqi } (3.81)
 
δφ i
6) The total variation given by Hamilton’s principle ( = 0) consists of the

separate contributions from each of the finite elements:
= i =0 (3.82)
“chapter3(new)” — 2005/11/16 — page 94 — #28

ROTATING BEAMS 95
where
t2
i = (δUi − δTi − δWi ) dt (3.83)
t1
The mathematical operations defined by the statement of Hamilton’s principle

(when the expressions for the displacements and virtual displacements in terms
of the nodal variables are substituted) give rise to the following form for the ith
element:
[δqi ][Mi (qi )]{q̈i } + [δqi ][Ci (qi )]{q̇i } + [δqi ][Ki (qi )]{qi } − [δqi ]{Qi } = i
(3.84)
where [Mi (qi )], [Ci (qi )], and [Ki (qi )] represent the inertia, damping, and stiff-
ness matrices, respectively, and {Qi } represents the element load vector for the ith
element.
7) The global matrices are obtained from the assembly of the element matrices.
We denote the global degree-of-freedom vector as {q} and the global virtual
displacement vector as {δq}. Likewise, we denote the global load vector as {Q}. All
three of these vectors are formed by combining the respective components from
each element. When the components of the statement of Hamilton’s principle
are added together, the following expression results:
[δq][M(q)]{q̈} + [δq][C(q)]{q̇} + [δq][K(q)]{q} = [δq]{Q} (3.85)
Because the virtual displacements {δq} are arbitrary, this equation then leads to
the final desired equations of motion:
[M(q)]{q̈} + [C(q)]{q̇} + [K(q)]{q} = {Q} (3.86)
8) The matrix equation of motion (3.86) is typically quite banded, and compu-
tational techniques should be employed to eliminate the need for storing elements
that are always equal to zero and to effect the eigensolution of such banded sys-
tems. The method outlined earlier has elements of a collocation technique in that
a solution to a differential equation is obtained at discrete points with approximate
analytic functionality in between. Further, the use of Hamilton’s principle in this
manner defines a generalized Galerkin method of solution wherein the assumed
shape functions replace the use of normal modes in the more conventional use of
the Galerkin method.
3.4 Approximate Methods

3.4.1 Yntema Charts
The principal approximate method available for estimating the natural frequen-
cies in bending of a rotating beam is that afforded by the charts developed by
Yntema. As formulated for two alternate assumed (simplified) configurations, these
charts enable one to obtain estimations of the nonrotating natural frequencies and
“chapter3(new)” — 2005/11/16 — page 95 — #29

Fig. 3.17 Nonuniform beam approximation for use with the Yntema charts (no tip
weight).
the rise factors for the rotating beam in the in-plane and out-of-plane directions.
Similar to the development given earlier in Sec. 3.1.5 and 3.1.6, the formulations
implicit in the Yntema charts include the assumption that the mode shapes are
invariant with rotor speed. Consequently, the results given for rise factor include
no variation with rotor speed.
Configuration 1. The beam has only linear variations in mass distribution m

and bending stiffness EI. Thus, each of these variables is defined in terms of its
value at the root of the beam ( )0 and at the tip ( )t . For this case the root of the
beam is assumed to begin at a radial station equal to (ēL), as shown in Fig. 3.17.
The natural frequency is assumed to be given by a hyperbolic equation (as
described in an earlier subsection). The charts then present data by which the non-
rotating natural frequency ωNRn and the Southwell coefficient Kn can be determined
as charted functions of the root-to-tip ratios each of mass distribution and bend-
ing stiffness, for each of the first three natural bending modes. Specifically, the
nonrotating natural frequencies and Southwell coefficients are given by
#
EI0
ωNRn = an (3.87)
m0 L 4

K0n + ēK1n (out-of-plane modes)
Kn = (3.88)
K0n + ēK1n − 1 (in-plane modes)
The quantities an , K0n , and K1n are given in the charts formulated by Yntema for
each of the first three bending modes (n = 1, 2, 3); excerpts of these charts are
presented in Appendix B.
Configuration 2. The beam has constant values of mass and stiffness distri-
butions, but has a variable concentrated tip weight MT , as shown in Fig. 3.18.
This configuration implicitly addresses those blade designs wherein the mass
characteristics are dominated by the tip weight. For this case (using the original
nomenclature) the nonrotating natural frequency is given by
#
EI0
ωNRn = θn2 (3.89)
m0 L 4
“chapter3(new)” — 2005/11/16 — page 96 — #30

ROTATING BEAMS 97
Fig. 3.18 Uniform beam approximation for use with the Yntema charts (with tip
weight).
The Southwell coefficient Kn is determined using values of K0n as in the case of

the nonuniform beam without tip weight (except that ē is assumed to be zero and
hence K1n is not given). Both θn2 and K0n are given in the charts as functions of
the ratio of tip mass to beam mass (MT /m0 L). Note that Fig. B.3 in Appendix B
presents the offset correction factors for hinged blades for the zeroth, or pendulum
mode. Because this particular mode is independent of the blade stiffness, it is
possible to present variations in both tip mass ratio r as well as the ratio defining
linear varying mass distribution mt /m0 .
3.4.2 Equivalent Distributions

One of the difficulties of using the Yntema charts is that, with the exception
of the configuration 2 results, which assume a concentrated mass, the assumed
mass and distributions are either uniform or linearly variable with span. However,
most blade designs tend to have their greatest complexity at the inboard end,
resulting in mass and stiffness distributions that are typically highly nonlinear and
usually stepped. Because the mass characteristics of the blade are most sensitive
to “outboard” masses, the inboard complexity of the given design will not impact
greatly on the mass characteristics. Consequently, the problem of determining
equivalent mass distributions is determined by suitably averaging the outboard
mass distribution. Clearly, some engineering judgment must be made here, but the
scope of the problem is facilitated by the fact that mass distributions of typical
blade designs tend to be reasonably smooth and approach the linear distributions
implicit in the Yntema charts.
Equivalent EI distributions. The material in this section presents some

approximate methods for reducing actual EI distributions to ones equivalent to
those used by Yntema.
EI equivalencies over short sections of the blade. Over any section of the
blade of total length , wherein the EI distribution is undergoing rapid variations
with span, an approximation can be formulated based on the maintenance of
similar rotation per moment characteristics. To this end, we note that the basic
beam bending equation is given by the following usual form:
M = EIw (3.90)
“chapter3(new)” — 2005/11/16 — page 97 — #31

Or, in terms of rotation angle θ ,

M
θ = w = w dx = dx (3.91)
0 0 EI
which over a series of small span lengths i (all within the basic length , wherein
the moment can be expected to have a relatively constant value), the relationship
can be written by equating the sum of the bending rotations to the total angular
rotation:
Mi i M
θ= θi = =M = (3.92)
EIi EIi EIequiv
i i i
Therefore,
$
i i
= ; → EIequiv = $ i (3.93)
EIequiv EIi i (i /EIi )
i
EI distributions over the entire length of the blade. In some cases we require
that an arbitrary EI distribution over the entire blade span be reduced to an equiv-
alent linear variation, as discussed earlier. One basis for calculating an equivalent
linear variation would be to equate the bending strain energy of the real blade result-
ing from the application of a load at the blade tip, with that strain energy resulting
for the blade with the linear EI variation. In effect, this amounts to having both the
real blade and the linear EI equivalent blade have the same tip deflection per load
characteristics. This equivalency condition does not, however, produce a unique
answer in that the linear EI variation blade needs the values of EI at both the
root EI0 and the tip EIt [or, equivalently, a root EI(EI0 ) and a tip-to-root EI ratio,
REI (=EIt /EI0 )]. A further equivalency condition would be to require that both the
real blade and the equivalent blade have the same tip slope per load characteristics.
The following material presents derivations resulting in a final useful equivalency
relationship. The starting point is to solve the basic (nonrotating) beam equation
wherein a linear variation of EI is assumed and a moment distribution is taken
appropriate to a load at the beam tip. Let us further define the spanwise location
in terms of a nondimensional variable η(= x/L) and formulate expressions for the
resulting EI and moment distributions:
EI = EI0 [1 − (1 − REI )η] (3.94a)

M = FL(1 − η) (3.94b)
Then
% &
M FL (1 − η)
w = = (3.95)
EI EI0 1 − (1 − REI )η
“chapter3(new)” — 2005/11/16 — page 98 — #32

ROTATING BEAMS 99
Subsequent integration yields the following expressions for slope and deflection
at the end of the beam (θt and wt , respectively):
FL 2 /EI0
θt = [REI nREI + 1 − REI ] (3.96)
(1 − REI )2
FL 3 /EI0
wt = [−REI
2
nREI + 1/2(1 − 4REI + 3REI
2
)] (3.97)
(1 − REI )3
These two equations can then be combined to give the following expression for
REI in terms of the blade tip deflection and slope:
1/2 − wt
REI = (3.98)
θ t − wt
where
EI0
wt ≡ wt (3.99a)
FL 3
EI0
θ t ≡ θt 2 (3.99b)
FL
The previously stated equivalency condition can also be used to obtain an
expression for the tip value of EI:
1/2 − EI0 wt /FL 3
EIt = (3.100)
θt /FL 2 − wt /FL 3
The use of the preceding equations still does not constitute a unique solution
for EI0 and EIt , as required for using the Yntema charts. For any given application
the (nonrotating) static deflection characteristics of the blade (wt and θt ) would
be known, as well as the length L. However, the equivalency condition stated
earlier only gives REI once a selection for EI0 is made. Note that the appropriate
value of EI0 to be used is not necessarily the EI0 value for the original blade.
Furthermore, although the nondimensional tip slope will generally be greater than
the nondimensional deflection, the range of root EI available is restricted only
to values that result in a finite, positive value for the tip EI. Let us develop this
further. What the engineer needs is some measure of the ratio of tip stiffness to root
stiffness (REI = EI t /EI 0 ) for a blade design that has two fairly distinct spanwise
portions with distinct values of EI. This configuration is depicted in Fig. 3.19,
wherein the inboard (elevated) value of EI is related to the outboard value, EI1 , by
the factor β. This inboard extend is arbitrary and defined by the factor η1 .
Again, we assume a unit tip load and form expressions for the tip deflection and
slope:
' (
EI1 1 η13 1
wt = η1 − η1 +
2
+ (1 − η1 )3 = W (3.101)
(1)L 3 β 3 2
EI1 η1 η1 1
θ t = 1 − + (1 − η1 )2 = (3.102)
(1)L 2 β 2 2
“chapter3(new)” — 2005/11/16 — page 99 — #33

Fig. 3.19 Idealization of a blade with two distinct bending stiffnesses.
Then, with a rearranging of Eq. (3.99) the preceding expressions can be inserted
to give the required ratio:
η1 ) * 1
β 1 − η1 + η1 3 + 3 (1 − η1 )
2 3
W
REI = = ) 2 * 1
1
2 +W − 1
2 + β −η1 /2 + η1 /3 + 3 (1 − η1 ) − 2 (1 − η1 )
1 3 3 1 2
(3.103)
Figure 3.20 presents the evaluations of this expression for REI for a range of values
of β and a practical variation of η1 :
Fig. 3.20 Equivalent stiffness ratios for use with the Yntema charts.
“chapter3(new)” — 2005/11/16 — page 100 — #34

ROTATING BEAMS 101
Fig. 3.21 Alternate equivalent EI distributions for a nonlinear (stepped) EI distribu-

tion (Example 3.2).
Example 3.2
Consider the case of a blade (beam) that has, over the inboard third of its length, a
value of EI equal to 5 and, over the outboard 2/3 of its length, a value of EI equal
to 1. Then, assuming a unit tip load F and unit length, the tip deflections can be
readily calculated from simple beam theory to have the following values:
wt = 0.14568 and θ t = 0.27778
The equivalency equation, Eq. (3.100), expressing EIt in terms of EI0 , gives
EI0 = 3.4322 − 0.9068EIt (3.104)
The original and various resulting equivalent EI distributions are shown in

Fig. 3.21. If the original EI value of the tip (EIt = 1) is used, the root value must
be 2.52. If the tip value is reduced to 0.5, then the root value must be 2.98, and if
the tip value is completely reduced to zero then the root value is 3.43. Note that
the original value of root EI(EI0 = 5) cannot be used because it would result in a
negative value of EIt .
3.5 Two-Bladed Rotor

An enduring rotor configuration is that of the two-bladed type because of a
number of considerations. Two-bladed rotors enable compact storage of the rotor-
craft, have relatively simpler hubs, and are the least expensive. Two-bladed rotors,
as a rule, are stiff in-plane and as a result do not need lead-lag dampers, as do
“chapter3(new)” — 2005/11/16 — page 101 — #35

Fig. 3.22 Pictorial representation of the coupling of the rotor in-plane bending with
shaft lateral flexibility.
three-or-more articulated rotors, to preclude pitch-lag and ground resonance insta-

bilities (see Chapters 10 and 11). They are usually the configuration of choice for
small helicopters and gyroplanes. The one significant dynamic characteristic of
two-bladed rotors is that they are anisotropic in that they have different inertia
and stiffness properties parallel and normal to the axes of the blades. Because
of this anisotropy, the lateral stiffness of the main rotor drive shaft and/or other
flexibilities in the rotor support structure can play a significant role in the in-plane
dynamics of the rotor. As shown in Fig. 3.22, the asymmetric elastic in-plane
bending of the rotor blades creates an excursion of the rotor center of gravity that
can couple with the flexibility of the main rotor shaft.
An examination of the dynamic properties of the two-bladed rotor starts by
nondimensionalizing the key parameters. Here we again consider a uniform rotor
blade with radius R, mass distribution m0 , and in-plane stiffness EI edge . The rotor
speed is nondimensionalized by a reference frequency ω0 and the shaft stiffness
by the blade stiffness:
#
EIedge
ω0 = (3.105)
m0 R4
EIedge
K0 = (3.106)
R3
Figure 3.23 presents the pertinent dynamic characteristics that arise from the
coupling of the rotor dynamics with the lateral stiffness of the rotor shaft. A key
characteristic is that the in-phase mode is unstable in divergence for lateral stiff-
nesses below a certain value. Also in evidence is the sharp increase in natural
frequency for the out-of-phase mode with increasing lateral stiffness while the in-
phase mode approaches an asymptotic value. For reasons relating to ground reso-
nance stability, the rotor is generally designed to be subcritical. That is, the resulting
“chapter3(new)” — 2005/11/16 — page 102 — #36

ROTATING BEAMS 103
Fig. 3.23 Typical out-of-phase and in-phase frequency variations with lateral stiffness
for the two-bladed rotor (/ω0 = 1).
natural frequency must be greater than the rotor frequency (i.e., /ω < 1).
The converse of subcriticality is supercriticality wherein /ω > 1. Figure 3.24
presents a map of the various dynamic operational conditions as determined by
nondimensional rotor speed and lateral stiffness.
Fig. 3.24 Map of significant dynamic operational conditions for the two-bladed rotor.
“chapter3(new)” — 2005/11/16 — page 103 — #37

The figure clearly demonstrates that the effective lateral stiffness of the shaft
and/or support structure must be sufficiently robust and the blades sufficiently
stiff to ensure that the coupled frequencies remain subcritical. Note that even for
infinite lateral stiffness supercritical operation will occur for (/ω0 )2 ≥ 15.03.
Also of interest is the boundary for the in-phase coupled frequency reaching a
value of twice rotor speed (ω = 2). Such a condition leaves the rotor susceptible
to resonance excitation from the two-per-rev component of the in-plane airloads
in forward flight and should be avoided.
3.6 Blade Torsion Dynamics

3.6.1 Geometric Description of the Blade Torsion Kinematics
Figure 3.25 presents the basic kinematic description of the beam in torsion.
Of particular importance is the assumption that the elastic axis is a straight line.
Furthermore, we assume for the moment that the blade torsion motion is uncoupled
from bending. In later sections we will deal with the features of this coupling. As
depicted in Fig. 3.25, (x, t) is the geometric pitch angle at the station x and is
comprised of the control angle θ0 , the built-in twist angle distribution θB (x), and
the elastic torsion deformation θe (x, t)
(x, t) = θ0 (t) + θB (x) + θe (x, t) (3.107)
Fig. 3.25 Rotating beam geometry defining torsional kinematics.
“chapter3(new)” — 2005/11/16 — page 104 — #38

ROTATING BEAMS 105
3.6.2 Basic Elastic Description for Blade Torsion Equilibrium

The blade torsion description begins with an equilibration of the loadings
with the elastic restoring moment. Because of the relatively high aspect ratio
of rotorcraft blades, the high degree of deformation possible admits a wide range
of possible torsional excitation from the applied loadings. Consequently, a full
accounting of the higher-order loadings is required. It has been found, moreover,
that this accounting generates essential nonlinearities. Furthermore, the elastic
torsional restoring moment itself is characterized by nonlinearities caused by
couplings with bending and with terms that arise from the warping of typically
noncircular blade cross sections. These coupling terms have been well addressed
by Hodges and Dowell, who developed the full nonlinear equations of motion.
Their developments were found to linearize to the pioneering work in this area of
Houbolt and Brooks. These earlier results have been combined with other formu-
lations to establish the basic elastic torsional dynamic equation presented herein.
First, however, some basic ideas and nomenclature must be established.
One principal difficulty in formulating the correct equation of motion for the
blade in torsion is the fact that the elastic axis of the blade is not a straight line, but a
time-variable space curve caused by the elastic bending deformations of the blade
in the chordwise direction ve and in the flatwise direction (normal to the edgewise)
we . Another difficulty in arriving at the necessary formulation is the requirement
for using several coordinate systems to describe the inertial position of a point
on the deformed elastic axis. In particular, the description of the inertia loadings
requires that the accelerations relative to inertial space (the 0 system) be expressed
in terms of motion in a coordinate system that includes motion of the nonrotating
rotor hub (the 1 system), the rotor rotation (the 2 system), the hinge offset position
(the 3 system), the rigid lead-lagging motion (the 4 system) and, finally, the rigid
flapping motion (the 5 system). As shown in Fig. 3.26, the coordinate system used
to define the elastic bending motion (denoted the 5 system: x5 , y5 , z5 ). This system
is taken to be that wherein the elastic deformations are isolated from the rigid-
body flapping and lead-lag motions β and ζ , respectively, in the case of articulated
blades and from the built-in coning βB in the case of hingeless blades. Additionally,
this coordinate system then conveniently provides a definition for the in-plane y5
and out-of-plane z5 displacements, respectively.
Figure 3.26 indicates the local directions of the edgewise and flatwise bending
deformations that are required for determining the actual displacements of the
elastic axis in the five-coordinate system. This determination is complicated by
the fact that the blade has a time-varying (total) pitch angle and, because of
blade twist (resulting from both θB and θe ), the transformation of edgewise and
flatwise deformation to the five-coordinate system cannot be made using a simple
trigonometric (rotational) coordinate transformation on the displacements. As will
be shown later, the transformation must be made on the curvatures:
y5 = ve cos − we sin (3.108)

z5 = ve sin + we cos (3.109)
This coordinate system transformation has the advantage that the force boundary
conditions at the tip of the blade (zero moment and shear) are preserved. With these
“chapter3(new)” — 2005/11/16 — page 105 — #39

Fig. 3.26 Schematic of the blade five-coordinate system.
definitions and concepts the basic equation of motion for the blade in torsion can
then be addressed. Based on the observation already discussed that, because of the
relatively high aspect ratio of rotor blades, the blade elastic torsional deformation
takes on the character of a “space curve,” attention must be paid to the resulting
nonlinearities. The nonlinearities derive from the action of forces (blade loadings)
and moment arms that are provided by blade bending. The detailed derivation of
the resulting basic elastic description for the blade in torsion is beyond the scope
of this text, and the reader is referred to the cited literature. For present purposes
the basic torsion equation, which contains the essential characteristics resulting
from the space-curve character, is presented without derivation as follows:
+
−[GJθ + Tka2 θ + EB1 (θB )2 θ ] (elastic terms)
+
= [qx5 + y5 qy5 + z5 qz5 ] (“distributed” moment loadings)
+ y5 Fz ( px5 , pz5 , qy5 ) − z5 Fy ( px5 , py5 , qz5 )

(“nonlinear” moment loadings resulting from curvatures
× force loadings) = [−(EIE −EIF )ve we
+ ··· + (higher-order terms)] (3.110)
where px5 , py5 , and pz5 represent the five-coordinate system components of the
force loading distribution, and qx5 , qy5 , and qz5 similarly represent the components
of the moment loading distribution. The Fy and Fz functions are complicated
integral functions of the loadings and deformations and are not given herein for
clarity. The reader is referred to the works of Bielawa and of Loewy, Rosen,
and Mathew for detailed treatments of this approach. Note that the nonlinear
“chapter3(new)” — 2005/11/16 — page 106 — #40

ROTATING BEAMS 107
moment loadings are expressed as either an integral formulation or in a compact

differential form. This particular nonlinearity forms the subject matter of numerous
investigations into the nonlinear aeroelastic behavior of rotors. As such, we will
return to the subject of rotor nonlinearities later; for now we concentrate on the
linear behavior of blades in torsion.
Linearized torsion equation. A zeroth-order linearization for the blade in

torsion is achieved by deleting the nonlinear terms identified in Eq. (3.110)
(linearization by truncation) and substituting the appropriate expression for the
torsional loading qx5 :
2
mkm θ̈e − {[GJ + Tka2 + EB1 (θB )2 ]θe } + 2 m[(km
2
2
− km
2
1
) cos 2θ]θe
= Mapp = Mθ + (Tka2 θ ) − 2 m[(km
2
2
− km
2
1
) sin θ cos θ] (3.111)
where
θ (x, t) = θ0 (t) + θB (x) (3.112)
km1 , km2 ≡ mass radii of gyration of the blade section about the major (chord-
wise) neutral axis and about an axis perpendicular to that axis and
through the elastic axis, respectively
(note that the polar moment
of inertia is defined by km = km 2 + k2 )
1 m2
GJ ≡ primary (St. Venant) torsional stiffness
T ≡ local tension
ka ≡ polar radius of gyration of the tensile stress carrying area about
the elastic axis (note that this term is of questionable validity for
isotropic materials with circular sections and built-in twist)
EB1 ≡ incremental torsional stiffening (coiled spring effect) = (bending
modulus of elasticity E) × (section constant defined in Houbolt and
Brooks):

ηLE t2
B1 = tη2 η2 + − ka2 dη (3.113)
ηTE 6
where t is the chordwise variable thickness, η is a chordwise position integration

variable going from the trailing edge (TE) to the leading edge (LE), ka is as defined
earlier, and Mθ (x, t) is the applied moment loading (e.g., aerodynamic moment
loadings).
Boundary conditions are as follows:
Zero torque at the tip:
θe (L) = 0 (3.114a)
Cantilever root elastic deformation:
θe (0) = 0 (3.114b)
“chapter3(new)” — 2005/11/16 — page 107 — #41

Elastic root pitching motion:
Kθ θe (0) = GJθe (0)

(3.114c)
= root moment
Significance of various terms is given here:
mkm2 θ¨ ⇐ inertial (torsional) moment

e
[GJθe ] ⇐ usual St. Venant torsional stiffness

[Tka θe ]
2 ⇐ “bifilar” stiffening, ∼2
2 m[(km 2 − k 2 ) cos 2θ ]θ ⇐ propeller moment (tennis racquet effect)
2 m1 e
[EB1 (θB )2 θe ] ⇐ incremental (coiled-spring-like) elastic stiffening
Simplified rigid-body oscillations. In the limit, as the blade bending stiffness

approaches infinity, it can be shown that, at least for the articulated rotor blade, the
blade motion reduces to that of rigid-body flapping about the articulation hinge.
In this configuration the resulting natural frequency is completely deterministic,
depending solely on the hinge (offset) length, the radius of gyration about the
hinge point, the mass center location, and the rotor rotational speed . In fact,
the resulting frequency is seen to be directly proportional to rotor speed. In a like
manner the torsional characteristics of a rigid blade in torsion can be similarly
defined. To this end we make the following assumptions:
1) The torsional response is constant along the blade span and equal to the root
measured value:
θe (x, t) = θR (t) (3.115)
2) The root restraint is zero. This corresponds to a blade that has lost its pitch
control attachment to the swash plate.
3) The blade cross section has a high aspect ratio and thus is quite thin compared
with its chord. This is expressed mathematically as
2
km 2
2
km 1
−→ km
2
≈ km
2
2
(3.116)
4) Both the impressed pitch angle θ (= θ0 + θB ) and the torsional response

motion are “small”: θ, θR 1. The resulting dynamic equation then becomes
2
mkm [θ̈R + 2 θR ] = Mθ − mkm
2 2
θ (3.117)
One principal conclusion to be drawn from this result is that the natural freq-
uency of the blade for rigid-body torsion is the rotor frequency itself. The
significance of such a value of frequency is that the 1P (1/rev) control loads that
are required to impart cyclic pitching motion to the blade are significantly reduced
because for 1P motion the dynamic system is being driven at its natural frequency
and hence obtains a dynamic attenuation of the resisting (dynamic) control loads.
“chapter3(new)” — 2005/11/16 — page 108 — #42

ROTATING BEAMS 109
Thus, the main control forces to be taken up by the control system are the non-
1P propeller moments (principally the steady loads) and all harmonics of the
aerodynamic damping in pitch:
L% &
dM
Mθ = dx ∼ −θ̇ (3.118)
0 dx aero
Note that this aerodynamic damping is not generally small and constitutes a
principal source of the 1P control loads.
3.6.3 Solution for Natural Frequencies

Blade discretization. The starting point for calculating the torsion natural
frequencies is the linearized dynamic equations Eq. (3.111). An efficient method
for obtaining the natural frequencies and mode shapes for the linearized torsion is
the Holzer–Myklestad technique described in Sec. 3.3.1. From a practical stand-
point the blade can be divided into a reasonable number of spanwise segments,
each of which has length Ln and generally distinct, but constant properties, shown
in Fig. 3.27.
The conventional use of this technique assumes sinusoidal motion; thus, over
each segment the equation of motion for the sinusoidal amplitude θ̄e takes the
following form:

−ω2 mkm 2
θ̄e − GJeff θ̄e + 2 m km
2
cos 2θ θ̄e = 0 (3.119)
Fig. 3.27 Pictorial representation of a typical torsional blade segment.
“chapter3(new)” — 2005/11/16 — page 109 — #43

This is a straightforward linear second-order differential equation whose solution

is given by
θ̄e (x) = A cos λx + B sin λx (3.120)
where
#
2 − 2 k 2 cos 2θ)
,
m(ω2 km m α
λ= ≡ (3.121a)
GJeff β
2
km = km2
2
− km1
2
(3.121b)
The torque over each segment is given by
Q(x) = β θ̄e = βλ [−A sin λx + B cos λx] (3.122)
The A and B coefficients are determined by the deflection and torque conditions
at the two ends of the segment. Eliminating these coefficients gives rise to the
following matrix equation:
1

cos λLn sin λLn θ̄e θ̄
βλ = e (3.123)
−βλ sin λLn cos λLn Q n−1
Q n
Because the Holzer–Myklestad technique is best implemented by working from

the blade tip conditions inward to the blade root conditions where alternate boun-
dary conditions are to be applied, Eq. (3.123) has to be inverted to give the final
most useful form:

θ̄e cos λLn − βλ
1
sin λLn θ̄e θ̄
= = [Tn ] e (3.124)
Q n−1 βλ sin λLn cos λLn Q n Q n
This equation is then used to cascade multiply from one segment to the next to form
a matrix relationship relating the boundary conditions at the blade root, {**}0 to
those at the tip {**}N :

θ̄e θ̄e θ̄
= [T1 ] [T2 ] · · · [TN ] = [T1N ] e (3.125)
Q 0 Q N Q N
The boundary conditions at the blade tip are simply that the deflection is unity
and that the torque is zero [QN = 0]. Those at the blade root are determined by
the presence of a finite spring rate Kθ :
Infinite spring rate: θ̄e (0, t) = (θ̄e )0 = 0[(θ̄e )N = 1] (3.126a)

Finite spring rate: [GJeff ]1 θ̄e (0, t) = Q0 = Kθ (θ̄e )0 (3.126b)
“chapter3(new)” — 2005/11/16 — page 110 — #44

ROTATING BEAMS 111
Typical torsion modal characteristics. From the dynamic equations given

earlier for both the elastic and rigid-body cases, it is apparent that the rotating
beam (rotor blade) is significantly stiffened by centrifugal forces in torsion just
as it is in transverse bending. Indeed, the natural frequency characteristics of the
blade in torsion can be described in approximate terms as being comprised of a
nonrotating natural frequency part and a rotor speed dependent part governed by
an appropriate rise factor. Thus, for any given jth torsion mode the rotor speed
variation is expressed as
ωθ2j = ωNR
2
θ
+ Kθj 2 (3.127)
j
The mode shapes appropriate to torsional vibrations are governed by the

boundary conditions already identified. Figure 3.28 depicts typical mode shapes
Fig. 3.28 Typical torsional mode shapes for a rotating beam.
“chapter3(new)” — 2005/11/16 — page 111 — #45

for configurations that are, respectively, constrained by a rigid (cantilevered)

attachment at the root and attached by means of a discrete root torsion spring.
Note that, in the case of the flexible root constraint, the mode shape must reflect
a realistic torsion moment as indicated by a first derivative in deflection having the
same sign as the deflection itself.
3.7 Coupling Effects

In the preceding material the various types of blade modal response [flapping (or
flatwise) bending, in-plane (or edgewise), or torsion] were all treated separately,
or uncoupled. In reality, however, all of these types of response are coupled to
some degree with each other. A complete description of the necessary additional
formulations needed to achieve this coupling with complete accuracy is outside
the intended scope of this text, and only some of the more important aspects of the
various types of coupling are described.
3.7.1 Sources of Coupling

Coupling effects arise in rotor blades principally from four main sources:
1) general and usual misalignments of the blade section principal axes (both
inertial and elastic) from the spin plane caused by all components of blade pitch
angles and built-in twist, and θB , respectively;
2) noncoincidence of the various elastomechanical centers within the blade
sections (mass center, elastic axis point, neutral axis point);
3) skewness of the blade’s radial axis from the spin plane caused by built-in
coning or flapping; and
4) nonlinear effects of combined flatwise and edgewise bending.
The coupling caused by misalignment of the blade section principal axes from
the spin plane (as a result of blade section pitch and twist angles) is the principal
(linear) coupling mechanism between out-of-plane and in-plane bending. This
coupling is directly controlled by the blade section pitch angle , as shown in
Fig. 3.29. The two basic ways of accounting for this type of coupling are to use
either elastic coupling terms or inertia coupling terms.
3.7.2 Elastic Coupling

As in an earlier section, let us denote the in-plane and out-of-plane directions
as y5 and z5 , respectively, and the edgewise and flatwise directions as ve and we ,
respectively. Then with an elastic coupling scheme the elastic bending moments
within the blade section are related to the respective curvatures by the following
expressions:
My5 = (EIF cos2 + EIE sin2 )w5 + (EIE − EIF ) sin cos v5 (3.128)
Mz5 = (EIE − EIF ) sin cos w5 + (EIF sin2 + EIE cos2 )v5 (3.129)
where EIE and EIF are the bending stiffnesses in the edgewise (or chordwise) and
flatwise directions, respectively.
“chapter3(new)” — 2005/11/16 — page 112 — #46

ROTATING BEAMS 113
Fig. 3.29 Typical blade section with arbitrary bending and torsional deformations.
The internal elastic moments defined by Eqs. (3.128) and (3.129) can then
be combined with the inertial and tension loads to produce the following set of
dynamic equations for bending in the two directions:
d
mv̈5 − [v5 T ] − m2 v5 + (EIE − EIF ) sin cos w5
dx

+ (EIF sin2 + EIE cos2 )v5 = Fy5 (x, t) (3.130)
d
mẅ5 − [w T ] + (EIE − EIF ) sin cos v5
dx 5

+ (EIF cos2 + EIE sin2 )v5 = Fz5 (x, t) (3.131)
This type of formulation gives rise to mode shapes and frequencies that are
generally referred to as coupled modes, inasmuch that each mode has both in-plane
and out-of-plane components:
γv (x)
v5 (x, t)
= q (t) (3.132)
w5 (x, t) γw (x) n n
n
where, for free vibration
qn (t) = q̄n exp(iωn t) (3.133)
The principal difficulty with this type of coupling scheme is that the use of v5 and
w5 as components of blade modal motion implies that they are being represented
by coupled modes that are necessarily determined at some fixed value of the total
pitch angle . Although this requirement is appropriate for the dynamic analysis
of propeller blades, it is untenable for helicopter rotor blades because the latter
must necessarily undergo substantial variations in (cyclic) pitch even in steady
flight conditions.
“chapter3(new)” — 2005/11/16 — page 113 — #47

3.7.3 Inertial Coupling with Zero Twist

With an inertial coupling scheme the blade deflections are typically defined
in the edgewise and flatwise directions and the elastic bending moments remain
uncoupled with each other:
MF = EIF we (3.134)
ME = EIE ve (3.135)
This type of kinematic description is often referred to as the use of uncoupled
modes. The coupling that ensues between these uncoupled bending components ve
and we arises from the condition that the inertia loads are neither proportional to nor
aligned with the edgewise and flatwise directions. Consequently, moments applied
to the elastic sections (from sections at adjacent blade stations) will have inertial
load components in both directions that will then couple these two-degrees-of-
freedom components. Stated another way, the inertia loads are most conveniently
expressed in terms of the in-plane and out-of-plane motions of the blade section
v5 and w5 , respectively. Therefore, the use of uncoupled modes requires that
these components of blade motion be expressible in terms of the edgewise and
flatwise motions ve and we , respectively. Reference to Fig. 3.29 shows that the
transformation of the flatwise and edgewise motions to in-plane and out-of-plane
motions can be accomplished relatively easily for the case of a radially constant
pitch angle, such as would arise from a control angle without a built-in twist (i.e.,
= θ0 ):
v5 = ve cos θ0 − we sin θ0 (3.136)
w5 = ve sin θ0 + we cos θ0 (3.137)
In this instance, because the blade pitch angle would be constant over the span,
Eqs. (3.136) and (3.137) could be used to formulate valid coupling inertia loads.
The internal elastic moments are now defined by Eqs. (3.134) and (3.135) and
the inertia and rotational terms must be transformed using Eqs. (3.136) and (3.137).
When these operations are performed, the following set of dynamic equations for
bending in the two directions are obtained:
d
m(v̈e cos θ0 − ẅe sin θ0 ) − [(v cos θ0 − we sin θ0 )T ]
dx e

− m2 (ve cos θ0 − we sin θ0 ) + (EIE − EIF ) sin θ0 cos θ0 w5

+ (EIF sin2 θ0 + EIE cos2 θ0 )v5 = Fy5 (x, t) (3.138)
d
m(v̈e sin θ0 + ẅe cos θ0 ) − [(ve sin θ0 + we cos θ0 )T ]
dx

+ (EIE − EIF ) sin θ0 cos θ0 v5

+ (EIF cos2 θ0 + EIE sin2 θ0 )v5 = Fz5 (x, t) (3.139)
“chapter3(new)” — 2005/11/16 — page 114 — #48

ROTATING BEAMS 115
To complete the formulation, these equations must be trigonometrically

resolved by appropriately multiplying by cosine and sine of θ0 . After these
resolutions are performed, the final forms of the inertia coupled equations become
d
mv̈e − [v T ] − m2 (ve cos2 θ0 − we sin θ0 cos θ0 )
dt e
+ [EIE ve ] = Fy5 cos θ0 + Fz5 sin θ0 = FE (x, t) (3.140)
d
mẅe − [w T ] − m2 (−ve sin θ0 cos θ0 + we sin2 θ0 )
dt e
+ [EIF we ] = Fz5 cos θ0 − Fy5 sin θ0 = FF (x, t) (3.141)
This type of formulation is typically referred to as using uncoupled modes
because the equations use uncoupled modes to describe ve and we :

ve = γv(m) (x)qv(m) (t) (3.142a)
m

we = γw(m) (x)qw(m) (t) (3.142b)
m
where, for free vibration

qw(m) (t) = q̄w(m) exp[iω(m) t] (3.143a)
qv(m) (t) = q̄v(m) exp[iω(m) t] (3.143b)
However, in the general case rotor blades do have twist. Hence, the edgewise
and flatwise directions are not generally the same from one section to the next.
The trigonometric transformation defined by Eqs. (3.136) and (3.137), which
deals with displacements, consequently does not work with the inclusion of twist.
This is because it is impossible to satisfy the force boundary conditions using ve
and we elastic deformation functions (mode shapes) even though the deformation
functions themselves do. This difficulty can be amply demonstrated by differenti-
ating Eqs. (3.136) and (3.137) to obtain the second and third spanwise derivatives
wherein now includes a spanwise variation caused by twist.
3.7.4 Inertial Coupling with Nonzero Twist

The use of elastic coupling already described has the advantage that the elastic
couplings that arise from the rotation of the elastic principal axes because of pitch
are expressed in terms of the total pitch angle , which inherently includes the
part caused by built-in twist θB . Because of the modeling advantages afforded
with the use of uncoupled modes, a way of including the coupling effects caused
by twist would therefore be highly desirable. One way of including twist in an
uncoupled modal approach is to use a trigonometric transformation similar to that
given earlier, not on displacement, but on curvature:
v5 = ve cos − we sin (3.144)
w5 = ve sin + we cos (3.145)
“chapter3(new)” — 2005/11/16 — page 115 — #49

Thus, if ve and we are expressed in terms of summations of normal uncoupled

modes

Nv
ve = γvm (x)qvm (t) (3.146)
m=1

Nw
we = γwi (x)qwi (t) (3.147)
i=1
then multiple applications of integration by parts result in approximations for the

required transformations of the v5 and w5 displacements in terms of the flatwise
and edgewise modal displacement functions, which are accurate to second order
in twist rate:
v5 = (ve + v− V ) cos − (we − w− W ) sin + O(θ 3 ) (3.148)

w5 = (ve + v− V ) sin + (we − w− W ) cos + O(θ 3 ) (3.149)
where v, w, V , and W are integral functions of the modal displacement and

slope functions and the twist rate. These functions are broadly denoted as deflection
correction functions and are defined by the following integral expressions:
First order in twist:
x̄ x̄ x̄1
v= θ we dx̄1 + θ we dx̄2 dx̄1 = v (1) + v (2) (3.150)
0 0 0
x̄ x̄ x̄1
w= θ ve dx̄1 + θ ve dx̄2 dx̄1 = w(1) + w(2) (3.151)
0 0 0
Second order in twist:

x̄ x̄ x̄1
V= θ w dx̄1 + θ w(2) dx̄2 dx̄1 (3.152)
0 0 0
x̄ x̄ x̄1

W= θ v dx̄1 + θ v (2) dx̄2 dx̄1 (3.153)
0 0 0
where (− ) denotes nondimensionalization with respect to the blade radius R and

( ) denotes differentiation with respect to x̄. Note that in these expressions for
deflection correction vectors the total twist rate θ can be arbitrarily nonlinear and
even discontinuous. Furthermore, the total twist rate is comprised of a built-in
component θB , as well as the elastic component θe . Thus, the resulting equations
for v5 and w5 (with the inclusion of the deflection correction terms) [Eqs. (3.148)
and (3.149)] can be seen to include linear contributions from both flatwise and
edgewise bending (acting in the presence of built-in twist), as well as nonlinear
contributions from elastic torsion (acting in the presence of elastic flatwise and/or
“chapter3(new)” — 2005/11/16 — page 116 — #50

ROTATING BEAMS 117
edgewise bending). Equations (3.150–3.153) can be greatly simplified if the built-

in twist rate θB is constant. In this case the expressions for the deflection correction
functions caused by built-in twist become:
x̄
vB = 2θB we dx̄1
0
x̄
wB = 2θB ve dx̄1
0
(3.154)
x̄ x̄1
VB = 3(θB )2 ve dx̄2 dx̄1
0 0
x̄ x̄1
WB = 3(θB )2 we dx̄2 dx̄1
0 0
3.7.5 Frequency and Mode Shape Characteristics

The inclusion of coupling effects in the combined blade bending-torsion dynam-
ics mainly modifies the details but not the general trends of frequencies that would
otherwise be defined by the uncoupled configuration. In particular, instead of hav-
ing two or more uncoupled frequencies of the same value at any given rotor speed,
the frequencies will tend to coalesce and form separate discrete values in the vicin-
ity of such a rotor speed. The coupled response mode shapes usually retain their
principal characteristics with respect to the predominant deflection components
inherent in the modes (e.g., first “flatwise-like”). However, as coupling factors
become more pronounced, the modal response will increasingly include additional
components of motion as well. For some cases wherein two or more frequencies
are coalescing (as a result of the uncoupled frequencies being nearly equal), the
specific predominant character of any of these modes can be lost, and the accurate
designations of such a mode can become quite ambiguous.
Figures 3.30 and 3.31 demonstrate typical coupling characteristics. Both fig-
ures present modal characteristics for a representative hingeless rotor blade with
progressive amounts of coupling, as defined by 1) zero pretwist and zero control
pitch angle (the uncoupled case), 2) zero pretwist and a control pitch angle of 22.5
deg, and 3) both a linear pretwist value of −40 deg (from the blade root to the
tip) and, again, a control pitch angle of 22.5 deg. Figure 3.30 presents the modal
frequencies for the first three “flatwise-like” modes (1F, 2F, and 3F) and the first
“edgewise-like” mode (1E). Note that the 1, 2, and 3/rev (1P, 2P, and 3P) lines
are superimposed on the figure as well. Figure 3.31 presents the radial distributions
of the out-of-plane and in-plane components w and v, respectively, for the second
lowest frequency (1F) mode set given in Fig. 3.30 at a rotor speed of 300 rpm.
Generally the usefulness of the inclusion of such coupling effects in the anal-
ysis of the blades’ natural frequencies is the increased accuracy they provide in
ascertaining the modal frequencies. Such accuracy is warranted in order to pre-
clude any of the natural frequencies from intersecting one of the P lines (resonance
conditions) or coalescing with another modal frequency so as to create a condition
conducive for an instability
“chapter3(new)” — 2005/11/16 — page 117 — #51

Fig. 3.30 Typical variation of blade frequencies with the coupling effects of pitch and
twist rate.
3.7.6 Coupling with Torsion

Blade bending couples significantly with torsion because of both linear and non-
linear effects. Because of a variety of reasons, blade designs continue to evolve
with spanwise variations in practically all key structural parameters: both bending
and torsion stiffnesses, and the chordwise locations of the mass center, the neutral
axis and the shear center. Of particular importance are the chordwise locations
Fig. 3.31 Typical variation of blade mode shape components as a result of the
coupling effects of pitch and twist rate—first flatwise mode (1F) at 300 rpm.
“chapter3(new)” — 2005/11/16 — page 118 — #52

ROTATING BEAMS 119
of the mass center and the shear center. Both of these parameters produce lin-
ear coupling of the bending degrees of freedom with the elastic torsion modes.
Furthermore, practical design considerations of rotor blades have driven blade
designs to assume long beam-like structures of high aspect ratios (certainly com-
pared to fixed wings). Furthermore blades have relatively low bending stiffnesses
because so much stiffening is available from the inherent tension field. The further
requirement for minimizing rotor weight combines with these structural charac-
teristics to produce conditions wherein the rotor blade can flex in a manner as to
produce couplings not normally present in fixed wings. This situation has gener-
ated an important source of nonlinear coupling wherein blade flatwise bending
together with edgewise bending can produce significant torsion moments. Brief
examinations of both of these types of torsion coupling forms the remainder of
this chapter.
Linear coupling caused by shear center and mass center offsets. The
variability of the shear center location along the span of the blade inherently defines
an elastic axis that is now a space curve rather than a straight line. Although the
ability to analyze such a structural configuration is well within the capabilities of
current finite element analyses, we seek a more direct (albeit approximate) way
of including these effects. The motivation for such an approach is the need to
construct aeroelastic stability and vibration analyses that are sufficiently compact
that we might understand the physics at work and thereby correct unacceptable
dynamic problems in a rotor design when they occur. The thrust in this text is a
strong proclivity to use modal approaches. We will use such an approach to deal
with blades having a spanwise variable elastic axis with mass center offsets.
It can be reasoned that as far as the torsion kinematics are concerned a space-
curve elastic axis can be the result of either being built-in or caused by elastic
bending deformations. Thus, we can use the material from Sec. 3.7.4 as a starting
point: substitute ySC in place of ve and use the elastic torsion twist rate θe as the
requisite twist rate in the expressions for the deflection correction functions. The
displacements of the mass center will then include the mass center offset yCG in
addition to the already present elastic bending ve . With the retention of only the
linear terms in the bending and torsion variables, and using Eqs. (3.148), (3.149),
and (3.151) as a starting point, the v5 and w5 displacements of the mass center can
first be written as
v5CG = (ve + yCG ) cos − (we − wSC ) sin (3.155)

w5CG = (ve + yCG ) sin + (we − wSC ) cos (3.156)
Then, after separating out the elastic torsion from , the following final
expressions for the mass center displacements are obtained:
v5CG = (ve + yCG ) cos θ − [we + (yCG γθj − wSCj )qθj ] sin θ (3.157)
w5CG = (ve + yCG ) sin θ + [we + (yCG γθj − wSCj )qθj ] cos θ (3.158)
“chapter3(new)” — 2005/11/16 — page 119 — #53

Fig. 3.32 Pictorial representation of the kinematics of mass center displacement

caused by spanwise variations in chordwise mass center and shear center locations.
where ve and we are given by the usual representation [i.e., Eqs. (3.146) and
(3.147)] and the following definition is made:
x̄ x̄ x̄1
wSCj = γθj ySC dx̄1 + γθj ySC

dx̄2 dx̄1 (3.159)
0 0 0
The displacements of the mass center as defined by Eqs. (3.157) and (3.158)
are depicted in Fig. 3.32. The elastic bending displacements have been omitted
for clarity. Note that in this development the reference axis is assumed to coincide
with the elastic axis at the blade root. This is not unreasonable given the symmetry
of the pitch bearing and cuff assembly at the root of the blade.
Nonlinear coupling caused by combinative blade bending. The descrip-

tion of blade torsion response, as given by Eq. (3.114), includes the nonlinear
moment loading resulting from blade curvatures interacting with force load dis-
tributions perpendicular to the curvatures. As shown in Fig. 3.33, this moment
loading accrues from both flatwise and edgewise direction curvatures combining,
respectively, with the out-of-plane and in-plane (force) load distributions.
A convenient way of formulating this moment loading is to represent the force
load distributions by the internal elastic moment resisting them. The simplified
expression for this nonlinear moment loading then becomes
(qx5 )NL = − (EIE − EIF ) ve we
This coupling term has been found to be important as an additional source

of coupling of torsion with both flatwise and edgewise bending by virtue of the
“chapter3(new)” — 2005/11/16 — page 120 — #54

ROTATING BEAMS 121
Fig. 3.33 Pictorial representation of the nonlinear torsion moment resulting from the
combinatorial action of edgewise bending and flatwise bending.
linearization that occurs about the steady deflected values of both components of
bending. Minimizing the impact of this coupling typically requires stiff torsion
blade designs. Alternatively, the coupling has been used to provide stability
augmentation for some forms of aeroelastic instability phenomena.
References
Section
3.1 Bisplinghoff, R. L., Ashley, H., and Halfman, R. L., Aeroelasticity,
Addison–Wesley, Cambridge, MA, 1955.
Bramwell, A. R. S., Helicopter Dynamics, Wiley, New York, 1976.
Johnson, W., Helicopter Theory, Princeton Univ. Press, Princeton, NJ,
1980.
Payne, P. R., Helicopter Dynamics and Aerodynamics, Macmillan,
New York, 1959.
Rao, S. S., Mechanical Vibrations, Addison–Wesley, Reading, MA,
1995.
3.2 Friedmann, P. P., and Straub, F., “Application of the Finite Element
Method to Rotary-Wing Aeroelasticity,” Journal of the American
Helicopter Society, Vol. 25, No. 1, 1980, pp. 36–44.
Hurty, W. C., and Rubinstein, M. F., Dynamics of Structures, Prentice–
Myklestad, N. O., Vibration Analysis, McGraw–Hill, New York, 1944.
3.3 Bailey, C. D., “Exact and Direct Analytical Solutions to Vibrating Sys-
tems with Discontinuities,” Journal of Sound and Vibration, Vol. 44,
No. 1, 1976, pp. 15–25.
“chapter3(new)” — 2005/11/16 — page 121 — #55

References (continued)
Hodges, D. H., “Vibration and Response of Nonuniform Rotating
Beams with Discontinuities,” Journal of the American Helicopter
Society, Vol. 24, No. 5, 1979, pp. 43–50.
3.4 Payne, P. R., Helicopter Dynamics and Aerodynamics, Macmillan,
New York, 1959.
Yntema, R. T., “Simplified Procedures and Charts for the Rapid Estima-
tion of Bending Frequencies of Rotating Beams,” NACA TN-3459,
June 1955.
3.5 Hodges, D. H., and Dowell, E. H., “Nonlinear Equations of Motion
for the Elastic Bending and Torison of Twisted Nonuniform Rotor
Blades,” NASA TN D-7818, 1974.
Houbolt, J. C., and Brooks, G. W., “Differential Equations of Motion
for Combined Flapwise Bending, Chordwise Bending and Torsion
of Twisted Non-uniform Rotor Blades,” NACA Rept. 1346, 1958.
3.6 Bielawa, R. L., Johnson, S. A., Chi, R. M., and Gangwani, S. T.,
“Aeroelastic Analysis for Propellers,” NASA CR-3729, 1983.
Loewy, R. G., Rosen, A., and Mathew, M. B., “Application of the Prin-
cipal Curvature Transformation to Nonlinear Rotor Blade Analysis,”
Vertica, Vol. 11, No. 1/2, 1987, pp. 263–296.
Problems
3.1 A helicopter rotor blade has been found to have nonrotating first flatwise and
edgewise mode frequencies of 1.6155 and 4.8864 Hz, respectively. Addi-
tionally, it is known that the first flatwise mode frequency (at a nominal rotor
rotation speed of 5 Hz) is 5.6 Hz and that the first edgewise mode has a
resonant 2P frequency at = 2.4903 Hz.
(a) Estimate the variations of the modal frequencies in the rotor speed range
of nominal ±10%.
(b) Are these frequency characteristics acceptable from a structural dynamic
point of view? Substantiate your answer.
3.2 Consider a massless rotating cantilevered beam of length L with a constant
bending stiffness distribution EI and a concentrated mass M attached at its
free end. The beam rotates about an axis passing through its point of cantilever
attachment with a constant rotor speed thereby imparting a constant tension
T to the beam.
(a) Show that its (out-of-plane) natural frequency is given by

tanh λL −1/2
ω = T /M L − (3.160)
λ
where

λ= T /EI (3.161)
“chapter3(new)” — 2005/11/16 — page 122 — #56

ROTATING BEAMS 123
(b) Obtain an approximation to the Southwell rise factor for small

values of λ.
3.3 Derive the expressions for displacement and slope transfer matrix elements
for the elastic beam segment given in Sec. 3.3.1 (p. 83), that is,
 L

 z
 
 
z R

=[ ? ]

S
n+1 
 
M n
3.4 Consider a model (articulated) rotor blade that has a total rotor radius of
56.224 in, a flapping hinge of 3 in, and a rotor speed of 69.32 rad/s. The
geometry of this blade and the distributions of the blade’s structural properties
are given, respectively, in Fig. 3.34 and Table 3.3.
Using the Yntema charts, obtain estimates of the first three flatwise and
first edgewise natural frequencies.
3.5 As discussed in Sec. 3.6.2, because of flexibilities in the rotor control system
the blade torsion modes can typically have finite values at the blade root
(x = 0). One way of approximating these root values and the resulting
frequencies and mode shapes from the modal characteristics obtained with
the rigid root boundary condition is to use the Galerkin method. Let the
normal mode shapes and natural frequencies of the blade with the rigid root
boundary condition be denoted by γθn and ωθn , respectively. We wish to
obtain the similar characteristics for the flexible root condition γ̂θn and ω̂θn ,
where it is assumed that the two mode shapes can be related to each other by
the following relationship:

θ (x, t) = [cn γ̂θn (x)] exp(iω̂θn t) (3.162)
n
where

cn γ̂θn (x) = (1)a0n + ajn γθj (x) (3.163)
j
The objective is to calculate the a0n and ajn coefficients for the nth mode.
Mathematically, this objective defines an eigenvalue problem. Note that the
γθj (x) mode shapes are orthogonal to each other but not to the pseudotorsion
mode γθ0 (x)(= 1). These constants can be related to the root torsion flexibil-
ity (spring stiffness constant) Kθ by relating each of the coefficients to the
magnitude of the vibratory root torsion moment:
M θ = −Kθ θ(0) = −GJeff θ̄ (0) (3.164)
“chapter3(new)” — 2005/11/16 — page 123 — #57

Fig. 3.34 Planform of model articulated (S-76) rotor blade.
or for the nth mode,

ajn γθj (x)
(n)
M̄θ = −Kθ a0n = −GJeff (3.165)
j
where
GJeff = GJ + Tka2 + EB1 (θB )2 (3.166)
Table 3.3 Tabulation of structural properties for model S-76 rotor blade
Inboard Section Section Stiffness, lb-ft2

section length, mass, Iθ ,
r/R ft slugs Flap Chord Torsion lb-s2
0.0534 0.322 0.0510 101944.0 104166.7 6763.9 0.570 × 10−3

0.1222 0.166 0.0110 9326.4 69444.4 1269.6 0.143
0.1577 0.333 0.0062 9326.4 2777.8 432.1 0.050
0.2288 0.333 0.0062 74.3 2777.8 236.1 0.050
0.2999 0.333 0.0062 74.3 2777.8 88.9 0.050
0.3710 0.333 0.0062 74.3 2777.8 88.9 0.080
0.4421 0.333 0.0062 81.3 2777.8 91.6 0.080
0.5132 0.333 0.0062 75.7 2777.8 93.1 0.080
0.5843 0.333 0.0062 81.3 3777.8 94.4 0.080
0.6554 0.333 0.0062 81.3 2777.8 94.4 0.080
0.7265 0.333 0.0062 81.3 2777.8 94.4 0.080
0.7976 0.333 0.0062 86.8 2777.8 92.4 0.080
0.8687 0.207 0.0054 33.3 694.4 95.4 0.117
0.9128 0.073 0.0024 33.3 694.4 27.1 0.117
0.9283 0.336 0.0045 21.5 347.2 22.0 0.117
“chapter3(new)” — 2005/11/16 — page 124 — #58

ROTATING BEAMS 125
Using a standard application of the Galerkin method, set up an eigenvalue

problem statement for determining the coupled natural frequencies for the
nth torsional mode ω̂θn and the coefficients defining the coupled mode shapes
a0n and ajn . (Note that there should be as many coupled or flexible root torsion
modes as there are “assumed” or rigid root torsion modes.) Therefore, also
formulate and incorporate in the eigenvalue analysis a suitable constraint
equation to relate a0n directly to ajn . Thus, we require a matrix such that
[a0 a1 · · · aj ]Tn = [Tn ][a1 · · · aj ]Tn (3.167)
3.6 Consider a massless cantilevered beam of length L with a tip mass of weight
W . The beam has a constant high-aspect-ratio rectangular cross section so
that it can be assumed to have infinite stiffness in the direction normal to
the span; thus, it can be characterized by a single (finite) flatwise bending
stiffness EI. Furthermore, the beam has a linear built-in twist and is mounted
so that 1) at the root section, the chord (widthwise, i.e., edgewise dimension
of the beam) has a pitch angle θ0 , which for present purposes can be assumed
to be “small”; and 2) the tip section chord pitch angle is zero [i.e., (0) = θ0 ,
and (L) = 0]. Assuming the beam to be infinitely rigid in torsion, calculate
approximate tip deflections of the beam in the vertical z5 and horizontal
y5 directions because of the weight of the tip mass. Express the component
deflections in terms of the beam length L, the nondimensional factor WL 2 /EI,
and the root pitch angle θ0 .
“chapter3(new)” — 2005/11/16 — page 125 — #59

4
Gyroscopics
In the preceding chapter the basic vibrational characteristics of a flexible rotating

beam (representative of a rotor blade) were developed with the emphasis placed
on the additional stiffening provided by the centrifugal forces. In this chapter we
explore another aspect of the dynamics of rotation wherein the axis of rotation itself
is rotated. The tools resulting from this development address a basic dynamic
phenomenon inherent in rotorcraft structural components in an ever-increasing
variety of operational conditions.
4.1 Rotational Motion of a Solid Body

4.1.1 Angular Momentum and Dynamic Equation
Consider a system of particles such as that shown in Fig. 4.1.
Newton’s law applied to the kth mass particle is as follows:
mk p2i RCk = F(ext)k (4.1)
where pi is the time derivative with respect to the (fixed) inertial coordinate system.
We can then take the cross product of RCk with this equation where RCk is the
vector from the mass center of the solid body to the kth mass particle, noting
that
pi (RCk × mk pi RCk ) = RCk × mk p2i RCk (4.2)
After taking the cross product and summing over all of the particles, we obtain

pi (RCk × mk pi RCk ) = RCk × F(ext)k (4.3)
k k

= Hc = Mc
where H c is the angular momentum of the system of particles about its mass center.
The equation of dynamic equilibrium then becomes
pi H c = M c (4.4)
127
“chapter4(new)” — 2005/10/18 — page 127 — #1

Fig. 4.1 Rotational kinematics of a mass particle.
Two difficulties arise with this equation for a rigid body:

1) The angular momentum, as given earlier, is obscurely defined.
2) Differentials with respect to inertial space are awkward when one deals with
rapidly and complexly rotating rigid bodies.
4.1.2 Alternate Form of Angular Momentum

To eliminate the difficulties identified earlier, the theorem of Coriolis can be used
to reduce the expression for angular momentum to a more tractable form. Thus,
let us define the angular velocity of the body-fixed coordinate system relative to
the inertial coordinate system as ωib . Then
pi RCk = pb RCk + ωib × RCk
(4.5)
= 0, for a rigid body
Therefore,

Hc = mk RCk × (ωib × RCk ) (4.6)
k
and using the formula for the vector triple product yields

Hc = mk [(RCk · RCk ) ωib − (RCk · ωib )RCk ] (4.7)
k
From this expression it can be concluded that the angular momentum and angular
velocity H c and ωib , respectively, are not generally aligned in the same direc-
tion. A more tractable form of the angular momentum vector can be obtained by
expanding out the general expression for H c in terms of the components of the
particle position and the angular velocity. Let
 
xk 
{RCk } = yk (4.8)
 
zk
“chapter4(new)” — 2005/10/18 — page 128 — #2

GYROSCOPICS 129
and
 
ωx 
measured about
{ωib } = ωy ←− (4.9)
  body axes
ωz
These expressions are then substituted into the definition for H c , and the summation
is taken. The result of this summation can then be expressed by the following matrix
equation for the body-fixed components of H c :
    
Hx  Ixx −Ixy −Ixz ωx 
Hy = −Iyx Iyy −Iyz  ωy (4.10)
   
Hz −Izx −Izy Izz ωz
where Ixx , Iyy , and Izz are the mass moments of inertia of the rigid body about
arbitrarily defined (orthogonal) x, y, and z axes of the body, respectively. The
off-diagonal terms are the products of inertia, for example,

Ixy = mk xk yk (4.11a)
k

Iyz = mk yk zk (4.11b)
k

Izx = mk zk xk (4.11c)
k
If the principal axes are chosen as the body-fixed coordinate frame, the products
of inertia vanish. Thus, if the resulting principal axis moments of inertia are then
denoted with a single subscript, the angular momentum vector can be expressed
in the following simplified form:
H c = Ix ωx i + Iy ωy j + Iz ωz k (4.12)
This result can then be used with an invocation of the theorem of Coriolis to obtain
the desired more tractable form of the dynamic equilibrium equation:
pb H c + ωib × H c = M c (4.13)
From this vector differential equation the following set of equations (Euler’s equa-
tions for rotational motion of a rigid solid, as expressed in body-fixed coordinates)
can then be written:
Ix pωx + (Iz − Iy )ωz ωy = MCx (4.14a)

Iy pωy + (Ix − Iz )ωx ωz = MCy (4.14b)
Iz pωz + (Iy − Ix )ωy ωx = MCz (4.14c)
“chapter4(new)” — 2005/10/18 — page 129 — #3

Fig. 4.2 Vector representation of rotational quantities.
These equations are highly nonlinear, and a general closed-form solution does
not exist. Successful solutions usually involve a suitable linearization appropriate
to the problem at hand. The following section, dealing with simplified general
gyroscopic motion, presents development for just such a case where linearization
is possible. In preparation of this development, Fig. 4.2 presents a summary of the
use of vector representations of rotational quantities.
4.2 Simplified Gyroscope Equation

In this section we consider a general gyroscopic element; examples include
a rotor in a gyroscopic instrument, a spinning top, or a spinning propeller.
Such a gyroscopic element is basically defined to be a rigid body (either with
polar symmetry or inertial isotropy about one of its principal axes) in rotational
motion.
4.2.1 Mathematical Development

The “rotational” motion of the basic gyroscopic element is typically character-
ized by being principally of high rotation rate about one of the principal axes (that
with polar symmetry or isotropy). With this generalization in mind, we first make
“chapter4(new)” — 2005/10/18 — page 130 — #4

GYROSCOPICS 131
the following definitions:

H c = H s + H ns (4.15)
where
H s ≡ the spin angular momentum (vector) of the spinning rotor (Hs = Js ωs )
H ns ≡ the nonspin angular momentum (i.e., that which is left over when ωs = 0),
= Ix ωx i + I y ωy j + Iz ωz k
Note that Ix , Iy , and Iz include all mass that can move, that is, more than just that
of the rotor. With these definitions we then make the following assumptions:
1) The rotor spins about an axis of symmetry or isotropy.
2) The spin velocity is constant, that is, dωs /dt ≡ 0.
3) The magnitude of the spin angular momentum is much greater than the
nonspin, that is, Hs Hns .
With these assumptions the vector form of Euler’s equation becomes
pH + pH ns + ωib × H s + ωib × H ns = M (4.16)
s
→0 (. . .) (. . .)
The resulting equation then becomes
pH ns + ωib × H s = M (4.17)
This equation is then the simplified gyroscope equation. The most signifi-
cant characteristic of this equation is that it is linear in the components of the
angular velocities defining the nonspin angular momentum. The significance of
the three terms of this equation is described as follows: The variable pH ns is the
source of the characteristic dynamics or transient response of the gyro and gives
rise to nutation in two-degrees-of-freedom gyros. The expression ωib × H s is the
source of the precessional characteristics of the gyro. The variable M is the applied
moment defined in terms of body-fixed coordinates (unit vectors imbedded in the
nonspin body).
Simple direction rule: A gyro rotor precesses in such a way so as to align the spin
momentum vector with the applied moment M.
The precessional motion thus described is demonstrated in Fig. 4.3.
4.2.2 Basic Dynamic Equations for Two-Degree-of-Freedom Gyro

Although the development just given represents an elegantly simple mathe-
matical representation for the dynamics of a gyroscopic element, its direct
applicability to a given gyroscopic element is not obvious. As an example of
its application, let us consider the derivation of the basic equations of motion for
a simple gyroscopic element using both the Euler equations and the simplified
gyroscope equation.
“chapter4(new)” — 2005/10/18 — page 131 — #5

Fig. 4.3 Basic law of precessional motion of a practical gyroscopic element.
Euler’s equation. Euler’s equations are defined only in an axis system embed-
ded in the rotating solid. Therefore, let the XYZ axis system be attached to the
rotating gyro element, but defined in terms of the rotation rates θ̇x and θ̇y , describing
the motion of the disk as a moving but not spinning solid:
“chapter4(new)” — 2005/10/18 — page 132 — #6

GYROSCOPICS 133
Then the following relationships result:

ωx = θ̇x cos t + θ̇y sin t (4.18a)
ωy = −θ̇x sin t + θ̇y cos t (4.18b)
ωz = (4.18c)
The components of the applied moment are given by
MCx = Mx cos t + My sin t (4.19a)
MCy = −Mx sin t + My cos t (4.19b)
MCz = 0 (4.19c)
Simple differentiation yields
ω̇x = (θ̈x + θ̇y ) cos t + (θ̈y − θ̇x ) sin t (4.20a)
ω̇y = −(θ̈x + θ̇y ) sin t + (θ̈y − θ̇x ) cos t (4.20b)
Substitution of these relationships into the first of Euler’s equations then yields
I[(θ̈x + θ̇y ) cos t + (θ̈y − θ̇x ) sin t]
+ (J − I)(−θ̇x sin t + θ̇y cos t) = Mx cos t + My sin t (4.21)
Then, upon equating like coefficients of cos t and sin t, we obtain the following
desired equations:
I θ̈x + Jθ̇y = Mx (4.22a)
−Jθ̇x + I θ̈y = My (4.22b)
The second of Euler’s equation can be seen to produce the same set of equations
as Eqs. (4.22a) and (4.22b) and thus adds nothing new; the third equation loses
meaning for this problem.
Simplified gyroscope equation. For this case let

H s = Jk (4.23a)
H ns = I(θ̇x i + θ̇y j) (4.23b)
pH ns = I(θ̈x i + θ̈y j) (4.23c)
(ωib )ns = θ̇x i + θ̇y j (4.23d)
Then

i j k

ωib × H s = J θ̇x θ̇y 0 = J(θ̇y i − θ̇x j) (4.24)
0 0
M = Mx i + My j (4.25)
“chapter4(new)” — 2005/10/18 — page 133 — #7

Entering the preceding expressions into the simplified gyro equation yields
[I θ̈x + Jθ̇y ]i + [I θ̈y − Jθ̇x ] j = Mx i + My j (4.26)
from which the same equations can be obtained as those with Euler’s equations.
These equations are general and quite basic. They constitute a fundamental
analytical tool for the inclusion of gyroscopic effects in any given modeling
problem. As demonstrated earlier, these equations are obtainable by more than
one method and thus should appear in the formulation with whatever method
might be used.
4.3 Precession and Nutation Characteristics

4.3.1 Precession: The Particular Solution
Gyroscopic elements have a number of important characteristics that can be
gleaned from the equations of motion derived in the preceding subsection for the
simple gyroscopic element. Let us rewrite the equations of motion in terms of
the angular velocities about the x and y axes, ωx and ωy , respectively:
I ω̇x + Jωy = Mx (4.27a)

−Jωx + I ω̇y = My (4.27b)
These equations are clearly linear coupled equations, and solutions can be obtained
using any of a variety of standard methods. In general, the complete solution
consists of the sum of a homogeneous solution and a particular solution. The form
of the particular solution depends on the specific form of the applied moments
Mx and My . The homogeneous solution depends on the roots of the characteristic
equation, which results from setting the characteristic determinant equal to zero.
Consider first, however, the particular solution that can be obtained from setting
the derivative terms equal to zero and assuming some specific functionality to the
applied moments Mx and My :
1
ωy = Mx (4.28a)
J
−1
ωx = My (4.28b)
J
The first important characteristic of gyroscopic elements can be readily seen from
setting the components of applied moment equal to zero: In the absence of applied
moments the gyroscopic element retains its orientation in space; that is, the compo-
nents of (nonspin) angular velocity are zero. The second important characteristic
of gyroscopic elements is that of precession, wherein the components of non-
spin angular velocity are equal to their respective opposite components of applied
moment in such a way that the angular momentum vector of the gyroscopic element
tends to align itself with the applied moment.
“chapter4(new)” — 2005/10/18 — page 134 — #8

GYROSCOPICS 135
4.3.2 Nutation: The Homogeneous Solution

The third important characteristic of gyroscopic elements is that of nutation,
which describes the transient behavior (usually oscillatory) of the gyroscopic
element. Consider the characteristic equation for the basic gyroscopic element
equations given earlier:

Iλ J

−J Iλ = 0 −→ I λ + J = 0
2 2 2 2
(4.29)
or, recalling that the polar moment of inertia J is ideally twice the diametral
moment of inertia I, gives
λ2 = −42 (4.30)
The roots of this simple quadratic form are then written as
λ = ± i2 (4.31)
This resulting oscillatory root pair then defines the nutation frequency ωN of the
element:
ωN = 2 (4.32)
4.3.3 Dynamics of a Spinning Top

A further example of the nutational characteristics of gyroscopic elements is
afforded by examining the dynamics of a spinning top or gyroscopic element,
which pivots about a (fixed) point on the spin axis located a distance below the
center of gravity. As shown in Fig. 4.4, the spin axis z , as well as the unit vector
along this axis k, are inclined to one side by the angle θ; this vector, as well as the
x axis, lie in the precessionally rotating (but not spinning) x-z plane. The weight
of the element creates an applied moment that can be expected to precess the spin
axis of the top in a more or less circular arc about the vertical axis.
In the following development we take the axis system to be rotating about the
z axis with an angular velocity comprised of the sum of a steady precessional
rate z and a perturbational angular rate ψ̄. ˙ The x and y axes are defined so
that the inclination angle of the spin axis with respect to the vertical θ is always
measured about the y axis, that is, the plane of the inclined spin axis and the
vertical axis is always normal to the y axis. The x axis is then taken normal to
the y and z axes. Additionally, an auxiliary coordinate system is attached to the
(precessionally rotating, but not spinning) principal axes of the top (x , z axes);
this auxiliary coordinate system is formed from the basic coordinate system by a
θ angle rotation about the y axis.
Using the auxiliary coordinate system as a basis, we can write the mathe-
matical quantities required by the simplified gyroscopic equation. For this example
we expect that the spinning top will exhibit equilibrium conditions wherein the
inclination angle θ will be comprised of an equilibrium value θ0 , and a pertur-
bational value θ̄, and the precessional rate z will be generally finite and not
“chapter4(new)” — 2005/10/18 — page 135 — #9

Fig. 4.4 Dynamics of a spinning top.
necessarily small. As a result, although the resulting equations can be linearized

with respect to the perturbational quantities, those nonlinear terms involving the
equilibrium values must be retained. With these points in mind, the following
required mathematical quantities can be written:
Spin angular momentum:
H s = Jk (4.33)
Angular velocity of the (nonspin) body axes:
˙ sin θi + θ̄˙ j + ( + ψ̄)

ωib = −(z + ψ̄) ˙ cos θ k (4.34)
z
Nonspin angular momentum:
˙ sin θ i + Iˆθ̄˙ j + J( + ψ̄)

ˆ z + ψ̄)
H ns = −I( ˙ cos θ k (4.35)
z
where Iˆ and J are, respectively, mass moments of inertia about axes through the
pivot point transverse to and along the spin axis.
Applied (gravity) moment:
M = m(g + 2 2z cos θ) sin θ j (4.36)
“chapter4(new)” — 2005/10/18 — page 136 — #10

GYROSCOPICS 137
Note that, for this configuration, Iˆ = Id + m2 , where Id is the diametral moment
of inertia of the gyro disk. Application of the simplified gyroscope equations and
the linearization of the equations about a constant value of the inclination angle θ0
then yields the following two differential equations of motion:
ˆ
−I(sin θ0 ψ̄¨ + z cos θ0 θ̄)
¨ + Jθ̄˙ = 0 (4.37a)
Iˆθ̄¨ + J(ψ̄˙ + z ) sin θ0 = mg sin θ0 + I

ˆ 2z sin θ0 cos θ0 (4.37b)
Deleting the transient terms in the second of these equations and factoring out the
common sin θ0 term from the equation gives a quadratic relationship defining the
equilibrium inclination angle θ0 in terms of the precession angular velocity z :
ˆ 2z cos θ0
z J = mg + I (4.38)
Define two nondimensional parameters, η and , as follows:
Î mg z
η≡ ; ≡ ; ¯z ≡
(4.39)
J J2
Then the preceding equation can be solved for the two equilibrium precessional
rates possible for the top:
1
¯z =
1 ± 1 − 4η cos θ0 (4.40)
2η cos θ0
Alternatively, the cosine of the inclination angle can be expressed in terms of the
equilibrium precessional rate:
¯ z − )/η
cos θ0 = ( ¯ 2z (4.41)
¯ 2z → 0; hence,
Note that for high spin rates, cos θ0
¯ z = = mg/J2
(4.42)
Using the nondimensional parameters already defined and nondimensionalizing

the time derivatives with respect to the spin angular velocity , we can write the
two perturbational equations of motion in the following somewhat simplified form:
η sin θ0 ψ̄ + (η cos θ0 − 1)θ̄ = 0 (4.43a)

sin θ0 ψ̄ + ηθ̄ = 0 (4.43b)
where the prime denotes differentiation with respect to nondimensional time

τ (= t). The usual eigensolution techniques, wherein the perturbational responses
θ̄ and ψ̄ are assumed to be of exponential form, are then applied:
θ̄ = ˆ eλτ ; ψ̄ = ˆ eλτ ; τ = t (4.44)
“chapter4(new)” — 2005/10/18 — page 137 — #11

Expansion of the characteristic determinant gives the following characteristic

equation for the eigenvalue λ:
η2 λ4 + (1 − η cos θ0 )λ2 = 0 (4.45)
It can be verified that the solution for λ yields two zero values and an imaginary
pair. The zero pair of roots correspond to the precessional characteristics of the
top, whereas the imaginary pair gives the final desired expression for the (non-
dimensional) nutation frequency ω̄N :
1
ω̄N = 1 − η cos θ0 (4.46)
η
where η is the ratio of moments of inertia, as defined earlier. Note that in the limit
as the pivot to c.g. distance goes to zero, the top becomes a simple gyroscopic
element as described in the preceding subsection. For this case the η parameter
goes to 21 , and the parameter goes to zero. These values then produce a nutation
frequency equal to two times the rotation frequency , as expected.
4.4 Gyroscopic Characteristics of Rotor Blades

4.4.1 Blade Flapping Motion
Moment control of rotorcraft rotors is typically maintained by means of a
cyclic variation of the pitch angle part of the blade angle of attack. These cyclic
angles of attack in turn produce cyclic blade loadings that accelerate the blades in
a harmonic fashion. Whereas the blades see the loadings as oscillatory excitations
with period 2π/ , a nonrotating observer would see the cyclic loadings as a con-
stant moment applied to the rotor disk at some (vectorial) orientation angle in the
rotor plane. Also, because the rotor blades have a natural frequency of approxi-
mately unity (1/rev), they will respond to a 1/rev excitation 90 deg (phase-lagged)
after the excitation and “secularly” in amplitude, that is, the amplitude of the flap-
ping response will grow linearly with time. This response would be viewed by
the nonrotating observer as an ever-increasing “tilt” of the rotor disk at a point
90 deg from the point of maximum effect from the constant moment. Thus, the
disk would appear to “precess” in accordance with the rule of gyro precession
given in a preceding subsection. These events are illustrated in Fig. 4.5.
4.4.2 Rotor Modes

The interpretation just given of the rotor response in flapping β, that is, that of
a tilting disk as viewed in the fixed system, can be formalized by the use of rotor
modes. When only the tilting of the rotor disk is considered, the response of the
jth blade can be expressed as follows:
βj (t) = βc (t) cos(t + ψj ) + βs (t) sin(t + ψj )

= βc (t) cos θj + βs (t) sin θj (4.47)
“chapter4(new)” — 2005/10/18 — page 138 — #12

GYROSCOPICS 139
Fig. 4.5 Precessional characteristics of rotors.
where
θj = t + 2π j/b, j = 1, 2, . . . , b (no. of blades) (4.48)
and where βc and βs represent the fore and aft and lateral tilt angles, respectively,
of the rotor disk and are arbitrary functions of time. Indeed, as material in sub-
sequent chapters will show, aeromechanical formulations involving rotor mode
descriptions of both flapping and edgewise motions of the blade will generate
gyroscopic-like, skew-symmetric terms in the equations of motion.
4.4.3 Whirl Modes

The gyroscopic characteristics of rotors can be further defined using the afore-
mentioned concept of rotor modes. First, the earlier equation defining the rotor
modes, that is, Eq. (4.47), can be rewritten using the exponential terms of the
trigonometric functions:
βj (t) = βc (t)/2[exp(iθj ) + exp(−iθj )]

− iβs (t)/2[exp(iθj ) − exp(−iθj )] (4.49)
Then assume that the rotor disk as defined earlier is now undergoing sinusoidal
motion, so that the two tilt angles βc and βs are expressed as
βc (t) = β̄c eiωt ; βs (t) = β̄s eiωt (4.50)
“chapter4(new)” — 2005/10/18 — page 139 — #13

Finally, let us assume for simplicity that the amplitudes of these two angles are
equal to each other, but that their relative phase angle φ is different:
β̄c = β̄; β̄s = β̄eiφ (4.51)
From a physical point of view, β̄ represents the amount of tilt that one would
see. These concepts can be further visualized by considering the motion of the
head of an arbitrary constant amplitude vector maintained normal to the disk N.
As a top view (see Fig. 4.6) of the rotor shows, positive values of β̄c and β̄s
produce component projections of this vector, respectively, in the forward x and
lateral (to port) y directions, where the x-y plane represents the undeflected rotor
plane.
If now the first time derivative of βs is related to βc and a value of time is
selected in order to make βc have a purely real positive value, then, depending on
the value of the relative phase angle φ, the two-dimensional motion of the head of
the vector can be determined:
βs = β̄s eiωt = β̄eiωt eiφ = (cos φ + i sin φ)βc (4.52a)

β̇s = iωβ̄eiωt eiφ = ω(− sin φ + i cos φ)βc (4.52b)
The relationship between βc , βs , and β̇s can then be examined for different assumed
values of φ:
Fig. 4.6 Vectorial representation of a cyclically flapping rotor.
“chapter4(new)” — 2005/10/18 — page 140 — #14

GYROSCOPICS 141
Case 1 (φ = 0): For this case the two-component projection vectors are not
only in phase, but are equal to each other. Thus, the motion of the head of the
vector would be that of a degenerate ellipse (line segment) with semimajor axis
equal to β̄. This type of motion could be described as a back and forth “wobbling”
of the disk:
Case 2 (φ = −π/2): For this case the real-valued components of βs and β̇s
would be zero and positive respectively; hence, the βs component of the N vector
would be moving in the negative y direction. The head of the vector would therefore
describe a circular motion in the same direction as the rotor rotation. This type of
motion is called progressive or forward whirl:
Case 3 (φ = π/2): For this case the real-valued components of βs and β̇s
would be zero and negative respectively, and the motion of the head of the N
vector would be again circular, but in a direction opposite to the rotor rotation.
This type of motion is called regressive whirl:
“chapter4(new)” — 2005/10/18 — page 141 — #15

A mathematical description of whirl modes can be obtained by substituting the

assumed simplified expressions for the sinusoidal motion descriptions of βc and
βs into the equation for blade flapping:
β̄
βj (t) = Re{exp[i(θj + ωt)] − i exp[i(θj + ωt)]eiφ
2
+ exp[i(−θj + ωt)] + i exp[i(−θj + ωt)]eiφ } (4.53)
or, more compactly, as
β̄
βj (t) = Re{(1 + sin φ − i cos φ) exp[i(θj + ωt)]
2
+ (1 − sin φ + i cos φ) exp[i(−θj + ωt)]} (4.54)
Let us now examine this equation for each of the two types of whirl. First, consider
the case of progressive whirl (φ = −π/2). For this case the preceding equation
simplifies to the following expression:
β̄
βj (t) = Re{2 exp[i(−θj + ωt)]}
2
= β̄ cos[(ω − )t − ψj ] (4.55)
As conceptualized earlier, the rotor disk tilt angles are measured in the fixed or
nonrotating coordinated system. Therefore, the oscillation frequency ω represents
a fixed coordinate system frequency and is what would be seen by the nonrotating
observer. In the following material this frequency will be denoted by a subscript
F. Thus, the whirling motion (either progressive or regressive) of the N vector
or, more specifically, the rotor disk itself would be observed as occurring with a
circular frequency of this value.
Note, however, that the blade oscillating in the rotating coordinate system would
still be governed by its natural frequency, as defined in the rotating frame ωR :
βj (t) = β̄j cos ωR t (4.56)
“chapter4(new)” — 2005/10/18 — page 142 — #16

GYROSCOPICS 143
Thus, after the frequencies in these two expressions are equated, the following
frequency relationship emerges:
ωF − = ωR −→ ωF = ωR + (4.57)
Following the same argument for regressive whirl (φ = π/2), a similar frequency
relationship can be deduced:
ωF + = ωR −→ ωF = ωR − (= |ωR − |) (4.58)
Note: For this case the fixed coordinate frequency (as expressed by this equation)
can reach zero or even negative values. The interpretation to be given to these cases
is based on the expressions given earlier for relating βc to β̇s . As the fixed coordinate
system frequency approaches zero, the observed whirling motion comes to a halt
and the rotor assumes a steady deflected position. As the calculated frequency goes
negative (as would most likely be the case with increased rotor speed), the whirling
motion would begin again, but in the advancing direction. Thus, for this condition
the rotor would demonstrate two advancing whirl modes; one would be the original
progressive whirl mode, and another, at a significantly lower frequency, would be
what originally was the regressive whirl mode. In the preceding argument the
concept of a “negative frequency” can be used only in the context of a complex-
valued representation for the assumed sinusoidal motion. In reality only positive
values of frequency can be observed: the use of a negative frequency is justified
only for defining the direction of the whirling motion.
It is through the concept of whirl modes that the third identifiable gyros-
copic characteristic of rotor blades, that is, nutation, can be defined. For the case
of rigid blade flapping (in the limit as the offset approaches zero), the (rotating)
blade natural frequency is 1/rev, and the two resulting fixed system whirl mode
frequencies for the rotor are (by the previously given equations) zero for the
regressive mode (precession) and 2/rev for the progressive mode (nutation). Thus,
generalization leads to assigning the progressive mode as the nutation equivalent
characteristic of the rotor.
4.4.4 Reactionless Modes

The principal usefulness of the concept of whirl modes as described earlier is
that they provide a convenient tool for coupling the rotor dynamics with the pylon
or airframe dynamics irrespective of the number of blades. This is because the
whirl modes characteristically involve the generation of hub moments and shears
that can excite the airframe. A complement to the whirl mode description is the
so-called “umbrella” or collective mode, which involves the in-phase flapping
motion of all blades. Such a rotor mode would also be expected to couple with the
airframe by virtue of the vertical shears and torques generated by such a collective
motion of all the blades in flapping.
In addition to the collective and whirl rotor modes, there exists the possibility
of higher rotor modes that involve modulation of higher harmonic functions:
βj (t) = βc(m) (t) cos m(t + ψj ) + βs(m) (t) sin m(t + ψj ) (4.59)
“chapter4(new)” — 2005/10/18 — page 143 — #17

where m = 2, 3, 4, . . . . Such a rotor mode description defines the reactionless

rotor modes for the reason that they indeed produce no net loads on the airframe.
Physically, if the amplitudes βc(m) and βs(m) were constant, such modes would
appear to the fixed coordinate system observer as a “warped” tip path plane.
References
Section
4.1 Scarborough, J. B., The Gyroscope, Theory and Applications, Inter-
science, New York, 1958.
Wrigley, W., Hollister, W. M., and Denhard, W. G., Gyroscopic Theory,
4.2 Wrigley, W., Hollister, W. M., and Denhard, W. G., Gyroscopic Theory,
4.3 Scarborough, J. B., The Gyroscope, Theory and Applications, Inter-
science, New York, 1958.
Wrigley, W., Hollister, W. M., and Denhard, W. G., Gyroscopic Theory,
4.4 Hohenemser, K. H., and Yin, S.-K., “Some Applications of the Method
of Multiblade Coordinates,” Journal of the American Helicopter
Society, Vol. 17, No. 3, 1972, pp. 3–12.
Problems
4.1 An element of good tail rotor design is providing adequate clearance for
teetering (or flapping, as appropriate, depending on the number of blades)
angle responses caused by yawing maneuvers in hovering flight. Consider
the spinning tail rotor to be a rigid disk (gyroscopic element) as shown in
the following:
“chapter4(new)” — 2005/10/18 — page 144 — #18

GYROSCOPICS 145
where J is polar moment of inertia; X, Y , Z is the coordinate system defining

undeflected position; and x, y, z is the coordinate system attached to the
deflected position.
(a) Using the preceding representation, together with the simplified gyro
equation, derive the two differential equations of motion for the teeter angles,
α(t) and β(t), where
1) The angular velocity of the x, y, z coordinate frame relative to the
inertial frame is given by
ω = β̇i − (α̇ + ψ̇) j
2) The spin angular momentum is Hs = Jk.
3) The components of the applied (aerodynamic) moment, M = Mx i +
My j, are
Mx = M1 α − M2 β̇
My = M1 β + M2 (α̇ + ψ̇)
where M1 and M2 are the effective aerodynamic spring and damping
rates, respectively, as available using rotor aerodynamic (strip) theory or
experimental evaluations.
(b) Find the steady-state teetering angle components ᾱ and β̄ caused by
a constant helicopter yaw rate ψ̇, equal to 100 deg/s, for a tail rotor with
the following specifications:
J = 30 lb-in.-s2 ; M1 = 100,000 lb-in./rad

= 250 rad/s; M2 = 400 lb-in.-s/rad
4.2 Consider the modal differential equation for the first elastic flapwise bending
mode of a main rotor blade (as defined in the rotating coordinate system):
R
Iβ [β̈1 + (ωNR
2
1
+ Kβ1 2 )β1 ] = β1 (t) = 0 γβ1 (x)p(x, t) dx
where β1 is the generalized forcing function for this mode caused by
perturbational load distributions [as arising from either aerodynamic (control
inputs) and/or inertial load considerations].
(a) By representing both β and β1 as Fourier series, each truncated
after only the first harmonic components, [β1c (t), β1s (t), 1c (t), and 1s (t)],
that is,
β(t) = β1c (t) cos t + β1s (t) sin t
β1 (t) = 1c (t) cos t + 1s (t) sin t
Derive two differential equations for the Fourier components of flapping
motion β1c and β1s .
(b) Compare the equations obtained in part 1 with the equations obtained
for the basic gyroscopic element; in particular, explain differences and/or
additional terms obtained.
“chapter4(new)” — 2005/10/18 — page 145 — #19

4.3 The equations of the basic gyroscopic element can be considered to be

reversible to the extent that impressed rotation (precession) of the spin axis ω
“creates” reactionary gyroscopic moments in the structure. These moments
are determined by the same vectorial equation as developed for the basic
gyroscopic element (with the nutational terms deleted). Using this prop-
erty, develop the additional terms in the equations derived in problem 4.2 to
account for the responses resulting from a pitching rate, impressed on the
rotor axis, that is, ωy = θ̇y J, where J is a unit vector in the nonspinning rotor
plane, pointing towards the advancing blade (ψ = 90-deg position). (Note:
Remember that the generalized excitation β1 is a mode shape weighted
integration, whereas moment is a radial location weighted integration.)
4.4 Consider the design of a helicopter utilizing an advancing blade concept rotor
system. This rotor system is characterized by having two counter-rotating,
coaxial rotors, which are separated from each other vertically by a distance
Z and which are comprised of “stiff” hingeless rotor blades. The rotor
system must be designed to accommodate hard landings wherein rapid nose-
down pitch rates can occur as a result of a hard initial impact on the tail
landing gear.
Using the results of problem 4.3, determine the minimum safe rotor ver-
tical spacing Z needed to preclude the rotors from striking each other
due to the precession induced by a hard landing pitch rate of −200 deg/s
(nosedown). Assume 1) that the rotors respond in flapping quasistatically,
wherein the nutational dynamics (transients) are disregarded, and 2) that the
aerodynamic damping and stiffness are not factors.
The operational data are given:
1) The upper rotor rotates in the American sense (counterclockwise
when viewed from above) and thereby has a positive right-hand angular
momentum, whereas the lower rotor has a negative angular momentum.
2) The blade properties are as follows:
Rotor radius:
R = 30 ft
Rotor speed:
= 20 rad/s
First blade flapping (rotating) natural frequency:
ωβ1 = 1.6P
First flapping mode generalized inertia:
1
L 3 mγβ21 dx = Iβ1 = 1398 lb-ft-s2
0
First flapping mode coupling inertia:
1
L 3
mγβ1 x dx = Sβ1 = 1677 lb-ft-s2
0
“chapter4(new)” — 2005/10/18 — page 146 — #20

GYROSCOPICS 147
4.5 Consider an approximation to the wing-rotor system of the tilt-rotor concept

in the vertical nacelle (hover) configuration. The dynamic system consists
of four degrees of freedom: the first beam bending and torsion modes of the
(cantilever mounted) wing qw and qθ , respectively, and the (inertial) rolling
and pitching rotational deflections of the nacelle θy and θx , respectively:
Assumptions:
1) The fuselage has zero compliance so that the wing root is a “ground”
point for the cantilever wing mount.
2) The nacelle is joined to the wing tip by means of a roll spring Ky , which
couples with the wing in bending, and a pitch spring Kθ , which couples with
the wing in torsion.
3) The wing elastomechanical properties can be expressed in terms of
generalized masses and natural frequencies.
4) The mass of the nacelle and rotor are included in the wing generalized
mass.
5) The rotor is rigid and acts as a gyroscopic element, whose “spin”
angular momentum is given by J.
6) The nacelle has mass mN , located a distance h1 (vertically) from the
wing tip attachment point, and a moment of inertia about that c.g., IN , both
in pitch and roll.
7) The rotor has mass mR , located at the hub a distance h2 from the wing
tip attachment point, and a diametral moment of inertia IR .
(a) Derive the four differential equations of motion for the dynamic system
thus defined.
“chapter4(new)” — 2005/10/18 — page 147 — #21

(b) Derive a simplified two-degrees-of-freedom (two-equation) set appro-

priate to the case when both the springs restraining the nacelle Ky and Kθ ,
→ ∞.
[Hint: Use the basic gyroscopic element equations as a building block in
a free-body diagram of the system components. Use the equations to relate
the internal moments (either as carried by the springs between the wing tip
and the nacelle or directly) to the rotor’s gyroscopic moments.]
“chapter4(new)” — 2005/10/18 — page 148 — #22

5
Drive System Dynamics
A distinctive characteristic of rotorcraft vis-à-vis fixed-wing aircraft is the

necessity to deliver rotational power to the rotors via a system of shafts and
gear boxes over lengths approaching the length of the aircraft itself. This is espe-
cially true for tandem helicopters and tilt-wing and tilt-rotor VTOL aircraft. Even
single rotor helicopters rarely escape the need to deliver power to the tail rotor
via a shaft. Thus, relatively long shafting defines not only the requirement to ana-
lyze the system for torsional vibrations, but for lateral (bending) vibrations as
well. This chapter addresses both of these issues from the standpoint of analyz-
ing vibratory responses. A subsequent chapter addresses the problem of torsional
instabilities.
5.1 Shaft Critical Speeds

5.1.1 Introduction
No matter how well a rotor shaft might be balanced, there will always remain
some infinitesimal amount of mass imbalance (i.e., the mass center of the shaft
system lies off the rotation axis) and/or initial curvature. If the shaft now rotates
at a frequency equal to one of the bending natural frequencies of the shaft, the
small imperfection in mass or curvature would be sufficient to excite the sys-
tem in this mode. This condition commonly referred to as shaft critical speed
operation is a form of resonance that can involve whirling motion in either the
rotational or contrarotational directions. Furthermore, because of the presence of
rotational elements having significant rotary inertia, gyroscopic effects can play
a major role. The first order of business is to identify the characteristics of these
resonances.
5.1.2 Jeffcott Rotor

Although simplistic in form, the Jeffcott rotor is a useful tool for understand-
ing the basic concepts and response characteristics. Ehrich has investigated this
concept in great detail, and the following material is only an abridged treatment
of this subject. Consider the rotating shaft carrying a disk at its midspan location
with fixed end points, as shown in Fig. 5.1.
The presence of both mass unbalance and initial curvature or bowing requires
the investigation of both of these conditions separately.
149
“chapter5(new)” — 2005/11/16 — page 149 — #1

Fig. 5.1 Jeffcott rotor on rigid bearing supports (after Ehrich, 1992).
Mass eccentricity. For this case the elastic axis is considered to be initially
straight but flexible in bending, so as to provide effective springs Kr in the x and y
directions, as shown in Fig. 5.2. Let
G ≡ center of gravity of the disk
b ≡ geometric center of the disk
O ≡ center of rotation
Kr ≡ elastic restoring spring rate of the shaft
Br ≡ equivalent viscous damping (air resistance and bearing damping)
Fig. 5.2 Geometry of a flexible shaft elastically deformed in the transverse direction
as a result of an unbalanced attached disk.
“chapter5(new)” — 2005/11/16 — page 150 — #2

DRIVE SYSTEM DYNAMICS 151
The forced motion dynamic equations for the shaft/disk system are
M ẍ + Br ẋ + Kr xb = eM2 cos t (5.1a)

M ÿ + Br ẏ + Kr yb = eM2 sin t (5.1b)
for which the steady-state solution is given by
1 2
wb1 = u/e = (x + y2 ) = r 2 [(1 − r 2 )2 + 4ζ 2 r 2 ] (5.2a)
e
φ = tan−1 [2ζ r/(1 − r 2 )] (5.2b)
where

r = /ωn = M/Kr (5.3)
The steady-state solution for this system, as given by Eq. (5.2a), is graphed in
Fig. 5.3.
Initial shaft curvature. For purposes of analysis, we assume the mass center
of the disk to be coincident with the elastic center (i.e., point G is coincident with
point P, with e = 0), but the disk center now has an initial offset from the straight
line point O. For this case the whirl amplification factor wb2 is given by
u 1
wb2 = = (5.4)
s (1 − r )2 + 4ζ 2 r 2
2
Fig. 5.3 Response caused by mass unbalance vs rotor speed for the Jeffcott rotor on
rigid bearing supports.
“chapter5(new)” — 2005/11/16 — page 151 — #3

This function is graphed in Fig. 5.4. A comparison of Fig. 5.3 with Fig. 5.4
shows the following:
1) The two types of unbalance have completely opposite behavior at rotor speeds
both above and below the resonant frequency point, r = /ωn . For rotor speeds
below the resonant frequency, the mass unbalanced rotor has the advantage of
lower response levels, whereas the initial curvature rotor has the advantage at
the higher rotor speeds.
2) At the resonance point the response levels are essentially the same for the two
cases and demonstrate the typical results for all resonant systems.
Flexible supports with mass eccentricity. A further relaxation of assump-

tions is afforded by including separate stiffnesses in both the shaft, as before, and
in the support bearings, as shown in Fig. 5.5.
The distinction here is that the two sources of stiffness are now in the rotating
coordinate system and in the nonrotating one. As developed by Ehrich, steady-state
solutions for whirl amplification factors are presented. The factor for the bearing
support points wa is defined by the whirl radius of the support bearing points, and
that for the attachment point of the disk wb is defined by the whirl radius of the
center of the disk, both normalized by the mass eccentricity. Expressions for these
two factors are as follows:
u σ2
wa = = (5.5)
e [1 − (1 + κ)σ 2 ]2 + 4σ 2 ζs2 (1 − σ 2 κ)2
Fig. 5.4 Response as a result of initial curvature vs rotor speed for the Jeffcott rotor
on rigid bearing supports.
“chapter5(new)” — 2005/11/16 — page 152 — #4

Fig. 5.5 Jeffcott rotor on flexible bearing supports (after Ehrich, 1992).
where σ is a new speed ratio defined by the natural frequency for a rigid rotor
(Kr = ∞, i.e., all of the stiffness accrues from the support bearings):

σ = =√ (5.6)
µ Ks /M
and an appropriately defined new damping ratio η is determined by the support
dampers Bs :
Bs Bs
ζs = = √ (5.7)
2Mµ 2 MKs
A flexibility ratio κ is defined as
Ks
κ= (5.8)
Kr
The whirl amplification function for the center of the disk, wb is then given by

wb = wa (1 + κ)2 + (2σ ζs κ)2 (5.9)
An alternative design approach would be to define the speed ratio in terms of

the undamped rigid bearing natural frequency. The two speed ratios are related to
each other by
√
r = κσ (5.10)
“chapter5(new)” — 2005/11/16 — page 153 — #5

Fig. 5.6 Peak response characteristics caused by mass unbalance vs damping—

Jeffcott rotor on flexible supports (after Ehrich, 1992).
Because the preceding development entails a wide range of parameters and it

is to be expected that high responses will be obtained only near the resonance
conditions, a short-hand way of describing the response characteristics is in order.
Figures 5.6 and 5.7 provide useful information regarding the variation of the peak
value of the whirl amplification factor and the speed ratios at which it occurs.
Figure 5.6 highlights the important role of damping in minimizing the shaft
critical response: In particular, an optimum value of damping can be defined based
on the support to rotor stiffness ratio κ.
5.1.3 Subcritical and Supercritical Operation

Clearly, a successful design must be one wherein the rotor speed is adequately
removed from any lateral bending frequency. If the operational rotor speed is below
the principal bending frequency (r < 1), the rotor is defined as being subcritical,
whereas operation at a rotor speed higher than the bending frequency (r > 1) is
defined as supercritical operation.
Subcritical operation. The main reason for adopting a subcritical shaft

design is its inherent simplicity. Subcritical shaft systems are “tried and true”
and have no peculiar dynamic problems other than their need to be balanced to
minimize vibration. However, their disadvantage is significant for helicopter appli-
cations; subcritical designs are heavier and/or bulkier than equivalent supercritical
“chapter5(new)” — 2005/11/16 — page 154 — #6

Fig. 5.7 Speed ratio at peak response vs damping—Jeffcott rotor on flexible supports
(after Ehrich, 1992).
designs. Furthermore, the trend has been to operate at higher shaft speeds, making
subcritical designs more difficult.
Supercritical operation. The advantages claimed for supercritical shafts are

substantial:
1) Lighter weight: Shafts can be longer (fewer bearing supports).
2) Lower vibration: At supercritical speeds, as shown in the mathematical devel-
opment just given, the mass center of gravity G tends to approach the rotation
center O.
3) Lower support spring rates are required.
4) There is less sensitivity to structural vibration.
As with most attractive things in life, however, successful supercritical shaft-
ing comes at a price. First, in order to operate supercritically, the shaft must at
some time during run-up operate at the critical speed without excessive response.
This is accomplished first by providing external dampers to the shaft and second
by passing through the critical speeds quickly! A second problem area of super-
critical shafting is the issue of stability. Bishop presents theoretical formulations
to account for the catastrophic instabilities some manufacturers have had with
supercritical shafting. Bishop’s findings state that the presence of internal shaft
damping (structural damping) is a contributing factor to the instability. However,
the motion can be made stable again with adequate amounts of external damping;
hence, a second requirement exists for external lateral dampers. Bishop’s findings
“chapter5(new)” — 2005/11/16 — page 155 — #7

can be summarized as follows: For the kth lateral mode to which the shaft speed
is supercritical, the motion will be stable provided
ζ2 (ζint + ζext )
= rk < = (5.11)
ωk ζ1 ζint
where
ωk ≡ natural bending frequency of the kth lateral bending mode
ζ1 ≡ damping ratio of the shaft in the lateral bending mode without external
damping
ζ2 ≡ damping ratio of the shaft in the lateral bending mode with external damping
5.1.4 Gyroscopic Effects

The shaft-mounted disk used as an example in the Introduction to this section
(as drawn) has the property that, when the shaft is deflected, the disk will vibrate
or whirl in its own plane. When the disk is placed near one of the bearings or
on the end of an overhung shaft, however, it will not whirl in its own plane.
When this is the case, the disk will generate a centrifugal moment stiffening that
will give a higher critical speed. This phenomenon is customarily referred to as
the pseudogyroscopic effect, which is actually a special case of a critical speed
condition.
The example presented in Fig. 5.8 illustrates the effect more fully. Consider
a disk attached to the end of a constant section cantilevered shaft of length L.
The effects of the disk are thus seen to be an incremental force proportional to δ
and an incremental moment proportional to φ:
F = m2 δ; M = −Id 2 φ (5.12)
Fig. 5.8 Cantilevered shaft with attached disk.
“chapter5(new)” — 2005/11/16 — page 156 — #8

From basic elasticity relationships for a beam, the following influence coefficient
matrix is obtained:
3
δ L /3EI L 2 /2EI F
= 2 (5.13)
φ L /2EI L/EI M
When these two sets of equations are combined, the following equation set for δ
and φ is obtained:
[m2 L 3 /3EI − 1]δ − [Id 2 L 2 /2EI]φ = 0 (5.14)

[m2 L 2 /2EI]δ − [Id 2 L/EI + 1]φ = 0 (5.15)
A nontrivial solution for δ and φ requires that the determinant of coefficients be
zero (typical statement of an eigenvalue problem), which then yields
1/2
R = 2(1 − 1/3D) + 2 (1 − 1/3D)2 + 1/3D (5.16)
where R is the frequency with the disk effect divided by the frequency without the
disk effect, and
D ≡ Id /mL 2 (5.17)
This result, together with the results for other similar configurations, is pre-
sented in Fig. 5.9. Note that each of the curves is asymptotic to the value that
Fig. 5.9 Change in the natural frequency of a rotating shaft caused by the
pseudogyroscopic effect of a disk.
“chapter5(new)” — 2005/11/16 — page 157 — #9

would be obtained if the disk had remained in its initial plane, thereby providing
a zeroslope restraint to the lateral bending of the shaft.
5.1.5 General Transverse Vibration Analysis—Lumped-Mass

Approach
If the assumption is made that the dynamic system consists of concentrated
masses (with rotary inertia) tied together with massless, but elastic shafts, a gener-
alized analysis procedure can be defined. To give more generality, we can assume
that the elastic beams can be defined by Timoshenko beam theory. Again, as devel-
oped by Ehrich, the system is assumed to consist of a number of rigid disks each
having mass mi , diametral moment of inertia Idi , and axial moment of inertia Ji ,
connected sequentially by massless elastic shafts. Allowing for gyroscopic effects,
each mass has four degrees of freedom: ui and vi , transverse displacements in the
x and y directions, respectively, and rotations αi and βi also about these respective
axes. Each of the disks is then put in equilibrium with the forces and moments
from the elastic shaft segments on either side of it. With this assumed modeling
each of adjoining disks can be represented by four element displacement vectors:
 
 i
 u

 vi 
{qi } =

 α
 i 
βi
and
 


ui+1 
v  
i+1
{qi+1 } = (5.18)

α 
 i+1 
βi+1
These vectors also define the states at the i and i + 1 ends of the elastic seg-
ment connecting the adjoining disks. A complete elastic description of the segment
makes use of the pictorial representations given in Figs. 5.10a and 5.10b, which
show the transverse deformations in the x − z and y − z planes, respectively:
a) bending in the x − z plane and b) bending in the y − z plane.
The force-displacement for the connecting beam segment is then given by

{Qi } {qi }
= k (i)
(5.19)
{Qi+1 } {qi+1 }
Using a Timoshenko beam assumption results in both shearing and bending

compliances, and the [k (i) ] matrix then becomes
 
(i) (i)
ki,i ki,i+1
 
k (i) =   (5.20)
(i) (i)
ki+1,i ki+1,i+1
“chapter5(new)” — 2005/11/16 — page 158 — #10

Fig. 5.10 Transverse deformation of a flexible shaft segment (after Ehrich, 1992).
where
 
12 0 0 6Li
EIi  0 12 −6Li 0 
(i)
ki,i = 3   (5.21)

Li (1 + 12εi ) 0 −6Li 4Li (1 + 3εi )
2 0 
6Li 0 0 4Li (1 + 3εi )
2
 
−12 0 0 6Li
T EIi  0 −12 −6Li 0 
(i) (i)
ki,i+1 = ki+1,i = 3  
Li (1 + 12εi )  0 6Li 2Li (1 − 6εi )
2 0 
−6Li 0 0 2Li (1 − 6εi )
2
(5.22)
“chapter5(new)” — 2005/11/16 — page 159 — #11

(i) (i)
The [ki+1,i+1 ] matrix is equal to the [ki,i ] matrix, but with the polarity of the
6Li terms reversed. Also, the factor εi = EIi /(κGAi Li2 ) is a measure of the relative
shear compliance in the ith segment. The value of κ to be used is not univer-
sally accepted for circular sections. Young gives κ values of 0.968 and 0.90, for
solid circular and thin-walled hollow circular sections, respectively. Mindlin and
Deresiewicz give it a value of 0.847 for solid sections. Ehrich suggests that if 12εi is
negligible compared with unity, the simpler Bernoulli–Euler beam representation
(with εi set to zero) can be used. The additional elements required to formulate
the four scalar equations defining the dynamics for each of the disks are the mass
matrix [mi ] and the gyroscopic matrix [gi ]:
 
m 0 0 0
0 m 0 0 
[mi ] = 
0 0 Id 0 
(5.23)
0 0 0 Id i
 
0 0 0 0
0 0 0 0
[gi ] = 
0 0 0 J
(5.24)
0 0 −J 0 i
Putting all of the pieces together then produces the following matrix equation
for the ith disk:

(i) (i)
[mi ]{q̈i } + [gi ]{q̇i } + ki,i {qi } − ki,i+1 {qi+1 }

(i−1) (i−1)
− ki,i−1 {qi−1 } + ki,i {qi } = 0 (5.25)
Example 5.1 Jeffcott Rotor with Gyroscopic Dynamics

Equation (5.25) can be applied to the Jeffcott rotor wherein there are two elastic
shaft segments on either side of the single disk, as shown in Fig. 5.11. This equation
can be simplified for the Jeffcott rotor by setting the end moments to zero, relating
the endpoint rotations to the motion at station 2, and by imposing zero displacement
constraints at the endpoints. These simplifications result in two uncoupled equation
sets defining the translational motion (as was presented earlier) and the rotational
motion:

M 0 ü kT 0 u
+ = {0}
0 M v̈ 0 kT v

Id 0 α̈ 0 J α̇ kR 0 α
+ + = {0}
0 Id β̈ −J 0 β̇ 0 kR β
where the translational and rotational stiffness coefficients are respectively given by
6EI 6EI
kT = ; kR = (5.26)
(1 + 3ε)L 3 (1 + 3ε)L
“chapter5(new)” — 2005/11/16 — page 160 — #12

Fig. 5.11 Pictorial representation of the Jeffcott rotor with gyroscopic dynamics and
a comprehensive elastic description of the shaft segments (after Ehrich, 1992).
The equation set defines four imaginary paired eigenvalues, that is, four
frequencies:

kT J J 2 kR
ω1 , ω2 = ± and ω3 , ω4 = ± +
M 2Id 2Id Id
Fig. 5.12 Jeffcott rotor whirl speeds and modes and shaft critical speeds (after Ehrich,
1992).
“chapter5(new)” — 2005/11/16 — page 161 — #13

The positive and negative frequencies pertain to motion in opposite directions.

In the case of frequencies ω3 and ω4 , the positive frequency is associated with a pos-
itive rotor speed and represents a corotating whirl motion in the same direction
as the rotor speed. Conversely, the negative frequency represents a counter-rotating
whirl opposite to the rotor speed. These frequencies are shown in Fig. 5.12 along
with the intersection points with the rotor speed, which define the shaft critical
speeds. The critical speed point associated with the corotating whirl frequency
actually defines the shaft critical speed with pseudogyroscopic effects discussed
in a preceding subsection. It is to be understood that the lumped-mass approach
presented in this subsection is an approximation. A more rigorous continuum
model for transverse vibration analysis is presented in Ehrich, but is beyond the
intended scope of the present text.
5.2 Torsional Natural Frequencies of Shafting Systems

5.2.1 Introduction
From a forced-response standpoint problems normally associated with system
torsional vibration can best be solved by avoiding excitation of the drive system at
any of its natural frequencies. Typical sources of drive system excitation include
blade passage frequency loads from either the main or tail rotors (or propellers),
Hooke’s joint couplings, and mechanical couplings with torsion-related airframe
subsystems (e.g., transmissions and/or engines). Although the general problem of
natural frequency calculation has been covered in earlier sections and despite the
present general availability of applicable finite element analysis programs, signifi-
cant details of the torsional vibration problem of drive systems require examination.
Basic preliminary elements of the torsional vibration problem are the proper
definitions of the torsional masses and stiffnesses of discrete elements, as well
as the identification of practical methods of modeling and analyzing distributed
properties, gears, and branched systems.
5.2.2 Element Equivalencies

To find the torsional natural frequencies and mode shapes of a drive system,
it is necessary to reduce the various components (shafts, gears, rotors, etc.) to
equivalent rotary inertias and torsional springs.
Rotary inertia equivalences. Basic rotary inertia is expressible as J = Mk 2 ;

k is the polar radius of gyration. The use of mass and polar radius of gyration to
describe torsional inertia provides a convenient basis for calculating this quantity
for a variety of solid shapes. Figure 5.13 presents formulas for calculating polar
radii of inertia for a sampling of basic (homogeneous) solids representative of
typical drive train components.
Torsional stiffness equivalences. The basic torsional stiffness for any

constant circular solid section shaft is given by
K = π D4 G/32L (5.27)
“chapter5(new)” — 2005/11/16 — page 162 — #14

Fig. 5.13 Polar radius of gyration equivalences for basic solids.
where
G ≡ shear modulus of elasticity [= E/2(1 + ν)]

ν ≡ Poisson’s ratio
L ≡ length of the shaft element
For shafts with varying sections, one technique is to form equivalent lengths
of reference diameter shaft (D). These lengths are added together to form a single
constant section length of shaft (with diameter D), and the formula given earlier
is then used. Thus, for basic shaft types (of length L1 ) the equivalent lengths are
given as follows:
“chapter5(new)” — 2005/11/16 — page 163 — #15

Solid cylindrical shaft of diameter D1 :
L = L1 [D/D1 ]4 (5.28a)
Hollow cylindrical shaft of inner diameter d1 :
L = L1 [D4 /(D14 − d14 )] (5.28b)

(Solid) tapered shaft:
L1 D4
L= [1/d13 − 1/D13 ] (5.28c)
3(D1 − d1 )
Cylinder of material G1 (different from the reference material G ):
L = L1 [D4 G/D14 G1 ] (5.28d)
Effect of gears. The torsional system is most conveniently analyzed if all of

the elements are rotating at the same speed N. The conversion through the gears is
obtained using the following formulas based on the gear ratio (Nj /N = nj ), where
Nj is the shaft speed of the shaft element to be converted:
Moments:
M/Mj , nj (5.29)
Angles:
θ/θj , nj−1 (5.30)
Inertias:
J/Jj , nj2 (5.31)
Stiffnesses:
K/Kj , nj2 (5.32)
where
Nj ≡ rpm of jth shaft (original system)
N ≡ rpm of all shafts (equivalent system)
See Fig. 5.14 for an example.
5.2.3 Basic Natural Frequency Calculations for Simple Systems

For some simple torsional systems either closed-form or graphical methods of
solution for the natural frequencies are possible. Although most practical problems
typically involve several degrees of freedom, these problems can often be reduced
“chapter5(new)” — 2005/11/16 — page 164 — #16

Fig. 5.14 Example of a simple geared shafting configuration.
to simpler systems by means of the equivalences considered in the preceding

section. Three such basic systems are considered herein.
Two-inertia (free-free) problem. Because torsional systems typically do not

have an infinitely rigid point of attachment, the two-inertia problem with a single
connecting stiffness is the simplest basic form. The natural frequency of the single
free-free mode is given by

K(J1 + J2 )
ω= (5.33)
J1 J2
The node point is given by the length L1 :
L1 = J2 L/(J1 + J2 ) (5.34)
“chapter5(new)” — 2005/11/16 — page 165 — #17

Three-inertia (free-free) problem. The next most complicated of the basic,

simple systems is the three-inertia problem, again with the assumption of a free-free
constraint (see Fig. 5.15). This system has only two natural frequencies that are
determined from the solution of the following (essentially quadratic) characteristic
equation:

J1 J2 J3 J1 J2 J1 J3 J1 J3 J 2 J3
ω4 − + + + ω2 + (J1 + J2 + J3 ) = 0
K1 K2 K1 K1 K2 K2
(5.35)
Two-inertia problem (with a zero compliance end). For cases wherein the
free-free constraint is not met and there are two inertias, two stiffnesses, and an
infinite effective impedance constraint at one end (i.e., zero compliance), then there
Fig. 5.15 Simplified representation of a three-inertia shafting configuration.
“chapter5(new)” — 2005/11/16 — page 166 — #18

will again be two natural frequencies. The appropriate characteristic equation can
be written using the results of the preceding three-inertia case by taking the limit
as J1 approaches ∞:

J2 J3 J2 J3 J3
ω4 − + + ω2 + 1 = 0 (5.36)
K1 K2 K1 K1 K2
Shafts with significant mass. In some cases the moment of inertia of the
shaft system resides principally in the shaft itself, in which case either an appro-
priate simplified technique must be used, or recourse must be made to the use of
higher degree-of-freedom techniques such as those considered in subsequent sec-
tions. Consider the (three-) inertia system given in Fig. 5.16, which is comprised
of the two concentrated inertias at the ends J1 and J2 , as well as the polar moment
of inertia of the connecting shaft J0 .
Make the following definitions:
G
a2 ≡ (5.37)
ρ
where ρ is the density of the shaft, and
m = J1 /J0 ; n = J2 /J0 (5.38)
The a parameter can be seen to be the propagation speed of shear waves in the shaft
material. The dynamics of the shaft (with inertia) is then defined by an appropriate
differential equation. The mating of this shaft with discrete inertias at the ends of
the shaft is accomplished using appropriate boundary conditions:
GJθ0 (0) = −J1 ω2 θ(0); GJθ0 (L) = J2 ω2 θ(L) (5.39)
Fig. 5.16 Simplified shaft configuration having significant distributed inertia.
“chapter5(new)” — 2005/11/16 — page 167 — #19

Timoshenko presents the basic classical solution technique to this problem in the
form of the solution βj of the following transcendental equation:

tan βj aβj
(mn) [βj tan βj ] − = m + n; ωj = = J2 ω2 θ(L) (5.40)
βj LGJθ0 (L)
where the solution for jth natural frequency ωj is thus given in terms of βj . Note
that more than one solution exists to the aforementioned transcendental equation.
Because the equation represents the dynamics of a distributed inertia system with-
out discretization, it therefore contains the information for all of the natural modes.
Consequently, the various solutions obtainable represent the modal frequencies of
the system.
Solution using tabulations of functions. Until the advent of ubiquitous

computational resources (both personal computers and/or programmable hand
calculators), the solution for the eigenvalues βj was typically obtained iteratively
using a table of functions for [x tan x] and [tan x/x]. These functions are graphi-
cally depicted in Fig. 5.17. In this type of solution, the value of β is varied until a
value is found that satisfies the transcendental equation already given, using (lin-
ear) interpolation to solve for βj (once bracketing values have been found). Indeed,
for purposes of obtaining a rough, quick calculation for the first mode, the use of
such a table can be quite useful.
Fig. 5.17 Plots of the functions x tan x and tan x/x.
“chapter5(new)” — 2005/11/16 — page 168 — #20

Solution using numerical iteration. Apart from the tedium and inconvenience
of iteratively extracting a solution to Eq. (5.41) from tabulated data, the method is
inherently limited in the number of modes it can extract. As plotted in Fig. 5.17, the
functions are given only for abscissa values extending from zero to π . Although
the tan function is periodic with period π , the [tan x/x] and [x tan x] functions are
not. Thus, if the tabulated functions are available only from zero to π , only the first
mode can be extracted with the use of tabulated values. For shaft configurations
that are relatively short, this might be sufficient, but for longer ones, wherein
higher modes would still be significant, an alternate approach is warranted. One
method would be to modify Eq. (5.41) to admit the higher modes and solve the
modified equation in some fashion. To this end, we first rewrite the eigenvalue
problem in a somewhat different form, recognizing that the tan function is repetitive
with a period of π . The solution β can therefore be expressed as a sum of a
periodic part kπ and a principal-value part β̃ (i.e., the total solution is given by
β = kπ + β̃). The basic transcendental equation is then recast as a function of β̃
and put in the standard form for any of a variety of numerical iteration solution
schemes:
tan β̃
f (β̃) = mn(kπ + β̃) tan β̃ − − (m + n) = 0 (5.41)
kπ + β̃
Four or more inertias. For more complicated systems that cannot be sim-
plified down to three inertias or less, recourse must be made to more systematized
methods such as the Holzer–Myklestad , as is introduced in Section 3.2.1. Consider
an idealization of an elastic dynamic torsional shafting system, as approximated by
a series of multiple torsional elements each consisting of a concentrated torsional
inertia interconnected by massless springs, as shown in Fig. 5.18.
The iterative solution for the natural frequencies follows along the same lines
as is described for the blade torsion methodology. Again we seek a transfer matrix
approach that will require statements of equilibrium and continuity at the separate
Fig. 5.18 Idealized multiple-degree-of-freedom torsional shafting system.
“chapter5(new)” — 2005/11/16 — page 169 — #21

node points. Let

Tn ≡ amplitude of torque in the nth shaft segment
θn ≡ amplitude of angular displacement of the nth inertia
vn ≡ flexibility of the nth shaft (= 1/Kn )
The pertinent equations for the nth element then become
θn+1 = θn − vn Tn (5.42a)
Tn+1 = Tn + Jn+1 ω θn+12
(5.42b)
where the boundary conditions are imposed, respectively by the starting equation
(at the right-hand end):
T1 = J1 ω2 (θ1 ) = J1 ω2 (1) (5.43a)

and by the convergence criterion (zero torque at the left end):
TN = 0 (5.43b)
Although not strictly needed for obtaining a solution, all of Eqs. (5.42) and (5.43)
can be recast in a transfer matrix format:

θ 1 −vn θ
= (5.44)
T n+1 Jn+1 ω2 (1 − vn Jn+1 ω2 T n
This formulation is thus seen to be very similar to the methodology for calculat-
ing the blade torsion natural frequency described in Section 3.5.4. A frequency
sweep is necessary to find that value that drives the torque at the left-hand side
to zero.
5.2.4 Branched Drive Systems

In the material discussed thus far, it was implicitly assumed that all of the ele-
ments of the drive system were connected in series. However, for most helicopter
applications the engine(s) are driving more than one inertia element (main and
tail rotors and possibly a propeller) through various branch points (transmissions
and/or gear boxes).
Analyzing such systems requires two prior steps:
1) Reduce the branched system to an equivalent system all of whose elements
rotate at the same revolutions per minute.
2) Set the net sum of the inertia torques at each of the branch points to zero.
Once these steps have been taken, the frequencies are calculable using either
the Holzer–Myklestad method or a formal matrix eigenvalue statement solu-
tion. Consider the application of these two techniques to the example given in
Fig. 5.19.
At the branch point,
ω2 [(J4 θ4 + J3 θ3 ) + (J1 θ1 ) + (J2 θ2 )] = 0 (5.45)
“chapter5(new)” — 2005/11/16 — page 170 — #22

Fig. 5.19 Simplified shafting configuration having a branched gear system.
Holzer–Myklestad method. Using the development given in the preceding

subsection, we can write the pertinent equations as follows:
T1 = J1 ω2 θ1 = J1 ω2 (1) (5.46a)
T2 = J2 ω2 θ2 (5.46b)
T3 = T1 + T2 + J3 ω θ3 2
(5.46c)
θ3 = θ1 − v1 T1 = θ2 − v2 T2 (5.46d)
θ4 = θ3 − v3 T3 (5.46e)
T4 = T3 + J4 ω2 θ4 (5.46f)
where vk = 1/Kk .
These equations can be then combined to yield the following more tractable
forms:
T1 = J1 ω2 θ1 = J1 ω2 (1) (5.47a)
θ2 = ω̄22 (ω̄12 − ω2 )/ω̄12 (ω̄22 − ω2 ) (5.47b)
θ3 = θ1 − v1 T1 = 1 − v1 T1 (5.47c)
T3 = ω 2
θ3 [J3 + 1/v1 (ω̄12 −ω 2
) + 1/v2 (ω̄22 − ω )]
2
(5.47d)
θ4 = θ3 − ν3 T3 (5.47e)
T4 = T3 + J4 ω2 θ4 (= 0, as one of the boundary conditions) (5.47f)
where
ω̄12 = 1/v1 J1 ; ω̄22 = 1/v2 J2 ; ω̄42 = 1/v3 J4 (5.48)
“chapter5(new)” — 2005/11/16 — page 171 — #23

Note that ω̄1 , ω̄2 , and ω̄4 are the frequencies that each of the respective branches
would experience if the branch point were a node (i.e., θ3 had zero torsional
deflection).
1) If ω̄1 = ω̄2 , one natural mode will have a node at the branch point, and the
resulting natural frequency would be given by ω = ω̄1 = ω̄2 .
2) Similar situations can arise between the other two paired combinations of
branches.
3) If ω̄1 = ω̄2 = ω̄4 and J3 = 0, then there are only two distinct nonzero
frequencies (ω = ω̄1 = ω̄2 = ω̄4 ), and those are given by
ω2 = ω̄12 (J1 + J2 + J3 + J4 )/J3 (5.49)
4) If ω̄1 = ω̄2 = ω̄4 and J3 = 0, then θ3 ≡ 0, and the inertia torques as a result
of J1 , J2 , and J4 must be in balance.
5) The situations depicted in points 3 and 4 define variants of a nodal drive
configuration.
Matrix eigenvalue method. Using either a Lagrangian approach or apply-

ing net moment equilibrium for each inertia yields the following basic (matrix)
equation for the four-inertia system:
 
   K1 0 −K1 0  
J1 0   
θ̈   
θ1
0 
1
     0 K2 −K2  
 J2  θ̈2   θ2
  +
3  =0 (5.50)
 J3 θ̈3 −K1
 −K2 −K3 θ3
 
Ki

    i=1

 
0 J4 θ̈4 θ4
0 0 −K3 K3
The condition of zero net torque at the branch point is given as a constraint equation
conveniently expressed in matrix form:
   

 θ1 
 1 0 0    

θ   −J /J θ1  θ1 
2  1 2 −J3 /J2 −J4 /J2 

=  θ3 = [T ] θ3 (5.51)

 θ3  0 1 0    

   θ4 θ4
θ4 0 0 1
When this constraint equation is combined with the preceding set of dynamic
equations and the second of the resulting equations (second row of the matrix
equation) is discarded, we obtain
     
J1 0 0 θ̈1 
 K1 −K1 0 θ1 
0  
J3 0  θ̈3 + −K̃1 K̃B −K̃3  θ3 = {0} (5.52)
   
0 0 J4 θ̈4  0 −K3 K3 θ4
“chapter5(new)” — 2005/11/16 — page 172 — #24

where
K̃1 = K1 − J1 K2 /J2 (5.53)

3
K̃B = Ki + J3 K2 /J2 (5.54)
i=1
K̃3 = K3 − J4 K2 /J2 (5.55)
When sinusoidal motion is assumed, this equation can be reduced to a standard

matrix eigenvalue problem of the form
[[K] − λ[I]]{} = {0} (5.56)
where λ = −ω2 .
This matrix equation is nonsymmetric but can still be solved using any of a
variety of standard matrix eigenvalue techniques. Note that a symmetrical form
of the matrix equation can be obtained by instead retaining the second row
after the substitution of the constraint matrix T and then premultiplying by the
transpose of T(= T T = T −1 ). The resulting matrix equation is then 3 × 3 and
symmetric.
5.2.5 Drive System of a Typical Helicopter

All of the elements described in the preceding sections come together in a
typical helicopter. As shown in Fig. 5.20, the engine, any attached accessories, the
main rotor, and the tail rotor all form a multibranched drive system. Furthermore,
each branch consists of several distinct inertias connected by elastic elements of
varying sizes and flexibilities.
Although the exact details of any one particular design can differ from the
configuration shown in Fig. 5.20, the methodology presented subsequently should
suffice for obtaining a suitable dynamic analysis of most configurations. The basic
procedure for analyzing the generic system shown in Fig. 5.20 consists of the
following steps:
1) Make element equivalencies to have all elements rotating at the same
rotational speed.
2) Within each branch identify and calculate the properties of each basic unit,
to consist of a concentrated inertia connected with a flexible element, which might
or might not have distributed inertia properties.
3) Within each branch connect the elements from the free end back to the branch
root point.
“chapter5(new)” — 2005/11/16 — page 173 — #25

Fig. 5.20 Schematic representation of the drive system of a typical helicopter.
4) Formulate and apply the constraint(s) that connect all of the branch roots
together.
5) Use a robust eigenvalue routine to extract the roots of the resulting matrix
eigenvalue formulation.
Figure 5.21 demonstrates how each branch should be idealized into discrete
elements. As the figure implies, the details of the elastic elements can be quite
general; it can be a shaft with distributed inertia and stiffness, or a massless tor-
sional spring. Figure 5.22 gives the mathematical quantities needed to define the
dynamics of the element.
The methodology we use to formulate the dynamic description is a variant of the
Ritz/Hamilton method articulated by Bailey, which takes an energy approach. For
each element we separately formulate the kinetic and potential (strain) energies.
First, however, the torsional deflection over each (kth) element is taken as a power
series:

I
(k)
θk (η) = ai ηi (5.57)
i=0
where η is a the local spanwise coordinate over the length of the element and the
value of I to be used in the summation is selectable to as high a value as is needed
“chapter5(new)” — 2005/11/16 — page 174 — #26

Fig. 5.21 Idealization of a typical branch into a series of cascaded elements.
(k)
for convergence of the final eigenvalues. The ai coefficients represent the modal
variables in the final eigenvalue solution. The kinetic and potential energies are
then expressed as

1 t2 lk
T= ρJ0k θ̇k2 (η) dη + Jk θ̇k
(O)
dt (5.58)
2 t1 0
Fig. 5.22 Mathematical description of the basic torsional element.
“chapter5(new)” — 2005/11/16 — page 175 — #27

lk
1 t2 (I) 2
(GJ)k θk2 (η) dη + Kk θk − θk
(O)
U= dt (5.59)
2 t1 0
where ρJ0k is the distributed inertia, having the units of (F–S 2 ); Jk , the end
concentrated inertia, has the units of (F–S 2 –L) and (GJ)k , the distributed torsional
inertia, has the units of (F–L 2 ). After assuming sinusoidal motion, the application
of the Ritz–Hamilton method requires that the variation of the difference of these
energies be zero, that is,
δU − δT = 0 (5.60)
For the kth element this variation procedure results in the following expression:

t2 lk
δU − δT = (GJ)k θk (η)δθk (η) dη
t1 0

(O) (I) (O) (I)
+Kk θk − θk δ θk − θk dt

t2 lk
(O) (O)
+ω 2
ρJ0k θ (η)δθk (η) dη + Jk θk δθk dt = 0 (5.61)
t1 0
(k)
All of the variations in Eq. (5.62) are essentially variations on the coefficients ai .
Thus, Eq. (5.62) can then be expressed in a more useful variational form:
 (k) 

 a 
 0 
 

! " 1 a(k) 
 
δa0(k) , δa1(k) , . . . , δaI(k) (−ω2 ) lk ρJ0k η̄i+j dη̄ + Jk [I] 1


.. 
0 
 . 

 (k) 
 
aI
 


0 


 
!
" 1 1 a1(k) 

(k) (k)
+ 0, δa1 , . . . , δaI (GJ)k η̄i+j+2 dη̄ + Kk [I] .. =0
l 0 
 . 


 

 (k) 
aI
(5.62)
“chapter5(new)” — 2005/11/16 — page 176 — #28

Because each of the δai(k) is arbitrary, Eq. (5.62) can be written in the following
matrix form:
 (k) 

 a0  

 

 (k) 
# $ a1 

−ω[M̃i, j ]k + [K̃i, j ]k ..  = 0 (5.63)

 . 

 

 (k) 
 
aN
Note that the stiffness matrix for the kth element [K̃]k has a zero first row and
a zero first column. At this point all of the spanwise properties have been taken
care of, and the remaining chore is to properly string all of the elements together.
The first construction is to string all of the elements in each branch “element to
element”. Two relationships are needed; first, to go from element to element, the
following expressions must hold:
(I) (O)
θk = θk−1

I
∴ ao(k) = an(k−1) 2 ≤ k ≤ Nm (5.64)
i=0
(m)
Secondly, all of the (m) branch root deflections θR must be tied together to a
common trunk point deflection θT .
(m) (1)
θR = [a0 ](m) = θT (5.65)
These relationships are accomplished with appropriate constraint matrices. First,

however, we add additional indexing to the modal variable coefficients. The
coefficients and the inertia and stiffness matrices must now have an index m to
denote to which branch they belong. With this additional indexing the stringing
together of the elements within a branch takes the following form:
 (1,m)   (m) 
 a0   θR   (m) 

    

 (1,m) 
  
 


(1,m) 




θR  


 a1   
a1   
 (1,m) 


  
  
 
a 


 ..     ...   
 1.  
. ..
= = [Tm ] (5.66)

 a0(2,m) 
 
 a0(2,m) 
 
 

 
 
 
 
 (2,m) 



 
(2,m)  
 
(2,m)  
 a 


 a 
 
a 
 

1




1 
 
 1 
  .
. 
 ..    .  .  .
. .
“chapter5(new)” — 2005/11/16 — page 177 — #29

or, more specifically,
 (1,m) 

 a0 


 (1,m) 
 1 | | |

(m) 

 
  θR 


a1 
  1 
 


 ..    | 0 | 0 | 0 
 (1,m)  

 
  
 a 


 . 
  .. 
 1 


 
  . | | | 
 . 


 (1,m) 
  
 .
. 



aI 
  | | | 
 


 − − − 
 
1

 (1,m)  

 
  − − − − − − − − − − − − − − − − − − − − − 
 a 

a
 (2,m)  
 
 I 


 
  0 1 · · · 1 | 0 · · · 0 | | 
 − − − 
(2,m) 
0

 
  
 


 a1
(2,m) 
 0 0 · · · 0 | 1 | 0 | 0 
 a 


 
  

1


 .   .  .
. 
..  · · · | . . | |  .
= 0 0 0
 (2,m) 

 (2,m)   0 0 · · · 0 |  

 a 
  1 | | 
 aI 


 I 
  
 − − − 


 − − − 
  − − − − − − − − − − − − − − − − − − − − − 
 

 . 
   . 
 . 


 . 
  0 | · · · | . . | 0 
 . 


 .  
 

. 


 − − − 
  | | · · · | 
 − − − 

 (Nm ,m) 
   0 0 0 
 


a 
  
 (N m ,m) 


 0 
  0 | 0 | | 1  1
a 


 
(Nm ,m)   
 


a1 
  . 

.
. 


 
  | | | . . 
 . 


 
 
 
(Nm ,m) 
 .
. 

 . 
 | | | 1 a

 (Nm ,m)   I
aI
 (m) 
 θR 
 

 (1,m) 


 



a1 


 ..  

 

 .  


 (1,m) 


 aI 


 


 − − − 

 

 a1(2,m) 
 

 

 .. 
  (m)
. θ
= [Tm ] = [Tm ] R(m) (5.67)

 
 aI(2,m) 
 

A

 
− − −
 


 
 ... 
 


 


 − − − 


 

 (Nm ,m)  


a 



1



 . 
 . 

. 

 (Nm ,m) 
 
aI
“chapter5(new)” — 2005/11/16 — page 178 — #30

Thus, in Eq. (5.67) the mode shape vector has been compacted to include the
branch root deflection θR(m) plus all of the coefficients relating to the higher powers
deflection for the entire branch A(m) . Equation (5.63) can then be combined with
Eq. (5.67) to give the dynamic description for the mth branch:
 (m) | | 
−ω2 [M̃](m)
1 + [K̃]1 0 ··· 0
 | | 
 − − − − − − − − − − − − − − − − − − − − − − − − − − − − −− 
 
 | 
−ω2 [M̃]2 + [K̃]2 |
(m) (m)
 0 ··· · · ·
 | | 
 
 − − − − − − − − − − − − − − − − − − − − − − − − − − − − −− 
 .. .. 
 | |
−ω2 [M̃]3 + [K̃]3 · · ·
(m) (m)
 . . 
 | | 
 − − − − − − − − − − − − − − − − − − − − − − − − − − − − −− 
 
| .. | .. ..
0 . . .
| |
 (m)   (m) 
 θR   θR 
× [Tm ] −− = −ω2 [M (m) ] + [K (m) ] [Tm ] −− = {0} (5.68)
 (m)   (m) 
A A
The final mathematical construction is to tie all of the branch root deflections
to the same trunk deflection θT . Also, as formulated by the preceding develop-
ment, the trunk point has not been attached to any inertia. Thus, the dynamics
of the trunk point must be included as a separate equation. The following matrix
operation achieves the required connections of the branch root points and prepares
the equation set for the trunk inertia equation:
 
 θT 

 


 (1) 
  

 θ 
 1 0 ···  


R 
 θT 

 (1) 
 1 0 ···  

 A 
   

  0 I 0 ···  (1) 

 (1) 
   A 

θR  1 0 ···   
= · · ·  A (2)
= [U]{X} (5.69)
A   0 I 0 


(2)

 1 0 ···  (3) 


    A 

 (1) 

  ··· 0 · · ·

 


 θR    I  ..  

 
 .. .

 A 
(3)  .

 

 .. 
 
.
The final form of the matrix eigenvalue problem is obtained by making the
Eq. (5.69) substitution and then premultiplying by the transposes of the U and
“chapter5(new)” — 2005/11/16 — page 179 — #31

combined Tm matrices:
1 0 ··· T
0 [T1 ] 0 · · ·
 .. 

T [T2 ] · · ·
[U]  . 

 ... ..
[T3 ] · · ·
 . 
.. ..
. .
 
−ω2 JT | 0 | ··· |
− − − − − − − − − − − − − − − − − − − − − − −
 
 | −ω2 [M (1) ] | | 
 
 | 
 0 +[K (1) ] | 0 | ··· 
− − − − − − − − − − − − − − − − − − − − − − −
 
 | | −ω2 [M (2) ] | 
×
 ... 
 | 0 | +[K (2) ] | 0 ··· 
− − − − − − − − − − − − − − − − − − − − − − − 
 
 | | | −ω2 [M (3) ] 
 
 | .
. | | 
 . 0 +[K (3) ] 
− − − − − − − − − − − − − − − − − − − − − − − 
| | | ..
.
1 0 ··· 
0 [T1 ] 0 · · ·
 .. 
 [T2 ] · · ·
×
.
.  [U]{X} = {0}
 (5.70)
 .. ..
 . [T3 ] · · ·
.. ..
. .
or, more compactly as
[−ω2 [M] + [K]]{X} = {0} (5.71)
Equation (5.71) is now in the final desired matrix eigenvalue form. The com-
bined mass and stiffness matrices comprising Eq. (5.72), [M] and [K], respectively,
are both symmetrical, and hence the equation is amenable to a variety of standard
matrix eigenvalue solution routines.
5.3 Special Devices

5.3.1 Pendulum Dynamic Absorbers
In a manner similar to the operation of the vibration absorber discussed in
Section 2.3, a pendulum dynamic absorber eliminates vibratory responses of
elements in a torsional dynamic system. However, the lineal vibration absorber
“chapter5(new)” — 2005/11/16 — page 180 — #32

achieves its frequency characteristics using an elastic spring and is thus effective
at only one frequency of excitation, actually contributing to amplified responses
above and below this frequency. On the other hand, the torsional pendulum
absorber will attenuate responses at a frequency that is a constant factor of the
shaft rotation frequency regardless of that frequency!
Let the drive system have a torsional natural frequency of n, and let it be
excited at that frequency by a torque of amplitude T0 . Consider a pendular mass
m attached to the drive system as shown in Fig. 5.23. The equation of motion for
the pendulum response θ is
mL(L θ̈nA + 2 RθnA ) = −m(R + L)L θ̈n (5.72)
where θn is the torsional response of the shaft to be attenuated. Let the responses
be sinusoidal:
θ̈n (t) = −(n)2 θ̄n sin (nt) (5.73a)

θnA (t) = θ̄nA sin (nt − φ) (5.73b)
then
θ̄n 2 [R/L − n2 ]
= (5.74)
θ̄nA [n2 2 (R + L)/L]
Fig. 5.23 Schematic of a torsional pendulum absorber.
“chapter5(new)” — 2005/11/16 — page 181 — #33

Consequently, θn = 0 if R/L = n2 . This represents the tuning condition required

to attenuate the torsional vibrational responses. The vibratory torque amplitude
generated by the motion of the pendulum is easily calculated to be
A
T0 = mn2 2 (R + L)Lθ n (5.75)
In practice, for a given application, the value of L required is usually small

[O(L) = 0.1 in.], and the value of mass required to keep θnA of the order of mag-
nitude of 10 deg is impractically large. These considerations make a “simple”
pendulum design difficult. For cases of this type, a bifilar type of torsional
pendulum absorber becomes practical (see Fig. 5.24).
Integration of the pendulum absorber into a torsion drive system. The

development in the preceding subsection serves to demonstrate why the pendulum
absorber works, but in practice the dynamics of the pendulum absorber must
be integrated into a more complicated drive system. This can be accomplished
in two different ways. Figure 5.25 shows how an equivalency can be made for
including a pendulum absorber into a simple system that can be analyzed using a
Holzer–Myklestad method.
One factor that must be considered is that the dynamic description for the
pendulum absorber, as given by Eq. (5.74), assumes that the absorber response
variable θnA is a relative variable and not inertial. That is, it is measured relative
to the attachment angle θn and not relative to inertial space, as are all of the other
Fig. 5.24 Bifilar configuration for a torsional pendulum absorber.
“chapter5(new)” — 2005/11/16 — page 182 — #34

Fig. 5.25 Equivalency of pendulum absorber to a single-degree-of-freedom element

attached to the drive train at a branch point.
drive train rotational degrees of freedom. For this reason, the additional inertia
provided by the absorber must be included with the base element, as shown in
Fig. 5.25.
Using Eqs. (5.72) and (5.75), we can form the appropriate attachment, torque
TnA , and the equation for the absorber:
TnA = −mL(R + L)ω2 θ̄nA (5.76)

%
R R
θ̄nA = ω2 + 1 θ̄n 2 − ω2 (5.77)
L L
One difficulty with this formulation is that it offers the solution the opportunity
to go singular for frequency values equal to [R/L]1/2 . One remedy for this is
to introduce a small amount of damping in the absorber equation of motion and
perform the Holzer–Myklestad method in complex arithmetic.
A second method for including the absorber dynamics in a general drive train
dynamic analysis is to use the matrix synthesis method outlined in Sec. 5.2.5 for a
typical helicopter. The basic methodology outlined in that section can be applied
to this case by including an additional equation, treating the base inertia as a trunk
point, and including a branch on either side of the absorber base inertia. Thus, the
“chapter5(new)” — 2005/11/16 — page 183 — #35

equation set would consist of a partitioning of the following form:

I 0 0 T
[U]T 0 T1 0
0 0 T2
 2 
−ω (Jn + mL 2 ) −ω2 mL(R + L)
 −ω2 mL(R + L) mL 2 [−ω2 + 2 R/L] 0 0 
 
 −ω2 [M (1) ] 
 
× 0 0 
 +[K (1) ] 
 
 −ω2 [M (2) ]
0 0
+[K (2) ]
 

 θn 
I 0 0  θA 
 
n
× 0 T1 0 [U] = {0} (5.78)
A(1) 
 
0 0 T2  
 (2) 
A
5.3.2 Hooke’s Joint

A common (and relatively inexpensive) type of universal joint is the Hooke’s
joint depicted in Fig. 5.26. The input and output rotation speeds are related by the
following expression:
2 cos β
= (5.79)
1 1 − sin2 β cos2 θ
where
1 ≡ angular velocity of the driving shaft

2 ≡ angular velocity of the driven shaft
θ ≡ angular displacement of the driving shaft from the position, where the
pins on the drive shaft yoke lie in the plane of the two shafts
β ≡ angular misalignment between the axes of the two shafts
Fig. 5.26 Kinematics of a deflected Hooke’s joint.
“chapter5(new)” — 2005/11/16 — page 184 — #36

Characteristics of the Hooke’s joint
1) The input and output angular velocities through the joint are not in a constant
ratio. For small shaft angles (β 1), this ratio contains a significant second
harmonic (2P) in the angular frequency of the driver end of the joint.
2) A velocity ratio of unity can be obtained at any angle β using two Hooke’s
joints with an intermediate shaft, as long as the input and final output shafts are
parallel and the pins on the ends of the intermediate shaft are oriented parallel.
An example of the use of a Hooke’s joint and the usage of Eq. (5.75) is the
calculation of the 2P torque in a tail rotor drive shaft, which is driving a (rigid) tail
rotor connected by means of a Hooke’s joint (gimbal) (see Fig. 5.27). Assumptions
that can be made to define the dynamics are as follows:
1) The flapping angle β is small.
2) The perturbational torsional deflections inboard and outboard of the gimbal
are θ1 and θ2 , respectively, and are also small.
Using these assumptions, we can write the velocity ratio equation between the
input and output ends of the shaft:
θ̇2 +
= cos β(1 + sin2 β cos2 t) (5.80)
θ̇1 +
Differentiation yields
θ̈2 = θ̈1 cos β(1 + sin2 β cos2 t)

+ (θ̇1 + ) cos β[− sin2 β sin 2t] (5.81)
Fig. 5.27 Simplified representation of a gimbaled tail rotor driven by a torsionally

flexible drive shaft.
“chapter5(new)” — 2005/11/16 — page 185 — #37

where the underlined terms are small relative to the terms with which they are
added. Therefore,
θ̈2 = cos β(θ̈1 − 2 sin2 β sin 2t) (5.82)
The torque applied to the end of the shaft is then given by

GJs
T1 = − cos βJR θ̈2 = elastic restoring torque, = θ1 (5.83)

Substitution for θ̈2 and making the small-angle assumption on β yields
θ̈1 + ω2 θ1 = 2 β 2 sin 2t (5.84)
where
GJs
ω2 = (5.85)
JR
The solution to this linear equation presents no difficulties, and we can write the
particular solution as
−2 β 2 sin 2t

θ1 = (5.86)
42 − ω2
from which the acceleration can be calculated:
42 β 2 sin 2t
θ̈1 = (5.87)
42 − ω2
Then the torque in the shaft can be expressed as follows:

42
T1 = −JR 2 β 2 − 1 sin 2t (5.88)
42 − ω2
Thus, the amplitude of the 2P (twice/rev) torque is given by
JR β 2 ω2 2
|T1 | = (5.89)
|42 − ω2 |
References
Section
5.1 Bishop, R. E. D., “The Vibration of Rotating Shafts,” Journal of
Mechanical Engineering Science, Vol. 1, No. 1, 1959, pp. 50–65.
Den Hartog, J. P., Mechanical Vibrations, McGraw–Hill, New York,
1940.
“chapter5(new)” — 2005/11/16 — page 186 — #38

Ehrich, F. F., Handbook of Rotordynamics, McGraw–Hill, New York,
1992.
Mindlin, R. D., and Deresíewicz, H., “Timoshenko’s Shear Coefficient
for Flexural Vibrations of Beams,” Proceedings of Second National
Congress Applied Mechanics, ASME, New York, 1954.
Tse, F. S., Morse, I. E., and Hinkle, R. T., Mechanical Vibrations, Allyn
& Bacon, Boston, 1964.
Tuplin, W. A., Torsional Vibration, Pitman, London, 1966.
Young, W. C., Roark’s Formulas for Stress and Strain, McGraw–Hill,
New York, 1989.
5.2 Myklestad, N. O., Vibration Analysis, McGraw–Hill, New York, 1944.
Timoshenko, S., Vibration Problems in Engineering, Van Nostrand,
New York, 1937.
5.3 Doughtie, V. L., and James, W. H., Elements of Mechanism, Wiley, New
York, 1954.
Schwamb, P., Merrill, A. L., and James, W. H., Elements of Mecha-
nisms, Wiley, New York, 1938.
Problems
5.1 Derive the expression for the frequency ratio R given in Section 5.1.4.
5.2 Derive the transcendental equation statement for the eigenvalue problem
defining the natural frequency characteristics of the simple three-inertia
problem wherein the shaft connecting two outer inertias also has significant
inertia.
5.3 Deduce an equation for determining the torsional natural frequencies for a
shaft with two end inertias and significant torsional inertia wherein one of
the end inertias approaches infinity. (This corresponds to a case of a built-in
condition at one end of the shaft with significant inertia and a finite inertia
at the other end.)
5.4 Set up the matrix partitioning and additional equation for the case of two
trunks (sheaves) connected by a stiffness KC as shown in Fig. 5.20.
5.5 Consider the tail rotor drive shafting of a certain helicopter wherein the (hol-
low) shaft is made of aluminum, is 12 ft long, and has an outside diameter
of 3 in. and a wall thickness of 0.1 in. This shaft drives a tail rotor that turns
at the same speed as the shaft and has a polar moment of inertia of 14 lb.-
“chapter5(new)” — 2005/11/16 — page 187 — #39

s2 -in. For each of the two cases of 1) inboard attachment inertia approach-
ing infinity (rigid attachment to the remainder of the drive system, i.e., main
rotor, transmission, engine, etc.), and 2) zero inboard attached inertia (rep-
resentative of a failed coupling), calculate the first three torsional natural
frequencies. Additional useful information is given here: weight density of
A1, w = 0.101 lb/in.3 ; shear modulus of A1, G = 3.9 × 106 lb/in.2 .
“chapter5(new)” — 2005/11/16 — page 188 — #40

6
Fuselage Vibrations
In the preceding chapters the emphasis is placed on the diverse methodolo-

gies for obtaining the dynamic responses of rotorcraft structures, with particular
attention being paid to the rotating substructures (rotors and shaft systems). In
this chapter we address perhaps the most important and directly discernible con-
sequence of these responses: fuselage vibrations. To approach this problem, we
must proceed from the concept of dynamic responses of substructures to that of
the dynamic loads impacting on them, with emphasis on the influence of the rotor
loads on the fuselage. Furthermore, because the rotor and the fuselage are both
dynamic structural subsystems, they, in effect, talk to each other, and the concept
of rotor-fuselage coupling must therefore be addressed. Finally, it must be stressed
that the general problem of designing rotorcraft with fuselage vibrations that are
not only acceptable but “jet smooth” is still unsolved.
This state of irresolution is caused by two principal difficulties:
1) Precise knowledge of the loads (in particular, those of aerodynamic ori-
gin) acting directly on the rotor and acting indirectly on the fuselage is far from
complete.
2) The ability to calculate the structural dynamic characteristics of the airframe
in terms of natural frequencies, mode shapes, and mobilities does not yet exist with
sufficient accuracy. Consequently, at present heavy reliance must be made on the
use of various vibration alleviation devices to bring any given rotorcraft airframe
design to a point of at least having a moderately acceptable vibration level.
6.1 Dynamic Loads

6.1.1 Mode Deflection and Force Integration
Basic ideas. For any elastic structure the following relationship holds:

elastic restoring forces the applied forces
= (6.1)
and moments and moments
189
“chapter6(new)” — 2005/10/18 — page 189 — #1

or, more simply stated,
Felastic = Fapplied loads (6.2)
For any dynamic system knowledge of either of these quantities constitutes

knowledge of the dynamic loads. Methods of dynamic loads calculation that are
based on knowledge of the left-hand side of this equation are generally referred to
as mode-displacement methods, whereas methods based on the right-hand side of
the equation are alternatively referred to as force-integration, force-summation, or
mode-acceleration methods.
Illustrative examples. The following examples illustrate the concepts behind

these two basic methods of dynamic loads analysis:
Example 6.1
Given a simple spring-mass system, as shown in the following sketch, knowledge
of the dynamic load at the point P, FP , is required.
Solution
The differential equation for the single-degree-of-freedom system is, by now,
familiarly given by
mz̈ + kz = F(t) (6.3)
“chapter6(new)” — 2005/10/18 — page 190 — #2

FUSELAGE VIBRATIONS 191
and the load at point P is alternatively given by
FP = kz (via elastic restoring force)

FP = F(t) − mz̈ (via applied forces)
These two alternate methods for formulating the load at point P define the basis for
the mode-deflection and force-integration methods, respectively, for calculating
dynamic loads. The procedure for calculating the dynamic load can then be stated
as follows:
1) Solve for the response z(t).
2) Calculate the load FP at the point P using either of the two expressions given
earlier.
Consider a step function for the applied load F(t):

0; t < 0
F(t) = (6.4)
1; t ≥ 0
The resulting response, for quiescent initial conditions, is given by

z(t) = (1/k)(1 − cos ωt)
where ω = k/m (6.5)
z̈(t) = (ω2 /k) cos ωt
The load at point P, FP (t), is then calculated using the two methods:
Mode-displacement method:

1
FP (t) = kz = k (1 − cos ωt) = 1 − cos ωt (6.6)
k
Force-integration (summation) method:
FP (t) = F(t) − mz̈

= 1 − m(ω2 /k) cos ωt = 1 − cos ωt (6.7)
Thus, for the single-degree-of-freedom system the two methods are exactly
equivalent.
Example 6.2
Find the shear and bending moment at the midspan of the nonuniform cantilevered
beam shown in following sketch:
“chapter6(new)” — 2005/10/18 — page 191 — #3

Solution
1) Differential equation of motion is
mz̈ + (EIz ) = F(x, t) (6.8)
the solution of which can be expressed using natural modes:
∞

z(x, t) = γn (x)qn (t) (6.9)
n=1
where the forced-response equation for qn (t) is given by
Mn (q̈n + ωn2 qn ) = n (t) (6.10)

and where the nth generalized mass and excitation Mn and n , respectively, are
defined as
L
Mn = mγn2 (x) dx (6.11)
0
L
n(t) = γn (x)F(x, t) dx (6.12)
0
2) Loads by the mode-displacement method are as follows:

a) Bending moment Mb :
∞
∞

L
Mb , t = EIw = EI(x)γn (x) qn (t) = (Mb )n qn (t)
2
n=1 x=L/2 n=1
(6.13)
“chapter6(new)” — 2005/10/18 — page 192 — #4

b) Shear S:
∞
∞

L
S ,t = −[EIw ] = − [EI(x)γn (x)] qn (t) = Sn qn (t)
2
n=1 x=L/2 n=1
(6.14)
3) Evaluation of the loading coefficients Mbn and Sn is as follows: Because

the shear and moment are directly definable using the second and third span-
wise derivatives of the modal deflection shape function, it is tempting to try to
form appropriate derivatives and thereby evaluate these loading quantities directly.
However, practical considerations, involving the loss of accuracy with numerical
differentiation especially with EI variations that are abrupt and not easily differ-
entiable, preclude the use of such an approach. The loading coefficients can be
alternatively evaluated with reasonable accuracy using either of two main methods
described in the following, depending on how the mode shapes have been obtained.
a) Semianalytical method: With this method it is assumed that the nth natural
bending mode shape and corresponding modal frequency are known for the beam.
The appropriate starting point is the relationship between internal shear and load
distribution:
dS
= −f (x, t) (transverse loading) (6.15)
dx
For unit displacement of the nth bending mode, the appropriate loadings are the
inertia loads fn (x, t) generated by the modal displacements. These loadings are in
turn equilibrated by the internal bending stresses:
fn (x, t) = −mz̈n = ωn2 [m(x)γn (x)]qn (t) (6.16)
Then the required shear coefficient can be obtained by an integration of the

spanwise derivative (as given earlier) to give the following integral representation:
L
Sn (x) = ωn2 m(x1 )γn (x1 ) dx1 (6.17)
x
Similarly, because dM/dx = −S(x), then

L L
Mbn (x) = ωn2 m(x2 )γn (x2 ) dx2 (6.18)
x x1
where, for this example, x = L/2.

b) Lumped-mass representation: This method of representing the modal shear
and moment coefficients consists of simply using the shear and moment distribu-
tions that fall out from the use of the Holzer–Myklestad technique for obtaining
the modal solution. Recall that the solution obtained from this approach gives, for
each spanwise station, a solution vector (eigenvector) of the following form:
{· · · }Tn ≡ {z, ψ, S, M}Tn (6.19)
“chapter6(new)” — 2005/10/18 — page 193 — #5

By this approach the modal shear and moments are thus directly available.
4) Loads by the force integration method are discussed here: For the given
example beam the force integration method can be implemented to predict the
internal shear by integrating the total loading, both the external loading F(x, t), as
well as the motion-induced loading, that is, the loading produced by the beam in
responding to the external loading:
L
L
S ,t = [F(x, t) − mz̈] dx
2 L/2
L ∞
L
= F(x, t) dx − Gn q̈n (t) (6.20)
l/2 2
n=1
pseudostatic shear
where the Gn modal coefficients are related to the already defined Mbn modal shear
coefficients:
L
Gn (x) = m(x1 )γn (x1 ) dx1 = Sn (x)/ωn2 (6.21)
x
This representation of the internal shear thus consists of two distinct parts: 1) the
part that would be calculated if the external applied loading F(x, t) were applied
as a static loading and the beam were not allowed to respond structurally; and
2) a part that is caused by the inertial loading induced in the beam by virtue of
its dynamic response to the applied loading. It is for this reason that the force-
integration method is often referred to as the mode-acceleration method. In a like
manner the force-integration method can be used to calculate the bending moment
in the beam:
L
L L
Mb ,t = x− [F(x, t) − mz̈] dx
2 L/2 2
L L
∞
L
= x− F(x, t) dx − Hn q̈n (t) (6.22)
L/2 2 2
n=1
pseudostatic moment
where the Hn coefficients are defined in a manner similar to that for the Gn
coefficients:
L
Hn (x) = (x1 − x) m(x1 )γn (x1 ) dx1 = Mbn (x)/ωn2 (6.23)
x
“chapter6(new)” — 2005/10/18 — page 194 — #6

The following observations can be made:

1) The pseudostatic shears and moments are the shears and moments that would
result if the beam were infinitely rigid.
2) The terms involving the summations over the modes represent “corrections”
to the infinitely rigid results arising from beam flexibility.
3) For infinite summations the two methods would give identical results. How-
ever, for finite summations the results will always be different (to some degree),
but would approach each other as the number of modes is increased.
4) The relations given earlier, ωn2 Gn = Sn and ωn2 Hn = Mbn , are not generally
true for the rotating beam case.
Example 6.3
The following classic example taken from Bisplinghoff et al. concerns the wing-
root bending moments and shears of an idealized fuselage-wing combination.
Consider an unrestrained uniform wing attached to a central fuselage mass with a
disturbance force loading having the form of a half sine wave. For this simplified
dynamic structure we require the transient distributions of the maximum shear and
bending moment in the unrestrained wing (see Fig. 6.1).
Solution
1) Time-history response is as follows: The vertical response of the slender
beam can be represented by a superposition of natural mode shapes of the freely
Fig. 6.1 Uniform wing attached to fuselage mass subject to half sine wave external
force distribution (after Bisplinghoff et al.).
“chapter6(new)” — 2005/10/18 — page 195 — #7

vibrating beam as follows:

∞

z(ξ , t) = q0 (t) + φi (ξ )qi (t) (6.24)
i=1

rigid elastic
translation deformation
The excitation is taken to be a half sine wave whose describing frequency is

taken as a parameter:

sin t; 0 ≤ t ≤ π/
f (t) = (6.25)
0; t > π/
The equations of motion for the generalized coordinates q0 and qi are as follows:
M q̈0 (t) = 0 (t) (6.26)

M(q̈i + ωi2 qi ) = i (t) (6.27)
where the mode shapes φi (ξ ) are so normalized that
mφi2 ( y) dy = M = MF + m ; i = 0, 1, 2, . . . , n (6.28)
0
Furthermore, the generalized excitations can be written in terms of the time

function f (t):
i = Ci f (t); C0 = ; Ci = φi dy, i = 1, 2, . . . (6.29)

0
(See Bisplinghoff et al. for detailed development and formulas for the evaluations
of Ci .) The time-history solution for w(ξ , t) then becomes the problem of solving
for the generalized coordinates qi (t):

 1

 t − sin t ; 0 ≤ t ≤ π/
M
q0 (t) = (6.30)

 t
 ; t > π/
M



Ci
(sin t −

0 ≤ t ≤ π/

 sin ωi t);

 M(ωi − )
2 2 ωi


 Ci
qi (t) = sin t + sin (t − π/2) (6.31)

 M(ωi2 − 2 )






 − [sin ω i t + sin ω i (t − π/2)] ; t > π/
ωi
“chapter6(new)” — 2005/10/18 — page 196 — #8

2) Calculation of the dynamic loads is as follows: From the time-history

responses obtained earlier, the internal loads in the wing can then be calculated by
each of the two methods:
a) Mode-displacement method: As with the examples given earlier, the shear
and bending moment distributions can be represented, respectively, by summations
of (spanwise-dependent) shear and moment coefficients multiplying the (time-
dependent) modal displacements:
Shear:
∞

S(ξ , t) = [S (i) (ξ )]qi (t) (6.32)
i=1
Moment:
∞

(i)
Mb (ξ , t) = Mb (ξ ) qi (t) (6.33)
i=1
where
1
S (i) (ξ ) = m ωi2 φ(λ) dλ (6.34)
ξ
1
Mb(i) (ξ ) = S (i) (λ) dλ (6.35)
ξ
and where i = 1, 2, 3, . . . , n (see Fig. 6.2).

b) Force-integration method: The force-integration method can be applied as
given earlier to include the pseudostatic and motion-induced (mode acceleration)
parts explicitly:
Shear:
∞ (i)

S (ξ )
S(ξ , t) = Spseudo − q̈i (t) (6.36)
static i=1
ωi2
Moment:

∞ M (i) (ξ )
b
Mb (ξ , t) = (Mb )pseudo − q̈i (t) (6.37)
static i=1
ωi2
where
Spseudo = (M − m )(1 − ξ )q̈0 (t) (6.38)
static
(Mb )pseudo = ( /2)(M − m )(1 − ξ )2 q̈0 (t) (6.39)
static
Note that the pseudostatic shear and moment Spseudostatic and (Mb )pseudostatic ,
respectively, are obtained by treating the beam (wing) as a rigid body.
“chapter6(new)” — 2005/10/18 — page 197 — #9

Fig. 6.2 Shear and bending moment distributions for displacements of the vibratory
modes (R = 1) (after Bisplinghoff et al.).
3) Numerical calculations and comparison of results follow: Results for the

shear and moment coefficients for the first five symmetrical modes are given in
Fig. 6.2 for the case where the fuselage and wing masses are equal (R = MF /m =
1). The results of calculations of maximum shear and maximum bending moment
along the span by the two methods are shown in Figs. 6.3 and 6.4. These calcula-
tions were made for the case of TI /T1 = 0.5, where TI is the time to peak value
of the forcing function and T1 is the period of the fundamental natural mode.
“chapter6(new)” — 2005/10/18 — page 198 — #10

Fig. 6.3 Maximum shear distribution (TI /T1 = 0.5, R = 1) (after Bisplinghoff et al.).
Figure 6.3 shows the spanwise variation of maximum shear and the instants
of time that the maximum shears occur, as determined by the two methods. It is
apparent that the mode-acceleration (force-integration) method gives much bet-
ter convergence than the mode-displacement method, particularly in the vicinity
of the root, where the higher modes contribute quite substantially to the shear
when the mode-displacement method is used. In fact, it can be seen that, even
with five modes, satisfactory convergence of the shear is not obtained by the
“chapter6(new)” — 2005/10/18 — page 199 — #11

Fig. 6.4 Maximum bending moment distribution (TI /T1 = 0.5, R = 1) (after
Bisplinghoff et al.).
mode-displacement method. Figure 6.4 shows the spanwise variation of maximum

bending moment and the instants of time when the maximum bending moments
occur, again as calculated using the two methods. The significantly more rapid
convergence of the force-integration method is again readily apparent.
The following general comments on the two methods can be made:
1) In the limit (n → ∞) the two methods are equivalent.
2) The mode-displacement method is mathematically “cleaner” and will give
reasonably accurate answers if three or more modes are involved in the response
calculation (as in a system of discrete masses and springs).
“chapter6(new)” — 2005/10/18 — page 200 — #12

3) The force-integration method is mathematically more cumbersome, but

gives better convergence characteristics (i.e., the use of fewer modes). Further-
more, the pseudostatic loads, as tools for understanding the total loads, are directly
available.
4) In general, for structural loads analysis, when accuracy is important, the
force-integration method would be preferable.
6.1.2 Application to Rotating Beams

In the preceding sections the two basic alternate methods of calculating the
dynamic loads in an elastic structure (mode displacement and force integration)
were developed. In this section these methods are extended to cover the case of a
beam vibrating in a centrifugal force field (i.e., a rotor blade). For the subassemblies
on a rotorcraft not operating in such a centrifugal force field (such as the basic
airframe and horizontal lifting surfaces), the previously developed tools should
suffice. As with the development in the preceding section, consider descriptions of
the flapwise (out-of-plane) bending moment and shear at a spanwise station on a
rotor blade using the two basic methods (i.e., the mode-displacement method and
the force-integration method).
Loads by the mode-displacement method
1) Bending moment Mb :
∞

Mb (x, t) = EIz = EI(x)γn (x)qn (t) (6.40)
n=1
2) Shear S:
∞

S(x, t) = −[EIz ] = − [EI(x)γn (x)] qn (t) (6.41)
n=1
The only new element here is the problem of finding appropriate expressions for
the load coefficients for rotating beams when the second- and third-order spanwise
derivative forms of the coefficients are not used in the evaluations. Consider the
application of the semianalytical (integral) method (as given earlier) for evaluating
the coefficients. In particular, consider the inertial loads acting on the beam when
it is deflected in one of its flatwise natural modes, as depicted in the free-body
diagram given in Fig. 6.5. Upon equilibrating the inboard internal forces and
moments with those arising from the loads acting on the beam over its outboard
portion, one can then write the following expressions:
L
Mbn (x) = ωn2 (x1 − x)m(x1 )γn (x1 ) dx1
x
L
− 2 [γn (x1 ) − γn (x)]m(x1 )x1 dx1 (6.42)
x
“chapter6(new)” — 2005/10/18 — page 201 — #13

Fig. 6.5 Free-body diagram of an outboard portion of a modally deflected rotating

beam.
Thus, the modal bending moment coefficients can be written in the following
abbreviated form:
Mbn (x) = ωn2 Hn (x) − 2 H̄n (x) (6.43)
where H̄n (x) is now the contribution caused by centrifugal forces. In a similar
manner the modal shear coefficients can be written as follows:
Sn (x) = −[Mbn (x)]

L L
= ωn2 m(x1 )γn (x1 ) dx1 − 2 γn (x) m(x1 )x1 dx1 (6.44)
x x
Or, again, in a more compact, abbreviated form
Sn (x) = ωn2 Gn (x) − 2 Ḡn (x) (6.45)
The Gn (x) and Hn (x) coefficients are as defined in the preceding section, and the
new Ḡn (x) and H̄n (x) coefficients are summarized as follows:
L γn (x)
Ḡn (x) = γn (x) m(x1 )x1 dx1 = T (x) (6.46)
x 2
tension
L L
H̄n (x) = m(x1 )γn (x1 )x1 dx1 − γn (x) m(x1 )x1 dx1 (6.47)
x x

= T (x)/2
As in the case without rotation, the results of a Myklestad technique for obtaining
the mode shapes can also be used to give values for Sn (x) and Mbn (x) directly.
Loads by the force-integration method. The application of the force-

integration method to the rotating beam (rotor blade) follows the same general tack
“chapter6(new)” — 2005/10/18 — page 202 — #14

as that developed in the preceding sections. Again, let us consider the uncoupled
flapwise bending of the blade caused by out-of-plane loads:
L
Mb (x, t) = (x1 − x)[F(x1 , t) − mz̈1 ] dx1
x
L
− 2 (z1 − z)mx1 dx1 (6.48)
x
After breaking out the parts of the load distribution as a result of flapwise motion
and the part of the out-of-plane deflection caused by steady flapping (or, in the case
of hingeless rotor blades, built-in coning), the expression for out-of-plane bending
can be stated as follows:
L L
Mb (x, t) = (x1 − x)F(x1 , t) dx − β0 2 m x1 (x1 − x) dx1
x x

pseudostatic flapwise bending moment
∞

− [Hn (x)q̈n + 2 H̄n (x)qn ] (6.49)
n=1
Note that the bending moment relief caused by steady coning β0 becomes readily
apparent with this formulation. In a similar manner we can write a force-integration
expression for the flapwise shear:
L L
S(x, t) = [F(x1 , t) − mz̈1 ] dx1 − 2 z m x1 dx1 (6.50)
x x
And, again, after separating out the part of the loading caused by modal accele-
ration, the final requisite form of the flapwise shear can be written as
L L
S(x, t) = F(x1 , t) dx1 − β0 2 mx1 dx1
x x

pseudostatic flapwise shear
∞

− [Gn (x)q̈n + 2 Ḡn (x)qn ] (6.51)
n=1
The following points should be noted:

1) Pseudostatic loads for the rotating blade now contain “relief” effects caused
by coning (either built-in or steady equilibrium).
2) The force-integration method now involves mode displacements as well as
modal acceleration terms.
3) Although not substantiated herein, general comparisons between the two
methods regarding convergence are unaltered as a result of the inclusion of rotating
effects.
“chapter6(new)” — 2005/10/18 — page 203 — #15

We now complete this section with a consideration of the two loads analysis
methods as applied to the in-plane bending moment and shear of a rotating beam
(i.e., rotor blade).
Mode-displacement method. Semianalytical (integral) representations of the

load coefficients for the in-plane bending case can be written using the results of
the preceding section by noting that the principal difference between these two
cases is the presence of an additional centrifugal loading term in the transverse
(modal) loading expression:
fyn (x, t) = mγn (x)(ωn2 + 2 )qn (t) (6.52)

where it is to be noted that in this instance γn (x) is now the mode shape appro-
priate to in-plane motion. By inspection, the previously given modification to the
transverse loading then leads to expressions for the modal bending moment and
shear that can be respectively written as
Mbn (x) = ωn2 Hn (x) + 2 [Hn (x) − H̄n (x)] (6.53)

and
Sn (x) = ωn2 Gn (x) + 2 [Gn (x) − Ḡn (x)] (6.54)
Force-integration method. Again, by inspection and comparison with the

results for the case of flapwise bending, the force-integration formulation for
in-plane bending can be written as
L
Mb (x, t) = (x1 − x)[F(x1 , t) − m( ÿ1 + 2 y1 )] dx1
x
L
− 2 ( y1 − y)mx1 dx1 (6.55)
x
or
L L
Mb (x, t) = (x1 − x)F(x1 , t) dx − e0 2 m(x1 − x) dx1
x x

pseudostatic in-plane bending moment
∞

− {Hn (x)q̈n − 2 [Hn (x) − H̄n (x)]qn } (6.56)
n=1
Note that here the bending moment relief comes from e0 , the equivalent of a
steady-state lead-lag angle. In a similar manner we can write a force-integration
expression for the in-plane shear:
L L
S(x, t) = [F(x1 , t) − m( ÿ1 + 2 y1 )] dx1 − 2 y mx1 dx1 (6.57)
x x
“chapter6(new)” — 2005/10/18 — page 204 — #16

And, again, after separating out the part of the loading as a result of modal
acceleration, the final requisite form of the in-plane shear can be written as
L L
S(x, t) = F(x1 , t) dx1 − e0 2 m dx1
x x

pseudostatic in-plane shear
∞

− {Gn (x)q̈n − 2 [Gn (x) − Ḡn (x)]qn } (6.58)
n=1
The in-plane offset distance e0 , which provides in-plane load relief, is defined in
Fig. 6.6. Note that, in the case of rotor blades with in-plane articulation, the offset
distance e0 is simply related to the hinge offset distance e and the equilibrium lag
angle ζ0 :
e0 = e sin ζ0 (6.59)
In all of the preceding development, the portions of the transverse load distribution
caused by the inertial acceleration and the centrifugal force loadings were explicitly
broken out in order to implement the two basic types of internal load calculation.
In the next section the whole issue of loads caused by motion is more generally
developed.
6.1.3 Motion-Induced Loads

Inherent in the two methods of dynamic loads analysis is the prior require-
ment of a calculation of the actual dynamic responses of the elastic structure.
Thus far, the dynamic responses considered herein have generally consisted of
representations of the responses in terms of natural modes. The solutions for
the modal responses are typically achieved by solving single-degree-of-freedom
equations for the forced responses for each of the natural modes. These steps are
summarized as follows:
1) Let the transverse deflection of a blade element be expressed as

N
z(x, t) = γn (x)qn (t); N →∞ (6.60)
n=1
Fig. 6.6 In-plane offset distance providing centrifugal load relief.
“chapter6(new)” — 2005/10/18 — page 205 — #17

2) The resulting differential equations become

Mn (q̈n + ωn2 qn ) = n (t) (6.61)
where
Mn = mγn2 (x) dx (6.62)

structure
is the generalized mass for the nth mode, and
n (t) = mγn (x)F(x, t) dx (6.63)

structure
is the generalized excitation for the nth mode. The expression F(x, t) is the external
applied loading.
In all of the development heretofore, it has been assumed that the external
applied loading F(x, t) is a known explicit function of time. However, usually this
applied loading is also a function of the response(s) of the structure itself. The
problem is further complicated in aircraft structures in that this loading usually
arises from rate-dependent phenomena, which introduce significant complexities
to the dynamic description. Typically, motions of the aerodynamic surfaces create
additional pressure loadings, which abound with rate-dependent terms. For this
case the applied loading considered earlier F(x, t) is conceptually broken up into
two parts: a part dependent only on the external disturbances F D (x, t), and a part
dependent on the motion of the structure F M (z, ż, z̈, t). If natural modes are again
used to describe the structure and if orthogonality is used to uncouple the mass
and stiffness terms, the resulting equation then becomes
Mn (q̈n + ωn2 qn )
= M
n (q1 , q2 , . . . , qN , q̇1 , q̇2 , . . . , q̇N , q̈1 , . . . , q̈N , t) + n (t)
D
(6.64)
The principal observation to be made is that the M n terms now couple the
(previously) uncoupled natural modal equations. The system can no longer be
represented by a series of single-degree-of-freedom forced equations. Assuming
for simplicity that M n are functions of only the generalized velocities q̇i , then the
system of equations is expressible in the following matrix format:
     
M1 0   q̈1 
 ξ11 ξ12 · · · ξ1N   q̇1 

  
 
   
 

 M2   q̈2   ξ21 ξ22 · · · ξ2N   q̇2 
  . + . 
 ..  .   . .. ..   .. 
 .   .   . . ··· .  . 

    
 
0 MN q̈N ξN1 ξN2 · · · ξNN q̇N
    D 
M1 ω12 0 
 q1 
 
 1 (t)
 M ω2  
 
 D  
 2 2   q2   2 (t)
+ ..
 . =
  .   ..  (6.65)
 .  . 
   . 


     D  
0 MN ωN2 qN N (t)
“chapter6(new)” — 2005/10/18 — page 206 — #18

Before we discuss ways to solve such a matrix equation, let us use the preceding
ideas on some examples.
Example 6.4
Consider the two-dimensional airfoil section of an earlier homework problem that
is characterized by the presence of inertia coupling. Let the airfoil now experience
a freestream velocity of air of 100 mph. Find the equations of motion describing
the response of the airfoil section to a step gust velocity wG of 10 ft/s.
Solution
The differential equations of motion, assuming generalized forces, are given by
mz̈ + kz z − mxI θ̈ = fz(a) (6.66)

(a)
−mxI z̈ + (I + mxI2 )θ̈ + kθ θ = fθ (6.67)
(a) (a)
where the generalized forces fz and fθ are now assumed to be the result of
a disturbance gust penetration and motion dependency. Furthermore, let these
loads be defined using a simplified quasi-steady aerodynamic formulation wherein
the lift acts at a chordwise point xa in front of the elastic axis. (Note that this
quasi-steady formulation is deemed simplified because we are neglecting, for illus-
trative purposes only, the effects of pitch rate θ.) Then, with this simplification in
mind, we obtain the required generalized forces:

1 2 1 2 dcI 1
fz(a) = − ρV ccI = − ρV c θ + (wG + ż) (6.68)
2 2 dα V

(a) 1 1 dcI 1
fθ = ρV 2 ccI xa = ρV 2 cxa θ + (wG + ż) (6.69)
2 2 dα V
where wG is the disturbance vertical gust velocity.

Upon setting 1/2ρV 2 c(dcI /dα) = Ka , we can rewrite the differential equations
in the following form:

m −mxIz̈ Ka 1 0 ż kz 0 z
+ +
−mxI (I + mxI2 ) θ̈ V −xa 0 θ̇ 0 kθ θ

0 1 z Ka wG −1
+ Ka = (6.70)
0 −xa θ V xa
A solution for the time-history response of this system can be obtained using a
variant of the Galerkin technique wherein the physical variables z and θ are first
represented by a combination of the two natural modes as obtained from the zero
“chapter6(new)” — 2005/10/18 — page 207 — #19

air density eigenvalue solution (Ka = 0). Thus,

z 1 1 q1 1 1 q1
= = (6.71)
θ θ̃1 θ̃2 q2 −3.68 0.0218 q2

= []
Then, the preceding coordinate transformation is substituted into the system

equations of motion, and the resulting (matrix) equation set is premultiplied by
the transpose of the matrix. It can be readily appreciated that this invocation
of the Galerkin technique is consistent with and in fact utilizes the orthogonality
principle: The elastomechanical terms in the dynamic equations are then decou-
pled and appear only as diagonal terms. The resulting equation set then becomes

M1 0 q̈1 c11 c12 q̇1 k11 k12 q1
+ +
0 M2 q̈2 c21 c22 q̇2 k21 k22 q2

Ka wG f1
= (6.72)
V f2
It can then be verified that the various coefficients in this equation are given by
M1 = m + 7.36 mxI + 13.5(I + mxI2 ) (6.73a)

M2 = m + 0.0436 mxI + 0.000475(I + mxI ) (6.73b)
c11 = c12 = Ka (1 + 3.68xa )/V (6.73c)
c21 = c22 = Ka (1 − 0.0218xa )/V (6.73d)
k11 = ω12 M1 − Ka (3.68 + 13.5xa ) (6.73e)
k12 = Ka (0.0218 + 0.0802xa ) (6.73f)
k21 = Ka (−3.68 + 0.0802xa ) (6.73g)
k22 = ω22 M2 + Ka (0.0218 − 0.000475xa ) (6.73h)
f1 = −(1 + 3.68xa ); f2 = (−1 + 0.0218xa ) (6.73i)
The preceding transformed (modal variable) equation set is thus seen to be

similar in mathematical form to the original physical variable equation set, and
indeed, Laplace transform and Heaviside expansion techniques could be used to
obtain exact solutions for the responses (either z and θ , or q1 and q2 ). One primary
advantage of using the decoupling provided by the natural modal variables and
the use of the Galerkin technique is that it can reduce the number of degrees of
freedom that need to be carried in the problem description. However, in this simple
example such an advantage does not arise because we started off with such a small
number of physical degrees of freedom. In this case the advantage of the natural
mode/Galerkin approach lies rather in the ability to form reasonable approxi-
mations to the responses by neglecting the off-diagonal terms in the resulting C
“chapter6(new)” — 2005/10/18 — page 208 — #20

and K matrices and thereby to approximate a set of uncoupled single-degree-of-

freedom equations:
M1 q̈1 + c11 q̇1 + k11 q1 = (Ka /V )wG f1 (6.74)
M2 q̈2 + c22 q̇2 + k22 q2 = (Ka /V )wG f2 (6.75)
whose solutions are then given by
  
1   ζi 
qi (t) = 1 − exp(−ζi ωi t)  $ sin ωi t + cos ωi t  (6.76)
c11 1−ζ 2
i
where
$
1 (
ζi = cii Mi kii ; ωi = kii /Mi 1 − ζi2 ; i = 1, 2 (6.77)
2
In this example the calculation of the internal loads could be accomplished
using either of the two basic methods. The mode-deflection method would
have one advantage in the simplicity afforded by only calculating the modal
responses and then simply multiplying them by appropriate constants. The force-
integration method in this case (and generally for most aeroelastic problems)
requires the consistent calculation of all of the external loadings—both inertial and
aerodynamic—for an accurate solution. In this example this requirement involves
the use of the time histories of the modal displacements and velocities, as well as
the accelerations.
Example 6.5
Examine the in-plane response and resulting in-plane bending characteristics of
an articulated rotor with a lead-lag damper (see Fig. 6.7).
This example demonstrates how motion-dependent loads can enter the dynamic
description through the boundary conditions, in this case as a result of the loads
Fig. 6.7 Pictorial representation of the in-plane loading and root restraint of an
articulated rotor blade.
“chapter6(new)” — 2005/10/18 — page 209 — #21

imparted by the lead-lag damper. To this end, we denote the in-plane elastic
deformation by the variable v(x, t), and make the following simplifying assump-
tions:
1) The in-plane loads result solely from inertia considerations. The airloads,
represented by fy (x, t), do not depend on the in-plane responses, that is, the
transverse load distribution is given by
fy (x, t) = fyD − m[v̈ − 2 v] (6.78)
2) At the blade root the bending moment is related to the time derivative of the
slope and the damper rate of the lead-lag damper:
∂ 2v ∂ 2v
EI 2
= −Kζ (6.79)
∂x ∂x∂t
Solution
The basic differential equation of motion is that of the rotating beam for
in-plane motion (as given in Chapter 3) with the addition of the disturbance
transverse loading fyD :
[EIv ] − [v T ] = fyD − m[v̈ − 2 v] (6.80)
Assume a modal solution:

J
v(x, t) = γj (x) qj (t) (6.81)
j=0
The appropriate mode shapes are those for the “pinned” boundary condition at
the root. Note that, in this case the index j = 0 corresponds to the rigid-body
rotational mode [γ0 (x) = x] about the hinge point. We then invoke a standard use
of the Galerkin technique wherein we multiply the equation by the kth mode shape
and integrate over the blade. Consider the integration of the elastic stiffness term,
denoted for this purpose by Ksi , using integration by parts:
L
Ksi = γi (x) [EIv ] dx (6.82)
0
Two applications of integration by parts yields

L L
Ksi = − γi (x) [EIv ] + γi (x) [EIv ] dx
0
x=0

J
J
= γi (0)Kζ γi (0)q̇j (t) = bij q̇j (t) (6.83)
j=1 j=0
“chapter6(new)” — 2005/10/18 — page 210 — #22

where
bij = Kζ γi (0)γj (0) (6.84)
Thus, once again, the motion-induced loads at the root caused by the lead-lag
damper result in a coupling between the natural vibrations modes. The resulting
(matrix) differential equation is again of the form described earlier:
[M]{q̈} + [C]{q̇} + [ω2 M]{q} = {(t)} (6.85)
A solution can then be obtained for the response vector {q(t)} by either a rigorous
solution with the modal variables qi (t) being coupled by the fully populated C
matrix, or by an approximate one wherein the off-diagonal C matrix coupling terms
are neglected (bij = 0, where i = j), and then qi (t) are solved for independently.
Once the responses are obtained the loads in the blade can then be solved for
using either of the two basic load calculation methods. However, one problem
with the use of the mode deflection method is that, because pinned root boundary
condition modes were assumed, there is no mechanism present that can give a
nonzero bending moment at the blade root, consistent with the bending moment
imparted by the lead-lag damper. However, this difficulty is overcome with the
force-integration method.
One note of caution with the force-integration method follows: Because the
method implies a degree of balancing of the loads, the statement of the loads
calculation must be consistent with that for the response calculation. For instance,
if an approximate solution scheme is used, then the force-integration method will
give completely erroneous results. However, the mode-deflection method is not
typically prone to this type of inaccuracy.
6.1.4 Dynamic Magnification Factors

The two basic methods of load analysis discussed in the preceding subsections,
mode-deflection and force-integration methods, respectively, have two important
characteristics: they are both general and are hence adaptable to a wide variety
of applications, and they are both capable of yielding any degree of accuracy
desired (i.e., if sufficient computational effort is expended). However, they both
have the disadvantage of requiring considerable mathematical manipulation, which
is potentially expensive as well as time consuming. Consequently, a “ballpark
estimate” type of method is needed to complement the more exact methods and
to provide quick estimate answers that do not require the use of computers. The
use of dynamic magnification factors fulfills this need. The dynamic magnification
factor (DMF) is generally defined as follows:
(dynamic displacement)max
DMF = (6.86)
(static displacement)max
Thus, if the pseudostatic loads are known for a structure, the DMF gives an estimate
of the dynamic loads:
(dynamic load)max ≡ DMF · (pseudostatic load)max (6.87)
“chapter6(new)” — 2005/10/18 — page 211 — #23

The DMF is, in general, defined for only single-degree-of-freedom systems for a
variety of “standard” forcing functions. Some of these standard forcing functions
are considered in this section.
1) Step function (undamped spring-mass system):
Basic equation:
ẍ + ωn2 x = ωn2 1(t) (6.88)

unit step
function
The solution for the response (subject to zero initial conditions) is given simply as
x = 1 − cos ωn t (6.89)
which is graphed in Fig. 6.8. For this case the DMF is clearly equal to 2.
2) Step function (damped spring-mass system):
Basic equation:
ẍ + 2ζ ωn ẋ + ωn2 x = ωn2 1(t) (6.90)
whose solution is given by

) *
ζ
x(t) = 1 − exp(−ζ ωn t) sin ωD t + cos ωD t (6.91)
1 − ζ2
where the damped natural frequency ωD is given by

ωD = (1 − ζ 2 ) ωn (6.92)
Fig. 6.8 Response of an undamped spring-mass system to a unit step input.
“chapter6(new)” — 2005/10/18 — page 212 — #24

Fig. 6.9 Response of a damped spring-mass system to a unit step input.
In a similar manner the solution for this case is graphed in Fig. 6.9. For this case
the DMF is given by

DMF = 1 + exp(−π ζ / 1 − ζ 2 ) (6.93)
3) Sinusoidal function (damped modal response): Let us now focus our

considerations on the simple damped spring-mass system, in the form of the single-
degree-of-freedom representative of the kth modal response of a rotor blade. Let
the rotor blade be excited by the nth harmonic of (aerodynamic) loading based on
the fundamental frequency :
q̈kn + σk q̇kn + νk2 2 qkn = Ckn eint (6.94)
where νk = ωk ≡ kth modal (undamped) natural frequency, and σk is the kth

modal damping coefficient (resulting principally from aerodynamic loadings).
Assuming sinusoidal motion, we can write the (steady-state) solution as
qkn = qkn exp[i(nt + φkn )] (6.95)
where, by providing for a phase angle φkn , we can select qkn to be a real-valued
quantity. With this solution form the amplitude can be written as
Ckn
qkn = (6.96)
{[1 − (n/νk )2 ]2 + (nσk / νk )2 }1/2
The pseudostatic response is obtained by neglecting the acceleration and velocity
terms:
νk2 2 [qkn ]p−s = Ckn eint (6.97)
The DMF is then obtained by ratioing these two quantities:

1
DMFkn = (6.98)
{[1 − (n/νk )2 ]2 + [nσk / νk ]2 }1/2
“chapter6(new)” — 2005/10/18 — page 213 — #25

6.1.5 Approximate Blade Loads Using Dynamic

Magnification Factors
The concepts already developed can be applied to the rotor blade to form rudi-
mentary approximations to the blade bending moments and shears subject to the
following assumptions:
1) The hub does not translate or rotate. This forms the so-called infinitely rigid
hub approximation, which denies the very real effects of rotor-fuselage coupling.
2) Intermodal coupling caused by aerodynamic or other effects is negligible,
as per the decoupling approximation given in the preceding section.
3) Interharmonic coupling caused by forward-flight aerodynamic loadings is
negligible.
4) Pseudostatic bending moment and shear distributions have been calculated
a priori (or are available from some other source).
Based on these (actually somewhat shaky) assumptions, the maximum values
of the bending moments and shears can be expressed as follows:

N
K
Mb (x) Mb (x)
≤ (DMF)kn (6.99)
S(x) max S(x) max (pseudostatic)
n=1 k=1

harmonics ↑ ↑ modes
where (DMF)kn is as given earlier. With this formulation the following should be
noted:
1) (DMF)kn will be approximately unity (1.0) for modes whose natural frequ-
encies (vk ) are well above the harmonics of excitation and will be ≈0 for
modes whose natural frequencies are well below these harmonics. This follows
directly from the well-established frequency-response characteristics of a single-
degree-of-freedom dynamic system.
2) The formula tends to be conservative because it assumes that the maximum
displacements of each mode occur simultaneously.
These concepts can then be illustrated by considering the flapping excitation
of a rotor blade. Figure 6.10 shows the velocity components contributing to the
airloading at a typical blade section (whose flapping motion ż is again defined in
the downward direction).
Using a simple two-dimensional strip theory formulation of the aerodynamic
loads on this section, an approximate loading distribution can be written. Of prime
importance is the role of the induced velocity vi appearing in this formulation.
Induced velocity. The induced velocity is the component of the air velocity
impinging on the airfoil blade section that arises from the presence of the lifting
rotor itself. Generally, the induced velocity, as encountered by the lifting air-
foil section, is highly variable both in time and space (i.e., spanwise position).
The zero-frequency component of the induced velocity can be attributed to the
response of the air mass to the action of the rotor in producing a thrust. In a
similar manner time-variable components of the induced velocity arise from the
harmonic responses of the blade sections. These responses cause motion-induced
“chapter6(new)” — 2005/10/18 — page 214 — #26

Fig. 6.10 Velocity diagram for airloads at a typical rotor blade section.
harmonic loadings on the blade and must therefore also be accompanied by har-
monic changes of momentum in the air mass. At this point, such a description
is a necessarily overly simplified explanation of a complicated phenomenon that
encompasses a variety of other important fluid dynamic effects. But it is sufficient
to establish the case that the induced velocity vi can be considered to be comprised
of a steady part vi(0) and a multiharmonic part that is also a spanwise variable
(H)
vi . This latter component provides the major harmonic excitation of the blade
dynamics.
Upon taking the geometric blade pitch angle to be comprised of a steady part
θ0 and the (1P and higher) harmonic part θ1 , we can then write the (upwardly
directed) lifting airload distribution. [Here, we have assumed again that the effects
of pitch rate are negligible (for illustrative purposes) and that we have essentially
a hovering flight condition.] The resulting simplified expression is given by

1 ż − vi
fza (x, t) ≈ ρac(x) θ +
2
2 x
1
= ρac2 [x 2 θ0 − xRλ] ← (steady loads)
2
1 (H)
+ ρac[x 2 θ1 + xż − xvi ] (6.100)
2
where θ1 denotes harmonic feathering and perturbational torsion deflection; xż

denotes the (flapping) motion-induced component; and vi(H) denotes harmonic
inflow. Also, λ is the rotor inflow ratio, a nondimensionalization of the uniform
and constant part of the induced velocity plus any “ram” velocity of the rotor
through the air mass. Thus, for the kth flapping mode the more compact (modal)
form of the equation can be written as
Mk q̈k + vk2 2 Mk qk = M
k + k
D
(6.101)
“chapter6(new)” — 2005/10/18 — page 215 — #27

where
1
1
M
k = − ρacR4 γk (η)γj (η)η dη q̇j = −σkj q̇j (6.102)
2 0
1 1
k = − ρac R (A2k θ0 − A1λ ) − ρac R A2k θ1
D 2 4 2 4
2 2
1
1 (H)
− ρacR 4
γk (η)vi (η, t)η dη (6.103)
2 0
where, in the first part, A1k and A2k are integration constants involving the mode
shape γk and the nondimensional spanwise variable η(= x/R), as given in the
following:
1
A1k = γk (η)η dη (6.104a)
0
1
A2k = γk (η)η2 dη (6.104b)
0
Although the effect of steady coning β0 was not expressly included in the preceding
development (in order to focus on the effects of the aerodynamics), we can see that
the first part of D
k represents the steady blade loading that this coning relieves.
6.2 Harmonic Rotor Hub Loads

In the preceding section the basic concepts were developed for calculating
dynamic loads in a variety of structures with little attention to the specific form
of loading. In this section it is assumed that the structure is exclusively that of
a rotating rotor blade and the loading is the multiharmonic aerodynamic loading
typical of forward flight. From an analytic description of the harmonic inflow
and with the inclusion of the motion-induced aerodynamic loads, the dynamic
response of the elastic blade can be calculated using methods developed in earlier
sections. Using the concepts for dynamic loads of the preceding section, one can
then calculate the root shears and moments of a single blade and express them,
each in the form of a Fourier series.
6.2.1 Basic Reduction Formulation

Consider a rotor hub with an arbitrary number of blades b and harmonic root
shears and moments in both the in-plane and out-of-plane directions (see Fig. 6.11).
Root shears. On any single blade the shears at the root will be harmonic;
because the blades are assumed to be identical, they are all described by the same
Fourier series but with a different phase angle. Thus, for the jth blade the root
“chapter6(new)” — 2005/10/18 — page 216 — #28

Fig. 6.11 Definition of blade and hub shears and moments.
shears can be written as

∞

2π j 2π j
Syj = Sy0 + Sync cos n t + + Syns sin n t + (6.105)
b b
n=1
∞

2π j 2π j
Szj = Sz0 + Sznc cos n t + + Szns sin n t + (6.106)
b b
n=1
Integral order harmonics. Assuming constant periodicity, we can form the

Fourier coefficients of total hub shear by summing over the blades. Consider the
vertical Z component of the hub shear:

∞ 
b
2π j
Hz = bSz0 + Sznc cos n t +
 b
n=1 j=1

2π j
+ Szns sin n t + (6.107)
b
By means of trigonometric identities, one can write

+ ∞

Hz = b Sz0 + δn(kb) [Sznc cos nt + Szns sin nt] (6.108)
n=1
“chapter6(new)” — 2005/10/18 — page 217 — #29

where δij is the Kronecker delta:

1; i=j
δij ≡ (6.109)
0; i = j
In a similar fashion the longitudinal X and lateral Y components of the hub shears
can be formed:
+ ∞
Hx
b 2π j
= Sy0 + Sync cos n t +
Hy b
j=1 n=1

2π j − sin(t + 2π j/b)
+ Syns sin n t + × (6.110)
b cos(t + 2π j/b)
This can be simplified to
) ∞
Hx b −Sy1s Sy(n−1)s − Sy(n+1)s
= + δn(kb) cos nt
Hy 2 Sy1c Sy(n−1)c + Sy(n+1)c
n=1
*
−Sy(n−1)c + Sy(n+1)c
+ sin nt (6.111)
Sy(n−1)s + Sy(n+1)s
where, again, k = 0, 1, 2, . . . .
Root bending moments. Because root bending moments, like root shears,
are vector quantities, they are transferred through the hub to the fuselage by the
same trigonometric resolution equations as are the root shears. Thus, by noting the
similarity of the appropriate vectors, the nonrotating hub moments can be obtained
using the following substitutions:

Sy → −MbF 


Sz → MbI 


Hx → Mhx (6.112)


Hy → M hy 



Hz → −Q (negative of the rotor torque)
where the b and h subscripts refer to blade and hub, respectively, and where ( )F
and ( )I refer to flapwise and in plane, respectively.
The following summary can be made:
1) The hub acts as a filter for steady harmonic loads, that is, it admits only the
harmonics that are integral multiples of the number of blades b.
2) Although vertical admitted blade loads have the same harmonics as those of
the hub (k × b), the lateral admitted blade harmonics are +1 and −1 those of the
hub (again, k × b). Thus, the steady and kb harmonic vertical blade loads and the
first (kb + 1) and (kb − 1) harmonic in-plane blade loads are the only ones passed
(as zeroth and kb harmonic hub loads) by the hub.
“chapter6(new)” — 2005/10/18 — page 218 — #30

6.2.2 Nonintegral Order Harmonic Rotor Hub Loads

In the preceding section the assumption was made that the Fourier coefficients
describing the blade loads were constant, reflecting the nature of the aerody-
namic excitation: In the absence of hub motion and for aeroelastically stable
blades, the airloads can only assume integral values for the harmonics. How-
ever, although rotor blades are designed (or eventually made) to be stable, the
hub realistically undergoes motion that can and does couple with the aeroelastic
responses of the rotor blades. In the material given earlier, it can be seen that
the (n = kb) filtering action of the hub is caused entirely by the fact that the
blades are assumed 1) to be identical, 2) to be undergoing identical motion, and,
hence, 3) to have identical Fourier descriptions for their loads. However, the for-
mulation mathematically admits the case where these assumptions are met and
where the Fourier coefficients are themselves oscillatory (generally, at a nonin-
tegral harmonic frequency). This case realistically occurs when there is coupling
between the rotor and the fuselage at this nonintegral harmonic frequency and
when the rotor is undergoing some form of whirling motion (either progressive
or regressive) at one of the “admissible” harmonics (i.e., k = 0, 1, 2, . . .). Thus,
for example, assume that the hub is undergoing a nonintegral harmonic pitching
motion (as viewed in the nonrotating coordinate system) at some frequency ωF ,
given by
ωF = ωW + kb (6.113)
where ωW is arbitrarily defined to be the difference in frequency from the kb

value. For this motion to take place, the rotor would also have to be generating hub
pitching moments at this frequency as well. From the material of Sec. 4.4, this can
be seen to occur whenever the load distribution in the rotating coordinate system
has an oscillatory component at a frequency ωR (as seen in the rotating coordinate
system), where ωR is given by either of the following two values:
ωR = ωW + (kb ± 1) (6.114)
The simplest case would be where k = 0 and the rotor disk would appear to
be undergoing a whirling motion at a frequency of ωW , as is discussed in
Sec. 4.4. For the cases where k ≥ 1, the rotor would appear to be undergoing
a “traveling-wave” motion either in the progressive or regressive direction, or
a combination of the two directions. The oscillatory loads would correspond-
ingly have similar interpretations. These nonintegral harmonics, as was stated
earlier, are not generally excited or driven by the main rotor airloads because such
airloads are essentially integral harmonic in character. Such nonintegral loads
(and motions) must therefore come from some other source of fuselage excitation
and, in order to have significant response levels, must necessarily involve rotor
blade motions that are relatively devoid of (aerodynamic) damping. The follow-
ing section treats the area of non- (main) rotor sources of fuselage excitation;
the subject of what conditions lead to the rotor blades losing aerodynamic damp-
ing falls under the category of blade aeroelastic stability and is covered in later
chapters.
“chapter6(new)” — 2005/10/18 — page 219 — #31

6.3 Nonrotor Sources of Fuselage Excitation

Although the main source of vibration for the helicopter is the main rotor, as
defined by the shears and moments at the center of the main rotor hub, significant
other sources exist. These other sources can be classified principally with regard
to frequency content and source (i.e., aerodynamic vs nonaerodynamic). The most
pervasive sources of fuselage excitation are summarized in Fig. 6.12.
6.3.1 Integral Harmonic Sources

Excitations at multiples of the blade passage frequency kb essentially define
the principal integral harmonic vibration problem in rotorcraft fuselages. Further-
more, the rotor loads (shears and moments) at the main rotor hub are clearly the
greatest source of excitation at the blade passage frequency. However, other blade
passage frequency sources of excitation must be considered. As shown in Fig. 6.12,
the control rods take the blade feathering moments, which, like the other blade
loads, are multiharmonic. These loads are subject to the same type of trigonometric
resolution as are the hub loads; however, the point of load application is no longer
the center of the hub, but instead the swash-plate actuator attachment to fuselage
points.
Two other sources of integral harmonic excitation arise from the main rotor
and are of aerodynamic origin. The first is the harmonic pressure pulses, which
the blades impart to the top of the cockpit canopy because of their proximity of
passage. This is a relatively local fuselage excitation in that it will excite a part of the
fuselage that is not a principal structural element; such excitations will generally be
Fig. 6.12 Principal sources of fuselage excitation.
“chapter6(new)” — 2005/10/18 — page 220 — #32

sensed as local rattling of the canopy. This type of vibration is typically controlled
by providing adequate stiffness to the top of the canopy and sufficient mast height
to the rotor to give clearance.
The second source of integral harmonic excitation arising from the aerodyna-
mics of the main rotor is the system of trailing vortices constituting the far wake
of the rotor. These vortices impact on the two fuselage empennage aerodynamic
surfaces (i.e., the horizontal stabilizer and the vertical pylon fin). Such excitations
are quite dependent on the geometry and proximity of the wake structure relative
to these aerodynamic surfaces. As shown in Fig. 6.13, depending on the flight
condition, the trailing blade-tip vortex can pass within close proximity to the
horizontal stabilizer twice per rotor reduction before it is convected away from the
helicopter.
The analysis of the pressure excitations on the horizontal surface caused by the
near passage of the trailing vortices is a substantial exercise in fluid dynamics, and
the details of a solution are beyond the intent of this text. However, some of the
main characteristics of the solution can be identified. When vortices pass within
close proximity to a lifting surface such as a wing, nonpotential flow characteristics
of each element, such as the boundary layer of the wing and the finite core of the
vortex, can each interact with each other as well as with the potential flow charac-
teristics of each. Furthermore, a requirement of such a flow problem is knowledge
of the trajectories of the vortices, which constitutes a separate (substantial) poten-
tial flow problem in itself. Generally, this type of aerodynamic problem falls in
the category of wing-vortex interaction and is not yet fully solved. Because of the
ability of the trailing vortices to pass within close proximity to the surface twice per
Fig. 6.13 Division of the rotor wake into upper and lower parts, µ = 0.4, only tip
vortices shown for ψ = 30 deg (after Gangwani).
“chapter6(new)” — 2005/10/18 — page 221 — #33

rotor revolution and because of the highly impulsive nature of the pressure pulses,
these excitations are integral harmonic with substantial higher harmonic content
(i.e., ωE = kb, k = 1, 2, 3, . . .). The excitations caused by this main rotor wake
impingement on the empennage surfaces are aggravated by the susceptibility of the
fuselage vibrations to excitations at this relatively flexible portion of the fuselage.
6.3.2 Nonintegral Harmonic Sources

Nonintegral sources of fuselage excitation generally arise from non-(main) rotor
portions of the drive system and, in particular, the tail rotor and tail rotor drive shaft.
As shown in Fig. 6.12, the tail rotor, being a rotor, functions as an aerodynamic
source of excitations as does the main rotor. However, because of the gear ratio
of the tail rotor shaft to that of the main rotor, the tail rotor will not generally
operate at any whole number multiple of the main rotor speed much less at any
integral harmonic of the blade passage frequency b. Fortunately, however, the
frequencies associated with tail rotor excitation are considerably higher than the
principal vibration frequency (main rotor blade passage frequency) and constitutes
a source of structure-induced noise. Consideration of such sources of noise forms
the basis for the emerging technology of acoustoelasticity, which is beyond the
scope of this text.
6.3.3 Nonharmonic Sources

One source of vibration excitation potentially present with all aircraft is that
of buffeting, wherein a source of aerodynamic turbulent flow (e.g., separation
from some more forward portion of the fuselage) can impinge on any portion of
the empennage. A continuing trend with contemporary rotorcraft is an increased
horizontal stabilizer area. This increased tail area consequently increases the sus-
ceptibility of the empennage to any separation-induced turbulence or vorticity. The
nature of buffeting is that it is generally broadband; therefore, such vibrations are
not easy to isolate or to tune out, and the appropriate cure is to minimize the source
of the turbulence or vorticity by some sort of “aerodynamic” fine tuning.
6.4 Rotor-Fuselage Interactions

A straightforward approach to calculating the fuselage vibrations would be, first,
to calculate the harmonic loads that the rotor develops at the hub (in accordance
with Sec. 6.2) and, second, to apply them to the fuselage hub as the excitation
forces and then calculate the resulting responses of the elastic fuselage at selected
locations. The problems with this relatively simple procedure are twofold: First, as
stated, the task is still a difficult one. The harmonic loads are extremely complex
and consequently difficult to predict accurately, and the fuselage is a complex
structure and equally difficult to model with sufficient accuracy. Second, however
challenging a problem this procedure poses, it is still not sufficiently complete to
account for all of the elements dictated by the physics of the problem.
The missing element required for the successful vibration analysis of either
articulated or nonarticulated rotor helicopter designs is the proper description of
“chapter6(new)” — 2005/10/18 — page 222 — #34

the interactions occurring between the rotor and the fuselage. The importance of
this rotor-fuselage interaction is manifested in the fact that oscillatory loads being
transferred from the rotor to the fuselage via the hub can often be greatly amplified
from what they would be if the fuselage were considered to be infinitely rigid. And
yet, such “hub-fixed” harmonic rotor hub loads calculations have been accepted
methodology for many years with scant attention paid to the coupling mechanism.
Indeed, several important mechanical and aeromechanical instabilities result from
rotor-fuselage interactions including ground resonance and air resonance, wherein
the rotor can drive the fuselage in divergent oscillations. Furthermore, as could
be anticipated, the greater control power of hingeless and bearingless rotor con-
figurations tends to make interaction effects more pronounced than they would be
with an articulated rotor.
As developed in this section, the first major task in assembling the principal
elements needed for accurate fuselage vibration calculation is to gain accurate
knowledge of the vibration characteristics of the rotor and fuselage taken sepa-
rately. That is, we must know the natural frequencies and mode shapes of both the
rotor blades and the fuselage, as well as the (aerodynamic) forcing of the blades
and the fuselage. The second task is to gain commensurately accurate knowledge
of the dynamics between the rotor and the fuselage. In an interactional situation, by
definition, the fuselage and rotor are each developing and, to a degree, amplifying
oscillatory loads in the other at some common frequency, typically the blade pas-
sage frequency. Figure 6.14 depicts an idealization of the loading of the fuselage
by the rotor airloads while the rotor and fuselage are undergoing an interactional
coupling. In this depiction the applied (non-motion-induced) airloads acting on
the rotor blades are denoted { fapp1 } while the total integrated airloads acting on
Fig. 6.14 Schematic of the aerodynamically excited fuselage and the rotor-fuselage
interaction.
“chapter6(new)” — 2005/10/18 — page 223 — #35

Fig. 6.15 Conceptualization of the rotor-fuselage coupled dynamics.
the fuselage at the hub are denoted { fapp2 }. Note that in this figure the interaction
between the rotor and the fuselage is represented by the internal loads at the hub
{ fh } and the motions of the hub {qh }. The interactional process can then be further
conceptualized by the block diagram given in Fig. 6.15.
6.4.1 Impedances and Mobilities

General ideas. Two key features of the preceding representation of the
rotor-fuselage coupled dynamics are the two functional blocks representing the
fuselage-alone dynamics and the rotor-alone dynamics, respectively. The first
of these requires a functional description that relates the hub forces to the hub
responses via the fuselage dynamics, whereas the latter requires a functional
description that instead relates the hub forces to the hub responses. The former
is typically referred to as fuselage mobility and the latter is referred to as rotor
impedance. These quantities are multivariable and hence, are typically expressed
in matrix form. They can also be thought of as transfer functions. They are usually
defined for oscillatory motion, but the definition can be generalized to include
arbitrary motion as well.
The appropriate starting point for their descriptions is the set of matrix dynamic
equations used to describe the motion of the variables comprising these quantities.
Consider a general damped linear structure, which for oscillatory motion has the
following matrix equation of motion:

[K] − ω2 [M] + iω[C(ω)] {q} = { f } (6.115)
where [K], [M], and [C(ω)] are the N × N stiffness, mass, and damping matri-
ces, respectively. The matrix terms on the left-hand side of the equation define
a response-to-loads relationship and can be termed the displacement impedance
matrix [Z] or
[Z]{q} = { f } (6.116)
“chapter6(new)” — 2005/10/18 — page 224 — #36

On the other hand, the relationship of loads to dynamic responses can be

characterized by a similar relationship that defines the displacement mobility
matrix [Y ]:
{q} = [Y ]{ f } (6.117)
where [Y ] and [Z] are thus inverses of each other.
[Y ] = [Z]−1 ; [Z] = [Y ]−1 (6.118)
Note that the preceding expressions have meaning with systems wherein the
degrees of freedom {q} are explicit linear and rotational displacements, and the
forces { f } are the similarly defined forces and moments. Generally, [Y ] and [Z] are
complex valued and frequency dependent. If the response vector is composed of
displacements, velocities, or accelerations, then the transfer function (impedance
or mobility) is defined as the displacement, velocity, or acceleration (impedance or
mobility), respectively. Compliance, mobility, and inertance are sometimes used in
the literature to define the corresponding displacement, velocity, and acceleration
mobilities.
Application to rotor-fuselage coupling. These general concepts can then be

fruitfully applied to the problem of coupling the rotor-fuselage dynamics. Because
we are dealing with two dynamic subsystems that interact with each other, care must
be taken in the definitions of the interacting forces and moments as seen by each
of the respective subsystems. Note that mobilities and impedances are defined in
terms of loads and displacements that are taken to be positive in directions into each
of the structures at the interface. Furthermore, at the interface, the internal loads,
as seen by each of the respective substructures, are in directions that are opposite
to each other. Consequently, in order to be able to use consistent statements of
the component mobilities and impedances, as described earlier, the directions of
the load vectors, as seen by each substructure, must be defined with opposite
signs. Hence, in the preceding development { fH } will be defined by an additional
descriptor, a superscript indicating which subsystem is seeing the loads. Hence,
{ fH }R will be taken as the negative of { fH }F . These concepts can then be brought
together in the following manner:
Total hub loads (as seen by fuselage):
{ fH }F = { f M }F + { fapp2 } = −{ f M }R + { fapp2 }
= −[ZR ]{qH } + { fapp2 } (6.119)
Hub responses:
{qH } = [YH ]{ fH }F = −[YF ][ZR ]{qH } + [YF ]{ fapp2 } (6.120)
Then, upon eliminating {qH } and { fH }F from these expressions separately and in
turn, the following two sets of relationships are obtained:
Hub loads:
−1
{ fH }F = [I] + [ZR ][YF ] { fapp2 } (6.121)
“chapter6(new)” — 2005/10/18 — page 225 — #37

Hub responses:
−1
{qH } = [I] + [YF ][ZR ] [YF ]{ fapp2 } (6.122)
Thus, vibratory response of the fuselage can then be calculated once the three basic
quantities are known:
[YF ] ≡ fuselage mobility

[ZR] ≡ rotor impedance
{ fapp2 } ≡ integration of rotor loads { fapp1 } (includes effects of blade aero-
dynamics and aeroelasticity, with the assumption of a fixed hub)
Although this formulation is relatively straightforward, each of these three basic
quantities has important characteristics that must be addressed. The following
sections of this chapter address some of the more pertinent properties of each.
6.4.2 Characterization of Fuselage Dynamics

In practice, the calculation of the appropriate descriptor of the fuselage dynam-
ics, the fuselage mobility, must be calculated using a finite element modeling
(FEM) analysis with a frequency-response-function (FRF) calculation capability.
The displacement mobility matrix defined earlier is but one form of the FRF.
The high degree of complexity of the fuselage structuring precludes any form of
closed-form analytical solution for this quantity. Furthermore, typical helicopter
fuselage structures tend to have a variety of highly unsymmetrical (in a spatial
sense) vibrational modes with several natural frequencies below the blade passage
frequency of b (as defined in an earlier section). As we shall see subsequently, a
key to the successful calculation of fuselage mobility is the accurate solution for
the natural frequencies and mode shapes of the fuselage structure in the frequency
range straddling the blade passage frequency.
Presently, however, the application of state-of-the-art FEM modeling tech-
niques to this problem is not nearly accurate enough for successful vibration
calculations. Figure 6.16 typifies the caliber of correlation obtainable between
experimental frequency responses and those predicted using a typical state-of-the-
art FEM analysis. The results shown are for the responses at the aircraft caused
by excitations applied at a point on the tail in the vertical and lateral directions,
respectively. This figure shows that, although the correlation is reasonably good
in the frequency range of 2/rev, it is entirely unsatisfactory in the frequency range
of 4/rev, a blade passage frequency more typical of current helicopter design
practices.
6.4.3 Synthesis of Impedance from Modal Characteristics

It can be reasonably assumed that progress will continue to be made in achieving
more accurate FEM calculations. On the basis of this optimism, the following
material builds upon analytical FEM predictions that are for the present assumed
to be accurate. First, let us assume that the results of the FEM analysis consist of
the following modal parameters for each of the first N modes:
“chapter6(new)” — 2005/10/18 — page 226 — #38

Natural frequencies:
ω1 , ω2 , . . . , ωN
Mode shapes (generally vector quantities, i.e., having components in three

dimensions):
φ1, φ2, . . . , φN
Generalized masses:
M̃1 , M̃2 , . . . , M̃N
Fig. 6.16 Comparison of measured and analytic responses of fuselage c.g. as a result
of tail excitations.
“chapter6(new)” — 2005/10/18 — page 227 — #39

Fig. 6.16 Comparison of measured and analytic responses of fuselage c.g. as a result
of tail excitations (continued).
where the generalized masses are defined as

K
(k) (k)
M̃i ≡ mk φ i × φ i (6.123)
k=1
Note that k is the index over all of the fuselage mass elements as defined in the
FEM analysis; K is therefore typically a large number.
Indirect modeling of fuselage impedance using discrete masses. Our

ideas of fuselage impedance can be made more concrete by considering the flexible
“chapter6(new)” — 2005/10/18 — page 228 — #40

fuselage to be approximated by an equivalent solid rigid-body mass (with six

degrees of freedom), with (as yet unspecified) damping and stiffness matrices
defining the retention of the mass. We wish to size the mass, stiffness, and damp-
ing matrices defining the dynamics of the equivalent rigid-body mass so that the
mass has the same modal properties as the original elastic structure. Thus, let this
equivalent concentrated spring-mass-damper system at the hub be described by
the following differential equation:
[M]{q̈H } + [C]{q̇H } + [K]{qH } = {Fh } (6.124)
or
[ZF ( p)]{qH } = {FH } (6.125)
where [ZF ( p)] is the generalized fuselage impedance matrix, as a function of the
Laplace operator p. For our purposes, at present, the number of modes selected
must be equal to six. Correspondingly, let the modal matrix, containing only the
components of hub motion, be as denoted by ( ˆ ), as defined by the following
expression:
[] = [φ̂ 1 | φ̂ 2 | · · · | φ̂ 6 ] (6.126)
where
   
 x  Q1 

 
 
 


 y
 
Q2 

   
z 
Q  
3
{qH } = = [] = [] {Q} (6.127)

 θx 
 
 Q4 


 
 
 


 
θy  
 
Q5 
(6×6)
 
  
  
θz Q6
and where {Q} is a vector of generalized coordinates defining the modal activity.
By definition of the generalized mass, a diagonal (uncoupled) modal mass matrix
can be formed:
[]T [M][] = [M̃i ] (6.128)
Then, by assuming that the [] matrix is well behaved (i.e., is not near a singular
condition), the following expression can be written:
[M] = [T ]−1 [M̃i ][]−1 (6.129)
Similarly,
[C] = [T ]−1 [2ζi ωi M̃i ][]−1 (6.130)

[K] = [T ]−1 [ωi2 M̃i ][]−1 (6.131)
“chapter6(new)” — 2005/10/18 — page 229 — #41

It can be verified that these [M], [C], and [K] matrices will then yield a six-mode
eigensolution characterized by natural frequencies, mode shapes, and damping
values that are very near to those six selected modes of the original elastic fuselage
structure. The general (Laplace transform domain) fuselage impedance can then
be written as

[ZF ( p)] = [M] p2 + [C] p + [K] (6.132)
from which the (frequency-domain) fuselage mobility matrix can then be obtained:
−1
[YF (ω)] = (iω)2 [M] + iω[C] + [K]
−1
= [K] − ω2 [M] + iω[C] (6.133)
Two comments should be made regarding this method for calculating fuselage
mobility:
1) The method does require eigenvalue results for exactly six vibrational modes,
and the mode shapes must form a well-conditioned modal matrix.
2) The method yields a potentially convenient representation of the fuselage
dynamics for use in analyses wherein the coupling of the rotor with a rigid-
body (free-flying) fuselage is already available. This methodology thus extends an
otherwise rigid-body analysis into the modal response regime.
General modal approach. If the expression for the fuselage impedance is

written in a form utilizing the modal properties,

[ZF ] = [T ]−1 M̃i (ωi2 + 2iζi ωi ω − ω2 ) []−1 (6.134)
then the fuselage mobility can be written as follows:

−1
[YF ] = [] [M̃i (ωi2 + 2iζi ωi ω − ω2 )] []T (6.135)
Note that because no inverses are indicated in this expression, [] need not be
square. Therefore, any number of natural modes can be used (i = 1, 2, 3, . . . , N
where N is arbitrary).
Direct method using FEM. Although the calculation of eigenvalue proper-

ties (modal frequencies, generalized masses, etc.) is a staple calculation with most
FEM analyses, some FEM codes also perform forced-responses calculations. Thus,
for this case the mobility matrix is directly available:
, -
∂qF i
{qF } = { fH } (6.136)
∂fH j

directly obtainable from finite element
forced-response calculations
Some finite element analyses actually calculate the frequency-response functions

using a modal synthesis method anyway; thus, this method in effect amounts to
the method considered in the preceding subsection.
“chapter6(new)” — 2005/10/18 — page 230 — #42

6.4.4 Formulations in the Laplace-Transform Domain

The preceding formulations were developed in the frequency domain, wherein
it is assumed that the fuselage mobility matrix is required only at one specific
frequency, the blade passage frequency. An alternate approach to the problem is to
formulate it in the Laplace-transform domain. The advantages of this formulation
are, first, that it provides a direct means of calculating transients (time-history
responses as a result of both control inputs and/or as a result of instabilities)
and, second, that it provides a convenient basis for calculating the other dynamic
requirements for the fuselage vibration problem: rotor impedance and the inte-
grated rotor aerodynamic loads. Consider first the rotor (mode) equations of
motion:
[Z11 ( p)]{qR } = {FR } = {FR1 (applied airloads)}
+ {FR2 (qH , q̇H , q̈H )} (6.137)

= −[Z12 (p)]{qH }
or
+
qR
[Z11 ( p) · Z12 ( p)] −·− = {FR1 } (6.138)
qH
where the first subscript on the Zij submatrices refers to the equation “group.” In
this case, i = 1 refers to the rotor modal equations. The second subscript j refers to
the degree-of-freedom group. Again, the subscript j = 1 refers to the rotor modal
degrees of freedom, whereas the subscript j = 2 refers to the hub (fuselage) degree-
of-freedom group. Next, consider the equations of motion defining the hub or
fuselage degrees of freedom. (Note that, although the fuselage degrees of freedom
can be expressed in either modal or explicit finite element nodal descriptions, the
following formulation is most appropriate to a modal description for the fuselage):
[Z22 ( p)]{qH } = {FH } = {FH 1 (applied airloads)}

+ {FH 2 (qR , q̇R , q̈R )} (6.139)

= −[Z21 ( p)]{qR }
or
+
qR
[Z21 ( p) · Z22 ( p)] −·− = {FH1 } (6.140)
qH
These equations can then be combined to yield the total matrix equations of motion
for the rotor-fuselage coupled dynamics:
, -+ +
Z11 ( p) · Z12 ( p) qR FR1
−·−·− − ·− ·−·−·−−·− −·− = −·− (6.141)
Z21 ( p) · Z22 ( p) qH FH1
“chapter6(new)” — 2005/10/18 — page 231 — #43

This equation is then a general statement of the fully coupled rotor-fuselage interac-
tion problem and can be solved using standard inverse Laplace-transform methods.
In each of the equation group sets, the first of the applied loads is identified as
being of aerodynamic origin. Although this generalization is fairly accurate for
the rotor, there can typically be additional sources of nonrotor excitation for the
fuselage, as arising from the drive system and/or the empennage, for example.
Thus, {FH 1 } consists of the sum of the aerodynamic excitation components at the
(h) (nh)
hub {FH 1 }, and all of the remaining (nonhub) components {FH1 }:
(h) (nh)
{FH1 } = {FH1 } + {FH1 } (6.142)
Now, restricting our attention to only the excitations at the hub, let us consider the
fuselage modal equations of motion in the preceding rearranged form:
[Z22 ( p)]{qH } = {FH 1 } − [Z21 ( p)]{qR } (6.143)
whereupon, after substitution for {qR } (using straightforward matrix algebra tech-
niques), we obtain the following equation for the fuselage response variables:

[Z22 ( p)] − [Z21 ( p)][Z11 ( p)]−1 [Z12 ( p)] {qH }
= {FH1 } − [Z21 ( p)][Z11 ( p)]−1 {FR1 } (6.144)
Next let us separate out those contributions to [Z22 ] because of the fuselage and
the rigid-body rotor masses, respectively:
[Z22 ] = [Z22
F
] + [Z22
R
] (6.145)
Then we can rewrite the hub/fuselage equation as

F
[Z22 ( p)] + [Z22
R
( p)] − [Z21 ( p)] [Z11 ( p)]−1 [Z12 ( p)] {qH }
= {FH 1 } − [Z21 ( p)] [Z11 ( p)]−1 {FR1 } (6.146)
6.4.5 Methods for Evaluating the YF and ZR Matrices

Analytical methods: general considerations. The formulations of the pre-
ceding section can be used to effect analytical evaluations of the YF and ZR matrices,
as well as the integrated rotor loads vector {Fapp2 }. To implement the rotor-fuselage
coupling relationships, the responses of the fuselage must be defined by the spe-
cific responses at the hub. Upon setting the Laplace-transform variable s to iω, we
can then write the fuselage mobility as
F −1
[YF ] = Z22 (iω) (6.147)
It can be verified as an exercise that the rotor impedance is then expressible as

R
[ZR ] = Z22 (iω) − [Z21 (iω)][Z11 (iω)]−1 [Z12 (iω)] (6.148)
“chapter6(new)” — 2005/10/18 — page 232 — #44

The integrated aerodynamic loads and (hub-fixed) blade aeroelastic induced loads
can be written as
{ fapp2 } = {FH1 } − [Z21 (iω)] [Z11 (iω)]−1 {FR1 } (6.149)
Rotor impedance. Consider the following interpretation of the expression

just given for rotor impedance:
[ZR ] = [Z22
R
] − [Z21 ] [Z11 ]−1 [Z12 ] (6.150)
where [Z22R ] denotes the rotor acting as a rigid body, [Z ] denotes rotor (mode) to
21
hub motion coupling, [Z11 ]−1 denotes hub-fixed rotor modal response dynamics,
and [Z12 ] denotes hub motion to rotor (mode) coupling. Although the preceding
development serves to define the concepts involved, the actual calculation of rotor
impedance can be accomplished using a comprehensive rotor aeroelastic analy-
sis for which the dynamic effects of prescribed hub motion are included. Using
the aforementioned development as a guide, we can formally define the rotor
impedance matrix as follows:
, -
∂
[ZR ] = (hub loads as seen by the fuselage)
∂qHj
, -
∂
= [[Z22 ( p)]{qHj } + [Z21 ( p)]{qR }]
∂qHj
, , --
∂qR
= [Z22 ( p)] + [Z21 ( p)] (6.151)
∂qHj
Moreover, a more straightforward implementation can be obtained by direct

numerical differentiation. Assume that the hub motion is included in the rotor
aeroelastic analysis as an arbitrary fixed-amplitude, but sinusoidally time-variable
vector:
{qH } = {q̄H }eiωt (6.152)
The rotor impedance matrix is then formed in a column-by-column fashion in the

following manner. Let the impedance matrix be partitioned, using columns defined
by the vectors {Z̃Rj }.
[ZR ] = [Z̃R1 | Z̃R2 | · · · | Z̃Rj | · · · ] (6.153)
where
1
{Z̃Rj } = {FH2 (qR + qR , q̇R + q̇R , q̈R + q̈R )j − FH2 (qR , q̇R , q̈R )}
qHj
(6.154)
“chapter6(new)” — 2005/10/18 — page 233 — #45

Here, {FH2 } is the vector already defined of hub excitation terms arising from blade
aeroelastic responses. The vector {qR } denotes the rotor (or blade) degrees of free-
dom (aeroelastic responses) resulting from the fixed-hub condition ({qH } = {0}).
The vector {qR + qR }j denotes the modified rotor responses resulting from the
sinusoidal motion of the jth hub degree of freedom, with all other hub degrees of
freedom set to zero, that is, {qR + qR }j results from
 
 0 

 .. 


 
 . 

{qH } = q̄Hj eiωt (6.155)

 .. 

 

 . 
 
0
Integrated rotor aerodynamic loads. The calculation of { fapp2 }, the vec-

tor of integrated rotor aerodynamic hub loads, is essentially the basic hub-fixed
loads quantity that is routinely calculated by typical comprehensive rotor aero-
elastic analyses. Historically, this vector quantity has been used extensively as
the principal excitation loads vector for calculating fuselage vibrations in the
absence of rotor-fuselage coupling considerations. Because this quantity includes
the effects of blade aeroelastic responses and the unsteady rotor airloads, the loads
vector is subject to the many uncertainties presently inherent in these analytic
predictions.
Experimental evaluations. Of the three required dynamic quantities, only

the fuselage mobility [YF ] has been experimentally measured in any form. These
tests must necessarily be made at full scale, and more will be said about techniques
for doing this in a later chapter. The rotor impedance [ZR ] has been experimentally
measured only on a preliminary research basis, as reported by Cansdale et al. This
measurement is not readily available, at present, in a general form, complete with
parametric extensive variations for correlation with theory. The integrated rotor
aerodynamic hub loads have never truly been experimentally measured because
all rotor vibration tests, both at full scale and at model scale, have been made
with rotor support systems with finite hub mobilities; hence, the assumption of an
infinitely rigid hub has never been exactly met.
References
Section
6.2 Bramwell, A. R. S., Helicopter Dynamics, Wiley, New York, 1976.
New York, 1959.
“chapter6(new)” — 2005/10/18 — page 234 — #46

6.3 Gangwani, S. T., “Calculation of Rotor Wake Induced Empennage Air-
loads,” Journal of the American Helicopter Society, Vol. 28, No. 2,
1983, pp. 37–46.
6.4 Dowell, E. H., Curtiss, H. C., Scanlan, R. H., and Sisto, F., A
Modern Course in Aeroelasticity, Sijthoff & Noordhoff, Rockville,
MD, 1978.
New York, 1959.
Cansdale, R., Gaukroger, D. L., and Skingle, C. W., “A Technique for
Measuring Impedance of a Spinning Model Rotor,” Royal Aircraft
Establishment TR-71092, London, 1971.
Problems
6.1 Consider a simplified representation of a beam that is free at one end and
constrained by a pin connection and a rotary damper at the other. Let the free
end of the beam be excited by a force of amplitude F and step function time
variation:
(a) Write the equations of motion for the two degrees of freedom, z1 and z2 .
(b) Solve for the response vector [z1 (t), z2 (t)] both exactly and using a
Galerkin (modal) approach. Assume unit values for m, c, k, and .
(c) Considering the rotary damping to be applied exactly at the pin con-
nection (in the limit), calculate the bending moment just outboard of the
damper attachment point three ways:
1) using the exact solution and the (damper rate) × (angular velocity) at
the pivot point;
2) using the mode displacement method together with the modal response
solution; and
3) using the force integration method again together with the modal
response solution.
(d) Discuss how answers using the preceding three methods would
compare if the excitation function were an impulse function.
“chapter6(new)” — 2005/10/18 — page 235 — #47

6.2 Find the DMF for a damped single-degree-of-freedom system for the
following forcing function:
Sketch the variation of the DMF with the parameter (T ωn ) for damping ratios
of 0 and 0.5.
6.3 Consider a rotor blade in hover. Let
z(x, t) = γ1 (x) · q(t)

γ1 (x) = (R/3)[6(x/R)2 − 4(x/R)3 + (x/R)4 ]
ω1 = 41.47 rad/s
M1 = 2417 lb-in.-s2
R = 210 in.
c = 13 in.
= 355 rpm
B = 0.97
a = dc /dα = 0.1/deg
f (x, t) = −1/2 ρacx(xθ0 + ż) (excitation + flap damping)
“chapter6(new)” — 2005/10/18 — page 236 — #48

Find the maximum bending moment at (blade) station 120 caused by a

step input of collective of 2 deg [θ0 = 2 (deg) · 1(t)] at sea level (standard
conditions) by
1) the mode displacement method using
(a) Mb1 (120) = 46.89 in.-lb/(in. at tip), and
(b) formulas for Mb1 given in the text;
2) the force integration method, again using the formulas given in the
text; and
3) the dynamic magnification factor method.
“chapter6(new)” — 2005/10/18 — page 237 — #49

7
Methods for Vibration Control
In the previous chapter the basic tools are identified (if not as yet completely
formulated) for analyzing the helicopter/rotorcraft for vibration. The basic vibra-
tion analysis problem boils down to knowing the following items with sufficient
accuracy:
1) the aerodynamic environments of both the rotor and the fuselage;
2) the aerodynamic and aeroelastic response characteristics of the (rotating)
rotor in the form of integrated hub loads;
3) the structural characteristics of the (nonrotating) fuselage, in the form of
accurate frequency response functions ( fuselage mobilities); and
4) the rotor-fuselage interactions, in the form of rotor impedances.
Once each of these separate items can be predicted accurately, coupling tools
exist for combining them to give reliable vibration predictions. Such predictive
capability, if available sufficiently early in the design process, should lead to the
quantum reduction in vibration (with minimum weight) that is so highly prized.
However, the required knowledge to achieve this state of affairs is not yet at
hand, and aircraft design, fabrication, and certification must continue without the
benefit of this foreknowledge. Thus, methods that have necessarily been devised
for modifying the aircraft design “after the fact” will continue to be used to reduce
vibration to some tolerable level. This chapter describes major methods devised
and presently either used outright or seriously being considered for controlling
vibration.
7.1 Basic Modification Methodology

Generally, the structural modification methodology available to the designer is
of three basic types: 1) detuning the structural responses through passive struc-
tural changes, 2) using (passive or active) vibration alleviation devices, and 3)
altering the excitation forces to minimize the vibrational content. An important
consideration that should be kept in mind is that short of actually reducing the
aerodynamic excitations, any structural modification to the airframe, whether it be
in the form of actually adjusting structural members or adding devices, is this: to
exert any counteracting force to minimize response, the only basic resource avail-
able is to use inertial space as the hard point. Force monopoles are not available to
us. Thus, secondary mass must be made available to vibrate to create the required
counteracting forces.
239
“chapter7(new)” — 2005/11/16 — page 239 — #1

7.1.1 Design/Modification Variables

Inherent in the concept of structural modification is the notion that there does
indeed exist some finite number of parameters that can be varied regardless of
how far along the design has progressed (despite the fact that the design might
even be considered to be “frozen”). Normally such parameters would be referred
to as design parameters, but in the present context we can denote them modifi-
cation parameters as well (≡ pm ). Such a notion makes sense for cases wherein
systematic modifications must be made to a relatively mature design. As with
design parameters, which are conceptually most appropriate at an early stage in
the development process, modification parameters can furthermore be classified
as either detailed or derived.
Detailed parameters are parameters that are typically directly identifiable by
virtue of dimensions obtainable on a drawing or specifications for one or more of the
structural components. They could be linear dimensions, angles, densities, and/or
material selections. As such, these parameters can be readily related both to each
other and to individual ranges of variability by means of appropriate constraints. An
example of such a constraint might be the requirement that two or more component
masses must not occupy the same space. Constraints typically can be expressed
mathematically in terms of both equalities and inequalities.
Derived parameters are those parameters upon which the structural dynamics
are most naturally formulated mathematically and are typically combinations of
the detailed parameters. Examples of derived parameters are mass and stiffness
distributions, center of gravity (c.g.) and elastic axis (e.a.) offsets, aerodynamically
effective areas, and effective spring rates. For such modification parameters appro-
priate constraints are typically not directly available. Thus, analytic treatments
using derived parameters are often instructive but are not practical because of the
inattention of the real constraints posed by the detailed modification parameters.
7.1.2 Reduction of Dynamic Amplification by Detuning

On the basis of what we generally know of the responses of dynamic systems, we
can immediately identify one method for controlling vibration: reduce all instances
of dynamic amplification caused by close proximity to resonance conditions. This
can be achieved by detuning the structure in either of two basic ways: by chang-
ing existing unfavorable frequency placements or by changing unfavorable modal
orthogonality placements. In the former case the narrowband excitation problem
characteristic of the rotorcraft problem is to be controlled by keeping the struc-
tural natural frequencies well separated from the excitation frequencies, typically
multiples of the blade passage frequencies. In the latter case the problem is to be
controlled by maximizing the orthogonality of the structural mode shapes to the
spatial variability of the applied loadings, a process sometimes referred to as modal
shaping. In either case some knowledge of how more favorable conditions are to
be achieved is required; for this we seek appropriate vibrational trend information.
7.1.3 Extraction of Trend Information

In all cases the process of structural modification is to be guided by the principle
of reducing the dynamic magnifications through either frequency and/or spatial
“chapter7(new)” — 2005/11/16 — page 240 — #2

METHODS FOR VIBRATION CONTROL 241
(mode shape) detuning. For any given dynamic problem this process requires some
knowledge of how the resonant magnification varies with the set of design para-
meters appropriate to the problem. This variation of the resonance characteristics
constitutes trend information, which must somehow become known to guide the
modification process. The information can, of course, be obtained by calculat-
ing an in-depth ensemble of results for wide variations in the design parameters.
These results typically take the form of graphs showing the explicit variations of
the characteristics of interest for all feasible values of the design parameters. An
alternate method of dealing with trend information is to compute the derivatives
of the pertinent resonance characteristics directly either analytically, if possible,
or by direct numerical calculation of the partial derivatives constituting this trend
information. The following sections deal with methods for computing the trend
information.
Analytic evaluation. Evaluation of the variability of the undamped modal

characteristics with respect to the design parameters can be achieved using the basic
eigenvalue problem wherein the various dynamic matrices are found to be readily
expressible, explicit functions of the design/modification parameters. Thus, for
this case the eigenvalue problem can be expressed in terms of the mass and stiffness
matrices [M] and [K], respectively, both of which are analytic functions of the
design parameters pm :
[M( pm )] {Z̈} + [K( pm )] {Z} = 0 (7.1)
where the response vector {Z} is given by a modal decomposition using the first
J modes:

J
t
{Z} = {φ(x, pm )} eiωj ; where ωj = ωj ( pm ) (7.2)
j=1
Note that because the eigenvalue problem is defined in terms of the modifica-
tion parameters pm (m = 1, 2, . . . , M), both the modal matrix [] (a J × J matrix
whose columns are the mode shapes) and the J natural frequencies ωj are implicit
functions of the design parameters. The solution of the preceding eigenvalue prob-
lem for a given set of modification parameters pm = p̃m yields a set of natural
frequencies and mode shapes:
ωn = ωn ( p̃m ); {φn } = {φn ( p̃m )} (7.3)
Derivatives of the eigenvalues. If the preceding eigenvalue equation is differ-

entiated with respect to the mth modification parameter and the Galerkin procedure
is applied to the resulting equation, the following result is obtained for the derivative
of the ith natural frequency:
∂ωi [φi ][−Mpm ωi2 + Kpm ]{φi }

ωipm = = (7.4)
∂pm 2ωi M̃i
where M̃i is the ith generalized mass.
“chapter7(new)” — 2005/11/16 — page 241 — #3

Derivatives of the eigenvectors. Two principal methods exist for calculating

the derivatives of the eigenvectors with respect to the modification parameters.
One method is based on the representation of the derivative of the ith eigenvector
as a linear combination of the eigenvectors themselves. Although this method does
produce a calculation of the desired derivatives, it has the distinct disadvantage
that all of the eigenvectors for the system must be known, a requirement of tall
proportions for typical finite element modeling (FEM) solutions. Consider that
such FEM formulations invariably involve many degrees of freedom; consequently,
only a small number of the (lowest-frequency) modes are calculated. Furthermore,
a characteristic of FEM solutions is that the accuracy of the obtainable higher-
frequency modes, if they are calculated, is usually less accurate than the lowest
ones. As a result of this disadvantage, an alternate method has been formulated
that utilizes the derivative information of the given dynamic matrices, as before,
but only the eigenvector for the mode in question (i) . Again, we begin with the
basic statement of the eigenvalue problem, whose component matrices are analytic
functions of the modification parameters:

−ωi2 [M( pm )] + [K(pm )] {(i) } = [Z]{(i) } = {0} (7.5)
(i)
which can be partitioned with one arbitrary degree of freedom φk separated out
as follows:
  
| |  1 (i)
 
 Z11 |· Z1k |· Z13   −·−
−·−·− ·−·−·−·−·−·−  




 | | 
 Zk1 · Zkk · Zk3  φk = {0} (7.6)
 
−·−·− |· −·−·−·−·−·−
| 




 | | 
 −·−
Z31 · Z3k · Z33   

| | 3
Then the derivative of the eigenvector can be expressed as
 (i)

 v1 


−·−


∂(i)  
= 0 + ci {(i) } = {v (i) } + ci {(i) } (7.7)
∂pm 
 

−·−
 

 

v3
Let {v̂ (i) } be the nonzero compaction
 (i)
 v1 
−·−
 
v3
“chapter7(new)” — 2005/11/16 — page 242 — #4

Then the solution for this vector is obtained from the following simultaneous
equations:
   
| | |
 Z11 |· Z13  (i) Z
 11pm · Z 1kpm · Z13pm 
−·−·− · −·−·− {v̂ } =  −·−·− |· −·−·−·−·−·−· |  {(i) } (7.8)
 |   | | 
Z31 · Z33 Z31pm · Z3kpm · Z33pm
| | |
The constant ci is determined on the basis of how the eigenvector is normalized. If

φk is taken to be unity (normalization on the basis of the largest element), then the
constant ci is zero. In some instances the eigenvector is normalized on the basis
of the following relationship:
[(i)C ] [S] {(i) } = 1 (7.9)
where [(i)C ]T [i.e., = {(i) }C ] is the complex conjugate of {(i) } and [S] is some
symmetric matrix (often taken to be the mass matrix). For this case the ci constant
is determined by
ci = −Re([(i)C ] [S] {v (i) }) + 1/2 [(i)C ] [Spm ]{(i) } (7.10)
The derivatives of the [Zik ] partitionings, with respect to the parameters pm , require
not only the derivatives of the dynamic matrices [M] and [K] but those of the
eigenvalues as well, as per the prior development. Note also that the preceding
expressions for the eigenvalue and eigenvector derivatives are to be calculated
from the results of the eigenvalue solution for the particular set of modification
parameters pm = p̂m .
Numerical evaluation. Evaluation of the trend information is often required

when the variability of the system matrices with respect to the modification para-
meters is not explicit and therefore cannot be expressed as analytic functions. For
this case the calculation of the required trend information typically requires the
use of “brute force” numerical evaluations of the partial derivatives:
∂Y
= [Y ( p1 , p2 , . . . , pm + δ, . . .) − Y ( p1 , p2 , . . . , pm , . . .)]/δ (7.11)
∂pm
where Y represents any of the eigenvalue quantities considered in the preceding
subsection.
7.1.4 Cancellation Effects

Although the modal detuning method for vibration control described earlier
can be used to achieve reductions in vibratory response for any selected mode or
combination of modes, the method has the drawback that it cannot account for
vibration problems wherein significant cancellation effects are present. Cancel-
lation refers to the situation wherein the response of a critical point in the structure
“chapter7(new)” — 2005/11/16 — page 243 — #5

is the result of significant contributions from a number of modes all responding

to the same sinusoidal excitation. For this case the net response of the critical
point is the vector sum of the component modal responses that can often involve
substantial mutual cancellation. Thus, for this case a significant reduction of only
one modal component of the response can actually result in an increase of the net
response. For this reason detuning must be used in a much more comprehensive
manner than would be indicated by the preceding subsection.
7.2 Modification of Blade Dynamics

The most typical means used for modification of the blade dynamics is that
of detuning the design as described earlier. Generally this form of modification
is limited with respect to what is available to be changed. All too often the only
modification parameter available is the addition of weight. This form of blade
modification is less than optimal not only in that it is practical at usually only a
few nonoptimal blade locations, but also in that it counterproductively detracts
from the rotorcraft’s payload-carrying capability. Furthermore, as can be shown
from the basic behavior of rotating beams (as considered in a previous chapter),
the addition of nonstructural weight typically produces significant changes only in
the in-plane natural frequencies. Additional structural stiffening is usually required
for alteration of the flatwise and/or out-of-plane natural frequencies.
7.2.1 Aeroelastic Conformability

One approach to the favorable modification of blade dynamics is the so-called
technique of aeroelastic conformability, wherein the rotor blade is purposely
designed with specific nonlinear distributions of the basic elastomechanic para-
meters. These parameters include the basic spanwise mass and stiffness distri-
butions, as well as a variety of couplings between the various combinations of
flatwise, edgewise, and torsion modes. These couplings are introduced so that
“corrective” secondary aeroelastic responses might be generated by the blade itself
when it would otherwise respond in a mode tending to produce high vibratory hub
shears and moments.
Basic methods identified for achieving these couplings are 1) to alter the blade
elastomechanics, by devising appropriate spanwise distributions of the various
chordwise offsets centers (elastic axis, gravity center, tension center, etc.); 2) to
alter the aerodynamic loadings, by incorporating special aerodynamic refinements
(taper, twist, and special tip shapes); and 3) to devise new elastic couplings now
possible with the use of anisotropic composites. One difficulty of the aeroelastic
conformability approach is that the resulting blade response is typically a com-
posite of many contributions resulting from the several loads acting on the rotor
blade. This, in effect, is an example of reduction by cancellation as discussed in
an earlier subsection. Figure 7.1 typifies this phenomenon for the case of rotating
blade lateral shear.
This figure points to the difficulty in attempting to reduce any component of
hub loading by modifying any one component using a passive coupling effect.
That is, because of the high degree of cancellation present, reducing one compo-
nent without reducing all components commensurately might result in marginal
“chapter7(new)” — 2005/11/16 — page 244 — #6

Fig. 7.1 Polar diagram of predicted 3/rev rotating lateral shear for an articulated
blade in high-speed flight.
improvement or even a worsening. Furthermore, the introduction of certain

passive aeroelastic couplings to achieve the desired conformability could limit
the favorable cancellations to a small number of specific flight conditions wherein
the passive couplings would be “well tuned.” The concept is an attractive one, how-
ever, and could potentially achieve the desired reductions in vibration with little or
no increase in weight. The appropriate manner of applying such couplings is not
intuitively obvious in light of the considerations of the present cancellations. Thus,
any successful implementation must be achieved using an appropriate mathemat-
ical tool; presently, an attractive tool identified for this purpose is a formalized
optimization scheme applied to the rotor vibration problem.
7.2.2 Vibration Minimization Schemes

The process of designing for minimum vibration has a variety of routes to follow.
The aeroelastic conformability concept already described is often approached by
casting the problem in the form of a formal optimization process and then using
one of a variety of robust optimization algorithms currently available. Because
optimization schemes can be applied to more than just vibration, a presentation of
“chapter7(new)” — 2005/11/16 — page 245 — #7

such methodology is reserved for a subsequent chapter. One major hurdle to using
highly mathematical optimization schemes is that the results have to be “sold” to
the designer who has the responsibility for transforming the rotor design to the
real world.
An emerging technology is the use of appended devices based on special mate-
rials such as piezoelectric crystals and shape-memory alloys that are typically
activated electrically. The voltage levels are much higher than that typically used
for solid-state avionics. The advantage of using such devices, however, is that their
activation can be be made as harmonically variable as desired and can thereby sig-
nificantly impact on reducing rotor vibration inputs. The disadvantages are that
1) they require comparatively sophisticated implementation in going across the
nonrotating to rotating coordinate system boundary, and 2) a high degree of relia-
bility has to be maintained, which is difficult using sophisticated high-voltage-level
electronics.
7.3 Modification of Fuselage Dynamics

All of the remarks presented earlier regarding basic modification methodology
pertain to the fuselage problem as well as to the blade and rotor problem. Poten-
tially, the use of these basic techniques should be more straightforward for the
nonrotating fuselage than for the rotating blades and rotor. However, the fuselage
problem is complicated by the fact that this substructure is subject to a range of
variations in mass caused by the spectrum of operational loading conditions and
stores. Thus, a twofold need exists for analyzing the dynamic characteristics of the
fuselage structure for modifications that would essentially amount to variations
about some base set of characteristics. A number of general theoretical methods
currently exist for predicting total system structural dynamic characteristics from
those of the components and/or of modifications. However, relative to the broad
spectrum of applications for which such theories can be used the rotorcraft vibra-
tion problem is somewhat unique in that it is a narrowband process. For this reason
it becomes advantageous to analyze the problem in the frequency domain and to
work with mobilities (and conversely, with impedances) as the principal vehicles
for describing these characteristics. The following sections present material useful
for analyzing structures with this scheme in mind.
7.3.1 Modifications Using Analytical Testing

The somewhat ambiguously named technique of analytical testing relates to a
potentially powerful technique for analytically obtaining the modified vibratory
responses of a structure for which a basic set of (usually unsatisfactory) responses
have already been obtained from some form of testing. Thus, the method has the
potential for saving many expensive flight-test hours that would otherwise have
been spent in solving a vibration problem experimentally in a hit-or-miss manner
with relatively little coherent direction.
The specific steps one would use to solve a vibration problem using analytical
testing consist of the following:
1) Conduct a flight test with an assortment of (vibratory) response measur-
ing transducers (accelerometers) at several locations in the airframe, both where
“chapter7(new)” — 2005/11/16 — page 246 — #8

vibration is critical {∗ }F and where potential (interface) modifications might be

made {∗ }I . For both locations we denote these responses as the original or old
responses: {∗ }0 .
2) Conduct a ground-based shake test to measure Y FI , the mobilities of the
critical response locations relative to the (interface) points in the structure wherein
modifications are contemplated, and for which response test data were originally
acquired as per step 1. Additionally, the mobilities of the structure at the interface
points Y II are to be measured.
3) Some form of dynamic modification or correction is selected for attachment
at the interface point(s). This modification has a mobility associated with it Y C II ,
which can either be identified by analytical or experimental means.
4) The quantities identified in steps 1–3 can then be brought together to form
the modified responses at the critical response locations as a result of the addition
of the modification mobilities at the interface point(s):
{qF } = {qF0 } −[YFI ] [YII + YIIC ]−1 {qI0 } (7.12)

new old interface- at interface
interface mobility
7.3.2 Derivation of Basic Analytic Testing Equation

The basis of the preceding equation for analytical testing is not intuitively
obvious. To show that the preceding equation is indeed correct and, furthermore,
to illustrate the definitions of the component matrices in the equation, a formal
derivation of the equation is in order and is therefore presented in the following
material. First we describe the sinusoidal responses of the original (unmodified)
structure in terms of the applied loads { f }, the responses of the structure at the
points of interest {qF }, and at the interface points {qI }, and the mobilities at these
locations:

qF YFF YFI f
= (7.13)
qI YIF YII 0
Note that, as long as there are no applied loads at the selected interface location
(corresponding to {qI }) the equation is actually simply stated as {qF } = [YFF ]{f }.
In Fig. 7.2 we now denote the unmodified (fuselage) structure as body A and the
“add-on” modification as substructure B. As shown, there now exists an additional
load vector at the selected interface location {fI }. With the addition of substructure
B, the whole structure is now the union of A and B, and the interface location
loads are now internal loads of this combined structure. As with any investigation
of the internal loads between two substructures, the internal loads must be defined
in opposite directions relative to each of the subsystems. Thus, in the figure the
internal loads at the interface are depicted with load vectors as “seen” by each of
the respective substructures. Note that the mobilities for each of the substructures
at the interface location are defined in terms of the load and deflection vectors
both being directed inward to the substructure and in the same direction. Thus, the
deflection vector at the interface as seen by substructure A is denoted as {qI }a and
that as seen by substructure B as {qI }b . Clearly, these two vectors are negatives of
each other [{qI }a = −{qI }b ].
“chapter7(new)” — 2005/11/16 — page 247 — #9

Fig. 7.2 Schematic of the dynamic interaction between an existing structure and a
dynamic subsystem modification.
The basic approach to deriving the analytical testing equation is first to define the
interface load vector and introduce it into the mobility equation for the fuselage
(substructure A). When this load is so introduced, the new response is then the
required modified response:
a
qF YFF YFI f
= (7.14)
qI YIF YII fI
Note that, because the impedance matrix partitions were measured on the fuselage
alone, the deflection vector is identified as looking into substructure A [i.e., {∗ }a ].
The lower partition of the deflection vector can then be related to the deflection
vector looking into substructure B:
{qI }a = −{qI }b = −[YIIC ]{fI } (7.15)
Then the previous equation can be rewritten to isolate the interface load vector:
a
qF YFF YFI f YFF YFI f YFI
= = + { fI }
qI YIF YII fI YIF YII 0 YII
a
qF0 YFI
= + { fI } (7.16)
qI0 YII
The second partitioning of this equation can then be combined with the preceding
result to give
−[YIIC ]{ fI } = {qI0 }a + [YII ]{fI } (7.17)
“chapter7(new)” — 2005/11/16 — page 248 — #10

From this equation the required expression for the interface load vector can be
found:
{ fI } = [YIIC + YII ]−1 {qI0 }a (7.18)
The derivation is then completed by using this expression to eliminate the interface
load vector from the preceding equation. The preceding expression is for a structure
that is being excited at some point(s) separate from either the critical response
points or the interface points. Thus, in the case of a rotorcraft vibration problem
the “old” responses at the points F and I are those resulting from excitation located
at some other point H, which might correspond to one of the rotor hubs. An
important assumption inherent in the preceding formulation is that the excitation
at point H does not change because of the modification at the interface point.
This is not always the case with rotorcraft problems, and the analytical testing
methodology has been expanded to treat this more general case.
Example of usage. The AH-1G helicopter has been used extensively as a

test bed for fuselage vibration research. Based on the mobility measurements
made on this aircraft, it provides a realistic vehicle for exemplifying the basic
analytic testing methodology. The following example (Flannelly et al.) is that of
a hypothetical configuration of an AH-1G helicopter that had never flown with
rocket pods. An addition of rocket pods weighing 181 kg (400 lbw) each to the
outboard wing stations is contemplated.
For this hypothetical configuration we wish to introduce flight acceleration test
data taken of the clean aircraft (no rocket pods), on an as-available basis from
another project. We wish to predict what the fuselage responses would be with
the addition of the rocket pods. The addition of the rocket pods clearly defines a
structural modification (addition of mass). Data resulting from a set of selected
points within the fuselage for a variety of flight tests were analyzed. Conditions
of highest peak-to-peak vibration for the selected set of locations were assembled
and, in each class of flight condition, were analyzed with a harmonic analysis over
five to eight rotor revolutions.
The accelerances (acceleration-based mobilities) of the left and right wing
stores position for vertical motion constitute a 2 × 2 complex matrix. These are
at butt lines (BL) ±60 and are identified as Z200L and Z200R. The following
evaluation of this mobility matrix, for a frequency of 10.8 Hz, has the units of
(nondimensional) acceleration (i.e., vibrational load factor n)/force (in Newtons),
and herein specifically of n/N:

0.054/−8 deg 0.012/14 deg
[Ÿrr ] = × 10−3 (n/N) (7.19)
0.012/14 deg 0.040/8 deg
Analytic testing predictions are then made for the nine motion coordinates shown
in Fig. 7.3.
The matrix of mobilities for these nine coordinates (the j index) relative to
forcing at locations Z200L and Z200R (the r index) is given by Eq. (7.28).
This matrix has the units of load factor per Newton for sinusoidal motion
“chapter7(new)” — 2005/11/16 — page 249 — #11

Fig. 7.3 Location and direction of components of motion measured on AH-1G

helicopter.
at a frequency of 10.8 Hz:

Z200L Z200R
|− −|
Z200 | 0.033/– 4 deg 0.052/9 deg |
| |
Z90L | 0.017/22 deg 0.024/–140 deg |
| |
Z140L | 0.046/5 deg 0.018/–176 deg |
| |
Z396R | 0.046/–19 deg 0.068/11 deg |
[Ÿjr ] = | | (7.20)
Z400 | 0.100/– 49 deg 0.049/–97 deg | × 10−3
| |
Y380 | 0.193/144 deg 0.208/– 48 deg |
| |
Y440 | 0.091/–167 deg 0.086/8 deg |
| |
Y490 | 0.065/77 deg 0.084/–96 deg |
| |
Y517 | 0.168/130 deg 0.151/–53 deg |
|− −|
The vibration, in units of vibratory load factor, resulting from the change is obtained
from the basic analytic testing equation:
 1  −1
0
 1 
{q̈j } = {q̈j } − [Ÿjr ]  181 + [Ÿrr ] {q̈r }
1 G
(7.21)
0
181
where G is the gravitational acceleration (= 9.81 m/s2 ).
Structural synthesis using analytical testing. The modification of a struc-

ture can be thought of as a combination or synthesis of two lesser structures: the
original or unmodified structure and the structural modification. Thus, the concept
“chapter7(new)” — 2005/11/16 — page 250 — #12

of analytical testing can be expanded to include several concurrent “modifications”

or substructure syntheses, the totality of which forms the entire structure. This
extension of the analytical testing methodology was developed by Jetmunsen et
al. and is summarized in Appendix C.
7.4 Vibration-Suppression Devices

A specific modification that can be made to any vibration-prone structure is
some form of vibration-suppression device. Typically, these devices are passive in
nature and thereby require some finite response in order to generate counteracting
forces and/or moments. Such devices can be appended to either the rotor blade,
the rotor head, and/or the nonrotating fuselage. They can, furthermore, be classi-
fied according to whether they are amplitude reducers (which do not reduce the
excitation force level) or source alleviators (which work to reduce the excitation
force level). A key feature of the former device is the necessity for using mass
elements to “absorb” the dynamic load (see Chapter 2). Note that the use of these
mass elements is usually tempered by the need for minimizing weight as well as
the fundamental constraint of being able to minimize the static deflections under
load. Large deflections are often the result of the low spring rates typically required
to tune the selected absorption masses. As we will see, the distinction between the
two basic types of vibration suppression devices sometimes becomes fuzzy and
depends on where in the structure we are measuring the responses and/or where
we are measuring the excitations.
7.4.1 Rotor-Related Devices

Blade appended pendulum absorbers (pendabs). One of the simplest
of the rotor-related devices is the pendulum absorber (pendab), which is based
on the basic dynamics of the simple two-degree-of-freedom vibration absorber
considered in Chapter 2. The basic operation of a pendab is the generation of
opposing dynamic shear loads (caused by a near resonance condition) in order to
relieve the dynamic shears fed to the rotor hub. An example of a typical installation
of a blade pendab on an operational helicopter is presented in Fig. 7.4.
The distinction of a rotor-appended blade pendab from more conventional con-
figurations is that its stiffening comes almost exclusively from the centrifugal force
field; thus, the pendab is essentially “tuned” to a specific per rev frequency. This
characteristic can be appreciated by considering the pendab geometry given in
Fig. 7.5.
From concepts derived in a preceding chapter on blade dynamics, we can readily
find that the natural frequency of the pendab ωP is independent of pendab mass
and is given as

Rh
ωP = 1 + (7.22)
r
The essential operation of the pendab is to alleviate at their source the vertical hub
loads that interact with the fuselage. At the properly tuned condition the response
of the pendab is phased so as to counteract the vibratory blade out-of-plane shears
“chapter7(new)” — 2005/11/16 — page 251 — #13

Fig. 7.4 Installation of pendabs on MMB BK-117 helicopter. (Reproduced courtesy

of the Helicopter Division of Messerschmitt-Boelkow-Blohm Helicopter and Military
Aircraft Group.)
Fig. 7.5 Basic geometry of typical blade appended pendab.
“chapter7(new)” — 2005/11/16 — page 252 — #14

Fig. 7.6 Typical load attenuation characteristics of blade appended dynamic

pendulum absorber (pendab).
generated by the blade outboard of the pendab attachment location. The operation
of the pendab is most meaningfully described by means of the load attenuation
characteristics with variation of the pendab frequency as typically depicted in
Fig. 7.6.
Note that optimum pendab performance is usually achieved at a tuned frequency
ωP , which is somewhat less than the excitation frequency, in this case the blade
passage frequency b. Note also that for tuned frequencies above the excitation
frequency, the pendab actually amplifies the hub loads.
The advantages of using pendabs are as follows:
1) Pendabs are relatively simple, both to analyze and design and to install.
2) Pendabs are effective in substantially reducing (if not eliminating outright)
the hub shears at the blade root.
3) For reasonable blade-root shear attenuation, without excessive pendab
response, the pendab should be made as massive as possible. However, the addi-
tion of substantial nonstructural mass to the rotorcraft design impacts negatively
on the payload productivity.
4) The presence of damping in the pendab hinge is inevitable; however, this
damping is detrimental to the operation of the pendab.
5) Pendabs typically reduce only vertical blade-root shears at one nondimen-
sional frequency. A separate pendab must therefore be installed for each harmonic
of blade-root shear to be attenuated.
6) Because pendabs must be mounted directly to the blade, they change ori-
entation with blade pitch angle. With this orientation change coupling (usually
adverse) invariably occurs with the blade in-plane motion.
A principal disadvantage is that they create aerodynamic drag and are thereby
somewhat detrimental to performance.
“chapter7(new)” — 2005/11/16 — page 253 — #15

Fig. 7.7 Installation of rotor bifilar absorbers on the S-76 helicopter. (Reproduced
courtesy of the Sikorsky Aircraft Corporation.)
Rotor bifilar absorbers. Another passive device used to attenuate rotor

loads before they interact with the fuselage dynamics is the (in-plane) bifilar
absorber. An example of a typical installation of rotor bifilar absorbers on an
operational helicopter is presented in Fig. 7.7.
These devices are similar to the bifilar absorbers used to quiet torsional vibra-
tions considered in an earlier chapter. In this case the bifilar absorbers are used
to absorb rotor in-plane hub shears. To attenuate blade passage frequencies in the
fixed coordinate system b, the bifilar absorbers must be tuned to (b ± 1) in the
rotating coordinate system. Thus, because for a four-bladed rotor the 4P in-plane
hub shears arise from both 3P and 5P rotating blade-root shears, both 3P and
5P bifilar absorbers might be used concurrently to attenuate the 4P nonrotating
hub shears. From the blade geometry given in Fig. 7.8, it can be shown that the
bifilar absorber, like the pendab, is tuned to a specific per rev frequency. The tuned
frequency for this device is given by

Rh
ωBA = (7.23)
L
where L = (d1 − d2 ) and d1 and d2 are the diameters of the bifilar hole and roller
bearing mandrel, respectively.
As is demonstrated in Chapter 6 (Section 6.2), the blade in-plane harmonics
that are +1 and −1 of the number of blades are the harmonics that combine to
yield in-plane hub shears that excite the fuselage at the blade passage frequency
(= number of blades × the rotor frequency). For this reason there are two separate
bifilar absorbers shown in Fig. 7.7, one tuned to three per rev and the other to five
per rev.
“chapter7(new)” — 2005/11/16 — page 254 — #16

Fig. 7.8 Basic geometry of a rotor hub appended bifilar dynamic absorber.
The advantages claimed for the bifilar absorber are as follows:

1) The device is relatively simple. It incorporates roller bearings with rea-
sonable endurance and low inherent damping and is hub mounted with no blade
attachment problems and no couplings caused by blade pitch angle.
2) The attendant drag caused by a blade-mounted device is reduced.
Its disadvantages are as follows: 1) the device is relatively heavy, and 2) the
device can only attenuate in-plane hub loads, and does so only indirectly. That is,
the bifilar absorber is typically tuned to the blade passage frequency (+) or (−)
the rotor frequency. This tuning condition is invoked in order to reduce frequency
components of the in-plane loads, which result in blade passage frequency dynamic
hub loads in the nonrotating coordinate system. Hence, some of the load-reducing
motion of the bifilar absorber is wasted on the harmonic component of in-plane
load that does not contribute to the blade passage frequency component of the
nonrotating coordinate system hub load.
Blade higher harmonic control. Blade higher harmonic control (HHC) is

distinguished from the previous devices in that it is an active device. HHC is a
device/system that is aimed at the cancellation of the excitations of the blade at
the source. This device depends on the use of an active power source as well as
a computer to adjust the operation of the several mechanisms required to impart
higher harmonic values of blade pitch angle (over and above that of the usual
1P cyclic variation imparted by the swash plate). With HHC the blade pitch angle
harmonic content consists of cosine and sine components each of the nP, (n + 1)P,
and (n − 1)P frequency contributions. This is typically achieved by hydraulic
actuation of the swash plate at frequencies of (n − 1)P, nP, and (n + 1)P, as
shown in Fig. 7.9.
“chapter7(new)” — 2005/11/16 — page 255 — #17

Fig. 7.9 Mechanical implementation of an HHC system.
The basic hardware components of an HHC system are 1) some form of instru-
mentation (e.g., accelerometers) to measure the vibratory load factors g at some
selected locations in the airframe; 2) an onboard dedicated computer (controller)
to interpret the values of the measured vibratory accelerations and to calculate the
appropriate values of the six components of higher harmonic pitch control angles;
and 3) a means, usually a set of hydraulic actuators, for implementing the values
of HHC angles determined by the computer. The calculation of the appropriate
(optimal) values of HHC within the computer can be achieved by a number of
existing optimal control methodologies. The calculation of the gradient matrices
for establishing the sensitivities of the components of vibratory g to each of the
six components of HHC pitch angles is typically achieved using some form of
parameter identification (e.g., Kalman filtering). These features are schematically
represented in Fig. 7.10.
Vibration measurement. In addition to measuring the total vibration at the

selected airframe locations, means must be provided for extracting the nP vibration
content, both amplitude and phase. This can be achieved in a number of ways using
both analog and digital approaches.
Controller operation. The controller has two primary functions that must be
implemented in the computer at a sufficiently high sampling rate (typically several
times per second) to ensure an accurate and rapid response to a changing vibration
environment. First it must perform a parameter estimation to establish the vector
of responses and the gradient matrix, which are needed for the use of the optimal
control methodology. Specifically, the following equation defines the quantities
needed for implementing the optimal control theory:
{Zi } = {Z0i } + [T ]{u} (7.24)
where {Zi } is the vector of measured vibrational quantities (to be minimized),

{Z0i } is the {Zi } vector without HHC controls, and [T ] is a matrix of influence
“chapter7(new)” — 2005/11/16 — page 256 — #18

Fig. 7.10 Schematic representation of a higher harmonic control system for vibration
reduction.
coefficients relating changes in {Zi } to the vector of HHC control angles {u}. Both
{Z0i } and [T ] are typically evaluated using Kalman-filtering techniques.
The second primary function that the controller must perform is that of calculat-
ing the optimal controls that minimize some cost function based on the measured
components of vibration. It must be stressed that the vibration environment of the
rotorcraft is subject to constant change caused by changes in the flight condition
and/or the initiation of maneuvers. The usual implementation of the HHC system
therefore involves the use of an adaptive control algorithm to achieve the optimal
control settings. The vibration level(s) are sampled at several times per second in
order to produce accurate updates of both the zero control vibration level {Z0i }
and the gradient matrix [T ]. These updates are then continually used to determine
the values of the HHC pitch angles needed for an optimal or minimal value of the
vibration-related cost function.
The advantages claimed for HHC are as follows:
1) The method is effective. Experimental implementations have verified the
ability to produce substantial reductions in vibration. The advantage of an active
vibration control system is that it does not require some residual value of vibration
in order to drive it.
2) The controller is generic in that it need not be sized to any one specific
aircraft.
3) The use of HHC has been shown to have negligible or even beneficial effects
with regard to both blade vibratory stresses and aerodynamic performance.
However, the balancing disadvantages with HHC are as follows:
1) One pays a weight penalty with the HHC system. The extra weight associated
with the computer, hydraulics, and associated servovalves, and the additional
airframe strengthening is not negligible and could be a decisive factor in some
designs for which payload productivity is a primary consideration.
“chapter7(new)” — 2005/11/16 — page 257 — #19

2) Compared with the passive systems and devices available for the attenuation
of fuselage vibrations, the HHC system is quite complex and, therefore, subject to
considerations of reliability and maintainability.
3) Despite the increasingly more rapid computational speeds available with
microcomputers, the calculations are not instantaneous. Consequently, for some
high-frequency band the HHC system would not track random or otherwise rapidly
changing vibration levels well.
With the increasing compactness and speed of computation and the commen-
surate decrease in weight of microcomputers HHC approaches should become
more competitive with respect to weight, reliability, and maintainability. Wood
et al. report on successful flight tests of an OH-6A helicopter implemented with
a prototype HHC system that showed significant reduction in vibrations without
undue penalties in blade loads or aircraft performance. Production installations
of HHC systems will hinge on the vibration benefits obtainable vs the additional
system weight and complexity.
7.4.2 Fuselage-Related Devices

Devices that are attached to the fuselage and applicable to reducing vibrations
directly in the fuselage are all subject to evaluation with respect to basic dynamic
considerations. In particular, all such devices must operate both in a steady-state
or static-load environment and in a vibratory one. The static-load condition creates
the specification that static deflections through the device must be less than some
prescribed value. The vibratory load condition creates the specification that the
device have a useful life with minimal wear in the bearings, etc.
Transmissibility. The yardstick by which the performance of the various

devices can be assessed is that of transmissibility T . Transmissibility is alternately
defined 1) as the ratio of the acceleration of some secondary substructure, which is
to be vibrationally quieted, to the base motion acceleration of the primary structure
to which the secondary substructure is attached, and 2) as the ratio of vibratory
response amplitude to steady (zero-frequency) amplitude. As an example, let us
consider the characteristics of a simple passive isolator.
Simple passive isolator. The simple isolator is depicted in the following

sketch:
“chapter7(new)” — 2005/11/16 — page 258 — #20

Fig. 7.11 General transmissibility characteristics of a simple passive isolator.
The transmissibility characteristics of an isolator with respect to frequency of base

motion are as given in Fig. 7.11, where r equals the frequency ratio ω/ωn .
From the transmissibility results of Fig. 7.11, it would√appear that operation
at frequency ratio as high above the frequency ratio of 2 as possible would
be the appropriate design strategy. Note that a high frequency ratio translates to
a low natural frequency ωn . However, if the isolator is to operate in a vertical
configuration, as shown before, then the steady deflection caused by gravity (and
load factor n) must be considered. The steady deflection δsteady can easily be related
to the natural frequency ωn as follows:
nW = Kδsteady = ngM
(7.25)
δsteady = gn/ωn2
Thus, the need to limit the steady deflection places a constraint on the natural
frequency that can be designed into any device that relies on frequency tuning for
vibration attenuation. With that caveat in place, we now consider a basic device that
again relies on frequency tuning but has properties that are especially attractive.
Dynamic absorber. The dynamic absorber is a device that is best suited

to those conditions wherein the excitation is caused by a direct application of
vibratory load as shown in Fig. 7.12.
The addition of an auxiliary mass ma that can respond to loads applied to the
parent mass M is a form of the two-degrees-of-freedom system considered in
Chapter 2. (Auxiliary masses for this purpose are often quite available in the fuse-
lage; one such mass often considered for this purpose is the aircraft battery.) With
the inclusion of viscous damping in the absorber mass attachment for generality,
the response equation (transmissibility defined in terms of displacement) for the
“chapter7(new)” — 2005/11/16 — page 259 — #21

Fig. 7.12 Schematic representation of a dynamic absorber.
parent mass is given by

1/2
y1 (2ζa r)2 + (r 2 − f 2 )2
T = =

ysteady (2ζa r)2 (r 2 − 1 + µr 2 )2 + [µf 2 r 2 − (r 2 − 1)(r 2 − f 2 )]2
(7.26)
where r is the ratio of the

√ forcing frequency ω to ω0 (the separate natural frequency
of the parent mass√= K/M); f is the ratio of ωa (the separate frequency of the
absorber mass = ka /ma ) to ω0 ; and µ is the ratio of the two masses (=ma /M).
In application the form of excitation can be in the form of either a direct sinusoidal
force F or a sinusoidally varying base motion y0 . In either case we wish to define a
steady deflection ysteady . In the case of base motion, ysteady is just the amplitude of
the sinusoidally varying motion y0 . In the case of an explicitly applied sinusoidal
force, ysteady is what it would be if the force amplitude F0 were applied through
the base motion:
F0
ysteady = (7.27)
K
The auxiliary mass can vibrate in response to the applied load F(t) so as to
significantly reduce the response of the parent mass to the point of actually nulling
the response in a condition referred to as antiresonance. Indeed, as Eq. (7.26)
shows, when there is zero damping, the numerator reduces to zero, mathematically
defining the antiresonance condition. Figure 7.13 presents the results of using
Eq. (7.26) for a representative value of the mass ratio.
“chapter7(new)” — 2005/11/16 — page 260 — #22

Fig. 7.13 General transmissibility characteristics of a dynamic absorber.
The results of Fig. 7.13 present a mixed blessing however. The most effective
configuration for creating antiresonance is clearly one with very little or no damp-
ing, ζa [=ca /(ca )crit ]. In this condition, however, instead of one high response
condition where the forcing frequency is equal to the natural frequency, we now
have two resonant responses straddling the original resonance condition. The addi-
tion of damping does reduce the two straddling resonances but at the price of
diminishing the effectiveness at the antiresonance point.
The figure also shows an interesting and significant property of the system.
The two points labeled A and B are known as the fixed points and are actually
independent of the level of damping. The frequency location values of these points
are determined solely by the ratio of individual frequencies f (=ωa /ω0 ) and the
mass ratio. Rao presents formulations showing that the abscissas (i.e., frequencies)
of these points ωA /ω0 and ωB /ω0 , respectively, are then given by the solutions rA
and rB of the following quadratic equation:

1 + f 2 + µf 2 2f 2
r 4 − 2r 2 + =0 (7.28)
2+µ 2+µ
Tuned absorber. Of even greater importance than the frequency locations are
the transmissibilities (ordinates) of the two fixed points. As shown in Fig. 7.13,
where the frequency ratio is unity, these transmissibilities are quite dissimi-
lar. However, a tuning condition exists wherein the transmissibility values of
both points can be made equal. This condition defines a unique configuration
denoted the TVA (tuned vibration absorber) and a unique relationship between
“chapter7(new)” — 2005/11/16 — page 261 — #23

the frequency ratio and the mass ratio:
1
f = (7.29)
1+µ
For the tuned condition represented by Eq. (7.28), the transmissibility value for
both the A and B points is given by

2
TA&B = 1 + (7.30)
µ
Optimum damping. Examination of Fig. 7.13 further shows that at each of the
two fixed points a proper selection of damping can be made such that the transmis-
sibilities can be minimized to the values given by Eq. (7.30). Thus, for the tuned
absorber optimum damping values can be found such that the transmissibilities are
bounded by those values defined for the tuned condition. Although the bounded
values cannot generally be met for each of the fixed points simultaneously for the
same value of damping, the average damping value proves to be quite practical for
design purposes:

3µ
(ζa )optimal = (7.31)
8(1 + µ)3
The resulting transmissibility characteristics of optimally tuning the µ = 0.2

example of Fig. 7.13, using Eqs. (7.28–7.30), and with optimum damping are
presented in Fig. 7.14.
Fig. 7.14 Optimally damped tuned dynamic absorber characteristics for mass
ratio of 0.2.
“chapter7(new)” — 2005/11/16 — page 262 — #24

Fig. 7.15 Optimum damping ratios as functions of mass ratio for the tuned dynamic
absorber.
The figure shows that the differences between damping optimized separately
for points A and B are not much different from each other or from the average
value given by Eq. (7.28). Figure 7.15 presents the variations with mass ratio of
the optimum damping ratios, as separately optimized for points A and B, and for
the average.
Response of the absorber mass. Two response conditions are of interest to

the designer of a dynamic absorber. First off, the deflection caused by a steady
load must not exceed some maximum value. For a vertically mounted absorber,
Eq. (7.25) equally applies to this
√ case. Whatever separate natural frequency is deter-
mined for the absorber ωn [= (ka /ma )] will result in a steady vertical deflection
(y2 )steady , as given by Eq. (7.25). The second response quantity of interest is that
which occurs in operation as it interacts with the motions of the parent mass M.
Because it is only the spring ka and the damper ca driving the absorber mass, the
ratio of absorber mass response to parent mass response can be easily derived. For
any excitation frequency ratio r, this response ratio is given by
1/2
y2 (2ζa r)2 + f 4
= (7.32)
y (2ζa r)2 + (r 2 − f 2 )2
1
The use of this equation is of most use in a design that is both tuned with
respect to f and incorporates optimal damping. Figure 7.16 presents for a range
of frequency ratios straddling the resonance point the amplitude response ratio of
that of the absorber mass to that of the parent mass for these conditions.
The use of the results of Fig. 7.16 requires first obtaining the tuned response
of the parent mass |y1 |, which requires only the use of Eq. (7.30) for the tuned
condition.
“chapter7(new)” — 2005/11/16 — page 263 — #25

Fig. 7.16 Variation of the amplitude response ratio for the absorber mass with mass
ratio and selected excitation frequency ratios—the optimally damped, tuned condition.
Auxiliary mass with only a damper attachment. A related alternative to the

use of a tuned dynamic absorber is the use of an auxiliary mass, as before, but with
only a (viscous) damper attachment to the parent mass. (This type of vibration
control device can clearly be used only for vibration control in axes not normally
having components of gravity.) The use of a damper-only supported auxiliary mass
would be warranted for those cases wherein a design for a tuned configuration is
impractical for reasons of expediency, where an oversized mass can be used and
where a significant vibration reduction is not demanded. Equation (7.26) can be
rewritten to conform to this case by reducing f , the frequency ratio to zero and by
representing the damping by a nondimensional damping coefficient ξa , now defined
as ca /(ma ω0 ). The resulting expression for transmissibility is then given by
1/2
y ξa2 + r 2

T = = 2 2 (7.33)
yst. ξa (r − 1 + µr 2 )2 + r 2 (r 2 − 1)2
Figure 7.17 presents the transmissibility characteristics for the damper-only sup-
ported auxiliary mass, as defined by Eq. (7.31), for zero, infinite, and optimum
values of damping. The results are for a mass ratio of 0.2 for comparison with the
results presented for the dynamic absorber.
Note that a fixed point B that is independent of damping is also in evidence. The
infinite damping curve represents the case where the auxiliary mass is “locked”
to the parent mass, and the zero damping curve represents the case of the parent
mass without any auxiliary mass whatsoever. The optimum value of damping is
determined by finding that value of damping that gives zero slope at the fixed point.
The optimum damping is determined solely by mass ratio:

ca 2
(ξa )optimum = = (7.34)
ma ω0 optimum (2 + µ)(1 + µ)
“chapter7(new)” — 2005/11/16 — page 264 — #26

Fig. 7.17 Transmissibility characteristics for a damper-only supported auxiliary

mass, similar to results of Fig. 7.13 (after Reed).
Similar to the results for the dynamic absorber, the transmissibility at the fixed
point can be determined:
2+µ
TB = (7.35)
µ
The transmissibility defined by Eq. (7.35) thus defines the upper bounded value
obtainable with the optimum damping given by Eq. (7.34). Figure 7.18 presents
a comparison of the transmissibilities obtainable with optimum damping for the
use of an auxiliary mass either as a tuned dynamic absorber and as a damper-only
supported mass.
The figure clearly shows the more advantageous transmissibilities available
with the dynamic absorber vis-à-vis the damper-only attached mass. The figure
does show, however, that with a sufficiently large auxiliary mass comparable
transmissibilities can be achieved with the damper-only configuration.
Nodamatic isolation system. Another passive method for isolating the

vibratory responses of the fuselage is the use of the so-called nodamatic suspen-
sion system, whereby the isolation is achieved, first by mounting the high vibration
excitator, the rotor, and transmission on a beam and, second, by attaching this beam
to the fuselage at the node points of the beam, as shown in Fig. 7.19.
The nodamatic system is based on the principle that the reaction loads of a
beam at its node points (i.e., locations on the beam where, for a bending mode
whose frequency is close to the excitation frequency, there is zero beam motion)
are substantially reduced. This feature has the potential for giving load attenuation
over a range of frequencies. Although the nodamatic system is quite simple and
reasonably effective, it has the primary disadvantage of being relatively heavy.
“chapter7(new)” — 2005/11/16 — page 265 — #27

Fig. 7.18 Comparison of transmissibilities achievable with optimum damping of the

dynamic absorber and the damper-only attached auxiliary mass.
Dynamic antiresonant vibration isolator. The dynamic antiresonant vibra-

tion isolator (DAVI) would be used in situations wherein a simple isolator would
also be used. Figure 7.20 presents a schematic representation of the features of the
DAVI.
The transmissibility of the DAVI is given by the following equation:
!
ÿ1 K − ω2 I/r 2 + m2 (R/r) (R/r − 1)
T= = ! (7.36)
ÿ3 K − ω2 m1 + I/r 2 + m2 (R/r − 1)2
Fig. 7.19 Schematic representation of a nodamatic isolation system.
“chapter7(new)” — 2005/11/16 — page 266 — #28

Fig. 7.20 Schematic representation of kinematics of the dynamic antiresonant

vibration isolator.
The antiresonant frequency ωA is given by the following expression:
ωA = [K/I/r 2 + m2 R/r(R/r − 1)]1/2 (7.37)
where I is the mass moment of inertia of the arm containing the m2 element about
its mass center. Figure 7.21 contrasts the transmissibilities of the basic simple
isolator and of the DAVI for the same installation for which a static load of 100 lbf
produces a static deflection of 0.11 in. Note that for the set of parameters used in
this comparison the spring stiffness K is the same for both isolators. Consequently,
the resonant frequency for the DAVI is always lower than that for the simple
Fig. 7.21 Comparison of transmissibilities for simple and dynamic antiresonant

vibration absorbers.
“chapter7(new)” — 2005/11/16 — page 267 — #29

isolator. There then exists a range of frequencies, starting from some point between
the two resonant frequencies to a point above the DAVI antiresonant frequency,
wherein the DAVI will have a lower transmissibility than that of the simple isolator.
However, for sufficiently high frequencies the simple isolator will have superior
performance. Finally, for some configurations the tuned antiresonance condition
of the DAVI can be so narrowly frequency-banded as to be unusable, especially for
those cases wherein the excitation frequency is subject to relatively wide fluctua-
tions. A graphic demonstration of the ability of the DAVI to suppress vibration is
given in Figs. 7.22 and 7.23.
Note in these figures that the DAVI units (at the four corners of the platform)
exhibit significant motion in the arm parts. Although the lower base of the platform,
as well as the DAVI arms, register as blurs in the photographs, the upper (isolated)
platform is quite smooth.
IRIS rotor isolation system. The improved rotor isolation system (IRIS) is
a passive system developed by the Boeing Helicopter Company. This system is
shown schematically in Fig. 7.24 and is seen to be somewhat a combination of the
nodamatic and the DAVI systems. It would appear to possess the antiresonance
characteristics of the DAVI without the need for low damping bearings, but it would
also appear to share the relatively high weight penalty disadvantage associated with
the nodamatic system.
Fig. 7.22 DAVI platform installation showing antiresonant isolation. (Reproduced

courtesy of Kaman Aerospace Corporation.)
“chapter7(new)” — 2005/11/16 — page 268 — #30

Fig. 7.23 DAVI pilot seat isolation laboratory test, ±1.0 g at 10 Hz. (Reproduced
courtesy of Kaman Aerospace Corporation.)
Fig. 7.24 Schematic of the IRIS vibration-alleviation device.
“chapter7(new)” — 2005/11/16 — page 269 — #31

Fig. 7.25 Schematic of a focused pylon vibration-alleviation device.
Focused pylon. Whereas most of the devices discussed earlier are directed
at eliminating vibratory shears, the focused pylon device is directed at eliminating
vibratory moments. As shown in Fig. 7.25, the device consists principally of a
means of supporting an inertia with nonparallel linkages so that its motion is
constrained to occur about the point at which the support linkages are “focused.”
Usually, the focal point is taken to be the center of gravity of this inertia so that the
inertia will be forced to act as a “rotational” isolator. As with the simple passive
(linear motion) isolator, the focused pylon will function best with a minimum of
damping in the system.
References
Section
7.1 Nelson, R. B., “Simplified Calculation of Eigenvector Derivatives,”
AIAA Journal, Vol. 14, No. 9, 1976, pp. 1201–1205.
7.3 Flannelly, W. G., Fabunmi, J. A., and Nagy, E. J., “Analytical Testing,”
NASA CR-3429, 1981.
Jetmundsen, B., Bielawa, R. L., and Flannelly, W. G., “General-
ized Frequency Domain Substructure Synthesis,” Journal of the
American Helicopter Society, Vol. 33, No. 1, 1988, pp. 55–64.
Loewy, R. G., “Helicopter Vibrations: A Technological Perspective
(The AHS Alexander A. Nikolsky Honorary Lecture),” Journal of
the American Helicopter Society, Vol. 29, No. 4, 1984, pp. 4–30.
7.4 Desjardin, R. A., and Hooper, W. E., “Antiresonant Rotor Isolation for
Vibration Reduction,” Journal of the American Helicopter Society,
Vol. 25, No. 3, 1980, pp. 46–55.
Flannelly, W. G., “The Dynamic Antiresonant Vibration Isolator,”
Proceedings of the 22nd Annual National Forum of the American
Helicopter Society, May 1966.
“chapter7(new)” — 2005/11/16 — page 270 — #32

Paul, W. F., “Development and Evaluation of the Main Rotor Bifilar
Absorber,” Proceedings of the 25th Annual National Forum of the
American Helicopter Society, May 1969.
Taylor, R. B., and Teare, P. A., “Helicopter Vibration Reduction with
Pendulum Absorbers,” Journal of the American Helicopter Society,
Vol. 20, No. 3, 1975, pp. 9–17.
Halwes, D. R., “LIVE—Liquid Inertia Vibration Eliminator,” Proceed-
ings of the 36th Annual National Forum of the American Helicopter
Society, Paper 80-22, May 1980.
Reed, F. E., “Dynamic Vibration Absorbers and Auxiliary Mass
Dampers,” Shock and Vibration Handbook, 3rd ed., edited by
C. M. Harris, McGraw–Hill, New York, 1988, Chap. 6.
Rao, S. S., Mechanical Vibrations, Addison–Wesley, Reading, MA,
1995.
Wood, E. R., Powers, R. W., Cline, J. H., and Hammond, C. E., “On
Developing and Flight Testing a Higher Harmonic Control System,”
Journal of the American Helicopter Society, Vol. 30, No. 1, 1985,
pp. 3–20.
Problems
7.1 For a certain dynamic multiple-degrees-of-freedom (zero-damped) system
the mass and stiffness matrices are given to be the following functions of a
parameter x:
 
10 + 2x x−1 3
 x−1 x2 + 5 2 
[M] =  
√
3 2 10 5x
 
10x 3 x−5
 
[K] =  3 4x 2 x 2 − 21
x−5 x 2 − 21 2.4x 3
For a value of x = 5, the eigenvalue solution is as follows:
Natural frequencies:
[ω1 , ω2 , ω3 ] = [1.5511, 1.8792, 2.4688]
“chapter7(new)” — 2005/11/16 — page 271 — #33

Mode shapes:
 
1.0000 −0.5287 −0.2436
(1) (2) (3) 
[φ , φ , φ ] = 0.2393 1.0000 −0.0360 
0.0412 −0.0206 1.0000
Calculate (using analytical means) the derivatives of the natural frequencies

and mode shapes with respect to the parameter x.
7.2 Verify (using numerical means) the derivatives found in Problem 7.1. [Note
that access to a reliable eigenvalue solution (computational software) is
required for this problem.]
7.3 Using the basic analytical testing equation, we might wish to specify a
new (required) response vector {qFR }, thereby defining a specific change
in response; the problem is then to solve for the mobility (or inversely
impedance) at the selected interface location that will achieve the required
response change. Show that a solution for this required interface mobility
change is expressible as
YIIC = −{Q2 }{Q1 }# (7.38)
where
{Q1 } = [YFI ]−1 {{qFR } − {qF0 }} (7.39a)
and
{Q2 } = {qI0 } + [YII ]{Q1 } (7.39b)
and [ ]# ≡ pseudoinverse of a matrix:
Right pseudoinverse:
!−1
[B]# = [B]T [B][B]T (7.40a)
Left pseudoinverse:
!−1
[C]# = [C]T [C] [C]T (7.40b)
7.4 Derive (linear) simplified equations of motion governing the pendular motion
(βP ) of a rotor-appended pendular vibration absorber (pendab) attached to a
rotor blade whose flatwise bending flexibility is described by a single flatwise
bending mode generalized coordinate (qw1 ). The following assumptions can
be made:
“chapter7(new)” — 2005/11/16 — page 272 — #34

1) Let the blade out-of-plane bending w(x, t) be expressed with a single

modal response description Rγw1 (x)qw1 (t), wherein the modal variables are
taken without the pendab. Also, let the blade have infinite in-plane stiffness.
Hence, the modal equation with the addition of the pendab can then be
expressed in the following simple form:
Mw1 (q̈w1 + 2ζw1 ωw1 q̇w1 + ωw2 1 )

= (aero)
w1 +[ w1 ( Fw ) + w1 ( CF)]due to pendab (7.41)
(aero)
where w1 is the generalized modal excitation caused by the applied (har-
(a)
monic) airload distribution (pz ). The effects of motion-induced airloads
are assumed to be contained within the damping ratio ζw1 . Also, Fw and
CF are, respectively, the incremental force in the flatwise direction and
the incremental centrifugal force (caused by the pendab). Note that Fw is
taken to be a concentrated force at the hinge location (x = Rh ) and CF is
taken to be an incremental tension existing between the blade root and the
hinge location.
2) The pendab has a hinge located at a distance Rh from the center of
rotation and consists of a concentrated mass mP , connected to the hinge by
a massless rod of length r. Let the response of the pendab be represented
by the deflection angle βP relative to a (inertially) horizontal plane, and let
the dynamic description be governed by that of a simple spring-mass system
whose excitation comes from the base motion of the hinge location resulting
from the blade first flatwise modal bending response.
3) (a) Derive an expression for the hub shear load (per blade) in Problem
7.4 caused by the applied airload p(a)
z and the accelerations of both the blade
in bending q̈w1 and the pendab response angle β̈P .
(b) Show (using the results of part a) that, if it is assumed that the flatwise
responses change negligibly because of the effects of the pendab, then the
vertical hub shear per blade H can be expressed as
H = Hwithout pendab + mP rωP2 βP

(c) Using the results from parts a and b, develop a mathematical expla-
nation to show why, for an inboard installation of the pendab (where there
is typically a negative antinodal condition with γw1 ), the pendab reduces the
hub shear per blade for the case wherein the pendab natural frequency is
less than the excitation frequency.
4) Derive an expression for the mobility of a simple isolator (spring-
mass-damper) for use as a modification mobility (YIIC ).
5) Derive an expression for the mobility of a dynamic absorber (as per
Fig. 7.12) for use as a modification mobility (YIIC ).
6) An installation of a cockpit panel is experiencing unacceptable vibra-
tions in a four-bladed helicopter (R = 650 ft/s, R = 28 ft). Measurements
of the panel vibration level show it to have an amplitude of 0.5 in. It is also
learned that the panel is effectively cantilever mounted from a point in the
airframe whose vibration level is 0.049 in. The panel can be considered to
“chapter7(new)” — 2005/11/16 — page 273 — #35

be a point mass with an effective mass of 1.86 lb-s2 /ft. It is proposed to use
a 30-lb battery as a dynamic absorber to reduce the panel vibrations.
(a) What is the best vibration level in the panel that can be achieved with
this configuration?
(b) What attachment spring and damper rates are required to tune the
absorber mass (the battery) to achieve this level of attenuation?
(c) What is the maximum steady deflection of the absorber for a maximum
load factor n of 2.5?
(d) What is the maximum vibratory amplitude the absorber will
experience?
“chapter7(new)” — 2005/11/16 — page 274 — #36

8
Vibration Test Procedures
As the previous chapters have stressed, the ability to analyze rotorcraft for
vibration requires a good knowledge of the dynamic characteristics of the fuselage.
These characteristics generally include, as a minimum, the natural frequencies of
the airframe in the general frequency range of the blade passage frequency. Of
increasing importance are the mode shapes, the structural damping descriptors,
and the various mobility matrices [i.e., frequency-response functions (FRFs)] in
the frequency range of the fundamental blade passage frequency. Until such time
as the dynamicist has the tools (and the skill for using them) for accurately calcu-
lating these dynamic characteristics, some form of vibration testing will remain as
part of the development of new rotorcraft configurations. An adjunct to the direct
vibration testing of the airframe and airframe components is the testing for compo-
nent fatigue life. Just as in the case of natural frequency and mode prediction, the
prediction of component fatigue life is an inexact science that requires some form
of experimental verification. These two related types of vibration testing are sim-
ilar to the extent that they deal with structures operating in an environment that is
almost wholly in the frequency domain; therefore, some of the techniques overlap.
8.1 Basic Shake Testing

Present analytical methods for predicting fuselage vibration characteristics rely
heavily on finite element modeling techniques. As such, the structural details
of the structure are assumed to be idealized into “effective” linearized masses
and stiffnesses. In real life, however, things are more complex and are often not
all that overwhelmingly linear. Hence, present analytical techniques generally
give approximations that require validation by test. Furthermore, for those cases
wherein even the most rudimentary analytical predictions (natural frequencies and
mode shapes) are unavailable, some form of actual test data is the only other
source of this information. Shake testing is the appropriate form of testing that
fulfills these requirements. With the emergence of modern instrumentation for
conducting vibration testing, the measurable item of choice continues to be the
frequency response function (FRF). The market is well serviced with a variety of
equipment for making this measurement.
As shown in Fig. 8.1, shake testing basically consists of 1) shaking the structure
at discrete points using some form of variable-frequency electromagnetic or servo
driven hydraulic shakers; 2) measuring both the excitations and responses of the
structure (generally with strain-gauge-instrumented load cells and a multiplicity of
275
“chapter_8(new)” — 2005/10/20 — page 275 — #1

Fig. 8.1 Elements of a basic single-point shake test.
accelerometers); and 3) systematically analyzing the measured responses to obtain

the required engineering data, as identified earlier.
The simplest form of vibration testing is to excite the structure in a system-
atic fashion to obtain peak responses that are typically indicative of resonance
conditions and thereby identifiable with the natural frequencies of the structure.
Although in concept this procedure is relatively straightforward, problems arise
with regard to the most efficient setup of the equipment, its actual usage, the meth-
ods for reducing the data, and, perhaps most importantly, interpretation of the
results.
The principal vibration problems of rotorcraft are aerodynamically related, and
vibration is clearly an issue only in flight and, in particular, forward flight in
transition and at high speed. Thus, it would be highly desirable to conduct shake
tests in an airborne configuration. Such a procedure is not generally practical for
a number of reasons:
1) Any form of flight testing is time consuming and expensive.
2) Shake testing typically involves the issue of rotor-fuselage coupling, which
requires measurements of forces and responses at the rotor hub, quantities that are
very difficult to measure in flight.
3) The information that shake testing is to provide is generally required prior
to actual flight; thus, delaying shake testing until it can be done in flight is
counterproductive.
Hence, shake tests are conducted on the ground in a manner that simulates
as closely as possible the pertinent dynamics of the airborne configuration. The
“chapter_8(new)” — 2005/10/20 — page 276 — #2

VIBRATION TEST PROCEDURES 277
following sections of this chapter define and describe the various basic parts of
any shake test program.
Figure 8.1 presents the simplest form of shake testing: single-point excitation.
The application of force at only one point at a time does not, however, rule out
measuring responses at several locations. For rotary-wing aircraft, the use of single-
point excitation is quite appropriate in most cases. Single excitation at the rotor
hub closely simulates actual rotorcraft environment in that the rotor is the principal
source of excitation of the fuselage. This type of testing has the principal advantage
that it is a minimum effort approach. Single-point excitation does have somewhat
of a disadvantage when it comes to measuring modal characteristics in that this
type of testing inherently measures response amplitudes rather than mode shapes.
There are two principal reasons for this discrepancy. First, the excitation of the
structure at a given point will excite any mode that has a natural frequency close
to the excitation frequency, and so there will always be an admixture of modal
responses throughout the range of test frequencies. The second reason is that the
test structure will always have a degree of internal damping. This damping is
the mechanism whereby the structure dissipates the vibrational energy that the
shaker is imparting to it. This dissipation will distort the measurement of modal
characteristics because the points nearest to the excitation can be expected to
absorb the most energy. Thus, if the measurement of mode shapes is the desired
objective, there will be a limit to the accuracy that single-point testing can provide.
The alternative to single-point excitation is that of multipoint excitation wherein
a multiplicity of shakers are simultaneously driven to make the structure respond
in the desired modes. The full description of multipoint excitation systems is
beyond the scope of this text, but a description of the basic approach is given in a
subsequent section.
8.1.1 Test Setup

Fuselage configuration. For obvious reasons shake testing cannot be per-
formed with (at least) the main rotor rotating. Because the nonrotating charac-
teristics of the rotor blades are completely different from those at the operational
rotor speed, it is inappropriate to shake test with the rotor blades on the rotorcraft.
Furthermore, because the fuselage must be suspended to give as close an approx-
imation to the free-flight condition, it must be suspended by some suspension
device not generally used in flight. Typically a special-purpose hub replacement
fixture, which can attach conveniently to that device, must be fabricated. This hub
replacement fixture (or modified hub) must enable the fuselage to be elevated off
its landing gear and, furthermore, must account for the mass of the rotor blades
that are removed for the shake test. The structure to be shake tested should be
equivalent to the 100% weight of all of the actual items excluding the blades.
However, the equivalent blade weight to be attached to the hub replacement fix-
ture should be reduced from the full weight (a typical value is 60%). The purpose of
reducing the equivalent weight of the rotor blades is to provide the fuselage struc-
ture with an equivalent rotor impedance somewhat closer to that of the rotating
flexible blades than a concentrated mass equal to 100% of the rotor mass would
“chapter_8(new)” — 2005/10/20 — page 277 — #3

provide. In this manner the shake-tested fuselage will respond with resonances
and antiresonances reasonably close to those that the free-flying aircraft would
manifest.
Fuselage suspension. As discussed earlier, in order to have any relevancy,

the appropriate dynamics of the airborne configuration must be approximated to
a reasonable degree. Furthermore, because the nonrotating characteristics of the
rotor blades are completely different from those at the operational rotor speed, it
is inappropriate to shake test with the rotor blades on the rotorcraft. Thus, for such
testing the one basic practical method for making the free-flight approximation
that has evolved is to suspend the fuselage (from the hub) in a manner that results
in the required low-frequency configuration. As is described first, this has been
accomplished for many years using a system of bungee cords that when greatly
extended typically have a vanishingly small extensible stiffness.
Bungee-cord configurations physically consist of many coiled loops of cording
over two mandrels, each of which is attached to one of the two tension load points
(the fuselage hub and an elevated suspension hard point). Such configurations are
essentially very soft nonlinear springs. They tend to be bulky, entail large elastic
displacements, are subject to progressive failure (and, hence, degradation in load-
carrying capability), and require the equalizing of the loads in the many loops
before they can be used. However, one principal advantage of bungee configu-
rations is that they are relatively inexpensive and completely passive. Figure 8.2
presents a schematic arrangement of a typical bungee suspension system.
In recent years air-spring, or air-bag, devices have been finding application
to the shake-testing setup because of the convenience of being able to control the
spring rate and centering position of these devices. Figure 8.3 presents a schematic
of a typical installation of an air-bag-based suspension system. As the figure makes
quite clear, the use of air bags is most important for excitation test conditions
entailing significant vertical motion of the hub. If, for example, in the in-plane
configuration shown in Fig. 8.3, a fore and aft excitation excites a fuselage vertical
bending mode, the cable stiffness could significantly affect the response.
A comparison of Figs. 8.2 and 8.3 shows the air-bag configuration to be some-
what “cleaner” and more compact. However, special care must be taken with an
air-bag-type configuration in that the dynamic characteristics of an air bag are more
involved than those of the simple passive dynamics of the bungee-type support.
Consequently, appropriate modeling of this type of support becomes an issue in
attempting correlation with a finite element analysis of the configuration. An exam-
ple of a link chain supported helicopter fuselage wherein an air-spring suspension
system was employed is shown in Fig. 8.4.
General guidelines to be followed in the configuring of the fuselage suspension
system are as follows:
1) The frequencies of the suspended fuselage as a result of the spring rate of
the suspension device should be less than and well removed from the first natural
elastic mode frequency of the fuselage. Values within the range of 1–2 Hz, or half
the first elastic natural frequency, are generally acceptable.
2) The linkage between the suspension device and the airframe should be long
enough to provide a pendular frequency that is sufficiently removed from the rotor
operating frequency. Again, frequencies in the range of 1–2 Hz are acceptable.
“chapter_8(new)” — 2005/10/20 — page 278 — #4

Fig. 8.2 Schematic of a typical bungee fuselage suspension system. (Reproduced

courtesy of Sikorsky Aircraft Division, UTC.)
Fig. 8.3 Alternate operational modes of fuselage suspension for shake testing using
an air-bag configuration. (Reproduced courtesy of Bell Helicopter Textron, Inc.)
“chapter_8(new)” — 2005/10/20 — page 279 — #5

Fig. 8.4 Typical helicopter fuselage shake test setup using an air-spring suspension
system. (Reproduced courtesy of McDonnell Douglas Helicopter Company.)
3) The gantry structure to which the suspension system is attached should be

well isolated from the test airframe. Here the use of low-frequency air-spring
supports provide minimal feedback from the gantry modes of vibration into the
test airframe.
4) The suspension system should have a minimal amount of damping. This is
especially important in the blade passage test frequency range of interest. However,
this requirement is difficult to control with either type of passive suspension system.
5) The soft part of the suspension system should be placed as close as possible
to the overhead hoist to minimize excitation of the suspension system’s lateral
modes.
6) The part of the suspension system that is attached to and moves with the hub
must be considered in any FEM correlation study.
Excitation equipment (oscillatory). The primary method for exciting the

structure is with the use of any of a wide variety of commercially available electro-
magnetic and/or servocontrolled hydraulic actuators configured as shakers. Most
of these commercial units have dedicated power supplies that are driven by a user-
supplied controlled vibratory signal (voltage). More will be said about the required
driving vibratory signal in a later section.
“chapter_8(new)” — 2005/10/20 — page 280 — #6

The principal excitation quantities are the five components of load at the hub
[the three components (x, y, and z) of vibrational force, and the roll and pitch
components of vibrational moment] and at selected fuselage locations that are
either locations of secondary sources of vibration (e.g., the tail rotor hub) and/or
are to be considered to be interface locations for potential vibration suppression
modifications. Thus, the linkage between the actuator and the hub should contain
three principal elements: 1) where appropriate, an elastic attachment element such
as a coil spring (or torque tube for moments) to convert relatively large actuator
strokes into controlled forces (or moments); 2) a load cell fixture located close to the
airframe to measure the applied load; and 3) a drive rod or appropriate appliance to
connect the elements together. The actuator system should be configured for high
performance by the selection of a sufficiently large, high performance servovalve
for the hydraulic systems. The shakers should be capable of adequate excitation at
frequencies somewhat in excess of twice the maximum frequency of interest in the
fuselage structure. For main rotor hub excitations this frequency is approximately
2 × 25 or 50 Hz. For tail-rotor excitation purposes this frequency can run about
four times greater.
Excitation equipment (impulsive). One method of vibration excitation

available for general structural dynamics testing is the instrumented hammer for
impulsive load excitation. This system is based on the incorporation of an accu-
rate load cell in the head of a hammer so that a precise measure of the impulsive
energy imparted to the structure can be measured. Although this type of excitation
is widely used at present, it is generally not appropriate for testing of helicopter
airframes because it is often difficult to impart enough energy into a full-scale
airframe and still not damage the structure locally. An alternative to an impulsive
positive force is an impulsive release of an imposed steady load. This form of
excitation would be appropriate only for grounded structures, such as a test fix-
ture or grounded fuselage. One problem with such an excitation is that there is
no measurable load from which a frequency response function can be determined.
Nonetheless, it still provides a relatively inexpensive way of exciting a system to
determine at least its natural frequencies and equivalent damping levels.
8.1.2 Instrumentation
Basic measurement devices. The two types of basic measuring devices
needed for shake testing are load cells and accelerometers. The load cells are
part of the excitation equipment as discussed earlier and are used not only as
part of a mobility (transfer function) measurement, but also to ensure that suffi-
cient energy is imparted into the structure at all frequencies tested. With each of
these devices (accelerometers and load cells), measurement is made using either
of two physical principles. One principle is typically to use a flexible member
that is instrumented for elastic deflection: Either a conventional strain-gauge or a
piezoresistive device is used, which deflects in response to an applied or inertial
load. The second principle is to use the strain of a suitable crystalline material
directly (piezoelectric devices). Strain-gauged units have the advantages of being
relatively inexpensive and “tried and true.” They also have the advantage of being
able to measure both steady and vibratory loads or accelerations. Piezoelectric and
“chapter_8(new)” — 2005/10/20 — page 281 — #7

piezoresistive devices have the advantages of having lower compliance character-

istics than those of strain-gauged devices, having excellent sensitivity, and being
generally more rugged. Whereas piezoelectric devices have relatively poor steady
measuring capabilities, they have the advantage of compactness and the ability
to be configured for measuring multiple components of load and/or acceleration
with a single device. Although piezoresistive devices do not generally have the
multiple-axis capabilities of piezoelectric devices, they do have the advantage of
being usable in much the same fashion (circuitry) as strain-gauge devices and of
being able to measure both steady and vibratory components.
Advanced measuring devices. Whereas the required load cells are rel-
atively few in number and are relegated for use in the excitation system, the
number of accelerometers used in a shake test can be quite high (often more
than 100). The number of accelerometers is dictated by considerations of being
able to identify responses at key locations in the structure such as the pilot seats,
engines, and hubs; being able to describe fundamental modes; and being able to
identify local modes relating to specific components such as wings and pylons.
Another reason for the large number of required accelerometers is the fact that
most accelerometers typically measure only linear motion, and suitable rotational
motion accelerometers are relatively new and hence somewhat expensive. Thus,
rotational measurements are usually obtained indirectly using multiple accelerom-
eters appropriately positioned near each other. Another advancement currently
being made in vibrational measurements is the laser/fiber-optic vibrometer. Such
vibrometers operate by illuminating the vibrating structure with laser light and
using optical phenomena (Doppler frequencies and interference fringes) to mea-
sure vibrational amplitudes of both displacements and velocities. Their advantages
are that they can measure with significantly improved sensitivities and accuracies
relative to accelerometers, they do not need repeated calibration, and they can
measure the vibratory responses at any location on the structure that can be illu-
minated with the laser beams. Additionally, they can measure either in an absolute
motion mode or in a differential motion mode (using two optical sources). The
disadvantages of the new devices are that they are presently expensive, they are
not as rugged as comparable accelerometer-based measurement systems, and they
cannot directly measure load.
Transducer calibration. All quality transducers are sold with factory-tested

calibration factors, in terms of volts/measured quantity, or an equivalent. In par-
ticular, if the calibration factors of accelerometers and load cells are, respectively,
given as Sẍ and Sf , then the voltage outputs would be expressed as
Vẍ = Sẍ ẍ (8.1)

Vf = Sf f (8.2)
Acceleration and force will usually constitute the two quantities measured, as
shown in Fig. 8.1. Because the principal output from the shake test will be the
determination of frequency response functions [i.e., motion (acceleration)/force],
the voltage from the accelerometer will ultimately be divided by the voltage of
“chapter_8(new)” — 2005/10/20 — page 282 — #8

the force transducer. One practical way to ensure the proper calibration of the
transducers is to use a calibration mass, as shown in Fig. 8.5.
The idea is to excite the mass with the shaker and measure the voltages from
the two transducers. Because the mass is known (by weighing), the following
expression can be formed:

ẍ −1 Sẍ Vẍ Vẍ
=M = =S (8.3)
f Sf Vf Vf
If the voltages are divided before the calibration factors are employed, then the
preceding equation provides a way of evaluating the all-over calibration factor
S. If the calibration factors are used before the division, then either of the sepa-
rate calibration factors can be calibrated relative to the other using either of the
following expressions:

Vẍ
Sẍ = MSf (8.4)
Vf
or

Vf Sẍ
Sf = (8.5)
Vẍ M
where, in both of these equations, everything to the right of the equal sign is either
known or assumed to be correct.
Data acquisition. Data of the type that are typically acquired during shake
tests can be stored on magnetic tape (either in FM, direct analog, or digital forms)
and/or some similar form of high-density data storage device. State-of-the-art
Fig. 8.5 Calibration mass setup (after Ewins).
“chapter_8(new)” — 2005/10/20 — page 283 — #9

data acquisition systems typically provide for computer control of the test oper-
ation, data acquisition, and often some sort of online data reduction. A variety
of approaches is presently employed and is to a large extent driven by the ever-
increasing availability (and affordability) of computer-based equipment on the
market. Some of the major features of such equipment are described in the next
section.
8.1.3 Dynamic Signal Analyzers

Contemporary vibration testing is increasingly being accomplished using digi-
tally based equipment that is denoted either dynamic signal analyzers or spectrum
analyzers. Such equipment is typically precision-manufactured units embody-
ing state-of-the-art technologies in electronics and microcomputers, as well as
in advances in digital signal processing. Because of features of the signal pro-
cessing algorithms “built into” these analyzers, they are typically configured to
accept two channels of analog instrumentation signals, as shown in Fig. 8.1. These
analyzers generally have several physical characteristics in common:
1) Each of the channel inputs to dynamic signal analyzer are in the form of
analog voltages. However, once these inputs enter the analyzer, all subsequent
operations are performed digitally. To avoid aliasing or a folding over of the data
back onto itself, analog antialiasing (low-pass) filters are employed before the
input signals are converted to digital form. A variety of standard filters is available
and they are typically implemented with active operational amplifiers. Two such
filters that see a lot of usage are the Butterworth and Bessel filters, both of which
have smooth roll-off characteristics. Each of these filter types is defined in terms of
“families” that share the same basic characteristics, but depending on the number
of poles forming the denominators of the transfer functions n have different roll-off
rates, as shown in Figs. 8.6a and 8.6b. The frequencies in the figures are normalized
by ω0 , a user selectable frequency.
Note that all of the curves for the Butterworth filter pass through the same point
√
(= 0.5) for unit normalized frequency. The curves for the Bessel √ filter, however,
have been adjusted so that they all pass through the same 0.5 point for a unit
normalized frequency, for comparison. It is clear that higher-order filters produce a
sharper cutoff. This positive aspect must be balanced by the additional noise errors
introduced by the operational amplifiers (i.e., an operational amplifier is required
for every two poles in the denominator). Also, increased phase shift is introduced
by the each pole. It is thus imperative that the same type of filter must be used on
all channels of input to the analyzer. Although not universal, most models have
built-in antialiasing filters in the signal flow prior to the A/D conversion. Huelsman
et al. present details of the Butterworth and Bessel filters, along with details for
other popular filter types.
2) Each of at least two channels of analog data are sampled, usingA/D circuitry,
at as high a sampling rate as possible. Rates as high as several hundred kilohertz
are typical. Also, provision is made for mass storage of the digitized data in some
form of dynamic RAM device.
3) A microprocessor is incorporated for dedicated calculation of specific dig-
ital signal processing chores. Perhaps the most important of these is a fast-
Fourier-transform (FFT) calculation algorithm. (See Sec. 2.2.3: note that inverse
“chapter_8(new)” — 2005/10/20 — page 284 — #10

Fig. 8.6 Frequency response characteristics of varying orders of Butterworth and

Bessel low-pass filters.
“chapter_8(new)” — 2005/10/20 — page 285 — #11

Fig. 8.7 Typical block diagram of a dual channel dynamic signal analyzer, spectrum
averaging mode.
FFTs are digitally calculated with essentially the same algorithm and are thus
available with the same speed and accuracy characteristics.) This is important with
regard to all of the digital signaling processing required for modern frequency anal-
ysis. High-speed implementations of FFTs on microprocessors with architecture
specialized for digital signal processing (DSP), and FFT computations in partic-
ular, are now fairly standard, and later we shall see that this algorithm forms the
heart of the analysis and processing capability of these analyzers.
4) Means are provided for producing a variety of specialized excitation signals
for driving the shakers. Thus, in addition to a straightforward sweep of stationary
frequencies with incrementally indexed frequencies, a variety of alternate non-
stationary excitations are provided; these are discussed in greater detail in a later
section.
5) Another feature present in many models is built-in “firmware” relating to
modal analysis, wherein typical modal parameters such as natural frequency, mode
shape, and damping are calculated (e.g., see Sec. 2.4).
Figure 8.7 summarizes the basic functions of the spectrum analyzer.
8.1.4 Calculation of Frequency-Response Functions

The operations shown in Fig. 8.7 represent a systematic way of identifying
the linear behavior existing between the two input channels. The mathematical
measure of this linearity is the transfer function or frequency-response function
(FRF) of channel B relative to channel A. In the context of vibration testing, this
“chapter_8(new)” — 2005/10/20 — page 286 — #12

frequency-response function comprises one element of the mobility description

of the structure. The operations have been developed in the manner indicated to
maximize the utilization of averaging techniques and thereby filter out noise and
enhance accuracy. As shown in Fig. 8.7, this operation begins with the digitiza-
tion of N record lengths (of selected size) of time-history data (each consisting
of data at J discrete values of time) for each channel [a(tj )n and b(tj )n , where
j = 1, 2, . . . , J and n = 1, 2, . . . , N]. These record lengths are then immediately
Fourier transformed, as denoted by the F symbol, to produce the respective instan-
taneous spectra A(fk )n and B( fk )n (again, where k = 1, 2, . . . , J). These 2N spectra
are typically stored in the analyzer’s RAM units in preparation of the averaging
function.
Autospectra and Cross spectra. The averaging function is achieved by

forming three fundamental mathematical quantities: the autospectra, or averaged
power quantities, SAA ( f ), SBB ( f ) and the two cross spectra SAB ( f ) and SBA ( f ).
These spectral quantities are formed by averaging, over the record lengths and at
each frequency, the appropriate products of the Fourier-transformed time records,
that is, the instantaneous spectra A( fk )n and B( fk )n :
1
N
SAA ( fk ) = A( fk )n × A∗ ( fk )n (8.6)
N
n=1
1
N
SBB ( fk ) = B( fk )n × B∗ ( fk )n (8.7)
N
n=1
1
N
SAB ( fk ) = A( fk )n × B∗ ( fk )n (8.8)
N
n=1
1
N
SBA ( fk ) = B( fk )n × A∗ ( fk )n (8.9)
N
n=1
where ( )∗ denotes the complex conjugate.
Frequency-response functions. If the dual-channel process being mea-

sured were completely linear and free of noise, the frequency-response function
could be calculated from one record in the following straightforward manner:
B( f )
HAB ( f ) = (8.10)
A( f )
In practice, however, such ideal conditions are never met, and the “smoothing”
capability of the autospectra and cross spectra can be used to yield improved
measures of the frequency-response function. In particular, these spectra can be
used to define two alternate expressions for the frequency-response function:
(1) SAB ( f )
HAB ( f ) = (8.11)
SAA ( f )
“chapter_8(new)” — 2005/10/20 — page 287 — #13

and
(2) SBB ( f )
HAB ( f ) = (8.12)
SBA ( f )
Significant differences between these two versions of the frequency-response
(1)
function relate to the presence of noise in the system. It can be shown that HAB
tends to minimize noise in the output of the process. In the case of vibration testing,
this takes the form of sensor and amplifier noise. Conversely, it can also be shown
(2)
that HAB tends to minimize noise in the input of the process. Such noise can take
place when the system is vibrating near resonances and/or when the force levels
are low relative to background excitations. The principal importance of these two
estimates for the frequency-response function is that they provide a measure of
the linearity of the measured response. This measure is the coherence function
described in the next subsection.
Coherence function. The autospectra and cross spectra defined earlier

provide a convenient definition of the system coherence function as follows:
|SAB ( f )|2
γ 2( f ) = (8.13)
SAA ( f )SBB ( f )
It can be readily shown that the coherence is always equal to the ratio of the two
previously defined frequency-response functions:
(1)
HAB ( f )
γ 2( f ) = (2)
(8.14)
HAB ( f )
Thus, because the linearity of the process is equivalent to the equality of the two
frequency-response functions, the required measure of linearity is the nearness of
the coherence (function) to unity. Note that, as shown in Fig. 8.7, the calculation
of the coherence is an integral calculation within the dynamic frequency analyzer.
Commercially available equipment. With the advent of increasingly more

compact instrumentation design and the formulation of the fast Fourier trans-
form, a wide range of products is currently available for quality vibration testing.
Figures 8.8–8.10 show the physical aspects of three levels of state-of-the-art signal
analyzers. Figure 8.8 exemplifies what is available in the form of a stand-alone
unit with a maximum of optional features, capable of handling all of the functions
defined in Fig. 8.1. Figures 8.9 and 8.10 exemplify products that take advantage of
the ability of compact personal computers to address the chores of performing the
calculations outlined in Sec. 8.1.4, while retaining only those specialized functions
that the computer cannot easily perform and the analyzer can easily perform in
parallel with the computer.
8.1.5 Test Procedures

Narrowband low-frequency testing. A variety of procedures exists for con-
ducting shake tests with the equipment already described. Perhaps the simplest
“chapter_8(new)” — 2005/10/20 — page 288 — #14

Fig. 8.8 Stanford Research Systems Model SR785 Dynamic Signal Analyzer.
(Reproduced courtesy of Stanford Research Systems.)
Fig. 8.9 Pimento 4 Channel Vibration Analyzer. (Reproduced courtesy of LMS N.

America.)
“chapter_8(new)” — 2005/10/20 — page 289 — #15

Fig. 8.10 Brüel and Kjaer Model 3560-L Pulse Lite Pocket Analyzer. (Repro-
duced courtesy of Brüel and Kjaer Instruments, Inc.)
of procedures is that used in connection with assessing the so-called ground-

resonance characteristics of the helicopter. As developed in detail in a later chapter,
ground resonance is not really a resonance condition, but a true mechanical insta-
bility. For determining these characteristics, knowledge of the effective mass Meff
and effective stiffness Keff characteristics of the fuselage at the hub is required.
Because ground resonance is a relatively low-frequency phenomenon occurring at
a frequency less than rotor speed, the fuselage responds effectively as a rigid body.
For the evaluation of the ground-resonance characteristics, only the effective mass
and stiffness characteristics are required. To determine these characteristics, the
equipment already described can be employed, with the exception that the airframe
is not suspended off of its landing gear. The method for determining the required
characteristics is to measure the natural frequency of the fuselage (in the lateral
and fore and aft directions) with incremental additions of mass δM at the hub. As
shown in Fig. 8.11, if these characteristics are plotted as (1/ω2 ) vs (δM), then
Fig. 8.11 Graphical determination of generalized hub masses and stiffnesses for
low-frequency modes involving lateral hub motion.
“chapter_8(new)” — 2005/10/20 — page 290 — #16

the slope of the resulting line is proportional to (1/Keff ), and the effective mass is
given by Keff × (1/ω2 ) for (δM = 0).
Wideband frequency testing. The more usual purpose of shake testing is to

measure the fuselage dynamic characteristics pertinent to bands of frequencies in
ranges each centered about multiples of the blade passage frequency. To this end,
there are a number of basic ways to conduct shake testing:
Discrete (fixed) frequency excitation. This form of testing is the simplest and
represents the mainstay method for many years. The method consists very simply
of exciting the structure at a discrete frequency until a steady oscillatory response
has developed, taking a number of data records, and then proceeding to the next
frequency, etc. One technique that is used to increase the accuracy of the data reduc-
tion is to acquire data at a rate to ensure that the sampling rate exactly divides the
current oscillatory period by some integer (typically some power of two: e.g., 16).
With this kind of data sampling, inaccuracies caused by leakage effects are mini-
mized. The advantages of this type of testing are 1) it provides a direct control of the
accuracy [the presence of higher harmonics of response (caused by nonlinearities
and/or transients) can be monitored for subsidence], and 2) it is relatively sim-
ple and easily implemented. The disadvantages are that it can be time consuming
(and, commensurately, costly) and it does not allow for the use of the statistical
data checking algorithms commonly available with the more complex excitations.
Random excitation. This form of excitation consists of supplying random

noise at a prescribed level across the selected frequency range for the entire record.
This stimulus signal makes the record susceptible to leakage problems and some
form of “windowing” is typically required. Random excitation requires a source
of random input and a servocontroller that can accommodate a wide bandwidth
of required frequencies. To avoid aliasing problems, the bandwidth of the exci-
tation must be at least twice the bandwidth of frequencies required by the test.
The advantages of this method of excitation are that with suitable data-reduction
techniques it can excite all of the resonant frequencies at the same time, and it
also enables the use of various statistical data checking techniques (see Nagy).
The disadvantage of the method is that it does require advanced methodology and
more skill in utilization.
Frequency sweep excitation. This form of testing is characterized by continu-

ally increasing the frequency of excitation in some specific manner (constant rate,
logarithmic rate, etc.). Sometimes denoted sine chirp, this method of excitation
supplies a fast sine sweep over a selected frequency range that repeats with the
same period as the time record. The method has the advantage that it can, if used
properly, reduce the time of testing over that achieved by discrete frequency test-
ing. The method has the disadvantage in that, like that of random excitation, it
requires more complex data-reduction methodology (see Nagy and Jones et al.).
Burst chirp and burst random. Both of these excitation methods entail the
supplying of the signal only during a specified percentage of the time record and
“chapter_8(new)” — 2005/10/20 — page 291 — #17

not for the remainder of the record; usually there is no signal at the start of the
record and at the end. During the nonzero record, the signal is either swept sine or
random.
The more advanced methods of testing are clearly the “wave of the future,” as
noted in present design standards (e.g., ADS-27, which defines appropriate test
procedures for future military helicopter development programs). Additionally,
as new concepts of vibration become formulated, new methods of testing will
emerge. One example of such a new concept is that of chaos, which can be thought
of as being just a more complicated form of orderly motion subject to its own
appropriate mathematical descriptors. When such descriptors are well developed,
they will most certainly be implemented with dedicated algorithms in the frequency
analyzers of the future.
8.2 Other Test Objectives

8.2.1 Blade Frequencies and Mode Shapes
To a certain extent, the same basic techniques used for fuselage vibration testing
can also be applied to the rotor blades. Nonrotating frequencies and modes are
obtainable by fixing a rotor (blades and hub) rigidly to an appropriate support
fixture and shake testing it, as described earlier. In general, modal frequencies
tend to be measured more accurately than mode shapes. If possible, inclusion of
more than just one blade and shaker is preferable to ensure rigidity of the hub
attachment. Rotating natural frequencies are obtainable, at best with difficulty, on
whirl towers by either of two methods. The first is to vary rotor speed to obtain
a resonant crossing with one of the integral harmonics (P order frequency). This
method is complicated by the small amounts of higher harmonic excitation in
hovering whirl conditions and by the fact that blades receive significant levels of
damping caused by the airloadings. The second method is to oscillate the swash
plate to obtain aerodynamic excitation; this is complicated, in turn, by the structural
limitations of the control system and by the obtainable oscillatory response and
nonlinearity characteristics of the control system at high frequencies.
One new method for testing blades, at least at model scale, is to test them in
vacuum spin rigs and then to excite them structurally in a systematic manner with
piezoelectric crystals (i.e., in wafer form, attached to the blade surface). These
crystals are themselves excited with an ac voltage (typically of about 100 V AC),
and the blade responses are measured with suitable conventional instrumenta-
tion (e.g., strain gauges). This form of testing poses difficulties by virtue of the
need to provide electrical excitations to the crystals and to retrieve the electrical
instrumentation signals from the transducers going through a rotating to nonrotat-
ing coordinate system barrier. This can be accomplished using either a slip-ring
assembly or telemetric principles.
Multipoint excitation. As described in detail by Zaveri, multipoint excitation

(also called multishaker sine testing) involves the coordinated excitation of the
structure with multiple shakers in order to accurately isolate modal responses.
At the heart of this type of testing is the requirement for generating controllable
cosine and sine components of the sinusoidal shaker force at each of the shakers.
“chapter_8(new)” — 2005/10/20 — page 292 — #18

Usually, a sophisticated computer control algorithm must be employed to adjust

these components so that each shaker is forcing the structure in phase with the
velocity of its attachment point. Thus, all of the shakers must be completely in-
phase with each other or 180 deg out of phase. When this excitation configuration
is achieved, the structure is then virtually oscillating in an undamped condition,
with the internal damping absorbing the excitation energy. To achieve this state,
the first task is to properly select the excitation points in order to isolate the desired
modal responses. For example, if a symmetric body mode is desired then “mirror-
image” shaker locations must be selected, and those shakers must be excited in
phase.
Mathematical basis. The basis for this type of testing results from a con-
sideration of the general multiple-degree-of-freedom equation set defining the
dynamics of any structure:

[M] Ẍ + [C(ω)] Ẋ + [K] {X} = {F} eiωt = {Fs } sin ωt (8.15)
For sinusoidal motion the velocities will be in quadrature (90 deg out of phase)
with the displacements and accelerations. We seek a condition wherein all of the
components of the displacement vector {X} have the same phase and are then out
of phase with the accelerations. This condition defines a quadrature condition with
the velocities:
{X} = {Xc } cos ωt (8.16)

Ẋ = ω {Xc } sin ωt (8.17)
The requirement for these conditions is that the excitation force vector {F} be
exactly equal to and in phase with the damping terms:
ω [C(ω)] {Xc } = {Fs } (8.18)
At these conditions the response vector then defines the undamped natural mode
shapes, and the excitation frequency is the undamped natural frequency:

[K] − ω2 [M] {Xc } = {0} (8.19)
Implementation. The success of this method clearly hinges on the ability

to find the right components of the forcing vector to isolate each of the modes
under test. To do this, an iterative process is required. First off, initial tests must
be performed to obtain close approximations of the natural frequencies. Then,
for each of these frequencies, the shaker forces must be systematically adjusted
to produce responses that are in quadrature with the excitations. For this reason
provision must be made to obtain sine and cosine components of the responses and
the excitations. One method, proposed by Deck, is to minimize a criterion defined
as follows:

m
R= |Ai sin θi | (8.20)
i=1
“chapter_8(new)” — 2005/10/20 — page 293 — #19

where Ai is the velocity amplitude measured at point i, θi is the phase angle

between the velocity and force at that point, and n is the total number of
measurement/excitation points. The objective is to minimize this criterion by
proper adjustments of each of the shaker forces. Although there is a variety of
ways to iterate to minimize (preferably null) the R criterion, Deck presents a way
to achieve the minimization with only two changes in each of the forces. With
computer control of experimentation, even a nonoptimal way of achieving the cri-
terion minimization is practical. Deck proposes an additional index Z for providing
a measure of the modal isolation and resonance condition:
m
m
Z= |Ai sin θi | |Ai cos θi | (8.21)
i=1 i=1
Measurement of damping. As is described in Section 2.4.3 the locus of

complex-valued response amplitudes for single-excitation shaking results in a
spiral-like curve with sections that are partially coincident with circles when the
scan frequency is in the vicinity of a natural frequency (see Fig. 2.16). With the use
of multipoint excitation, however, each mode can be isolated by using the com-
plement of forces appropriate to each mode. If these forces are maintained and
a frequency sweep is then undertaken, the locus will be a simple circle and the
damping can then be accurately determined. Figure 8.12 provides the basis for this
determination.
Using the geometry of Fig. 8.12, we can determine the damping by either of two
methods. The first, attributed to Kenney and Pancu, entails using the frequency
values at the ends of a diameter that is normal to a diameter through one that has
Fig. 8.12 Construction for evaluating the loss factor (after Zaveri).
“chapter_8(new)” — 2005/10/20 — page 294 — #20

the maximum spacing of frequency on the the arc. Using this construction, the
loss factor (taken to be twice the damping ratio) is then given by
ωA − ωB
η = 2ζ = (8.22)
ω0
The second method is based on the experimental data having been taken at
equal frequency intervals (ω and the angle α is taken to be that angle that defines
a maximum arc length). The natural frequency ω0 would be midway between ω1
and ω2 . The loss factor would then be given by
2ω 2(ω2 − ω1 )
η = 2ζ = = (8.23)
ω0 tan(α/2) ω0 tan(α/2)
8.2.2 Fatigue Testing

Just as analyses for frequency and mode shape prediction must be tempered
by validation with test results, so too, and to an even greater degree, must fatigue
analyses be validated with test results. The test procedure here is basically very
simple: subject the structural components to repeated loads of a predetermined
intensity for a number of cycles and check for failure.
Loading spectrum. The intensity of the loads applied to the structural ele-
ment is defined by the stress (or load) time histories appropriate to the projected
service life of the element. However, to save time, because the part must be tested
to the service life, the time histories should be applied to the structural element at
an accelerated timescale; this is justified because it is the number of cycles that is
important for fatigue, and not time, per se.
Number of cycles to be tested. Although the chosen service life can define
a spectrum of load levels, each with a corresponding number of maximum cycles,
the structural elements are usually tested at N times the number of cycles to ensure
that the parts that survive the cyclic loadings are not lucky scatter points on the
S-N curves. This number N is defined as the scatter factor. The magnitude of
this number depends on the requirements (usually) of the governmental agency
involved, for example, the Federal Aviation Administration, for commercial cer-
tification, and upon the amount of accumulated data available on the material of
which the part is manufactured. Some typical values of scatter currently in use are
shown in Table 8.1.
Table 8.1 Scatter factors for typical aerospace metals
Material N
Steel 4
Aluminum 5–6
Titanium 8
Magnesium 8–10
“chapter_8(new)” — 2005/10/20 — page 295 — #21

References
Section
8.1 Broch, J. T., Mechanical Vibration and Shock Measurements, Larsen
& Son, Denmark (for Brüel and Kjaer), 1984.
Fabunmi, J. A., “Developments in Helicopter Ground Vibration Test-
ing,” Journal of the American Helicopter Society, Vol. 31, No. 3,
1986, pp. 54–59.
Huelsman, L. P., Active and Passive Analog Filter Design, An
Introduction, McGraw–Hill, New York, 1993.
Jones, R., Flannelly, W. G., Nagy, E. J., and Fabunmi, J. A., “Exper-
imental Verification of Force Determination and Ground Flying on
a Full-Scale Helicopter,” U. S. Army Applied Technology Lab.,
USAAVRADCOM-TR-81-D-11, Fort Eustis, VA, May 1981.
Nagy, E. J., “Improved Methods in Ground Vibration Testing,” Journal
of the American Helicopter Society, Vol. 28, No. 2, 1983, pp. 24–29.
Randall, R. B., Frequency Analysis, Larsen & Son, Denmark (for Brüel
and Kjaer), 1984.
“Standard Guide for Dynamic Testing of Vulcanized Rubber and
Rubber-Like Materials Using Vibratory Methods,”ASTM D5992-
96, 1996 (Reapproved 2001).
8.2 Kennedy, C. C., and Pancu, C. D. P., “Use of Vectors in Vibration
Measurement and Analysis,” Journal of the Aeronautical Sciences,
Vol. 14, No. 11, 1947, pp. 603–625.
Zaveri, K., Modal Analysis of Large Structures—Multiple Exciter
Systems, Naerum Offset, Denmark (for Brüel and Kjaer), 1985.
“chapter_8(new)” — 2005/10/20 — page 296 — #22

9
Stability Analysis Methods: Linear Systems
There are several ways of investigating instabilities. The fact that rotorcraft
dynamic issues must invariably be described by multiple-degrees-of-freedom
systems of equations drives our attention to mathematical techniques that can
deal with them. All too often these techniques, while giving accurate indications
of stability levels, also lead to a loss of understanding of the physics involved. At
the heart of any instability issue is the condition wherein there is an energy source
and one or more motion-induced force(s) that are in phase with velocities and grow
with response level. Perhaps the clearest example of this principle is the action of a
person on a swing. To initiate and increase the amplitude of the oscillatory motion,
the “person” must pull on the swing ropes at the bottom of the swing amplitude to
generate a gravity moment in phase with the angular velocity of the swing. This
approach to issues of instability forms the basis of examining the instabilities of
shafting (to be examined in a subsequent chapter), as well as forming the basis of
the method of force-phasing matrices, to be discussed in a subsequent section.
Other approaches to investigating instabilities require dealing with multiple-
degrees-of freedom equation sets and the use of sophisticated eigenvalue calcula-
tions. Within this broad class of methods are the methods associated with periodic
coefficients and with the marriage of frequency-domain data of one subsystem
with the analytic description of another. Both of these issues have application to
all aircraft aeroelasticity analysis, but show significantly greater application to
rotorcraft than to other types of aircraft.
In this chapter we restrict our attention to looking at stability analysis associ-
ated with linear systems, which precludes investigating a wide class of limit-cycle
unstable responses. The linear instability problem is perhaps most crucial, how-
ever, because the inception of unstable motion entails infinitesimal motion, which
inherently defines a linear process. In the real world the examination of test data
to analyze a vibratory response problem can become quite difficult because of
the distinction between resonant responses and unstable ones. Also, in extreme
catastrophic cases, there might not be any intact hardware or data left to examine!
9.1 Basic Concepts

We recall that the simplest of forced elastomechanical systems can be described
by the following differential equation:
m(ẍ + ωn2 x) = f (t) (9.1)

inertia stiffness applied load
297
“chapter9(new)” — 2005/11/16 — page 297 — #1

In the case of most aeroelastic and aeromechanical systems, the applied load
typically contains components that arise from aerodynamic loadings. Furthermore,
these loadings contain components that arise from a variety of external excitations
(i.e., because of a control angle input or a gust), and one or more that arise from
the response of the structure itself. These loads are classified as motion-induced
loads:
f (t) = f (t)dueθ + f (t)due blade response (9.2)

external motion-induced load
excitation (contains state variables)
in description
For most aeroelastic systems the motion-induced loads contain terms that are
proportional to both rate and deflection, the presence of which yields differential
equations of the following general and by now familiar form:
Mn q̈n + Cn q̇n + ωn2 Mn qn = n (t) (9.3)
If, in the preceding development, we take the applied load to consist of only that
caused by an external excitation sinusoidally varying at a frequency of ωn , while
omitting that caused by the motion-induced loads, then the resulting differential
equation has the following form:
ẍ + ωn2 x = A sin ωn t; x(0) = 0; ẋ(0) = 1 (9.4)
The solution to this equation is
x(t) = 1/2(A/ωn2 ) [sin ωn t − ωn t cos ωn t] + (1/ωn ) sin ωn t (9.5)
For this case x(t) grows as t.

If instead, we take the applied load to consist of only that caused by the motion-
induced loads and specifically that it be proportional to velocity, then the resulting
differential equation has the following form:
ẍ + ωn2 x = A ẋ; x(0) = 0; ẋ(0) = 1 (9.6)
The solution to this equation is

1 1 1 2
x(t) = exp At ωn2 − A2 sin ωn2 − A t (9.7)
2 4 4
For this second case x(t) grows as et .

Generalizations that can be drawn from these examples are as follows:
1) Responses that grow linearly with time typically result from one or more
applied external loads oscillating at the system resonant frequency.
2) Responses that grow exponentially with time are typically caused by one or
more motion-induced loads that are in phase with the response velocity such that
they feed energy into the system from some outside source.
“chapter9(new)” — 2005/11/16 — page 298 — #2

STABILITY ANALYSIS METHODS: LINEAR SYSTEMS 299
3) In practice, however, it is often difficult to distinguish between the two types

of response because of the presence of damping, nonlinearities, and time-varying
dynamic parameters, etc. Typically, a clear-cut exponential growth behavior might
only be in evidence over a short period of time before a (nonlinear) limit-cycle
mechanism might begin to predominate. However, all identified instabilities should
be regarded as being capable of producing responses that can grow exponentially
to the point of structural failure!
For most of the stability problems encountered, the governing differential equa-
tions are linear with constant coefficients. Of the two types of solutions to such
equations, transient (i.e., homogeneous) and steady state (i.e., particular), we will
be primarily concerned in the following chapters with only the former type of
solution.
9.2 Basic Tools for Linear Systems: Constant Coefficients

9.2.1 Characteristic Equations
The procedure for finding the transient solution and its attendant stability
information is generally as follows:
1) Set the external forcing function f (t)app equal to zero.
2) Assume an exponential solution:
x(t) = γ eλt (9.8)

where γ is an undetermined constant, herein referred to as the eigenfunction or
characteristic amplitude or mode, and λ is an undetermined coefficient, herein
denoted the eigenvalue or characteristic root.
3) Substitute the expression already given into the differential equation(s),
cancel the resulting eλt factor, and obtain algebraic equation(s) involving the char-
acteristic root(s) and linear combinations of the characteristic amplitude(s). In the
case of a single-degree-of-freedom system, this process will yield a single equation
of the following form:
(λn + a1 λn−1 + a2 λn−2 + · · · + an )γ = 0 (9.9)

By requiring that the characteristic amplitude shall be nonzero, this equation then
reduces to a single polynomial equation in λ. This resulting polynomial we denote
the characteristic equation.
In the case of a multiple-degrees-of-freedom system, the process will yield a
system of simultaneous equations in the characteristic amplitudes. Again requiring
that the characteristic amplitudes (i.e., the mode shape) shall be generally nonzero
and upon setting the determinant of the simultaneous equations to zero yields a
polynomial in λ that is also denoted the characteristic equation.
4) The characteristic root is typically complex:
λ = σ ± iω (9.10)
where σ is the characteristic damping coefficient and ω is the damped natural
frequency.
“chapter9(new)” — 2005/11/16 — page 299 — #3

Stability of the system is ensured when all of the characteristic roots of the
characteristic equation have negative real parts (all σ are negative).
Complex characteristic root plane. For systems that are described by linear
differential equations whose (constant) coefficients are all real-valued, the charac-
teristic roots must necessarily occur as either real roots or as complex conjugate
pairs:
λj = λR (9.11a)
or
λj·j+1 = σ ± iω (9.11b)
For the latter case of complex conjugate characteristic roots, the differential equa-
tion defining the motion of the corresponding characteristic amplitude can be
written in the following alternate form:
q̈ + 2ζ ωn q̇ + ωn2 q = 0 (9.12)
which is equivalent to the following characteristic equation:
λ2 + 2ζ ωn λ + ωn2 = 0 (9.13)
and, upon solving for λ,
λ = −ζ ωn ± i (1 − ζ 2 ) ωn (9.14)
Thus,
σ = −ζ ωn ; ω= (1 − ζ 2 ) ωn (9.15)
where ωn is denoted the undamped natural frequency for the mode. These
quantities are graphically described in Fig. 9.1. The substitution of an exponential
solution [Eq. (9.8)] renders the time derivative to be powers of the character-
istic root λ. For this reason the exponential solution shares some of the same
characteristics of the Laplace transform (see Section 2.4). The characteristic
root plane, as herein defined, can alternatively be referred to as the Laplace
transform plane.
Expansion of the characteristic determinant. Except for relatively simple

systems with only a few degrees of freedom, the solution for the characteristic
roots by first expanding the characteristic determinant to obtain the characteristic
equation and then solving the characteristic equation is not generally practical. The
literature is quite extensive with robust numerical methods (and accompanying
computer coding) for obtaining the eigenvalues of unsymmetrical matrices (e.g.,
Press et al.). However, for those cases where one desires to go the characteristic
determinant expansion route there are a number of tools available. The first chore is
to obtain the characteristic equation in the form of Eq. (9.9). A number of expansion
methods are presented by Tuma for this purpose. The limiting factor in the use
“chapter9(new)” — 2005/11/16 — page 300 — #4

Fig. 9.1 Complex plane and nomenclature defining characteristic roots.
of such methodology is the loss of accuracy caused by roundoff for increasingly

large systems.
Solution of polynomials. For single-degree-of-freedom dynamic systems,

the characteristic equation is simply a binomial for which a solution can be obtained
using the binomial theorem. However, for multiple-degrees-of-freedom systems
the characteristic equation is a polynomial of a degree higher than two; in most
applications the degree is greater than or equal to four. Beyond a degree of four,
there is no general explicit formula or set of formulas for the solution of high-degree
polynomials; hence, iterative solutions must be used.
Basic Newton–Raphson method. For the solution of characteristic equations

for real-valued roots, the Newton–Raphson method has generally good conver-
gence characteristics for real roots. This method is covered in a wide variety of
texts and the reader is referred to the literature (e.g., Press et al.).
Quadratic factorization. For most characteristic equations in structural

dynamics, the roots typically occur in one or more complex conjugate pairs.
Although the Newton–Raphson method could be used with complex arithmetic, it
is more efficient to use a factoring scheme that factors out quadratic factors that
can then be further factored using the binomial theorem to extract either the com-
plex conjugate pair or a pair of real roots. Let the given characteristic equation be
“chapter9(new)” — 2005/11/16 — page 301 — #5

expressed as
f (λ) = 0 = λn + a1 λn−1 + a2 λn−2 + · · · + an

= (λ2 + pλ + q) · (λn−2 + b1 λn−3 + · · · + bn−2 ) (9.16)
where one of the quadratic factors is indicated in Eq. (9.16). The Lin iteration
scheme, as described by Hildebrand, presents a method for sequentially factoring
out the quadratic factors.
Other methods. A variety of additional methods are given in both Press et al.
and Hildebrand that demonstrate differing convergence characteristics.
9.2.2 Stability Boundaries

Very often explicit knowledge of the roots of the system is not needed so much
as knowledge of the conditions under which the system becomes neutrally stable.
From the material presented earlier, the neutrally stable case results when one of the
characteristic roots has a real part exactly equal to zero (i.e., σ = 0). The functional
relationships of the system parameters, wherein the system is neutrally stable, are
denoted the stability boundaries; such boundaries define the ranges of parameters
that produce stable responses and those that produce unstable ones. There are two
basic ways of solving for the stability boundary using the characteristic equation:
the graphical method and the Routh–Hurwitz criterion, to be discussed in the
following subsections.
Graphical method. At the stability boundary, the characteristic root is purely

imaginary, λ = iω. When this expression is substituted into the characteristic
equation, that equation often becomes complex valued:
λ(iω) = (ω) + i (ω) = 0 + i0 (9.17)
Hence, each of the components of the complex valued characteristic equation must
separately be equal to zero also:
(ω) = 0; (ω) = 0 (9.18)
Take for example, a fifth-degree polynomial characteristic equation:
f (λ) = a0 λ5 + a1 λ4 + · · · + a5 = 0 (9.19)
Then
(ω) = a1 ω4 − a3 ω2 + a5 = 0
(9.20)
(ω) = a0 ω5 − a2 ω3 + a4 ω
“chapter9(new)” — 2005/11/16 — page 302 — #6

or
= (a0 ω4 − a2 ω2 + a4 )ω = 0 (9.21)
The assumption is that λ = iω will be fulfilled only when (ω) and (ω) both equal
zero for the same value of ω. In general, they will not both equal zero because
λ = iω only at certain combinations of the system parameters. It is precisely these
combinations of parameters that are desired. To this end, a reasonable approach
is to solve for ω using each of the component equations and to vary the pertinent
system parameter until the two values of ω are equal. This procedure is illustrated
in Fig. 9.2.
Routh–Hurwitz criterion. The Routh–Hurwitz stability criterion is an alge-

braic manipulation of the coefficients of the characteristic equation wherein a
condition of stability (or instability) is indicated by the signs of various derived
quantities. An important consideration when this criterion is used is that the coef-
ficients defining the characteristic (polynomial) equation are required. Thus, for
systems with several degrees of freedom means must be found for expanding the
characteristic determinant to obtain the required coefficients explicitly. This crite-
rion can be expressed in either of two ways: the Hurwitz determinant or Routh’s
criterion. Although the specific algebraic manipulations involved with the two are
somewhat dissimilar, each essentially determines the same stability information
from the explicitly expanded characteristic equation as given by
f (λ) = a0 λn + a1 λn−1 + · · · an−1 λ + an = 0 (9.22)
Routh array. Both variants of the criterion are amply described in the litera-
ture, but the Routh array finds more application to the dynamic systems of interest
Fig. 9.2 Graphical method for solution of characteristic equation.
“chapter9(new)” — 2005/11/16 — page 303 — #7

to us. The basic procedure is to construct the following array:
(λn ) a0 1 a2
* a4 . . .

↓ ↓
↓

(λn−1 ) a1 a3 a5 . . .
(a1 a2 − a0 a3 ) (a1 a4 − a0 a5 )
(λn−2 ) =A =B
a1 a1
(Aa3 − Ba1 )
(λn−3 ) ···
A
· ·
· ·
(λ0 ) ·
The array is continued until no nonzero terms remain. Although not necessary to
the use of the array, the associated power of the eigenvalue λ is indicated for each
row at the left of the array. The Routh array is then used in the following manner:
If all of the terms in the first column are of one sign, the equation has no roots with
positive real parts; the number of roots with positive real parts equals the number
of changes in sign.
Specific results for low-degree polynomials

Cubic:
a0 λ3 + a1 λ2 + a2 λ + a3 = 0 (9.23)
For stability,
i) a0 , a1 , a2 , a3 (all) > 0 (9.24a)

ii) a1 a2 − a0 a3 > 0 (9.24b)
Quartic:
a0 λ4 + a1 λ3 + a2 λ2 + a3 λ + a4 = 0 (9.25)
For stability,
i) a0 , a1 , a2 , a3 , a4 (all) > 0 (9.26a)

ii) a1 a2 a3 − a12 a4 − a32 a0 > 0 (9.26b)
Auxiliary equation. An interesting and useful feature of the Routh array is

that whenever some of the characteristic roots have polar symmetry relative to
the origin, then one of the rows of the Routh array will go to zero. It can be
appreciated that at this point the construction of the array ceases and the array
cannot then be used in the usual manner. However, the last nonzero row (assuming
that it is not the coefficients of the characteristic equation itself) can be used to
obtain root placement information. The coefficients of this preceding nonzero row
“chapter9(new)” — 2005/11/16 — page 304 — #8

can be used to form the auxiliary equation, which is the polynomial comprised
of the factorization of the characteristic equation containing the roots that are
symmetrically located about the origin. Generally, the auxiliary equation is of
even-numbered degree and takes for the power of its highest-degree root term the
power of the eigenvalue that would be written to the left of the array, as exemplified
in the representation just given. This auxiliary equation would then have only even
powers of the eigenvalue, a condition that would be necessary in order to have
the roots symmetrically placed about the origin. In some instances this auxiliary
equation is of small enough a degree that it can be solved to determine these
symmetrically placed roots.
Modifications for very lightly damped systems. For many cases wherein the
Routh array is used to obtain stability boundaries, variations of certain parameters
(usually damping) are to be made that include, in the limit, zero values. In such
instances either of two difficulties can develop. On one hand, the array can degen-
erate to the case discussed earlier wherein one row of the array is zero, in which
case the array construction terminates prematurely. On the other hand, one of the
calculated coefficients in the first column can take on a zero value thereby render-
ing the sign change criterion as undefined. For these two cases recourse can be
made to the use of a small fictitious parameter , which is then taken to be zero in
the limit (after the construction of the array).
As we have seen, the case of one of the rows of the array becoming zero
amounts to a condition wherein some of the roots are symmetrically located about
the origin. If, however, the roots are (artificially) displaced to the “right” in the
complex plane a distance , then the symmetry condition would be disrupted, and
the Routh criterion could then be applied. Thus, if in the characteristic equation
[F (λ) = 0] the following substitution is made
F (λ) = F (λ̄ + ) = 0 (9.27)
where λ̄ is now the “displaced” eigenvalue from the line σ = , then the new
characteristic equation in λ̄ will typically be “complete” in that the previously
zero-valued row will be finite. The coefficients will generally be proportional to
some power of . If the Routh array is then constructed as usual, the sign change
criterion can then be validly invoked with the provision that the parameter be
taken to zero in the limit. In such a way a sign change can be detected even though
the coefficients might be vanishingly small.
In the case of one of the coefficients in the first column becoming zero, this
coefficient can be replaced by , and the array construction can be continued as
usual. To be sure, some of the ensuing construction can involve divisions by . If
now in the limit, as goes to zero, the criterion is invoked, then some of the terms
can become vanishingly small and others become singular, but the sign change
criterion, as already defined, can still be applied with validity despite the size of
any of the terms.
9.2.3 Basic Frequency-Domain Methods

Adaptations of fixed-wing flutter calculations. One method of solving for
rotor aeroelastic stability is to adapt one of the frequency-domain types of solutions
“chapter9(new)” — 2005/11/16 — page 305 — #9

used for making flutter calculations of fixed wings (e.g., the so-called V -g method).
Such methods have the following general characteristics:
1) Sinusoidal motion is assumed: λ = iω. One advantage of working in the fre-
quency domain is that the unsteady aerodynamic formulations currently available
for use in rotary-wing aeroelastic calculations are in large measure adaptations
of fixed-wing formulations and are generally most accurately formulated in the
frequency domain.
2) Artificial damping is introduced into the system by making the usual simpli-
fied structural damping representation assumption but leaving the actual damping
level g and the frequency of oscillation ω as the eigenvalue components of the
problem. Generally, the equations can be rearranged in order to reduce the (flutter)
eigenvalue to a single complex quantity:
Z = (ωα /ω)2 (1 + ig) (9.28)

where ωα is one of the uncoupled natural frequencies of the structure (e.g., either
in bending or torsion).
3) The resulting eigenproblem is solved, using any of a variety of methods,
with a variation on some flight condition variable, such as flight velocity U or,
in the case of rotorcraft, rotor speed . The actual final aeroelastic instability
(flutter) calculation is obtained by plotting the variation of the g portion of the
flutter eigenvalue with the selected flight condition variable.
Instability is indicated when the g portion of the eigenvalue is positive and equal
to the actual value of structural damping coefficient assumed to be available. For a
two-degrees-of-freedom system, the behavior of these eigenvalue quantities would
be typified by the variations of the g and (ω/ωα ) portions of the two eigenvalues
(I and II) shown in Fig. 9.3.
Fig. 9.3 Structural damping and frequency required for neutrally stable motion vs
air speed.
“chapter9(new)” — 2005/11/16 — page 306 — #10

Classic Nyquist stability criterion. The Nyquist stability criterion is based

on the concept of treating differential equations as block diagrams involving
“transfer functions” and, in particular, of determining the behavior of closed-loop
block diagrams by analyzing their respective open-loop block diagrams. Thus, in
its most simple form the block diagram would consist of a forward loop with a
transfer function KG and a positive feedback loop with a transfer function H as
shown in Fig. 9.4.
Generally, the transfer functions in these loops are known because they are,
in fact, the terms of the differential equations defining the aeroelastic stability
problem. Once the transfer functions are known, then the following procedure is
followed:
1) Sinusoidal motion is assumed: d( )/dt is set equal to iω, and the amplitude
and phase of the resulting transfer functions KG(iω) and H(iω) are evaluated.
2) The locus of the product of KG and H, KGH, is plotted in the complex plane,
as the frequency ω is varied from 0 to ∞.
3) For positive feedback, as shown in the simplified diagram just given, the
system is determined to be unstable if this locus encloses the +1 point on the real
axis.
Single-degree-of-freedom system. An example of this technique is the analysis

of a simple single-degree-of-freedom system (spring-mass-damper) as defined by
the following differential equation:
mẍ + cẋ + kx = 0 (9.29)
which can be rewritten in the following expanded form:
mẍ + c1 ẋ + kx = y = c2 ẋ (9.30)
where c = c1 − c2 .
Fig. 9.4 Block diagram for a linear positive feedback system.
“chapter9(new)” — 2005/11/16 — page 307 — #11

For this form of the equation, the appropriate transfer functions are as follows:
x 1
KG ≡ = (9.31)
y k + c1 iω − mω2
y
H ≡ = c2 iω (9.32)
x
Thus, the combined open-loop transfer function is given by
c2 iω
KGH = (9.33)
k + c1 iω − mω2
As shown in Fig. 9.5, the locus of this product forms a loop that encloses or does
not enclose the +1 point on the real axis depending on whether c2 is respectively
greater or less than c1 .
From the preceding equations we know that when c1 is equal to c2 the total
damping c is zero and the system is neutrally stable. This condition corresponds
to the case wherein the locus passes exactly through the +1 point.
Fig. 9.5 Loci of open-loop transfer functions KGH, showing stability characteristics.
“chapter9(new)” — 2005/11/16 — page 308 — #12

Two-degrees-of-freedom system. The equations of motion for a two-degrees-

of-freedom system can be written in the following abbreviated form:
a11 (λ)x1 + a12 (λ)x2 = 0 (9.34a)

a21 (λ)x1 + a22 (λ)x2 = 0 (9.34b)
For this case the two transfer functions become
x1 /x2 = −a12 (iω)/a11 (iω) (9.35a)

x2 /x1 = −a21 (iω)/a22 (iω) (9.35b)
Thus, the required open-loop transfer function becomes
KGH = a21 (iω)a12 (iω)/a11 (iω)a22 (iω) (9.36)
which, again, can be evaluated to form the required locus. As developed more fully
in a subsequent section, this methodology can be generalized and applied quite
effectively to the general multiple-degrees-of-freedom problem.

Constant Coefficients
9.3.1 General Matrix Eigenvalue Problem
With the advent of readily available and increasingly more powerful digital
computer resources, as well as efficient matrix eigenvalue algorithms, a practical
way of solving stability (eigenvalue) problems is to solve for the eigenvalues
and eigenvectors directly rather than first obtaining the characteristic equation. In
most of the material presented in subsequent chapters, the standard form of the
multiple-degrees-of-freedom eigenvalue problem (eigenproblem) is a matrix form
analogous to the simple single-degrees-of-freedom system:
[M]λ2 + [C]λ + [K] {X} = 0 (9.37)
where [M], [C], and [K] denote, respectively, the inertia, damping, and stiffness
matrices. This form, however convenient for defining the physics of structural
dynamics, is not in a standard (canonical) form as typically formulated for any of
a wide variety of standard matrix eigenvalue extraction routines. For this purpose
we first rewrite the preceding equation in the following “semicanonical” form:

M 0 −C −K λX
λ − =0 (9.38)
0 I I 0 X
or, using an abbreviated notation,
[A]λ − [B] {Z} = 0 (9.39)
“chapter9(new)” — 2005/11/16 — page 309 — #13

With some matrix eigenvalue routines this semicanonical form is sufficient. With
others the full canonical form is required:
[D]λ − [I] {Z} = 0 (9.40)
where
[D] = [A]−1 [B] (9.41)
Note that the [A], [B], and [D] matrices are not generally symmetric, and the
resulting eigenvalues (roots) and eigenvectors (modes) are typically complex. For
the cases wherein these matrices are real-valued, the resulting complex roots and
modes will occur as complex conjugates.
9.3.2 Force Phasing Matrices

The theoretical development of the force-phasing-matrix (FPM) technique
follows from three simple ideas governing the unstable motion of any linear
multiple-degrees-of-freedom system:
1) The nature of any unstable system is that there exists destabilizing forces
acting on it that have components in phase with velocity. Thus, for unstable motion
these forces produce work on the system.
2) Within any such unstable dynamic system each degree of freedom has a
multiplicity of forces that have components correspondingly in phase with the
velocity of that degree of freedom. Such forces are herein denoted as driving
forces. That each degree of freedom has drivers in a condition of instability is
presented without proof, but follows heuristically from the properties of linear
differential equations.
3) For any instability involving two or more degrees of freedom, there will
exist a multiplicity of energy flow paths (i.e., vicious circles) wherein the two or
more degrees of freedom will mutually “pump” energy into each other.
The principal function of the FPM technique is to identify the force terms in the
equations of motion that, for an unstable mode, are so phased by the mode shape
(eigenvector) that they act as “drivers” of the motion. The technique is perhaps
nothing more than a formalization of the intuitive use an experienced dynamicist
would make of the eigenvector information. The basis of the technique can be seen
by considering the eigensolution of the basic form of eigenproblem given earlier
in this section, wherein the M, C, and K matrices are assumed to be constant. For
this case the general solution to the homogeneous differential equation is

K
{x} = {φ (k) }eλkt (9.42)
k=1
where λk denotes the kth eigenvalue and {φ (k) } denotes the corresponding kth
eigenvector. In general, both λk and {φ (k) } are complex-valued. The eigenvalues
λ (= σ ± iω), which give stability level and natural frequency information, are
obtained from the solution to the matrix eigenvalue problem.
“chapter9(new)” — 2005/11/16 — page 310 — #14

Upon inserting the preceding general solution into the eigenvalue equations,
each row of the resulting equation set represents the equilibrium of forces acting
on a corresponding degree of freedom. Each equation can be written as the sum of
the mass, damper, and spring forces of the nth diagonal element degree of freedom
along with the remainder of the terms lumped together as a combined excitation
force fn :
mnn λ2k φn(k) ... (inertia force)

+ cnn λk φn(k) ... (damper force)
+ knn φn(k) . . . (spring force)
(k)
+ (mnj λ2k + cnj λk + knj )φj = 0 (9.43)
j=n

= fn
For the usual, nonpathological case mnn , cnn , and knn are all positive numbers;
that is, each autogenous mass (i.e., when uncoupled from the others) is gener-
ally a stable spring-mass-damper system. Because the eigenvalue λk is generally
complex, the preceding equation can be interpreted as the sum of four complex
quantities or vectors in the complex plane that must, furthermore, be in equi-
librium. Assuming that for any complex pair the eigenvector with the positive
imaginary part is used throughout, we can state that the argument of the eigen-
vector θk is the angle by which the inertia force vector is rotated (counterclockwise)
relative to the damper force and the damper force is rotated relative to the stiff-
ness force. For unstable motion the real part of λk (i.e., σk ) is a positive number,
and, hence, θk will be less than π/2 (i.e., 90 deg). If a point in time is taken
when the velocity of the nth degree of freedom is pure real and positive (i.e., zero
dispacement), then the autogenous damper force i.e., −λk cnn φn(k) must corre-
spondingly have a negative real part. The same argument applies to the autogenous
inertia force.
Furthermore, if it is realized that the four vectors represent forces that are in
equilibrium and governed by the constraint already developed on θk , then the real
parts of the autogenous spring, damper, and inertia forces will all be imaginary
(the spring force) or negative; hence, the remaining lumped sum of all of the off-
diagonal terms must always have a positive real part. Figure 9.6 demonstrates this
argument; the four force vectors are shown in the complex plane for an unsta-
ble oscillatory mode [(ωk ) = σk > 0; θk < π/2] and pure imaginary stiffness.
Because the fn vector is a sum of all of the off-diagonal terms for the nth degree
of freedom, any of the individual terms within fn that has a positive real part must
be deemed a driver for that degree of freedom. Conversely, any of the individ-
ual terms within fn that has a negative real part can be deemed a quencher for
that degree of freedom. This fact thus provides the dynamicist with the comple-
mentary information regarding which parameters might be increased to stabilize
an instability.
The aforementioned interpretations of unstable motion can be quantitatively
implemented by forming herein defined force phasing matrices. For any unstable
“chapter9(new)” — 2005/11/16 — page 311 — #15

Fig. 9.6 Force vector diagram for nth degree of freedom, kth mode.
eigenvalue λk these matrices have a one-to-one correspondence to the original

M, C, and K system matrices defining the equations of motion: Any positive real
element in one of the force phasing matrices signifies that the corresponding system
matrix element is a driver.
Mathematical implementation. Force phasing matrices are essentially

formed by dividing each row of the original system of equations by a scalar quan-
tity that renders the diagonal damping term pure negative real. For the case of
constant coefficient matrices, this can be easily accomplished using the eigenvalue
and eigenvector information. The force phasing matrices corresponding to the M,
C, and K system matrices, for the kth eigenvalue can be written, respectively, as
(k)
(k)
αj
[PM ]ij = − [mij ] ⊗ (k)
(9.44)
iφi |kii |
(k)
(k)
βj
[PC ]ij = − [cij ] ⊗ (k)
(9.45)
iφi |kii |
(k)
(k)
γj
[PK ]ij = − [kij ] ⊗ (k)
(9.46)
iφi |kii |
where the ⊗ symbol denotes a Hadamard or term-by-term matrix multiplication.

(k) (k) (k)
The αn , βn , and γn vectors are formed from the results of the basic eigen-
solution:
{αn(k) } = λ2k {φn(k) } (9.47)
“chapter9(new)” — 2005/11/16 — page 312 — #16

{βn(k) } = λk {φn(k) } (9.48)

{γn(k) } = {φn(k) } (9.49)
Note that this formulation renders the diagonal terms in the stiffness force phasing
matrix to near zero values. In the case of an aperiodic instability (i.e., divergence),
the argument of θk will be zero. For the drivers to have a real part, the diagonal
spring force should be phased to be pure negative real. This would be accomplished
by substituting −1 in place of i in the denominators of Eqs. (9.44–9.46). Although
not strictly required for the methodology, the division by |kii | in the preceding
equations serves the useful normalization of the matrices relative to the diagonal
stiffness terms.
Use of force phasing matrices: energy flow paths. The interpretation and
usage of the force phasing matrices can be summarized as follows:
1) Identify the most active degrees of freedom from the eigenvector information
for the unstable mode in question.
2) Look for relatively large positive (+) values in the force phasing matrices
involving the most active degrees of freedom as identified from the eigenvector.
Such elements are the drivers for the unstable motion.
3) Of the drivers identified look for those that involve degrees of freedom that
mutually drive each other. Such drivers we denote critical drivers. As illustrated
in Fig. 9.7, such critical drivers would occur in the simplest form as off-diagonal
terms involving two distinct degrees of freedom, say, the nth and mth.
Fig. 9.7 Definitions of critical drivers and energy flow path.
“chapter9(new)” — 2005/11/16 — page 313 — #17

4) Thus, critical drivers would show up as relatively large (+) values in both
the ( )mn and ( )nm elements of one or more of the three force phasing matrices.
The interaction through these terms is defined herein as the energy flow path.
5) It is possible for damping terms on the diagonal to be zero, but unlikely
that the diagonal stiffness terms would be zero. The diagonal stiffness terms could
easily be negative, however. In such a case the [PK(k) ] force phasing matrix would
still return a near zero value for the diagonal terms. Thus, it would be appropriate
to examine the actual stiffness matrix to identify such a condition.
9.3.3 Kelvin–Tait–Chetaev Theorem

One characteristic of the dynamics of rotating systems is the presence of
gyroscopic-like moments. These moments typically show up in the equations of
motion in the form of skew-symmetric portions of the damping matrix. Thus, for
some dynamic phenomena the basic matrix equation of motion can be written in
the following specialized form:
[M] {Ẍ} + [C1 ] + [G] {Ẋ} + [K] {X} = 0 (9.50)
where [C1 ] is a symmetric damping matrix and [G] is a skew-symmetric matrix
(typically of gyroscopic origin).
The Kelvin–Tait–Chetaev (K-T-C) theorem addresses stability issues wherein
the system of equations defined by only the mass and stiffness matrices possesses
unstable motion. Thus, it is assumed that the following “truncated” matrix equation
possesses one or more unstable roots:
[M] {Ẍ} + [K] {X} = 0 (9.51)
Stabilizing effect of gyroscopic terms. The impact of the K-T-C theorem

can best be appreciated by noting the role the gyroscopic terms have in stabilizing
an otherwise unstable system. An unstable system, when defined by the preceding
matrix set of differential equations ([C1 ] and [G] both zero), can be made at least
neutrally stable by the addition of gyroscopic terms to the equations of motion.
This can be seen for a two-degrees-of-freedom system by including the gyroscopic
terms and using a coordinate transformation that diagonalizes the mass and stiffness
matrices. Such a transformation can be found that also leaves the gyroscopic
matrix skew symmetric. Such a transformation would result in a matrix eigenvalue
problem of the following form:

0 g12 k11 0
[I]λ2 + λ+ {X} = 0 (9.52)
g21 0 0 k11
where g21 = −g12 and k11 and k22 are both <0.
For this case the characteristic equation is
λ4 + (g12
2
+ k11 + k22 )λ2 + k11 k22 = 0 (9.53)
Solving for λ2 shows that both values of this quantity can be made negative for
sufficiently large g12 , thereby causing λ to have pure imaginary values (i.e., neutral
“chapter9(new)” — 2005/11/16 — page 314 — #18

stability). Furthermore, it can be shown that this reasoning can be extended to those
cases wherein either one or the other of the k11 or k22 values is negative and the
other is positive, and/or the total system is composed of pairs of degrees of freedom
that can be so reduced to diagonal form.
Basic K-T-C result for positive-definite matrices. For the cases wherein
both the [M] and [C1 ] matrices are positive definite and where the [K] matrix
has no zero eigenvalues, the K-T-C theorem states the following: For a nonzero
(positive-definite) damping matrix [C1 ], the number of unstable roots of the total
system is equal to the number of unstable roots of the truncated system. That is,
the addition of damping cannot stabilize an unstable system and will destabilize a
system neutrally stabilized by gyroscopic moments.
The basis of the theorem proof is that the presence of damping denies the system
from having purely imaginary roots (λ = ±iω). Therefore, any roots that reside
in the right half-plane with zero damping and gyroscopics must remain in that
half-plane with the addition of damping and gyroscopics.
Modified result for semi-positive-definite matrices. If the damping matrix

is semi-positive-definite, that is, the determinant of C1 is zero, then one useful
result from the K-T-C theorem is that if all of the roots of the truncated system
are stable (or unstable), then all of the roots of the total system are respectively
stable (or unstable).

Periodic Coefficients (Floquet Theory)
9.4.1 Sources of Periodic Coefficients
The structural dynamics and aeroelastics of helicopter (rotorcraft) rotor blades
are subject to dynamic forces whose descriptions contain periodic coefficients. The
resulting matrix differential equations of motion can be expressed generally as
[M] {Ẍ} + [C(t)] {Ẋ} + [K(t)] {X} = 0 (9.54)
where the [M] matrix is generally, but not always, constant and the damping and
stiffness matrices are periodic:

[C(t + T )] = [C(t)]
T = 1/ (9.55)
[K(t + T )] = [K(t)]
One of the hallmarks of rotary wing aeromechanical and aeroelastic formulations is

their abundance of equations of motion with periodic coefficients. Major sources of
these periodic coefficients are the aerodynamics of blades in forward flight, rotor
anisotropy, and the presence of nonlinearities. Each of these sources is briefly
described in the following material as a prelude to more detailed investigations in
later chapters.
“chapter9(new)” — 2005/11/16 — page 315 — #19

Forward-flight aerodynamics. In forward flight the mathematical descrip-

tion of the airloads acting on a typical airfoil section will contain terms periodic
in the rotor rotation period. For many rotorcraft aeromechanical and aero-
elastic problems efficient formulations of the dynamics are obtained by first
nondimensionalizing the time derivatives by the rotor rotation speed :
d( ) 1 d( )
(∗ ) = = (9.56)
dψ dt
This nondimensionalization can then be applied to the simplified aerodynamic
description of the typical airfoil section, as shown in Fig. 9.8, where the tangential
and perpendicular components of air velocity UT and UP are respectively given by
UT = R(x + µ sin ψ) (9.57)
∗
UP = − R(λ + w + µw cos ψ) (9.58)
where x and w are, respectively, the spanwise position r and the flapping dis-
placement (either rigid body and/or elastic), both nondimensionalized by rotor
radius R, and where λ and µ are, respectively the rotor inflow ratio and the rotor
advance ratio (≡ forward-flight speed nondimensionalized by rotor tip speed R).
The flapping airloads description, which gives the basic flapwise aerodynamic
∗
damping (i.e., the coefficient multiplying w), is then obtained from the following
formulation:
1 1
pza (x, ψ) = ρcU 2 cl = ρca(UT2 θ + UT UP ) (9.59)
2 2
Thus, the damping for the flapwise motion is seen to be proportional to (x +
µ sin ψ) and the aerodynamic spring proportional to [µ cos ψ(x + µ sin ψ)].
Rotor anisotropy. As will be developed in subsequent sections, many rotor

aeromechanical formulations require descriptions of the rotating rotor elasto-
mechanics in the nonrotating coordinate system (for coupling with a nonrotating
Fig. 9.8 Typical blade section showing velocity components.
“chapter9(new)” — 2005/11/16 — page 316 — #20

pylon or airframe). A characteristic of the coordinate transformation required to

achieve these descriptions is that rotor properties which are anisotropic in the
rotating frame appear as similar properties in the nonrotating coordinate frame,
but with periodic terms. For example, the diametral mass moment of inertia of a
two-bladed rotor is different depending on whether the diametral axis of rotation
is taken along the diameter defined by the axis of the blades or perpendicular to
the blade axis.
Normally, for all rotor systems, the blades are (or very nearly are) identical.
Hence, for three- (or more) bladed rotor systems, the elastomechanical properties
of the total rotor are isotropic and, hence, independent of the diametral axis about
which the properties are taken. For such systems the rotating to nonrotating coor-
dinate system transformation leaves the properties invariant with respect to rotor
azimuth ψ. However, even for three- (or more) bladed rotor systems anisotropy can
occur as a result of any difference between any of the blades (e.g., an inoperative
blade lead-lag damper or a damaged blade). Such anisotropy can again, as in the
case of two-bladed rotors, give rise to periodic terms in the equations of motion
when the rotating to nonrotating coordinate system transformation is invoked.
Nonlinearities. Many rotor aeroelastic formulations contain nonlinearities

that are of sufficiently high an order of magnitude that they simply cannot be
neglected. For such cases the appropriate matrix differential equation becomes
∗∗ ∗ ∗
[M] {X} + [C(ψ)] {X} + [K(ψ)] {X} = {F(ψ, X, X)} (9.60)
For this case the problem is twofold: First, there is the problem of determining
equilibrium (trim) responses [{X} = {X}trim = {X̄}], which are generally periodic.
Then there is the problem of determining the stability of the trim solution, that is,
determining whether the trim solution stays periodic and bounded. Thus, the total
response vector can be represented by the sum of the trim values and perturbation
values:
{X} = {X̃} + {δX} (9.61)
where {X̃(ψ)} = {X̃(ψ + 2π)}. When this representation of the trim responses
(assuming that they are somehow obtained) are inserted into the equations of
motion (including the nonlinearities), the following matrix equation of motion for
the perturbational responses is obtained:
∗∗ ∗ ∗ ∗
[M] {δ X } + [C̃(ψ, X̃, X̃)] {δ X } + [K̃(ψ, X̃, X̃)] {δX} = 0 (9.62)
In the preceding equation the [C̃] and [K̃] matrices are periodic, and the equa-
tion then represents a standard form of a set of linear equations with periodic
coefficients.
Trim solutions. The principal reason for making trim calculations is to obtain
the proper values of the control angles and rotor attitude angles that produce
the required values of the integrated hub loads (thrust, propulsive force, pitch-
ing moment, and rolling moment), that is, performance trim. A secondary reason
“chapter9(new)” — 2005/11/16 — page 317 — #21

for obtaining trim calculations is to obtain the equilibrium responses {X̃(ψ)}, that
is, response trim. Typically, all trim calculations must be made iteratively because
both the trim force and moment relationships and the equations of motion are non-
linear; some form of numerical solution of the equations of motion is usually the
only practical way of getting accurate solutions. These two trim calculations (per-
formance and response) are sometimes referred to as major and minor iterations,
respectively. Within any one (major) iteration can be several minor iterations. With
each major iteration the calculated integrated hub loads are compared with those
required (or specified), and then variations of the controls are made (using gradient
methods) in order to drive the errors in these quantities to specified tolerances.
The problem of obtaining response trim solutions {X̃(ψ)} is a nontrivial one and,
indeed, one that can be approached from a number of ways. One straightforward
approach to the problem is to integrate numerically the full set of (nonlinear) equa-
tions of motion (with or without periodic coefficients) until the solution converges
to periodicity. This approach, although quite simple, has two serious shortcomings.
First, for lightly damped systems this process can take a long time to converge.
Second, because the entire equation set is being used, the solutions thus obtained
inherently contain not only the desired trim solutions but the perturbation solutions
as well. Thus, if a solution is indeed obtained in this manner, then it must nec-
essarily be stable because the perturbations will have died out. Conversely, if the
solution refuses to die out or actually diverges, then a trim solution is not obtainable
by this method, and the perturbational solution must be deemed unstable. Conse-
quently, this approach is merely doing what the Floquet theory was intended to do
(i.e., check for stability), but much more expensively.
One particularly intriguing approach is that of Peters and Izadpanah. This
approach still uses some form of numerical integration to solve the (entire) non-
linear equations of motions for the forced response case [i.e., retaining the full
∗
description for {F(ψ, X̃, X̃)}]. Solutions are obtained repeatedly, but each time
only over one rotor revolution, assuming a specified set of response initial con-
ditions at the start of each revolution. From these separate responses a partial
derivative matrix can be formed so that a truly harmonic solution for the responses
can be calculated independently of the response stability. This approach has the
dual advantages that it will not be overpowered by any inherently unstable pertur-
bational solution, and that a transition matrix, useful in itself for stability analysis,
falls out as a byproduct. Although the method could be used as a superior basis for
forming a minor iteration, the two iteration loops can be combined into a single
iteration scheme. Thus, in essence, the method treats the aeroelastic responses at
the end of each rotor revolution as additional trim quantities and treats the initial
conditions of these responses as trim variables (along with the control and rotor
attitude angles).
9.4.2 General Characteristics of Floquet Theory

Although the physics of rotary-wing structural dynamics is most conveniently
expressed with a matrix differential equation of motion, using mass and (periodic)
damping and stiffness matrices, the mathematical tool for solving for the stability
characteristics (Floquet theory) requires at least a semicanonical form of the matrix
“chapter9(new)” — 2005/11/16 — page 318 — #22

equation. Thus, we rewrite the basic matrix equation in the following form:
+N

∗
[A] {Y } + [Bn einψ ] {Y } = 0 (9.63)
n=−N

Fourier series ≡ [B(ψ)]
For [B(ψ)] to be real valued, [Bn ] and [B−n ] must be complex conjugates.
Floquet theory gives as a general solution to this equation the following
expression:
{Y (ψ)} = [H(ψ)] [\exp(ηk ψ)\] {αk } (9.64)
where the [H] matrix is periodic and generally fully populated:
[H(ψ)] = [H(ψ + 2π )] (9.65)
where the kth characteristic exponent ηk is an analog to the usual eigenvalue

(characteristic root) arising in systems with constant coefficient systems and where
{αk } is a vector of undetermined constants, again an analog to the eigenvector or
mode shape arising in these systems. Specifically, the characteristic exponent for
systems with periodic coefficients contains the same type of stability information
as that for constant coefficients:
η = σ̄ ± iω̄ ± i2nπ n = 1, 2, . . . , ∞ (9.66)
Note that the characteristic exponents in Floquet theory are distinguished from
the characteristic roots of systems with constant coefficients by the presence of the
multiples of 2π in the imaginary or frequency part of the characteristic expo-
nent. This multifrequency content can be interpreted as follows: for transient
motion defined by any one of the characteristic exponents η, the response would
be observed as having a basic frequency content ω (=ω ), with a multiplicity
of secondary frequencies defined by the two frequencies ω and |ω − |, each
± all integral harmonics of the frequency , defined by the period of the time-
variable coefficients, (= 1/T ). This phenomenon is referred to as an aliasing
of the frequencies. It is a real phenomenon that can be observed from attempts
to frequency analyze the responses of dynamic systems that are characterized by
periodic coefficients in their system matrices.
Although the presence of the frequency aliasing makes the interpretation of the
imaginary parts of the characteristic exponents difficult, the real of the charac-
teristic exponent σ is completely and unambiguously a measure of the stability
of the system. Just as in the case of systems with constant coefficients, stabil-
ity is ensured when the real parts of all of the characteristic exponents σk have
negative values.
“chapter9(new)” — 2005/11/16 — page 319 — #23

9.4.3 Ripple Method of Solution

A meaningful (and actually an historical) link between the theory for systems
with constant coefficients and those with periodic coefficients is the so-called
ripple method of solution. The rationale for the ripple method derives from early
attempts to analyze the aeromechanical stability of rotor blades (Horvay) when our
modern Floquet theory techniques were not well-formulated, and, perhaps more
importantly, extensive computational resources were not available. In essence,
the ripple method assumes that over finite azimuth angle ranges the aerodynamic
properties of the blade sections are changing slowly enough so that the properties
can be taken to be constant. Thus, with this assumption the well-formulated linear
theory with constant coefficients could be implemented over each of these azimuth
ranges to define a solution pertinent to each range. Each of these solutions is then
“patched” together at its boundaries (initial and end conditions) by requiring that
the responses have continuous displacements and velocities. Stability is determined
by investigating how the responses match up beginning to end over one period
(rotor revolution). As we shall see, this method is actually a precursor to the
transition matrix approach presented in a later section.
The ripple method, moreover, also has practical meaning especially for the
rotorcraft aeromechanical problem. As shown in Fig. 9.9, in forward flight the
aerodynamic environment of the rotor disk has four distinct zones, each with
somewhat exclusive characteristics.
To be sure, the mathematically defined zones indicated in Fig. 9.9 (zone sep-
arations at ψ = π/4, 3π/4, 5π/4, and 7π/4) are but rough idealizations to the
indistinctly defined aerodynamic characteristics residing around the azimuth. As
an approximation, however, we note that in both zones 1 and 3 the aerodynamic
Fig. 9.9 Idealized characteristic aerodynamic zones of the rotor disk in forward flight.
“chapter9(new)” — 2005/11/16 — page 320 — #24

environment can be predominantly characterized by 1) being relatively low sub-

sonic and, hence, effectively incompressible, and 2) having substantial radial flow.
In zone 2 in contrast, the aerodynamic environment can be characterized by being
relatively high subsonic and even transonic near the tip; radial flow characteristics
are negligible in zone 2. Finally, all three zones, 1, 2, and 3, can be character-
ized as being in forward flow and having unstalled airfoil operation, that is, the
small-angle approximation on aerodynamic angle of attack is reasonably valid.
In contrast, although the aerodynamic environment in zone 4 is similar to that in
zones 1 and 3 in being incompressible, it is principally characterized by having
reversed flow in the inner portions of the blade and having large angles of attack to
the point of entering stall. Moreover, the nature of the stalled airfoil operation in
zone 4 is such that the dynamic effects on the airfoil stall cannot be categorically
neglected. The four zones thus define three distinct types of unsteady aerodynamic
load descriptions that unfortunately are most practically analyzable individually
with formulations that are awkward to apply and/or inaccurate in the other zones.
Consequently, the concept of defining discrete zones wherein the mathematical
modeling in each is generically uniform has practical justification.
The specifics of using the ripple method can best be appreciated by considering
the four zone conceptualization defined by Fig. 9.9. For each such zone let us
assume that a linear formulation can be made wherein the properties over each
zone are suitably averaged. Thus, the matrix equation of motion for any one zone
(i.e., the kth, where k = 1, 2, 3, 4) can be written as
[Mk ]λ2(k) + [Ck ]λ(k) + [Kk ] {X (k) } = 0 k = 1, 2, 3, 4 (9.67)
where the (k) system dynamic matrices are defined as

(k)
ψ2
2
[Mk ] = [M(ψ)] dψ (9.68)
π ψ1
(k)
(k)
ψ2
2
[Ck ] = [C(ψ)] dψ (9.69)
π ψ1
(k)
(k)
ψ2
2
[Kk ] = [K(ψ)] dψ (9.70)
π ψ1
(k)
where the integration limits are given by the following:
Inner integration limit:
(k) (2k − 3)π

ψ1 = (9.71a)
4
Outer integration limit:
(k) (2k − 1)π

ψ2 = (9.71b)
4
“chapter9(new)” — 2005/11/16 — page 321 — #25

The principles presented in Section 9.3 can then be introduced to yield the following
general solution of the preceding matrix differential equation:
{X (k) (ψ)} = [k ] [\ exp(λ(k) ψ)\]{C (k) } (9.72)
where the [k ] matrix of eigenvectors is n × 2n, where n is the dimension of {X}
and the number of degrees of freedom in the aeromechanical/aeroelastic formu-
lation. Both the columns of [k ] and the eigenvalues [λ(k) ] generally occur either
as complex conjugate pairs or as real-valued pairs. The column of undetermined
coefficients {C (k) } for each of the four zones is determined by directly equating
the displacements and velocities between zones 1 and 2, 2 and 3, and 3 and 4. The
solution is then completed by requiring that the displacements and velocities across
zones 4 and 1 (to complete the solution around the azimuth) be proportional with a
coefficient of proportionality . This coefficient of proportionality is, in fact, the
basic eigenvalue for this type of problem and is referred to as the characteristic
multiplier. These ideas are illustrated by the following equations, which define the
boundary conditions applicable to the solutions for each zone boundary:
π π
X (1) = X (2)
4 4
and

∗ (1) π ∗ (2) π
X = X (9.73a)
4 4

3π 3π
X (2) = X (3)
4 4
and

∗ (2) 3π ∗ (3) 3π
X = X (9.73b)
4 4

5π 5π
X (3) = X (4)
4 4
and

∗ 5π ∗ 5π
X (3) = X (4) (9.73c)
4 4

7π −π
X (4)
= X (1)
4 4
and

∗ (4) 7π ∗ −π
X = X (1) (9.73d)
4 4
“chapter9(new)” — 2005/11/16 — page 322 — #26

The solution scheme is thus seen to comprise two levels of eigenvalue solution. First
there is the eigenvalue solutions within each zone to determine the eigenvectors
[k ] and the eigenvalues [λ(k) ], and second, the eigensolution for the characteristic
multipliers j , resulting from combining the respective undetermined for each
zone through the aforementioned boundary conditions.
9.4.4 Characteristic Multipliers

The characteristic multipliers j , defined by the ripple method eigenvalue solu-
tion are, in fact, an intrinsic characteristic of solutions of equations with periodic
coefficients. They can be related to the characteristic exponents discussed earlier.
First, some qualitative discussion is in order. The role of the characteristic multi-
plier in defining stability can be appreciated again by considering the ripple method
and noting that any perturbational disturbance to the response of the system will
propagate around the disk from zone 1 to 2 to 3 to 4 and back to zone 1. If, when
the disturbance reaches zone 1 (from zone 4) it has grown in magnitude, then the
system can be readily seen to be unstable. Thus, the basic use of the characteristic
multiplier can be stated as the following: If any of the characteristic multipliers
has an absolute value greater than unity, then the system is unstable. The converse
is also true in that if all of the characteristic multipliers have a value less than unity,
then the system is stable. These ideas are formalized in the following sections.
9.4.5 Transition Matrix Solution

The basic objective of using Floquet theory is to evaluate outright or otherwise
obtain information of the quantitative properties of the characteristic exponents
ηk . Of three basic methods for achieving this objective (Hill’s method of infinite
determinants, various perturbational methods, and use of the transition matrix),
the last (transition matrix) method has been most widely implemented and utilized
at present. The transition matrix method follows from the properties of the general
solution given earlier and can be seen to be a generalization of the ripple method.
From the general solution characteristics just presented, it can be shown that for the
homogeneous solution case (stability analysis calculation) the solution at any point
in time is linearly related to the solution exactly one period T (herein 2π ) later.
This linearity is expressible by means of a constant matrix [(T , 0)], denoted
the transition matrix, which typically relates the response state vector at the end
of one period, {Y (T )} to that at the start of the period {T (0)}:
{Y (T )} = [(T , 0)]{Y (0)} = [I] {Y (0)} (9.74)
Once this matrix is found, the stability is ascertained by finding the eigenvalues
of [(T , 0)], which are again denoted the characteristic multipliers j . These
eigenvalues are related to the characteristic exponents that we seek by the following
relationship:
j = exp(ηj T ) (9.75)
“chapter9(new)” — 2005/11/16 — page 323 — #27

Thus,
1
ηj = σj ± iωj = n(j )
T

1 jI
= n|j | ± i tan−1 ± 2nπ (9.76)
T jR
For rotary-wing applications wherein the period is 2π , this expression becomes

1 1 jI
ηj = n|j | ± i tan−1 ± in (9.77)
2π 2π jR
which shows these quantities to have the same characteristic aliasing behavior as
that discussed earlier.
Two basic methods exist for calculating the transition matrix, both of which rely
on numerical integration of the equations of motion: the n-pass and single-pass
methods.
The n-pass method. In this method (n) time-history solutions are obtained
for the given periodic coefficient differential equations of motion. For each of
the (n) solution passes, one of each of the (n) initial conditions (including both
displacements and velocities) is systematically set to unity and all of the remainder
set to zero. By definition, the resulting response vectors obtained at the end of one
rotor revolution (period) are the respective columns of the transition matrix. Thus,
if there are N degrees of freedom in the dynamic system, then 2N passes must be
performed to obtain the transition matrix in this manner.
Single-pass method. In the former n-pass case the initial conditions were
incorporated explicitly in a one-at-a-time sequential manner. In the single-pass
method the initial conditions are incorporated all at once in a more implicit manner.
Although there are alternate methods for achieving a single-pass solution, one
basic requirement is that it leave the initial conditions in a completely general
form, that is, that the method be independent of startup conditions. Typically,
Runge–Kutta methods are ideally suited to this application, and one version that
is well suited to the single-pass method is the fourth-order Runge–Kutta method
with Gill coefficients briefly described in the following material. For this method
a canonical (first-order) form of the differential equations of motion is required.
Define the state vector {Y } as follows:
∗
{Y } = [X, X]T (9.78a)
which results in the following form of the matrix differential equation:

∗
{Y } = [A(ψ)] {Y } = { f (ψ,Y)} (9.78b)
With the single-pass method a numerical integration scheme must be used that
leaves the initial conditions as explicitly stated quantities rather than using them
“chapter9(new)” — 2005/11/16 — page 324 — #28

implicitly in some form of startup process. The Runge–Kutta methods generally

achieve this objective with the use of the usual integration interval h = ψ and
with one-half of this interval h/2 as well. The actual algorithm is then given as
follows:

h 1 1
{Yk+1 } = {Yk } + {k1 } + 2 1 − √ {k2 } + 2 1 + √ {k3 } + {k4 }
6 2 2
(9.79)
where the vectors {k1 }, {k2 }, {k3 }, and {k4 } are given by
∗
{k1 } = {Y k } = { f (ψk ,Y k )} (9.80a)

h h
{k2 } = f ψk + ,Y k + k1 (9.80b)
2 2

h h 1 √
{k3 } = f ψk + ,Y k + √ 1 − √ k1 − (1 − 2)k2 (9.80c)
2 2 2

h √
{k4 } = f (ψk + h,Y k + √ [−k2 + (1 + 2)k3 ] (9.80d)
2
One distinction of Runge–Kutta methods is that they are self-starting, that is,
solutions can be advanced in time [{ }k → { }k+1 ] without the need for knowledge
of the response in the past, that is, { }k−m . It is for this reason that a convenient
calculation for the solution vector at one point in time {Y (ψi+1 )} can be obtained
in terms of the solution vector at the previous point in time {Y (ψi )}:
{Y (ψi+1 )} = [K(ψi )] {Y (ψi )} (9.81)
where the matrix [K(ψi )] can be formed from the development given earlier. The
manner in which this matrix is used is as follows: first, it is formed for each of the
time steps (i = 1, 2, . . .) to form a sequence of the following matrices: [K(ψ0 )],
[K(ψ1 )], [K(ψ2 )], . . . . These are then cascade multiplied to form the required
transition matrix:
N
[(T , 0)] = [K(ψN−i )] (9.82)
i=1
The principal advantage of the single-pass method is that, for large-order dynamic
systems, it is computationally faster. Its principal disadvantage is that it locks
the solution algorithm into using a specific, fixed integration step size, and
methods that adjust the step size to achieve specified accuracy cannot be easily
used.
9.5 Nyquist Criterion for Multiple-Degree-of-Freedom Systems

The Nyquist criterion, as originally formulated for single-variable dynamic
feedback systems, is briefly described in an earlier subsection. The method can
be readily extended to multiple-degrees-of-freedom systems, as well. The use
“chapter9(new)” — 2005/11/16 — page 325 — #29

of a multiple-degrees-of-freedom Nyquist criterion is warranted for those cases

wherein the frequency-response characteristics of one or more of the separate
subsystems that eventually couple with each other are available, but from separate
sources or more accurately defined in the frequency domain.
9.5.1 Conceptual Formulation

The basis of this extension can be seen by reference to Fig. 9.10, which depicts
the loop closure between two multiple-degrees-of-freedom subsystems that couple
with each other.
It is necessary to reorganize the total system into the form shown in Fig. 9.10,
wherein the two separate state vectors X and Y and the transfer function dynamics
[G1 ] and [G2 ] defining the mutual feedback interactions can be clearly identified.
More specifically, one of the state vectors would consist of deflection responses,
and the other would consist of components of interacting force. Thus, one of the
transfer functions would be in the form of an impedance matrix and the other
would be in the form of a mobility matrix (multiple-degrees-of-freedom FRF).
The feedback is first deliberately cut, and the system responses are then assumed
to be undamped sinusoidal rather than the damped (or divergent) sinusoidal, as
would be obtained with the actual loop closure. Three simple relationships then
define the closure condition:
{Y } = [G1 (ω)] {X1 } (9.83)

{X2 } = [G2 (ω)] {Y } (9.84)
Lastly, conditions across the loop closure:
{X2 } = {X1 } (9.85)
Equation (9.85) mandates that the responses at the end of the opened loop X 2
shall be proportional to those at the start of the loop X 1 . The undetermined constant
of proportionality at this point is allowed to be arbitrary and complex-valued.
Fig. 9.10 Block diagram showing cut in loop closure.
“chapter9(new)” — 2005/11/16 — page 326 — #30

Taken together the aforementioned frequency-dependent matrix equations define

an (N × N) eigenvalue problem:
(ω){X1 } = [G2 (ω)] [G1 (ω)] {X1 } (9.86)
where the N frequency-dependent eigenvalues (ω) define a form of characteristic

multiplier eigenvalue and the path they undergo with varying frequency are denoted
characteristic loci. The multivariable Nyquist criterion can then be simply stated
as follows: The combined system (with loop closure) shall be unstable if any of
the characteristic loci crosses the real axis in the complex plane at a point on the
real-axis greater than unity (i.e., +1 ± i0). Figure 9.11 contrasts the conventional
eigenvalue method for indicating instability in the Laplace variable plane with the
method of characteristic loci.
This method has the particular advantage of full utilization of the dynamic
modeling of one or more of the subsystems that is defined most accurately in
the frequency domain. The method can use such modeling directly without the
necessity for reducing the data to approximately equivalent mass, damping and
stiffness matrices. One such situation would be the need to use experimentally
derived data for one of the subsystems, typically obtained in the form of FRFs. For
instance, the whole question of defining the equivalent viscous damping in one of
the subsystems is circumvented, as the actual damping in the subsystem would be
inherently measured and used without any need to characterize it. Furthermore,
the need to extract a mass and stiffness matrix for specified modes is also cir-
cumvented, as the only requirement would be to obtain the FRF data for that
subsystem within the relevant frequency range, regardless of the number of modes
present. Another situation arises when one of the subsystems is defined analytically
only in the frequency domain. Such is the situation, as is shown in Chapter 12,
with many unsteady aerodynamic theories of interest as applied to rotary-wing
flutter.
Fig. 9.11 Comparison of stability criteria.
“chapter9(new)” — 2005/11/16 — page 327 — #31

9.5.2 Practical Implementation Issues

As just formulated, the basic equation, Eq. (9.86), is a matrix eigenvalue equa-
tion whose solution yields discrete values of the characteristic loci j only at
discrete values of the scan frequency ωj . The search for the point of actual cross-
ing of the real axis (in order to assess stability) must then be accomplished using
supplementary techniques. First off, a suitable range of scan frequencies must be
established. A convenient method is to identify a median frequency ωM and then
define the scan range using selected frequencies below and above that frequency.
The selection of the median frequency and the frequency scan range must be deter-
mined in large part by preexisting knowledge of the physics of the instability under
investigation. In practice the characteristic loci typically define curves that move
clockwise in the plane with frequency. Thus, the crossing of the positive real axis
occurs from positive imaginary to negative.
It is then a straightforward programming exercise to evaluate the phase angles
(or alternatively, the imaginary parts) of the characteristic loci to identify the real-
axis crossing value. Specifically, a point in the frequency scan can be detected
(JJ ), wherein two consecutive values of the test quantity (j ) are negative [i.e.,
(j ) < 0, for j = JJ, JJ − 1] and the preceding two values are sequentially posi-
tive [i.e., (JJ−3 ) > (JJ−2 ) > 0]. The use of cubic-spline interpolation (Press
et al.) enables reasonably accurate interpolations of the function and its first deriva-
tive to be calculated, thereby enabling the use of a Newton–Raphson solution for
calculating the value of frequency and characteristic locus at the point of zero
phase ω = ω* and = (ω∗), respectively.
The use of cubic-spline fit interpolation also enables the (approximate) calcula-
tion of the second derivative of (ω) at ω*. As is subsequently described, the first
and second derivatives can be used to calculate effective modal stability exponents
σEFF , using analytic continuation. One postprocessing chore that is recommended
is to evaluate the maximum amplitude of each of the characteristic multipliers
in the scan frequency range to verify that a possible significant characteristic loci
does not get missed. The interested reader can devise any variety of postprocessing
operations with the characteristic loci to make the methodology most efficient for
the application at hand.
9.5.3 Relating Characteristic Loci to Conventional

Stability Indices
One useful characteristic of the conventional Laplace space eigenvalue methods
for ascertaining stability is that the eigenvalues can be put in the form of critical
damping ratios and equivalent undamped natural frequencies:
λ = σ ± iω = −ζ ωn ± i 1 − ζ 2 ωn
The critical damping ratio is a convenient index that “scales” the level of sta-
bility and by extension instability. As presented so far, using only Eq. (9.86), the
use of the method of characteristic loci yields only information as to whether the
total system is stable or unstable. For many applications a more definitive answer,
“chapter9(new)” — 2005/11/16 — page 328 — #32

as to how stable or unstable the system is, is required. For conditions near the
point of instability, it is possible to show equivalence between the two methods.
Essentially, the objective here is to recoup, at least approximately, the eigenvalue
results (i.e., λ = σ ± iω) from the characteristic loci results (i.e., = R ± iI ).
In a typical application of this methodology, the characteristic loci stability infor-
mation would be available only as numerical values of magnitude and phase of
for only discrete frequencies near the positive real-axis crossing. As described
earlier, numerical techniques can be used to obtain the actual real-axis cross-
ing values of ω and . However, in order to recoup, at least approximately, the
equivalent eigenvalue (characteristic exponent) results, a straightforward applica-
tion of analytic continuation can be performed about the real-axis crossing point
= 1 + ε + i0. This formulation begins by writing Eq. (9.86) in the following
more general form:
(λ1 ) {X} − [G2 (λ1 )] [G1 (λ1 )] {X} = {0} (9.87)
where, for the usual eigenvalue solution, λ1 = λ = σ ± iω, and (λ1 ) = +1.
However, the characteristic locus solution is known only for λ on the imaginary
axis: λ∗ = iω∗ , and (λ∗) = R + iI , where ω∗ is the frequency at the real-
axis crossing point (calculable using the techniques discussed in the preceding
subsection). The objective is to closely approximate the characteristic exponent
σ knowing the behavior of as a function of ω near the critical point. Note
first that, in the frequency domain, (iω*) will have the value of 1 + ε (where ε
is a calculable, but reasonably “small” quantity near the neutral stability bound-
ary). In the Laplace-variable domain, however, (λ) has the value of exactly 1
at the value of λ that we seek. Assuming that (λ) is an analytic function in
the neighborhood of (1 + i0), we can, therefore, expand (λ) in a Taylor series
taking the derivatives in the frequency (imaginary) axis direction and evaluating
them at the point λ = λ∗ = iω∗ . At the real-axis crossing point, we represent
the characteristic loci (λ∗ ) as (1 + ε). The analytic continuation equation then
becomes
d ∗ 1 2 d2 ∗
(λ) = 1 = 1 + ε + λ − λ∗ λ + λ − λ∗ λ + · · · (9.88)
dλ 2 dλ2
We note that, assuming continuity, the derivatives can be taken in any direction:
d dR dI dI dR
= +i = −i (9.89a)
dλ d(iω) d(iω) dω dω
2 2
d d d d R d2 I
2
= =− 2
−i (9.89b)
dλ d(iω) d(iω) dω dω2
and

λ − λ∗ ≡ σ + iω (9.90)
where ω = ω − ω∗ . Because (ω) would be known numerically only at a few
distinct points [corresponding to discrete values of ω near the (1 + ε + i0) point] in
“chapter9(new)” — 2005/11/16 — page 329 — #33

the complex plane, a numerical curve-fit calculation must be made to evaluate the
derivatives. One efficient method of fitting the points to analytic approximations
to the real and imaginary parts of (ω) is the use of cubic-spline interpolation.
Because spline fit interpolation schemes fit the data points, as well as the second
derivatives, they also provide an excellent method for approximating first and
second derivatives. The use of cubic-spline fits would thus limit the ability to use
the analytic continuation defined by Eq. (9.88) beyond the second derivative term.
The most accurate estimate of the damping and frequency σ and ω would therefore
be limited to what can be achieved by retaining only the second-order derivative
term in Eq. (9.88):
1 d2 2 d
2
λ − λ∗ + λ − λ∗ + ε ∼
=0 (9.91)
2 d(iω) d (iω)
Solution of Eq. (9.91) is perhaps most practically accomplished using the
binomial theorem with complex arithmetic. It is to be expected that the accu-
racy of Eq. (9.91) to give an analytic continuation would be most precise only
when ε is small (i.e., is close to 1 + i0). For cases where the zero-phase
crossing is at a value significantly greater than unity, the inverse of is more
useful. In this case the analytic continuation occurs about the point −1 (iω∗ ),
where
ε
−1 iω∗ = 1 − ε̂ ; ε̂ = (9.92)
1+ε
The appropriate equation for solving for (λ − λ*), using the inverse of , then
becomes

1 d2 −1 d −1
2
λ−λ ∗ 2
+ λ − λ∗ − ε̂ ∼=0 (9.93)
2 d (iω) d (iω)
Equation (9.93) is quite useful in that, for any unstable characteristic loci (i.e.,
∗ = 1 + ε), the deviation of −1 (iω∗ ) from the point 1 + i0 is always less
than unity, thereby affording a more accurate analytic continuation. Consequently,
Eq. (9.91) would more appropriately be used for stable characteristic loci and
Eq. (9.93) for the unstable ones.
9.5.4 Stability Margins

The inherent incompatibility between the damping ratio measure of stability
and the stability information provided by Nyquist methodology has been long
recognized and addressed for the simple single-degree-of-freedom feedback con-
trol systems (see Savant). Apart from the critical damping ratio, accepted indices
appropriate for using classic Nyquist frequency methods are the gain margin and
phase margin. Gain margin is defined as the factor by which the gain of the system
can be multiplied to produce marginal stability. For present purposes, the gain
margin could be taken to be the inverse of the value of the critical characteristic
locus at the real-axis crossing point 1/(ω∗ ). Similarly, phase margin is defined
“chapter9(new)” — 2005/11/16 — page 330 — #34

as the angle the locus makes with the real axis when the locus has unit value.
Either of these concepts can be meaningfully used to define a stability margin for
the multiple-degrees-of-freedom frequency-domain method.
Savant also defines an M criterion relating the nearness of the locus to the
singular point (in this case the +1 point). For present purposes, the M quantity
would most consistently be defined as the ratio of the distance from the origin to
the locus, to the distance of the locus from the +1 point:

(ω)
M = (9.94)
1 − (ω)
As presented by Savant, the M parameter is taken as an (inverse) measure of

the nearness of the locus to the critical point (here, the +1 point). Clearly, as
the locus nears the critical point, the system loses stability, and the M parameter
takes on larger values. Within the frequency variation, a maximum value of the
M parameter, denoted Mp , can be calculated and then related to an equivalent
damping ratio:

1
ζequiv. ≈ ± 1 + 1 + (1/Mp )2 (9.95)
2
where the sign is taken to be positive for a stable system and negative for an
unstable one.
9.5.5 Range of Applicability of the Methodology

Some aeroelastic systems can consist of a multiplicity of interacting subsystems.
For instance, the aeroelastic analyses of tilt-rotor and tilt-wing aircraft involve
defining the dynamic characteristics of four major subsystems: 1) the elasticity of
the wing and fuselage, 2) the unsteady airloads on the wing, 3) the elastomechanics
of the rotors mounted at the wing tips, and 4) the unsteady airloads on the rotor
blades. The elasticity of the wing and fuselage would typically be obtained first
using a finite element analysis, to be followed by a shake test for verification. These
data would be in the form of a multiple-degrees-of-freedom frequency-response
function (matrix). The unsteady airloads on the wing would typically be defined in
the frequency domain, using well-established unsteady aerodynamic theory. The
elastomechanics of the rotor(s) would typically be provided by an appropriate rotor
elastomechanical analysis (of which there are several of high quality). The rotor
unsteady airloads could be provided by simple quasi-steady theory formulated in
the time domain (see Chapter 10) or by a more comprehensive unsteady theory
(see Chapter 12) that would be defined in the frequency domain. All of these
subsystems can be consistently and validly combined in the frequency domain
with this methodology.
Because the methodology uses dynamic properties that can be measured “in
the real world,” it is not necessarily limited to conditions with constant coeffi-
cients. Systems with periodic coefficients wherein the periodicity does not reside
“chapter9(new)” — 2005/11/16 — page 331 — #35

in the experimentally measured subsystem can be meaningfully addressed with

this method without resorting to the use of Floquet theory (Bielawa). Similarly,
by recognizing that there can be measurable variation of FRFs with amplitude, the
effects of some nonlinearities can likewise be addressed to the point of predicting
limit-cycle behavior (Bielawa).
References
Section
Dowell, E. H., Clark, R. C., Cox, D., Curtiss, H. C. Jr., Edwards, J. W.,
Hall, K. C., Peters, D. A., Scanlan, R., Simiu, E., Sisto, F., and
Strganac, T. W., A Modern Course in Aeroelasticity, Fourth Edition,
Kluwer Academic Publishers, Norwell, MA, 2004.
Franklin, J. N., “On the Numerical Solution of Characteristic Equa-
tions in Flutter Analysis,” Journal of the Association for Computing
Machinery, Vol. 5, No. 1, 1958, pp. 45–52.
Fung, Y. C., An Introduction to the Theory of Aeroelasticity, Wiley,
New York, 1955.
Hildebrand, F. B., Methods of Applied Mathematics, Prentice–Hall,
New York, 1952.
Savant, C. J., Jr., Basic Feedback Control System Design, McGraw–
Hill, New York, 1958.
Scanlan, R. H., and Rosenbaum, R., Introduction to the Study of
Aircraft Vibration and Flutter, Macmillan, New York, 1951.
9.3 Hurty, W. C., and Rubinstein, M. F., Dynamics of Structures, Prentice–
MacFarlane, A. G. J., and Postlethwaite, I., “The General Nyquist Sta-
bility Criterion and Multivariable Root Loci,” International Journal
of Control, Vol. 25, 1977, pp. 81–127.
Zajac, E. E., “The Kelvin-Tait-Chetaev Theorem and Extensions,”
Journal of the Aeronautical Sciences, Vol. 11, No. 2, 1964,
pp. 46–49.
9.4 Horvay, G., andYuan, S. W., “Stability of Rotor Blade Flapping Motion
When the Hinges are Tilted: Generalization of the ‘Rectangular
Ripple’ Method of Solution,” Journal of the Aeronautical Sciences,
Vol. 14, No. 10, 1947, pp. 583–593.
Friedmann, P. P., Hammond, C. E., and Woo, T.-H., “Efficient
Numerical Treatment of Periodic Systems with Application to Sta-
bility Problems,” International Journal for Numerical Methods in
Engineering, Vol. 11, No. 7, 1977, pp. 1117–1136.
“chapter9(new)” — 2005/11/16 — page 332 — #36

Peters, D. A., and Izadpanah, A., “Helicopter Trim by Periodic Shoot-
ing with Newton-Raphson Iteration,” Proceedings of the 37th Annual
National Forum of the American Helicopter Society, Paper 81-18,
1981.
Saaty, T. L., and Bram, J., Nonlinear Mathematics, McGraw–Hill,
New York, 1964.
Ralston, A., and Wilf, H. S. (eds.), Mathematical Methods for Digital
Computers, Wiley, New York, 1960.
9.5 Bielawa, R. L., “Frequency-Domain Methodology for Practical
Ground Resonance Analysis and Test,” Proceedings of the 28th
European Rotorcraft Forum, Paper 69, 2002.
Bielawa, R. L., “Practical Analysis and Testing of Rotorcraft Insta-
bilities Using a Frequency-Domain Based Methodology,” Proceed-
ings of HeliJapan 2002, AHS International Meeting on Advanced
Rotorcraft Technology and Life Saving Activities, Paper T222-3,
2002.
Press, W. H., Flannery, B. P., Teukolsky, S. A., and Vetterling, W. T.,
Numerical Recipies in FORTRAN, the Art of Scientific Computing,
2nd ed., Cambridge Univ. Press, New York, 1992.
Savant, C. J., Jr., Basic Feedback Control System Design, McGraw–
Tuma, J. J., Handbook of Numerical Calculations in Engineering,
McGraw–Hill, New York, 1989.
Problems
9.1 A two-degree-of-freedom system, which is defined in part by two parameters,
p1 and p2 , respectively, has the following dynamic equations of motion:
ẍ1 + ẋ1 + p21 x1 + ẋ2 − (1/2 + p2 ) x2 = 0

ẋ1 − (1/2 − p2 ) x1 + ẍ2 = 0
Determine (and sketch) the map of the two parameters, p1 and p2 , for which
the dynamic system defined by the preceding equations is stable.
9.2 Using the Routh array method, determine the stability (including numbers of
unstable roots, if present) of each of the dynamic systems whose characteristic
“chapter9(new)” — 2005/11/16 — page 333 — #37

equations are respectively given by
a) λ4 + 198λ2 + 10,201 = 0
b) λ4 + 20λ3 − 2,000λ − 10,000 = 0
c) λ5 + 3λ4 + 3λ3 + 9λ2 + 2λ + 5 = 0
9.3 Recast the Ripple method in a standard canonical matrix eigenvalue form
using the linear solution statements for each zone together with the zone
boundary conditions. Note: let
{X (k) (ψ)} = [k ] [\exp(λk ψ)\]{C (k) } = [Gk (ψ)]{C (k) } (9.96)
Similarly, let
∗
{X (k) (ψ)} = [Hk (ψ)]{C (k) } (9.97)
9.4 A simplified (linear) equation of motion for the perturbational flapping

motion of a rotor blade in forward flight with a natural frequency of 1 (per
rev) is as follows:

∗∗ γ 4 ∗
β + 1 + µ sin ψ β + β = 0 (9.98)
8 3
Using a two-zone Ripple solution (zone boundaries at ψ = 0 and π ), deter-

mine the stability (characteristic multipliers) for this rotor for a Lock number
γ of 10 and an advance ratio µ of 0.8.
9.5 Given the following transition matrix for a two-degree-of-freedom dynamic

system with periodic coefficients,
 
1 0.5 0 0
 0.5 1 0 0 
[(2π , 0)] = 
0 0 0.5 0.25 
0 0 0.25 0.5
Determine the characteristic multipliers and assess the system stability.
“chapter9(new)” — 2005/11/16 — page 334 — #38

10
Mechanical and Aeromechanical
Instabilities of Rotors
The general stability characteristics of rotors can be broadly classified according

to the degree of interaction occurring between the inertial, elastic, aerodynamic,
and rotational forces or effects. More specifically, we will deal with the range of
rotor stability issues using the following outline of characteristics as an basis for
classification:
A. Degree of interaction of the aerodynamic forces
1. Low-frequency quasi-steady airloads are applicable.
2. Moderate- to high-frequency unsteady airloads are needed.
B. Linearity of unsteady aerodynamics
1. Subsonic, linear unstalled theory is applicable.
2. Subsonic or transonic, nonlinear stalled theory is needed.
C. Degree of interaction of the elastic forces
1. No elastic interactions occur.
2. Elastic forces play a major role.
D. Degree of interaction between the rotor blades
1. No interaction occurs; blades are uncoupled from each other.
2. Some interblade coupling through structural or aerodynamic forces.
E. Degree of interaction between the rotor and the nonrotating support structure
1. No interaction occurs; rotor hub can be assumed rigid.
2. Significant interaction occurs.
In the following sections this outline will be used as a matrix to explore rotor
instabilities of ever-increasing complexity and interactions between the major
forces. To be sure, completely real-world rotorcraft aeroelastic phenomena will
involve varying degrees of all of the aforementioned attributes. The topics selected
for description in this and subsequent chapters represent idealizations wherein one
or more of the characteristics identified earlier are strongly predominant. Detailed
refinements for analyzing any given design with maximum possible accuracy are
left to the reader/dynamicist as an ever-inexhaustible challenge!
10.1 Unsymmetrical Rotor Instability

The exposition of rotor instabilities begins in this chapter with a description of
perhaps the simplest form of rotor instability by virtue of its limited interaction
of forces: this instability involves no aerodynamics and can be observed in rotor
335
“chapter_10(new)” — 2005/11/16 — page 335 — #1

systems, which, although they do have some form of elastic shaft support, are
nonetheless relatively quite rigid. The key feature of the rotor is that it has
anisotropic inertias and/or stiffnesses either in the fixed coordinate system or
in the rotating shaft. Because the rotor description is relatively simple, the analysis
is applicable to the stability of unsymmetrical shafting.
10.1.1 Dynamic Description

As Fig. 10.1 shows, the phenomenon can be described using a rigid rotor that
is gimbal mounted about some point on the input shaft. The rotor has inertial
anisotropies defined by the ξ and η axes, which pass through the gimbal of the
rotor and are in the directions of the principal axes of the rotor. The motion of the
system is then described by the angular motions about the ξ and η axes, θξ and
θη , respectively. The rotor rotates about the input shaft with a constant rotor speed
and is elastically supported by the angular springs ka and kb . These springs are
grounded to a plane perpendicular to the drive shaft and attached to the points
a and b, which are respectively rotated relative to the ξ and η axes by the angle
φ. The ξ and η axes are attached to the rotating coordinate system so that they
make an angle ψ (= t ) with the x and y (nonrotating) coordinate system axes.
Figure 10.1 reflects the following assumptions:
1) The stability variables θξ and θη are “small”; hence, the stability equations
are linear.
2) The ξ , η, and ζ axes are in the directions of the principal axes of the rigid
rotor about the gimbal point.
Fig. 10.1 Unsymmetric rotor—schematic of dynamic description.
“chapter_10(new)” — 2005/11/16 — page 336 — #2

INSTABILITIES OF ROTORS 337
3) The gimbal point is characterized by constant rotor speed (i.e., it is free of

Hooke’s joint effects).
4) The axes of maximum and minimum shaft support stiffness a and b, respec-
tively, are located by means of a rotation angle φ behind the ξ and η axes. When
the stiffness is inherent in the rotating shaft, it is denoted as internal, and the angle
φ is a constant. When the stiffness is in the nonrotating support structure, it is
denoted as external, and the angle φ is equal to t.
5) The input to the gimbal rotates at constant rotor speed .
10.1.2 General Equations of Motion

If the differential equations for the rotor are written in the rotating coordinate
system for the variables θξ and θη , then Euler’s equations of motion for a rigid
body are directly applicable (see Section 4.1). Thus, the angular velocities and
applied moments can then be written as
ωξ = θ̇ξ − θη (10.1a)

ωη = θ̇η + θξ (10.1b)
ωζ = (10.1c)

Mξ ka cos2 φ + kb sin2 φ) (kb − ka ) sin φ cos φ θξ
=−
Mη (kb − ka ) sin φ cos φ (ka sin φ + kb cos φ) θη
2 2

1 (ka + kb ) + (ka − kb ) cos 2φ −(ka − kb ) sin 2φ
=−
2 −(ka − kb ) sin 2φ (ka + kb ) − (ka − kb ) cos 2φ

θξ
× (10.2)
θη
Using these expressions, together with the form of Euler’s equations given earlier,
we obtain the following basic equation set:

Iξ 0 θ̈ξ 0 1 θ̇ξ 2 (Iζ − Iη ) 0
+ (Iζ − Iξ − Iη ) +
0 Iη θ̈η −1 0 θ̇η 0 (Iζ − Iη )

θξ 1 (ka + kb ) + (ka − kb ) cos 2φ −(ka − kb ) sin 2φ
× +
θη 2 −(ka − kb ) sin 2φ (ka + kb ) − (ka − kb ) cos 2φ

θξ
× =0 (10.3)
θη
And make the following definitions:

Inertia inequality:
Iξ − Iη
i ≡ , −1 < i < +1 (10.4)
Iξ + Iη
“chapter_10(new)” — 2005/11/16 — page 337 — #3

Stiffness inequality:
ka − kb
s ≡ , 0≤ s ≤ +1 (10.5)
ka + k b
Axial inertia term:
Iζ
A≡ , | i| < A < 1 (10.6)
Iξ + I η
where the characteristic natural frequency (arbitrarily defined for nondimension-

alizing purposes) is given by
(ka + kb )
ω0 = (10.7)
(Iξ + Iη )
It can then be seen that the stability equations and the resulting stability characteris-
tics depend on the values of only five nondimensional parameters: (/ω0 ), φ, i , s ,
and A.
If the time derivative is nondimensionalized, d/dt = d/dτ , then the following
nondimensional form of the equations of motion results:
   
 ∗∗  ∗
(1 + i ) 0 θξ 0 1 θ ξ 
+ 2(A − 1)
0 (1 − i )  ∗∗  −1 0  ∗ 
θη θη

(2A − 1 + i ) 0
+
0 (2A − 1 − i )

(1 + s cos 2φ) − s sin 2φ θξ
+K =0 (10.8)
− s sin 2φ (1 − s cos 2φ) θη
where (∗ ) = d/dτ and K = (ω0 / )2 .
10.1.3 Stability Solution: Internal Stiffnesses

For the case of internal stiffnesses, the grounding points for the support springs
ka and kb rotate with the rotor; hence, φ = const. This case corresponds to a flexible
shaft without a universal joint. For a constant value of φ, the stability equations
have constant coefficients, and the stability solution is a straightforward exercise
using the techniques of the preceding chapter. The characteristic equation then
becomes

θξ θ̄ξ λt
αλ + 2βλ + γ = 0;
4 2
= e (10.9)
θη θ̄η
“chapter_10(new)” — 2005/11/16 — page 338 — #4

where
α =1− 2
i (10.10)
β = A2 − 2
i + (1 − A)2 + (ω0 / )2 (1 − i s cos 2φ) (10.11)
γ = (ω0 / ) (1 − 4
s ) − 2(ω0 / ) (1 − 2A + i s cos 2 φ)
2 2
+ (1 − 2A)2 − 2
i (10.12)
Because there is no energy dissipation (damping) for this configuration, stability

is defined in terms of whether or not the system is neutrally stable (λ = iω).
Therefore, λ2 must be a real, negative number. For this to happen, the following
must hold:
α > 0, β > 0, β 2 − αγ > 0, γ >0 (10.13)
It can be shown that, for ranges of the parameters involved, the first three conditions
are automatically satisfied. The condition γ > 0 is then the governing stability
criterion; this implies that the nature of the unstable motion is that of an aperiodic
divergence. The principal parameter subject to variation is the rotor speed . With
variations in this parameter, the rotor can have three different modes of behavior,
depending in part on the value of the rotor speed relative to two critical speeds
1 and 2 :
I. Stable for all speeds.
II. Stable at low speeds, but unstable for all speeds above the first of the critical
speed 1 .
III. Stable at low and high speeds, but unstable within a certain speed range
1 < < 2 .
The critical speeds 1 and 2 are obtained from the condition γ = 0 :
(/ω0 )21,2 = (1 − s )/((1 − 2A + i s cos

2
2φ)
± {[(1 − 2A) s + i cos 2φ]2 + i (1 − s )sin 2φ}
2 2 2 1/2
) (10.14)
Thus, a condition that a critical speed exists at all is that the square of the rotor
speed, as determined from the preceding equation, be a positive number. These
results are summarized in Table 10.1. Figures 10.2–10.5 summarize the results of
using the stability criteria defined by inequalities presented by Eq. (10.13).
Table 10.1 Critical speed stability characteristics of unsymmetric rotor
Case 21 22 1 2 Stability conditions
I <0 <0 Imaginary Imaginary Always stable

II >0 <0 Real Imaginary Unstable above 1
III >0 >0 Real Real Unstable between 1 and 2
“chapter_10(new)” — 2005/11/16 — page 339 — #5

Fig. 10.2 Stability diagram for rotors with stiffness inequality and gyroscope coupling
but symmetrical inertia properties (after Crandall and Brosens).
Fig. 10.3 Stability diagram for rotors with inertia inequality and gyroscopic coupling
but symmetrical elastic properties (after Crandall and Brosens).
“chapter_10(new)” — 2005/11/16 — page 340 — #6

Fig. 10.4 Stability diagram for rotors with equal elastic and inertia inequalities;
maximum inertia axis aligned with minimum stiffness axis (after Crandall and
Brosens).
Fig. 10.5 Stability diagram for rotors with equal elastic and inertia inequalities;
maximum inertia axis aligned with maximum stiffness axis (after Crandall and
Brosens).
“chapter_10(new)” — 2005/11/16 — page 341 — #7

Discussion of results. In each of the figures the rectangular prism encloses

the possible ranges for the three parameters varied. In Figs. 10.3–10.5 the parameter
A must always be greater than the absolute value of the inertia inequality i .
Figure 10.2 shows the instability boundary for the case of inertial symmetry, but
with unsymmetrical (internal) stiffness. This case corresponds to a disk driven by
a long flat shaft. When A < 1/2, the rotor has a type III instability with an unstable
speed range increasing with stiffness anisotropy. Note that this case confirms the
findings of the Kelvin–Tate–Chetaev theorem, which notes that an instability that
does not involve dissipative forces can be (neutrally) stabilized by the addition of
gyroscopic terms (i.e., the instability region closes for → ∞ and/or A ≥ 1/2).
Figure 10.3 shows the stability characteristics for an anisotropic rotor, but with
symmetrical stiffness. Here, in contrast to the preceding case, the “unstable valley”
does not close as → ∞. In this limit the rotor approaches the case of a free rigid
body rotating about a principal axis. It is known that such a configuration is unstable
if the axis of rotation is an intermediate principal axis, that is, Iη < Iζ < Iξ . In terms
of the parameters considered herein, this is expressed as
2 (1 − i ) < A < 21 (1 + i )
1
(10.15)
which is exactly the unstable triangular region shown in the limit as (/ω0 ) → ∞.
Figures 10.4 and 10.5 show the effect on stability of the angle φ. In general, it is
seen that the case that has the greatest spread between the two natural frequencies
is the least unstable.
10.1.4 Stability Solution: External Stiffnesses

For the case of external stiffnesses φ = t, and the coefficients are generally
periodic. However, for the special case of isotropic stiffness ( s = 0) the coef-
ficients are again constant, and the resulting stability boundaries are as given
in Fig. 10.3. Some investigators have treated the much more complicated case
wherein s = 0. For this case, the solution for the stability characteristics requires
the use of Floquet theory. The basic findings are best stated qualitatively.
Briefly, it has been found, both analytically and experimentally, that two addi-
tional instability regions are encountered at rotor speeds below the basic unstable
region that is characteristic of the system caused by inertia anisotropy i . Fur-
thermore, it is found that, for increasing degrees of stiffness anisotropy s , the i
related instability range occurs at higher rotor speeds, whereas the additional s
instability ranges occur at lower rotor speeds.
10.1.5 Effects of Damping

The equations of motion can be modified to include internal damping in a
heuristic manner by noting that internal damping behaves in a manner similar to
that for internal stiffness, except that
θξ −→ θ̇ξ , θη −→ θ̇η , ka −→ ca , kb −→ cb (10.16)
“chapter_10(new)” — 2005/11/16 — page 342 — #8

However, inclusion of external damping requires the use of coordinate trans-

formations between the rotating and nonrotating coordinate systems:

θx cos t −sin t θξ
=
θy sin t cos t θη
(10.17)
Mξ cos t sin t Mx
=
Mη −sin t cos t My
10.2 Quasi-Steady Aerodynamics

To investigate more involved forms of rotor instability, we must begin to intro-
duce the effects of aerodynamic loadings. In keeping with the logic of beginning
simple and building upon accumulated principles, the aerodynamic load descrip-
tions we now seek are the simplest that can be used for stability analyses. Because
we are dealing with beam-like structures and are interested in relatively low-
frequency phenomena, we seek an aerodynamic load description that can be
conveniently used within the framework of a modal or finite element solution.
By way of introduction, we make the following initial basic assumptions:
1) The airloads are defined within two-dimensional sections, that is, different
rotor blade sections are aerodynamically uncoupled from each other.
2) All angles are “small.” As appropriate, this assumption can be relaxed to the
extent that all perturbational angles of attack are small.
3) The two-dimensional airloads are determined solely by considering the
instantaneous angle of attack α, which is composed of the geometric inclination
angle of the chord line θ and the inflow angle φ. As appropriate, this assumption
can be relaxed to include the geometric angle-of-attack rate θ̇ .
10.2.1 Vector Diagram for Simple Quasi-Steady Formulation

A simple formulation describing the quasi-steady aerodynamic theory starts
with a vector diagram of the blade section (see Fig. 10.6), where the lift and drag
distributions are respectively given by
dL 1 1
= ρcU 2 c = ρcU 2 (aα) (10.18)
dr 2 2
dD 1
= ρcU 2 cd (10.19)
dr 2
and the two-dimensional angle of attack α is comprised of the sum of θ and φ.

The inflow angle φ is given by

−1 UP
φ = tan (10.20)
|UT |
“chapter_10(new)” — 2005/11/16 — page 343 — #9

Fig. 10.6 Vector diagram showing component velocities at a typical blade section.
and the perpendicular and tangential components of the velocity UP and UT are
respectively given by
UP = − Rλ − ż − Rµ cos ψz (10.21)
UT = R(x + µ sin ψ) + ẏ (10.22)
where µ and λ are, respectively, the advance ratio (nondimensional forward-
flight speed) and the inflow ratio. The in-plane and out-of-plane components of
the airloads can then be respectively related to the lift and drag distributions as
follows:
dFy dL dD
≈φ − (10.23)
dr dr dr
dFz dL
≈ (10.24)
dr dr
The inflow ratio λ in the expression just given for UP can be expressed in its
most general form as a function of both span x(=r/R) and azimuth ξ . Some typical
ways of expressing the inflow are given in the following.
Uniform inflow. The simplest assumption that can be made about the inflow
is that it is both constant and uniform over the rotor disk:
λ = λ0 (10.25)
Simple analytic variable inflow. A more realistic assumption would be

that the inflow is temporally constant, but has spatial variability around the rotor
disk and can be described with a fairly simple analytical representation. One
manner of accomplishing this approximation is to build on the previously given
simple model but, additionally assume that the variability is limited to linear span-
wise variability and first harmonic azimuthal variability (note that the symbol ξ is
used to denote azimuthal variation in a spatial sense, in contrast to the angle ψ,
which measures both nondimensional time and the azimuthal location of a rotor
blade):
λ = λ0 + x[λ1c cos ξ + λ1s sin ξ ] (10.26)
“chapter_10(new)” — 2005/11/16 — page 344 — #10

Air-mass dynamics. The principal assumption in the use of air-mass dynam-

ics is that the preceding expression for simple variable inflow is now time variable
as well. With ψ used to denote nondimensional time, the resulting expression for
inflow can then be written as
λ(x, ψ) = λ0 (ψ) + x[λ1c (ψ) cos ξ + λ1s (ψ) sin ξ ] (10.27)
where the time dependency of the inflow components is determined by differential

equations based on the instantaneous thrust, pitching moment, and rolling moment
of the rotor. More will be said about air-mass dynamics in a later chapter.
Nonanalytical variable inflow. For cases wherein accurate aeroelastic

responses of the rotor blade are required for steady flight conditions (constant
thrust and hub moments), the variable inflow is typically expressible only in terms
of numerically defined spanwise and azimuthal variability. Such descriptions are
obtained by a variety of solution schemes, all of which must take into account the
spatial displacement field of the vortex structure of the wake. Again, more will be
said of this problem in a later chapter.
As induced by the lifting blade airload. Whereas the air-mass dynamics

representation discussed earlier is primarily a low-frequency approximation (being
based on the time variability of integrated rotor loads), other approximations are
required for higher-frequency phenomena such as those that are associated with
blade vibrations. Such approximate formulations as can be made must take into
account the air velocities induced by the lifting sections of the blade, that is, blade
loads. This type of formulation forms the basis of the subject of unsteady rotor
aerodynamics to be covered in Chapter 12.
10.2.2 Aerodynamic Pitching Rate and Moment

Consideration of the quasi-steady aerodynamic pitching moment requires
attention to two basic issues. First, a distinction must be made regarding where
on the airfoil we define the inflow angle, because, for angular rates, the resulting
velocities normal to the chord (at any arbitrary chordwise location) depend on
the distance this point on the chord is from the rotation center. Second, rotational
pitching rates of the airfoil generate dynamic pitching-moment effects that are
not completely describable using just an effective inflow angle. Thus, stated in
basic terms, the two issues so defined are 1) the effect of pitch angle rate effects
on angle of attack and correspondingly on lift and drag and 2) the effect of pitch
angle rate effects on pitching moment itself. The tools needed for these two issues
come principally from results originally obtained for basic fixed-wing unsteady
aerodynamics.
Pitch rate effects on angle of attack. Depending on the chordwise location

about which the airfoil pitches, additional local velocities are developed normal to
the airfoil chord. From the results of incompressible two-dimensional unsteady air-
foil theory (see Bisplinghoff et al.), it can be demonstrated that the pitching airfoil
can be considered to behave quasi-statically, that is, the instantaneous (dynamic)
“chapter_10(new)” — 2005/11/16 — page 345 — #11

angle of attack can be used statically when the UP component of local inflow,
including pitch rate effects, is defined at the three-quarter chordwise location.
Thus, as shown in Fig. 10.7, two important aerodynamically relevant chord-
wise locations are defined: the quarter-chord point (i.e., the aerodynamic center),
wherein the airloads are considered to act, and the three-quarter-chord point,
wherein the angle of attack (including pitch rate effects) is defined.
Thus, the quasi-steady angle of attack αQS is again defined in terms of the
geometric pitch and inflow angles:

−1 UP
αQS = θ + tan (10.28)
|UT |
where, for rotations about an arbitrary point, a distance ySC from the quarter-chord
point (measured forward toward the leading edge), UP is then defined as follows:
UP = −Rλ − ż − Rµ cos ψz + (c/2 − yAC )θ̇ (10.29)
This definition of the angle of attack then gives the appropriate impact of pitch
rate on lift and drag (as defined by the lift and drag coefficients). Furthermore, this
reasoning can be extended to the case of the moment coefficient as well. Thus, for
pitching motion about the chordwise point ySC the aerodynamic coefficients for
unsteady motion would be approximated as follows:
   
 c   c 
cd = cd (10.30)
   
cmc/4 unsteady cmc/4 static, α = α
QS
Fig. 10.7 Pitching motion relevant geometry of blade section operating quasi-
statically.
“chapter_10(new)” — 2005/11/16 — page 346 — #12

Pitch rate effects on section moment. Apart from the effects of redefining
the angle of attack quasi-statically (for evaluating the steady moment coefficient),
an additional direct effect of pitch rate on the airfoil section moment must be
included. This effect takes the form of a direct pitch damping, which exists even
when the moment coefficient at the quarter-chord point cmc/4 is zero. Again, ref-
erence to classical two-dimensional unsteady airfoil theory gives the following
approximation for the total section pitching-moment distribution, as taken about
the rotation point ySC , defined in Fig. 10.7:
dMxSC 1
= ρcU 2 [cmc/4 (αQS ) + yAC c (αQS )]
dr 2

c3
− πρ U(1 − 2yAC /c) θ̇ (10.31)
8

direct pitch damping
This second term is not generally small, and, except for higher-frequency unsteady
effects, this term accounts for the principal source of damping available to the blade
in pitching and elastic torsional responses.
10.2.3 Table Look-Up Techniques

In the preceding material a general methodology is presented wherein the static
airfoil section properties can be used to approximate the unsteady load distributions
(subject to the condition of low frequencies). For such purposes simple approx-
imations are then used to approximate the lift, drag, and quarter-chord moment
coefficients:
c = aα; cd = cd0 ; cmc/4 = 0 (10.32)
For some cases, especially those for which a full-blown nonlinear analysis is
needed, the airfoil section properties are expressable in terms of the tabulated
values of the airfoil coefficients. Such tables typically express the three airfoil
coefficients as discrete functions of angle of attack α and Mach number M. For
these cases the quasi-steady approximation then takes the following form:
   
 c   c (α, M) 
cd ≈ cd (α, M) (10.33)
   
cmc/4 unsteady cmc/4 (α, M) static
where α = αQS and

M= [UP2 + UT2 ]/c∞ (10.34)
Note that the two component velocities UP and UT are as given earlier (i.e., with
the inclusion of the pitch rate dependency).
“chapter_10(new)” — 2005/11/16 — page 347 — #13

10.2.4 Reversed Flow

Examination of the tangential component of the air velocity at the blade section
shows that it can go to zero or negative values wherever the following condition
is met:
x ≤ −µ sin ψ; −→ π ≤ ψ ≤ 2π (10.35)
The development of quasi-steady airloads formulations, which both include

reversed-flow effects and which are suitable for aeroelastic analyses, poses special
problems relating to the integrations typically required to form the appropriate
differential equations. Rather than divide the rotor disk into forward- and reversed-
flow regions (which is done by some authors), the approach followed here is to
express the airload distributions analytically only in that spanwise region, as given
earlier, wherein reversed effects occur. This can be accomplished by defining a
function that accounts for the sign of UT , that is, sgn(UT ). For example, the lifting
airload distribution can be rewritten in terms of sgn(UT ) as follows:
dL
= 1/2 ρcc U 2 = 1/2 ρacαU 2
dr
= 1/2 ρacU 2 [sgn(U T )θ + φ]
≈ 1/2 ρac[UT2 sgn(UT )θ + |UT |UP ]
= 1/2 ρac sgn(UT )[UT2 θ + UT UP ] (10.36)
Similarly, while allowing for a more general form of cd , we can rewrite the drag
load distribution using the sgn(UT ) function:
dD
= 1/2 ρccd sgn(U T )UT2 (10.37)
dr
Thus, if the sgn(UT ) function can be expressed as an analytical function, it can

then be multiplied in as additional factors in these two load distributions, and the
expressions for UT and UP defined earlier can be used “as is.” To this end, we
define the following function ϒ such that
sgn(UT ) = 1 − ϒ (10.38)
where the ϒ function is as shown in Fig. 10.8. This function can be explicitly
stated mathematically as follows:


0; ψ < π + sin−1 (x/µ)
ϒ= ψ > 2π − sin−1 (x/µ) (10.39)


2; π + sin−1 (x/µ) ≤ ψ ≤ 2π − sin−1 (x/µ)
“chapter_10(new)” — 2005/11/16 — page 348 — #14

Fig. 10.8 Reversed-flow signing function.
This function can then be made more tractable using Fourier analysis. Upon
noting that the function is periodic, we can write an appropriate Fourier series
representation for it:
∞

sgn(UT ) = (1 + I0 ) + [Ins sin nψ + Inc cos nψ] (10.40)
n=1
where
2π 2π−sin−1 (x/µ)
1 (−2)
I0 = (−ϒ) dψ = dψ
2π 0 2π π+sin−1 (x/µ)
2 x
= −1 + sin−1 (10.41)
π µ
The general nth sine term in the series is then given by

2π −1
1 (−2) 2π−sin (x/µ)
Ins = (−ϒ) sin nψ dψ = sin nψ dψ
π 0 2π π+sin−1 (x/µ)

2 −1 x
= [1 − (−1) ]cos n sin
n
(10.42)
nπ µ
In a like manner the cosine term can be formulated:

2π −1
1 (−2) 2π−sin (x/µ)
Inc = (−ϒ)cos nψ dψ = cos nψ dψ
π 0 2π π+sin−1 (x/µ)

2 x
= [1 + (−1)n ]sin n sin−1 (10.43)
nπ µ
“chapter_10(new)” — 2005/11/16 — page 349 — #15

Note that all Ins equal zero for even values of n and all Inc equal zero for odd values
of n. Thus,

4 µ2 − x 2
I1s = ; I2s = 0; etc. (10.44)
π µ
and
4 x 2
I2c = µ − x2 ; I3c = 0; etc. (10.45)
π µ2
The resulting Fourier series has the distinct advantage that, because the coefficients
are only spanwise dependent, the inclusion of this form of reversed flow allows
for the numerical evaluation of various (spanwise) integrals of the airloads. Such
a situation would arise in a Galerkin procedure, for example, wherein generalized
excitations are required that involve integrals involving modal shape functions,
which are typically discretely evaluated and are thus available only at a finite
number of spanwise locations.
10.2.5 Perturbational Airloads

Despite the simplifications inherent in the small-angle and quasi-steady (strip
theory) assumptions, the expressions given earlier are for total airload distributions
and, consequently, are nonlinear as formulated. For use in linear aeroelastic anal-
yses, some form of linearization is required. A systemized method for linearizing
the airloads is presented in the following material. The basic starting point is the
definition of trim and perturbational variables. Generally, this can be accomplished
using the following relationships, wherein the subscript 0 refers to trimmed values
and δ refers to perturbational values:
UT = (UT0 + δUT ) sgn(UT0 ); UP = UP0 + δUP (10.46a)

θ = θ0 + δθ ; θ̇ = θ̇0 + δ θ̇ (10.46b)
However, before the perturbations are taken, let us first introduce refinements to
the drag coefficient and drag load distribution representations. Beyond the simple
representation given earlier, let the drag coefficient be additionally represented
by an nth-degree truncated power series variation in angle of attack. Moreover,
allowing for negative angles of attack and a desire to maintain analyticity, we
restrict our approximation to only even powers of angle of attack:

cd = cd0 + n
cdn αQS (10.47)
n=2k
The utility for using perturbational airloads is based on the premise that per-
turbations of the three components (i.e., lift, drag, and pitching moment) can be
expressed as perturbations in the four elemental kinematic descriptors of the air-
foil’s dynamic state, as given in Eqs. (10.46), namely, δUT , δUP , δθ , and δ (dθ/dt).
For any given aeroelastic problem, these basic perturbational descriptors can then
“chapter_10(new)” — 2005/11/16 — page 350 — #16

be defined in terms of perturbations of the actual degrees of freedom defining the

problem. To facilitate the mathematics, we take φ to be the simple ratio of UP and
UT , and then convert back to the use of φ, as defined by Eq. (10.20). Thus, after
the perturbations are performed the perturbational forms of Eqs. (10.23), (10.24),
and (10.31) can be written as follows:
Out-of-plane load distribution:

dFz dL
δ ≈δ
dr dr
1
= ρac(sgn UT0 )UT0 UT0 δθ + δUP + [2θ + (sgn UT0 )φ]δUT
2
(10.48)
In-plane load distribution:

dFy 1 cd
n−1
δ = ρac sgn(UT0 )UT0 UT0 sgn(UT0 )φ − n n αQS δθ
dr 2 a
n=2k

cd
n n−1
+ θ + 2 sgn(UT0 )φ − n αQS δUP
a
n=2k

cd cd
n n−1
+ φθ − 2 + φ n αQS δUT (10.49)
a a
n=2k
Pitching-moment distribution:

dMxSC 1
δ = ρcmc/4 c sgn(UT0 )UT0 δUT
dr 2

c dFz
+ yAC − [1 − sgn(UT0 )] δ
2 dr
π 3
− ρc θ̇0 δUT + sgn(UT0 )UT0 δ θ̇ (10.50)
8
where φ is given by Eq. (10.20) and αQS is given by Eq. (10.28).

Completion of the formulation requires that the perturbational quantities defined
earlier, δUT , δUP , δθ, and δ θ̇, be defined in terms of the system degrees of
freedom appropriate to the analysis at hand (δq1 , δq2 , δq3 , etc.). Thus, if a Galerkin
(modal) formulation is being used, then the final degrees of freedom would be the
(perturbational) modal response variables (δqv , δqw , δqθ , etc.). Such an operation
“chapter_10(new)” — 2005/11/16 — page 351 — #17

would amount to a linear (matrix) combination of the perturbational variables:

 
  

δq1 

δUT   δq2 
 

  
 
δU    
P
= [D] δq3 (10.51)

 δθ   
 δq4 


 
 
  



δ θ̇  .. 
.
The definition of the [D] matrix is thus entirely dependent on the formulated
solution type and can only be stated symbolically here. Note that the perturbational
components of the description for air-mass dynamics can also be incorporated by
this method, as they just become additional components of δUP .
10.3 Rotor Weaving

Rotor weaving is an aeromechanical instability characteristic of semirigid or
teetering rotors. Such rotors are relatively stiff in the chordwise direction, but
can have a wide range of flexibility in blade torsion and/or the blade pitch con-
trol system. Because of the inherent construction of these rotors, an effective
in-plane displacement can result from the combination of blade coning β0 and
cyclic feathering θ. This feathering consists of rotation of the entire rotor about
the hub universal joint hinge pin in the direction of the blades. The rotor-weaving
instability phenomenon is then caused by a coupling of this feathering motion with
the 1P dynamics of rigid blade flapping.
Fig. 10.9 Schematics of teetering rotor system.
“chapter_10(new)” — 2005/11/16 — page 352 — #18

10.3.1 Elastomechanical Description

Figure 10.9 presents a pictorial view of the teetering rotor with the degrees of
freedom and elastomechanical characteristics depicted.
To describe the kinematics of the in-plane motion resulting from blade feath-
ering motion adequately, the two possible hub gimbal configurations must be
identified. These two possible configurations result from the way in which the
inner gimbal ring attaches to the shaft and the outer gimbal ring, which comprises
the rotor hub proper (see Fig. 10.10).
Fig. 10.10 Top views of two possible hub configurations.
“chapter_10(new)” — 2005/11/16 — page 353 — #19

Fig. 10.11 Inward radial view of a typical blade section.
For a proper analysis of this rotor instability phenomenon, the quasi-steady

aerodynamic description described earlier must be used. Figure 10.11 depicts the
details of kinematics and airload resolution at a typical blade section.
In the formulation of the equations of motion for the rotor weaving phenomenon,
the difference between the type 1 and type 2 arises only in the aerodynamic loading
description. The inertia flapping and feathering moments for the two cases are
surprisingly the same.
10.3.2 General Equations of Motion

The general equations of motion are derivable using the inertia load distributions
for rigid blades, which are available from a number of sources (e.g., Houbolt and
Brooks). Therefore, the following equation set for the perturbational flapping and
feathering is presented as follows without derivation:
Flapping:
{ p̃2 + Cβ̇ p̃ + 1}β + {E3 cos θ0 p̃2 + [Cθ̇ − 2(β02 + E1 cos2 θ0
+ E2 sin2 θ0 + 2β0 E3 sin θ0 )] p̃ + [E3 cos θ0 + Cθ ]}θ = 0 (10.52)
Feathering:
{E3 cos θ0 p̃2 + [Dβ̇ + 2(β02 + E1 cos2 θ0 + E2 sin2 θ0 + 2β0 E3 sin θ0 )]p̃
+ [E3 cos θ0 + Dβ ]}β + {[E1 + E2 + β02 + 2β0 E3 sin θ0 ]p̃2
+ Dθ̇ p̃ + [(E2 − E1 ) cos 2θ0 − 2β0 E3 sin θ0
− β02 + Dθ + K̄θ ]}θ = 0 (10.53)
“chapter_10(new)” — 2005/11/16 — page 354 — #20

where p̃ ≡ d( )/dt and where E1 , E2 , and E3 are inertia-related integrals given

as follows:
R
1
E1 = m km
2
1
dr (10.54a)
Ib 0
R
1
E2 = m km
2
2
dr (10.54b)
Ib 0

1 R
E3 = myCG r dr (10.54c)
Ib 0
R
Ib = mr 2 dr (10.54d)
0
which equals the mass moment of inertia about the teeter hinge and where km1
and km2 are mass radii of inertia of the blade section about the major and minor
2
principal axes, respectively. (Thus, for most “thin” sections, mkm mkm2 .)
2 1
The aerodynamic related integrals are then defined as follows:
γ
Cβ̇ = ; [γ ≡ blade Lock number = ρacR4 /Ib ] (10.55a)
8
γ
Cθ = − (10.55b)
8

γ c 1 yA C 1
Cθ̇ = − − − β0 θ0 (10.55c)
2 R 4 3c 2
!
γ 1 2 cd " c yA C
Dβ̇ = − β 0 θ0 − λ 1 + 0 − (10.55d)
2 4 3 a R 3c

 γ cd0 1 1

− + λθ0 − λ2 ; rotor type 1
2 4a 3 2
Dβ = (10.55e)



0; rotor type 2
! "
γ1 c 2 1 yA C 1 ya
Dθ̇ = − −
22 R 4 3c 2 c

c 2 yA C 3 cd0 1
+ β0 θ0 − λ + β02 + θ0 λ (10.55f)
R 3 c 8 4a 3

γ 2 c yA C
Dθ = β0 θ0 − β0 λ − (10.55g)
2 3 R 3c
“chapter_10(new)” — 2005/11/16 — page 355 — #21

and where the nondimensionalized (mechanical) stiffness K̄θ is expressed as

follows:
K̄θ = Kθ /Ib 2 ≡ (E1 + E2 )(ωθ0 / )2 (10.56)
10.3.3 Stability Solution

For initial investigations let us set the collective angle θ0 equal to zero. As will
be shown later, this simplification does not change the results qualitatively and
introduces little quantitative error.
Divergence. The conditions for divergence stability are that the constant term
in the characteristic equation be positive. Thus, the divergence stability bound-
ary is given by setting that term to zero. This is accomplished by the following
determinantal equation:
# #
# 1 (E3 + Cθ ) #
# #=0 (10.57)
#(E + D ) [(E2 − E1 ) − β0 + Dθ + K̄θ ]#
2
3 β
or
[(E − E1 ) + Dθ + K̄θ ] − [β02 + E32 + Dβ Cθ ]

2
always (+) always (+)
− E3 (Dβ + Cθ ) > 0 (10.58)

always (−)
For E3 → 0 (coincident elastic and c.g. axes),
[(E2 − E1 ) + Dθ + K̄θ ] − [β02 + E32 + Dβ Cθ ] > 0 (10.59)
The divergence stability results can then be summarized as follows:

1) Aft positions of the c.g. from the elastic axis are destabilizing.
2) More divergence margin inherently exists in a hub design of type 2 (feath-
ering axis flaps with the blades) than in a design of type 1 (feathering axis remains
perpendicular to the rotor shaft), for which Dβ = 0.
Oscillatory instability (flutter). As was shown in preceding sections, the

stability boundary for oscillatory motion can be obtained using the Routh cri-
terion. After the characteristic equation is expanded, the instability boundary is
given by
a1 a2 a3 − a12 a4 − a32 a0 > 0 (10.60)
Application of this inequality leads to a quadratic equation for the required control
stiffness K̄θ . Solution of this equation gives this stiffness as a function of the
remaining parameters.
“chapter_10(new)” — 2005/11/16 — page 356 — #22

10.3.4 Stability Characteristics: Parameter Variations

In order to investigate the basic physics of the instabilities possible with
this approximation of a teetering rotor, we make the following simplifying
assumptions:
1) The thickness-wise mass distribution is zero, that is, the blade section is thin.
2) The mass distribution is constant along the span and has a chordwise
distribution, as shown in the following sketch:
3) The reference axis is taken to be the quarter-chord.

4) No damping in torsion exists except that as a result of aerodynamics.
5) Quasi-steady aerodynamics (strip theory) is applicable (as derived).
6) The inflow velocity is uniform and constant over the rotor disk and is
obtainable from simple momentum considerations.
These assumptions make all of the various equation coefficients dependent on
the type of rotor (1 or 2) and seven nondimensional parameters:
γ = blade mass ratio (Lock number)
θ0 = blade collective angle
σ = rotor solidity [= (2/π )(c/R)]
yCG /c = nondimensional chordwise mass offset
yAC /c = nondimensional aerodynamic offset
β0 = blade built-in coning angle
K̄θ = control system feathering stiffness
Assumptions 1, 2, 3, and 5 yield
2
E1 = 0; E3 = (yCG /c)(c/R)
3
3
E2 = (c/R)2 (10.61)
16
where (c/R) = π σ/2.
“chapter_10(new)” — 2005/11/16 — page 357 — #23

Assumption 6 yields the following relationship between inflow and collec-

tive angle:
$ 1/2
1 32Bθ0 64θ0 B
λ = √ (aσ/8) +1− +1 (10.62)
2 3aσ 3aσ
Then a basic (typical) configuration is defined wherein a = 0.101/deg,

cd0 = 0.0083, B (tip loss) = 0.97, σ = 0.05, and γ = 10. With these typical trim
parameter values, variations of all of the eight basic parameters can be made. The
results of these variations using the divergence and Routh criteria are shown in
Figs. 10.12–10.17.
General observations of the stability results:
1) In all cases coning angle β0 is destabilizing. Thus, an otherwise stable con-
figuration could encounter weaving divergence or oscillatory instability operating
at a sufficiently high load factor (i.e., CT /σ ).
2) For small coning angles β0 and low values of control stiffness parameter
(ωθ0 / ) (or high rotor speeds), the stability criterion is principally one of
divergence instability.
3) The divergence boundaries shown are inherently more accurate than the
flutter boundaries because the preceding assumptions, especially 4 and 5 relating to
damping and unsteady aerodynamics, are more nearly satisfied for the divergence
case. Control system damping (e.g., as a result of friction, discrete dampers, etc.)
would have a definite stabilizing effect on the oscillatory instability boundary.
4) During runup, the nondimensionalized stiffness (ωθ0 / ) varies from +∞
to O(8.0). Hence, divergence tends to be most important at operating rpm, and
flutter becomes important even during runup.
Fig. 10.12 Rotor weaving instability—effect of rotor type.
“chapter_10(new)” — 2005/11/16 — page 358 — #24

Fig. 10.13 Rotor weaving instability—effect of mass ratio (Lock number).
Fig. 10.14 Rotor weaving instability—effect of collective angle.
“chapter_10(new)” — 2005/11/16 — page 359 — #25

Fig. 10.15 Rotor weaving instability—effect of solidity.
Fig. 10.16 Rotor weaving instability—effect of chordwise mass offset.
“chapter_10(new)” — 2005/11/16 — page 360 — #26

Fig. 10.17 Rotor weaving instability—effect of chordwise aerodynamic offset.
5) One factor contributing to the oscillatory instability characteristics is the

ratio of (aerodynamic) feathering damping to effective feathering inertia in the
presence of coning. With zero coning the nominal feathering damping (obtainable
with zero aerodynamic offset) is approximately (γ /32)(c/R)2 , which corresponds
to a critical damping ratio of O(0.1). With increasing coning the effective feath-
ering inertia increases much more rapidly than does the aerodynamic feathering
damping, resulting in a deterioration of the effective damping as measured by the
damping ratio.
Specific trends identified by the parametric variations are as follows:
1) Effect of rotor type: The principal advantage of rotor type 2 over type 1 is that
the divergence instability is easier to stabilize with control stiffness. Because the
oscillatory stability boundaries shown for the two types are not greatly different,
the type 2 rotor configuration was used for all of the subsequently parametric
variations. All of the parameter variations are about the same reference set of
parameters identified in Fig. 10.12.
2) Effect of mass ratio: Aerodynamic effectivity, as measured by the mass
ratio, is seen to be stabilizing. As would be expected, the available damping is
aerodynamic and is therefore proportional to mass ratio. The divergence boundary
in this case is unaffected by mass ratio because there is zero collective angle and
aerodynamic offset.
3) Effect of collective angle: The divergence condition is seen to be monoton-
ically stabilized by collective angle. However, the oscillatory instability is first
destabilized and then stabilized by increases in θ0 . However, the variations show
this instability to be a weak function of collective angle. Increases in collective
angle cannot be made, generally, independent of coning angle. That is, a change
in collective angle will produce a change in coning.
“chapter_10(new)” — 2005/11/16 — page 361 — #27

4) Effect of solidity: The chord-to-radius ratio determines to a great extent

the amount of restoring propeller moment available for divergence stabilization.
Consequently, solidity is seen to be divergence stabilizing. The chord-to-radius
ratio also increases the feathering interia and aerodynamic damping approximately
in the same proportion. In the presence of some discrete control stiffness K̄θ , the
total effective feathering stiffness will increase proportionately less than will the
inertia (and damping) caused by solidity increases; hence, the effective feathering
damping ratio will increase with solidity. It can also be seen that the aerodynamic
coupling terms become reduced with solidity relative to the damping. These trends
are reflected in the strong stabilizing effect on the oscillatory instability shown in
Fig. 10.15.
5) Effect of mass and aerodynamic offsets: Mass and aerodynamic chordwise
offsets are seen to have strong inverse effects on both divergence and oscillatory
instability. Clearly, forward center of gravity and aft aerodynamic centers are
preferable.
10.4 Blade Pitch-Flap-Lag Instability

Using the outline of characteristics of rotor instabilities presented earlier, blade
pitch-flap-lag instability can be typified by its low-frequency content (linearized)
aerodynamics and the scant interaction between the blades and/or the hub dynam-
ics. This instability is a general one that has been identified in similar forms with
most principal rotor types. A review of the dynamic problem areas of articulated
rotors shows that a form of instability known as pitch-lag instability was found
to occur in the presence of blade feathering mechanical coupling with in-plane
motion. This instability is further characterized by blade motions being relatively
uncoupled from those of the hub. Hence, the classic analysis of this instability
that resulted is relatively simple, involving only simplified rigid-body rotational
motions of the blade about the articulation hinge (i.e., blade lead-lag motion and
rigid blade feathering).
Pitch-lag instability in articulated rotors is aggravated principally by the degree
of geometric pitch-lag coupling present and is typically controlled (stabilized) by
the already present and available lead-lag dampers in this type of rotor. This type
of instability was eventually recognized to be a special case of a more generic
type of instability, especially with the advent of hingeless rotor types, which are
devoid of discrete lead-lag dampers.
The development that follows is a general one that admits both articulated and
nonarticulated (hingeless) rotor types. This generalization requires that the blade
description include both rigid blade lead-lag and flapping motions as well as blade
bending modes. This general development additionally includes the kinematic
pitch angle change caused by lead-lag motion at the blade root. The addition of
these two considerations has defined a new more general class of rotor instability
that is generally referred to as pitch-flap-lag instability and, as a special case,
flap-lag instability.
10.4.1 Basic Considerations and Mathematical Development

Pitch-flap-lag instability is caused essentially by the in-plane forces on the blade
(Coriolis and/or aerodynamic) being in phase with the in-plane velocity, thereby
“chapter_10(new)” — 2005/11/16 — page 362 — #28

effectively producing negative damping. This is brought about by an interaction

of the pitch-lag coupling and the blade flapping response. This can be seen by
considering the following reasoning: Let the pitch-lag coupling be such that an
aft in-plane deflection will cause a closely in phase downward flapping deflection.
The in-plane Coriolis force is proportional to the blade coning β0 multiplied by
the flapping velocity and is positive aft for downward flapping velocity. In such a
configuration the blade will develop in-plane forces that have components in phase
with the in-plane velocity and in the same direction, or, in other words, negative
in-plane damping. The general (simplified) development of the dynamic equations
makes the following assumptions:
1) Quasi-steady aerodynamic theory in hover is applicable. The actual details of
the airloads follow directly from the preceding section on perturbational airloads.
For this development, the airloads are confined to only the out-of-plane and in-
plane components, as defined by Eqs. (10.48) and (10.49).
2) In its most general form the blade motion consists first of articulated motion
at the hub in the form of flapping and lead-lag motion βH and ζH , respectively, about
some offset point e0 . This point would be the explicit hinge point in articulated
rotors and some effective point in the case of hingeless rotors. In addition, the blade
is flexible in bending outboard of the hub hinge point in the flatwise and edgewise
directions we and ve , respectively. Throughout the formulation presented, the blade
is assumed to have no significant twist. These degrees of freedom are shown in
Fig. 10.18 for the blade in perspective view and in Fig. 10.19 at an arbitrary radial
station.
3) The blade has an equilibrium coning angle β0 , defined by the steady lifting
airloads and the root flapping restraint.
Fig. 10.18 Perspective view of blade defining the in-plane and out-of-plane degrees
of freedom.
“chapter_10(new)” — 2005/11/16 — page 363 — #29

Fig. 10.19 Section view of the blade deflection components at an arbitrary radial
station.
4) The blade root is constrained by flapping and in-plane (equivalent) springs

KβH and KζH that define whatever stiffness the hub imparts to the blade dynamics.
For articulated rotors these spring rates would be zero. Also, the blade is restrained
in the lead-lag direction by the lead-lag damper, whose damping is defined by the
damping coefficient Cζ . These constraints are expressed as follows:

MyH KβH βH
=− (10.63)
MzH KζH ζH + Cζ ζ̇H
5) Because of the (typical) geometric characteristics of the attachments of the

push rod to blade pitch horn and to the swash plate, an incremental mechanical
blade pitch angle αg is generated by motion of the blade root in the plane of the
rotor. This incremental pitch angle is defined as pitch-lag coupling Lζ . Although
pitch-flap Lβ coupling is equally possible, it usually plays a secondary role in
this instability, and results that would be obtained for nonzero values are omitted
herein for clarity. The range of pitch-lag coupling in a rotor system design is often
inherent and not subject to much practical variation, whereas the pitch flap can be
varied relatively easily. Thus, in complete generality, the incremental blade pitch
angle is given simply as
αg = Lζ ζH + Lβ βH (10.64)
6) Because the blades sustain root bending moments, the blade bending deflec-
tions ve and we are defined by the blade (cantilevered) normal modes in the flatwise
and bending directions, as developed in Chapter 3 and as coupled because of the
pitch angle θ0 , using Eqs. (3.145) and (3.146).
“chapter_10(new)” — 2005/11/16 — page 364 — #30

Using a standard application of the Galerkin technique, the four equations of

motion for the four variables defined in Figs. 10.18 and 10.19 can be derived.
The development that follows is based on a full-scale configuration that has
been nondimensionalized so as to give results that have universal application.
Although a complement of four equations has been formulated, for some con-
figurations only the two equations relating to hub motion will suffice. The
various Si integration constants (both nondimensional, “barred,” and dimen-
sional, “unbarred”) appearing in the equations are defined in Appendix D. The
nondimensionalization is completed by division by the rotor speed squared. This
renders the time derivatives to be derivatives with respect to nondimensional time
t(= ψ).
Hub deflections:
Flapping:

S̄1 K βH S̄12
p̃ + cβ ωβ p̃ + 1 + ē0 +
2
2
βH + 2β0 p̃ζH + cos θ0 p̃2 qw
S̄2 S2 S̄2

S̄46 S̄12 S̄46
+ sin θ0 p̃ qv + 2β0 −
2
sin θ0 p̃qw + cos θ0 p̃qv
S̄2 S̄2 S̄2

S̄12 S̄16 S̄46 S̄48
+ + ē0 cos θ0 qw + + ē0 sin θ0 qv
S̄2 S̄2 S̄2 S̄2
R
1 dFz γ 1 dF̄z
= r δ dr = x δ dx (10.65)
S2 2 e0 dr 2 ē0 dx
Lead-lag:

S̄1 K ζH S̄12
p̃ + cζ ωζ p̃ + ē0 +
2
2
ζH − 2β0 p̃βH − sin θ0 p̃2 qw
S̄2 S2 S̄2

S̄46 S̄12 S̄46 S̄16
+ cos θ0 p̃ qv − 2β0
2
cos θ0 p̃qw + sin θ0 p̃qv − ē0 sin θ0 qw
S̄2 S̄2 S̄2 S̄2
R
S̄48 1 dFy γ 1 dF̄y
+ ē0 cos θ0 qv = r δ dr = x δ dx
S̄2 S2 2 e0 dr 2 ē0 dx
(10.66)
Blade bending:
Flatwise bending:
S̄10 ! 2 " S̄12

p̃ + cw ωw p̃ + ω̄w2 − sin2 θ0 qw + (cos θ0 p̃2 βH − sin θ0 p̃2 ζH )
S̄2 S̄2

S̄12 S̄25
+ 2β0 (cos θ0 p̃ζH + sin θ0 p̃βH ) + p̃qv
S̄2 S̄2
“chapter_10(new)” — 2005/11/16 — page 365 — #31


S̄12 S̄16 S̄16 S̄25
+ + ē0 cos θ0 βH − ē0 sin θ0 ζH + sin θ0 cos θ0 qv
S̄2 S̄2 S̄2 S̄2

γ 1 dF̄z dF̄y
= γw δ cos θ0 − δ sin θ0 dx (10.67)
2 ē0 dx dx
Edgewise bending:
S̄49 ! 2 " S̄46

p̃ + cv ωv p̃ + ω̄v2 + sin2 θ0 qv + (sin θ0 p̃2 βH + cos θ0 p̃2 ζH )
S̄2 S̄2

S̄46 S̄25
− 2β0 (cos θ0 p̃βH − sin θ0 p̃ζH ) + p̃qw
S̄2 S̄2

S̄46 S̄48 S̄48 S̄25
+ + ē0 sin θ0 βH + ē0 cos θ0 ζH + sin θ0 cos θ0 qw
S̄2 S̄2 S̄2 S̄2

γ 1 dF̄y dF̄z
= γv δ cos θ0 + δ sin θ0 dx (10.68)
2 ē0 dx dx
where
d( ) 1 d( )
p̃ ≡ = (∗ ) = ; γ ≡ (Lock number) = ρacR4 /Ib (10.69)
dψ dt
and where cβ , cζ , cw , and cv represent structural damping coefficients for flapping,
lead-lag, flatwise bending, and edgewise bending, respectively.
The perturbational airloads are nondimensionalized according to
dF 1 dF̄
δ = ρac (R)2 δ (10.70)
dr 2 dx
and the matrix relationship between the basic perturbational descriptors and the
system variable perturbations is given by
   
p̃βH 
δ ŪT  0 x −γw sin θ0 γv cos θ0   
p̃ζH
δ ŪP = −x 0 −γw cos θ0 −γv sin θ0
  0 0 0 0  p̃qw 
 
δθ p̃qv
 
βH 
0 0 00 0  
ζH
+ −µ cos ψ 0 −γw µ cos ψ cos θ0 −γv µ cos ψ sin θ0
Lβ Lζ 0 0  qw 
 
qv
(10.71)
The use of such an implicit formulation allows for a more rigorous modeling
of the inflow. For purposes of generating useful results defining the stability char-
acteristics, the blade element momentum theory is used. This theory is standard
in rotor aerodynamic texts (e.g., Leishman) and treats the inflow by considering
“chapter_10(new)” — 2005/11/16 — page 366 — #32

the momentum and mass conservation laws on annuli of the rotor disk. This for-
mulation produces an inflow model with spanwise variation with a completely
arbitrary spanwise distribution of pitch angle. For the case of zero twist, the inflow
distribution is given by

−1 σ clα 32
λ(x) = UP = 1+ θ0 x − 1 (10.72)
R 0 16 σ clα
The resulting aerodynamic terms in the equations of motion result from spanwise
integrations and are thus implicit; they do reflect a more comprehensive modeling
of the perturbational airloads, however. This type of modeling with the attendant
spanwise integrations is actually easily implemented in a computer code.
The appropriate nondimensional natural frequencies for the blade in bending
are taken to be those for β0 = θ0 = 0:
ωw
ω̄w = ≡ first flatwise natural frequency (10.73a)

ωv
ω̄v = ≡ first edgewise natural frequency (10.73b)

10.4.2 Pitch-Lag Stability of Articulated Rotors

Rigid blade formulation. Classical pitch-lag instability was found to be a
concern early in the development of the helicopter and is one of the reasons for
attaching lead-lag dampers from the blades to the rotor hub. For articulated rotors
the instability is a low-frequency phenomenon and allows for the blade to be
considered “rigid.” The principal parameters affecting the stability are the pitch-
lag coupling and the lead-lag dampers. The general formulation presented in the
preceding section can be used to advantage for this case by omitting the blade
bending degrees of freedom and the corresponding dynamic equations. The result-
ing two-degrees-of-freedom equation set is simple enough to assess the stability by
expanding the characteristic determinant to get the quartic characteristic equation
and then to using the Routh stability criterion to obtain the stability boundaries
(see Section 9.2.2). The results presented herein were generated using a blade with
uniform mass and stiffness distributions with an offset ratio e0 equal to 0.05R and a
solidity σ equal to 0.05. The airfoil was assumed to be an NACA 0012-like airfoil
whose lift and drag coefficients are respectively given by
cl = cα α = 0.11α
cd = cd0 + cd2 α + cd4 α = 0.006 + (6.4421 × 10−5 ) α 2 + (9.4142 × 10−8 ) α 4
2 4
where the angle of attack α is in degrees. Also, the blade coning is calculated using
simple blade element theory:

γ 1 dF̄z [x, θ0 , λ(x, θ0 )]
β0 = x dx (10.74)
2ω̄β2 ē0 dx
“chapter_10(new)” — 2005/11/16 — page 367 — #33

where the nondimensional flapping frequency is given by
S̄1 KβH
ω̄β2 = 1 + ē0 + (10.75)
S̄2 S2 2
and, as defined in Appendix D, S1 and S2 are, respectively, the blade first and
second mass moments. For a uniform mass distribution m0 , Eq. (10.75) can be
simplified to
3 ē0 3KβH
ω̄β2 = 1 + + (10.76)
2 (1 − ē0 ) m0 R3 (1 − ē0 )3 2
For completeness the nondimensional lead-lag frequency is given by
S̄1 KζH 3 ē0 3KζH

ω̄ζ2 = ē0 + = + (10.77)
S̄2 S2 2 2 (1 − ē0 ) m0 R (1 − ē0 )3 2
3
Note that the barred (i.e., nondimensional) forms of all of the Si coefficients are
defined in Appendix D.
Figures 10.20a and 10.20b and 10.21 present the principal stability characteris-
tics for a nondimensionalized rotor configuration. Figure 10.20a is representative
of a completely articulated rotor with zero hub spring rates. Figure 10.20b shows
the similar characteristics as Fig. 10.20a, but for a case wherein the hub spring
rates KβH and KζH are finite, thereby raising the blade natural frequencies.
Stability Characteristics. A synopsis of the results shown in Fig. 10.20a is

as follows:
1) The results presented are for an articulated rotor wherein the hub stiffnesses
are zero.
2) Without any lead-lag damping the articulated rotor would be unstable for
practically any value of positive pitch-lag coupling. (Positive pitch-lag coupling
produces a pitch-down movement of the blades for aft lead-lag motion. This type
of pitch-lag coupling is generally the norm for leading-edge mounted pitch horns.)
3) With increasing lead-lag damping the stable region is widened, but is
diminished with increased collective pitch angle.
4) A secondary stability boundary exists for negative values of pitch-lag cou-
pling that is unaffected by damping. This lack of sensitivity to damping is because
this boundary is for a divergence condition. Again, increased collective angle is
detrimental to the range of stability.
Figure 10.20b presents results similar to those of Fig. 10.20a, but for a con-
figuration wherein finite hub stiffnesses raise the in-plane frequencies to values
typical of what a soft in-plane hingeless rotor would have. With finite values of
the hub stiffness springs, the nondimensional flapping and lead-lag frequencies
increased to 1.05P and 0.84P, respectively. This latter case is included as a prelude
to more comprehensive results for the nonarticulated or hingeless rotor. Generally,
the results of Fig. 10.20b indicate that even for configurations with positive lead-
lag coupling, some form of damping would still have to be introduced. However,
the higher stiffness of this configuration allows for a significant lessening of the
occurrence of the divergence with negative pitch-lag coupling. Another important
“chapter_10(new)” — 2005/11/16 — page 368 — #34

Fig. 10.20 Effects of system parameters on the pitch-lag stability boundaries—hover

condition.
parameter that impacts on aeromechanical instabilities is the Lock number. Figure

10.21 shows the impact of Lock number on the stability boundaries.
The destabilizing effect of increased Lock number (mass ratio) is clearly shown
in Fig. 10.21. Although not shown, this trend is also in evidence with the lower-
frequency articulated rotor. As a further introduction to the characteristics of
hingeless rotors, we again examine what can be learned from the simple two-
degrees-of-freedom modeling considered so far. Figure 10.22 presents the results
obtained by varying the lead-lag to flapping-frequency ratio from what would be
considered very “soft” conditions to substantially “stiff.” This variation is obtained
“chapter_10(new)” — 2005/11/16 — page 369 — #35

Fig. 10.21 Effect of Lock number on the pitch-lag stability boundaries (soft in-plane
hingeless rotor).
by simple variations of the lead-lag hub spring rate KζH . For hingeless rotors it
would be expected that there is generally less movement of the pitch horn/push
rod with lead-lag motion than what would be available to the fully articulated
rotor. Also for a hingeless rotor there is generally no lead-lag damper, and so the
damping parameter Cζ must be taken to be zero.
The figure clearly shows that the character of the stability essentially inverts
at the point wherein the lead-lag frequency is equal to the flapping frequency.
Fig. 10.22 Variation of pitch-lag stability boundaries with lead-lag frequency and
pitch angle—simplified hingeless rotor.
“chapter_10(new)” — 2005/11/16 — page 370 — #36

Consistent with the results presented in Figs. 20a and 20b, the effect of increased
collective angle, with the corresponding increase in coning angle, is destabilizing.
10.4.3 Pitch-lag Stability of Hingeless Rotors

Elastic blade formulation. The introduction of two new rotor concepts in
the 1960s, the hingeless rotor, and then a special case of the hingeless rotor, the
bearingless rotor, were major milestones in the general development of helicopter
technology. The principal driver for this development was the goal of producing a
simpler, more maintainable and, hence, less expensive rotor system. Two directions
were initially taken. One approach was to maintain a relatively stiff in-plane system
wherein the lead-lag blade frequency was greater than rotor speed (i.e., a subcritical
design). The major advantage of this system was the freedom from ground and
air resonance instabilities (covered in Chapter 11), and a perception of being a
generally low-risk approach. Another approach was to go with a soft in-plane
design, which would theoretically produce a lower vibration rotor, and perhaps
a lighter structure. The potential for rotor instabilities with these rotor types was
foreseen, and analytical development proceeded in part on the basis of extending
known articulated rotor problem areas.
As part of the development of the hingeless rotor, consideration had to be
given to the fact that the introduction of stiffnesses at the blade root now required
consideration of the blade elastic properties as well. Thus, for our purposes
the complete complement of equations governing the blade dynamics, that is,
Eqs. (10.65–10.68) must now be used for this examination. Furthermore, consid-
eration of the blades’ elastic properties now necessitates devising some form of
quantification of the blend of stiffnesses between the hub, as identified by the hub
spring rates, and the blade, as identified by the blade mass and bending stiffness
distributions.
Structural coupling factors. In the course of the investigations of the stability

characteristics of these new rotor systems, early investigators addressed the issue
of quantifying the stiffness blend. The pioneering analytical investigations (Ormis-
ton and Hodges, and later Peters) adopted a relatively simplistic elastomechanical
modeling, similar to that presented in Sec. 10.4.1. This modeling rejected the
potentially more rigorous capabilities of a modal approach in favor of an equiv-
alent rigid outer blade with an elaborate system of springs about an equivalent
hinge point. This system of springs was introduced to represent the flexibilities
of the blade outboard of the hinge that would have a different principal structural
axis orientation from that of the hub, as determined by the blade pitch angle θ0 .
Ormiston and Hodges further introduced a structural coupling factor R, which,
although being somewhat cumbersome to apply, had the useful characteristic of
quantifying the relative distribution of the blade’s principal flexibilities. Thus, a
value for R of zero corresponded to a completely stiff outer blade and soft hub
stiffness and an R value closer to unity corresponded to a soft blade with a stiff
hub. This concept can be extended to the development of Sec. 10.4.1, wherein we
now consider two separate coupling factors, one for the flapping/flatwise degrees
“chapter_10(new)” — 2005/11/16 — page 371 — #37

of freedom Rβ and one for the in-plane/edgewise degrees of freedom Rζ :

1/KβB KβH
Rβ = = (10.78)
1/KβB + 1/KβH KβH + KβB
1/KζB KζH
Rζ = = (10.79)
1/KζB + 1/KζH KζH + KζB
where the hub stiffness KβH and KζH are as defined in Fig. 10.18 and equivalent
“blade” springs are defined by considering the strain energy in the deflected blade:
1 2 2
1 d γw
KβB = 3 EIflat dx (10.80a)
R ē0 dx 2
1 2 2
1 d γv
KζB = 3 EIedge dx (10.80b)
R ē0 dx 2
Because the modes are the normal modes for the blade, a more accurate definition
for the equivalent blade springs can be alternatively obtained by considering the
inertia load parts of the modal equations.

S10 1 1 dγw 2
KβB = ωw2 2 − T (x) dx KζB (10.81a)
R R ē0 dx

S49 1 1 dγv 2
KζB = (ωv2 + 2 ) 2 − T (x) dx (10.81b)
R R ē0 dx
where T (x) is the distributed tension.
The structural coupling factors defined by Eqs. (10.77) and (10.78) thus serve
to define how the stiffnesses are distributed: low values denote stiff outer blades
with relatively soft hubs, and high values denote stiff hubs with relatively soft
outer blades.
Coupled frequencies. To facilitate the investigation further, a means must

be devised for calculating the coupled frequencies of the blade, where the coupling
is now caused by the combination of both hub and blade flexibilities. It is espe-
cially convenient to calculate the separate hub stiffnesses KβH and KζH and blade
stiffnesses EIflat and EIedge from specified values of the coupling factors. To this
end, we take pairs of equations from the equation set [Eqs. (10.65–10.68)], assume
zero pitch and coning angles, and form characteristic equations for the coupled
frequencies for the two resulting degrees-of-freedom systems. Thus, the charac-
teristic equations for the two systems are formed from the following determinant
equations:
Flapping system:
# #
# S̄12 S̄16 ##
# (ω̄β − ω̄ )
2 2 (1 − ω̄ ) + ē0
2
# #
# S̄2 S̄2 #
# #=0 (10.82)
# S̄12 S̄16 S̄10 2 #
# #
# S̄ (1 − ω̄ ) + ē0 S̄ (ω̄w − ω̄ )
2 2
S̄ #
2 2 2
“chapter_10(new)” — 2005/11/16 — page 372 — #38

Lead-lag system:
# ##
# S̄46 2 S̄48
# (ω̄ζ2 − ω̄2 ) − ω̄ + ē0 #
# #
# S̄2 S̄2 #
# #=0 (10.83)
# #
# − S̄46 ω̄2 + ē0 S̄48 S̄49 2
(ω̄v − ω̄2 ) #
# S̄ S̄ S̄2 #
2 2
where ωβ and ωζ are given by Eqs. (10.76) and (10.75), respectively.We first
rearrange the characteristic equation resulting from Eq. (10.82) to isolate the hub
flapping spring KβH :
S̄1 KβH [S12 (1 − ω̄2 ) + e0 S16 ]2

ω̄β2 = 1 + ē0 + = ω̄ 2
+ (10.84)
S̄2 S2 2 S2 S10 (ω̄w2 − ω̄2 )
Or, after some rearranging, invoking Eq. (10.81a), and relating the hub spring to
the blade spring using the coupling factor, we obtain an equation for the equivalent
blade spring KβB :

(1 − Rβ ) S̄1 [S12 (1 − ω̄2 ) + e0 S16 ]2
KβB − S2 ω̄ − 1 − ē0 −
2 2
=0
Rβ S̄2 S2 [−S10 ω̄2 + KCFw + R2 KβB ]
(10.85)
where
1 dγ 2
w
KCFw = R T (x) dx (10.86)
ē0 dr
Equation (10.85) is then seen to be essentially a quadratic equation for KβB

whose coefficients are implicit functions of the blade bending stiffness EIflat .
Equation (10.85) must be solved iteratively, wherein the effective coupled flapping
frequency ω [= (ωβ0 )C ] and coupling factor Rβ are first selected, and the bending
stiffness is then adjusted to the point where Eq. (10.85) is consistent with Eq.
(10.80a). In a similar manner the equations governing the solution for KζB and
EIedge are Eq. (10.80b) and the following equations:

(1 − Rζ ) S̄1 [−S46 ω̄2 + e0 S48 ]2
K ζB − S2 ω̄ − ē0 −
2 2
=0
Rζ S̄2 S2 [−S49 (1 + ω̄)2 + KCFv + R2 KζB ]
(10.87)
where
1 dγ 2
v
KCFv = R T (x) dx (10.88)
ē0 dr
and the coupled frequency ω for this case is now the effective coupled lead-lag
frequency (ωζ 0 )C .
“chapter_10(new)” — 2005/11/16 — page 373 — #39

Stability characteristics. The more rigorous investigation into the pitch-

flap-lag stability characteristics of hingeless rotors must now use the full set of
dynamic equations given by Eqs. (10.65–10.68). Using the calculations for the
effective coupled frequencies, we can then determine the effects of variable cou-
pling on the stability characteristics. The results of this investigation are presented
in Figs. 10.23a–10.23e. All of these figures present the stability boundaries for
variations in pitch angle and coupled lead-lag frequency for various combinations
of the structural coupling factors Rβ and Rζ : a) equal low values of Rβ and Rζ ,
b) low Rβ and moderate Rζ , c) moderate Rβ and low Rζ , d) equal moderate values
of Rβ and Rζ , and e) equal high values of Rβ and Rζ .
Figure 10.23a shows results for low values of both structural coupling factors.
These results are to be compared with those of Fig. 10.22, which are essentially
the same results but for zero Rβ and Rζ . Even the small amount of shifting of
the flexibility from the hub to the outer blade portion is seen to be somewhat
stabilizing. Figures 10.23b and 10.23c show the effects of alternately mixing low
(0.05) and moderate (0.50) levels of structural coupling between the flapping and
lead-lag systems. Figure 10.23b shows the moderate improvement in stability
attained by selectively softening the outer portion of the blade only in edgewise
bending stiffness. Figure 10.23c, however, shows the substantial improvement in
stability for negative values of pitch-lag coupling Lζ , but a degradation in stability
for positive values. Figures 10.23d and 10.23e, respectively, present the stability
characteristics for configurations that have (equal) moderate and high values of
the structural coupling factors. The beneficial effect on stability of softening of the
outer blade relative to the hub is clearly shown. Note that for all of these figures
the effective coupled flapping and lead-lag frequencies are each determined as per
Eqs. (10.85) and (10.87) while using differing combination of hub stiffness and
blade bending stiffness distributions, in accordance with the selected values of the
structural coupling factors.
10.4.4 Flap-Lag Stability Characteristics of Hingeless Rotors

In the early investigation of the dynamics of hingeless rotors, it was found that
a generalization of the pitch-lag instability problem evolved wherein instability
could be found to occur even without the explicit mechanical coupling afforded
by the geometric pitch-lag coupling Lζ . This form of the phenomenon was even-
tually deemed flap-lag instability because it was found to principally involve only
flapping and lead-lag motions of the blade. The equations presented earlier [Eqs.
(10.65–10.68)] are well formulated to predict this type of instability. All of the
development presented earlier for investigating pitch flap lag is applicable to
the flap-lag problem with the expedient of setting the pitch-lag coupling Lζ to
zero.
Hover condition. With the work of Ormiston and Hodges as a guide, varia-
tions can be made with both the effective coupled flapping and lead-lag frequencies,
as developed in Eqs. (10.83) and (10.85), to establish mappings of the resulting
stability boundaries. The selected method of modeling of the rotor blade (i.e., with
elastic bending modes supported by discreet springs) is intended to model only the
resulting dynamic responses that are roughly in phase. Mathematically, as part of
“chapter_10(new)” — 2005/11/16 — page 374 — #40

Fig. 10.23 Variation of pitch-lag stability boundaries with coupled lead-lag

frequency, pitch-lag coupling, and pitch angle—hingeless rotor with elastic blade
formulation.
“chapter_10(new)” — 2005/11/16 — page 375 — #41

Fig. 10.23 Variation of pitch-lag stability boundaries with coupled lead-lag

frequency, pitch-lag coupling, and pitch angle—hingeless rotor with elastic blade
formulation. (continued).
the eigenvalue solution out-of-phase modes are also predicted. These modes are
found to be of very high frequency and can be considered to be “phantom” modes
and not realistic. To suppress any unstable conditions for these modes, a small
amount of equivalent structural damping is introduced. In the results to follow, all
four of the degrees of freedom were given values of damping amounting to 0.1%
of critical (i.e., cβ = cζ = cw = cv = 0.002).
Figures 10.24 and 10.25 present stability boundaries as a function of blade
coupled frequencies and blade pitch angle, for the hingeless rotor as calculated
with the fully elastic formulations. Figure 10.24 presents results for low values of
the structural coupling factors (Rβ and Rζ = 0.02), commensurate with a relatively
stiff outer blade and soft hub attachment. On the other hand, the results of Fig. 10.25
are for a configuration of moderate structural coupling (Rβ and Rζ = 0.5).
“chapter_10(new)” — 2005/11/16 — page 376 — #42

Fig. 10.24 Variations of flap-lag stability boundaries with coupled frequency and
pitch angle for a hingeless rotor blade with low structural coupling.
Fig. 10.25 Variations of flap-lag stability boundaries with coupled frequency and
pitch angle for a hingeless rotor blade with moderate structural coupling.
The trends shown in these figures are consistent with the findings of Ormiston
and Hodges:
1) The instability is limited mostly to subcritical rotors (i.e., those with in-plane
frequencies greater than rotor speed).
2) The instability is most intense for those conditions wherein the in-plane
frequency is close to that for the flapping motion.
3) The effect of increased structural coupling is to translate the instability
boundaries to higher values of in-plane frequency.
“chapter_10(new)” — 2005/11/16 — page 377 — #43

4) The effect of collective angle above some minimum value is destabilizing.

Below that value the rotor is stable. Thus, stability could be an issue for high load
factor flight maneuvers.
Although not identified by Ormiston and Hodges, the effect of structural damp-
ing is very pronounced. For the generic configuration examined, even 0.5%
damping in the modes is sufficient to completely stabilize the motion.
Forward-flight condition. Although the coupled equation set formulated

earlier can be applied to the forward-flight condition, the results of Peters are
applicable and examined instead. Using a vigorous application of Floquet theory
to the problem, Peters established that the role of trimming the rotor is critical to
the calculation of credible stability levels in forward flight. Figure 10.26 presents
some of the key findings of Peters. In particular, besides presenting results for
grossly out-of-trim conditions, the results for trimmed conditions include zeroing
out the roll and pitch moments, and alternatively trimming or not trimming to
propulsive force, as represented by a fuselage equivalent drag area of 0.01 rotor
area.
The exact trim conditions for any particular rotor blade are quite configuration
and flight condition dependent, and a full mapping of possible trimming scenarios
is beyond the intended scope of this text. The results shown in Fig. 10.26 amply
make the point that a calculation of forward-flight stability levels should include
a rigorous accounting of trim.
Fig. 10.26 Effects of trim conditions and inflow modeling on blade flap-lag stabil-
ity in forward flight: ωβ = 1.15, ωζ = 1.4, R = 0, γ = 5, σ = 0.05, cd0 = 0.01, and
CT /σ = 0.2 (after Peters).
“chapter_10(new)” — 2005/11/16 — page 378 — #44

Nonlinear effects. The four-degrees-of-freedom formulation presented ear-

lier is a linear model of the stability dynamics, and the numerous nonlinear effects
that could be introduced were omitted for the sake of clarity. One nonlinearity
that bears discussion, however, is that afforded by the nonlinear torsion coupling
described in Sec. 3.6.6. For blade configurations that have sufficient torsional flex-
ibility, the effects of this nonlinear effect would be expected to impact on the
stability. Without a comprehensive examination of the effects of this coupling, a
heuristic argument can be made that the presence of a steady flapping moment
will introduce a form of pitch-lag coupling, quite similar to that examined earlier.
The results presented in Figs. 10.23d and 10.23e would indicate that hingeless
rotors would benefit from a negative value of pitch-lag coupling, such that rear-
wise motion (i.e., bending) of the blade produces positive pitch change. Generally,
this would require a steady positive flatwise bending. Thus, a favorable use of the
nonlinear torsion coupling requires that little or no built-in precone angle be used.
10.5 Rotordynamic Instabilities

A variety of sources of destabilizing phenomena exist with rotary shaft systems.
As developed in Chapter 5, rotary shafts can demonstrate whirling motion in both
the forward and retrograde directions, and instabilities can exist in both types of
whirl. As developed by Ehrich, Fig. 10.27 identifies the various forces at work in
the whirling motion of a rotary shaft.
Fig. 10.27 Force balance for a forward whirling or whipping shaft (after Ehrich).
“chapter_10(new)” — 2005/11/16 — page 379 — #45

Clearly, two rotations are in evidence, the rotor rotational rate , as well as a
whirling frequency ±ω, the sign depending on the direction of whirl. Most of the
forces acting on the rotor are either in phase or out of phase with the deflection
or act against the radial velocity of the motion, dr/dt (i.e., radial damping) or
at 90 deg to the radial displacement, as defined by Coriolis forces and available
damping of the whirling motion. Only the generation of a force in phase with the
whirling motion and proportional to the deflection of the motion Fθ , as shown
in the figure, can lead to instability. Note that this is a graphical interpretation of
the force-phasing matrix idea discussed in Chapter 9. A complete discussion of
all of the types of instability for rotary shafts is beyond the intended scope of the
present text. The interested reader is referred to the work of Ehrich, from which
the present material is excerpted. The intent herein is only to identify the various
destabilizing effects and the physics involved in generating forces that are in phase
with the direction of whirl.
10.5.1 Internal Damping—Hysteretic Whirl

The origin of this type of instability relates to the lag in internal stress relative
to the internal strain for a laterally deflected shaft. As shown in Fig. 10.28, the
hysteretic lag in the internal stress relative to the neutral strain axis develops the
component of internal stress that is in phase with the whirl velocity.
Reducing the onset of this type of instability poses a problem in that the presence
of stress hysteresis is a fact of material behavior. One basic remedy for avoiding
this instability is to limit lateral deflections with the use of subcritical operation.
Fig. 10.28 Hysteretic whirl (after Ehrich).
“chapter_10(new)” — 2005/11/16 — page 380 — #46

This solution, however, would limit the use of the lighter-weight designs possible
with supercritical operation. Another approach is to minimize the number of shaft
elements connected by couplings that provide damping in their interfaces.
10.5.2 Hydrodynamic Bearings and Seals

A necessary component of any journal bearing installation is its lubrication.
Of necessity, the journal will be of smaller diameter than its housing, and the
clearance will be occupied by the entrained lubrication. Because of its viscosity,
the lubricant circulates around the journal at approximately half the rotor speed
. As shown in Fig. 10.29, a whirling journal will create an asymmetric pressure
distribution around the circumference of the journal.
The hydrodynamic characteristics of the viscous fluid cause the pressure on
the upstream side of the clearance to be higher than on the downstream side. This
unsymmetrical distribution then creates a force component that is in phase with the
whirling motion. Instability ensues when this component exceeds already present
inherent damping forces. Simple relationships have been developed to estimate
the rotor speed for the onset of instability. The analysis begins with a calculation
of the whirl ratio w , which for any plain journal bearing is calculated using the
Fig. 10.29 Fluid bearing whip (after Ehrich).
“chapter_10(new)” — 2005/11/16 — page 381 — #47

effective cross-coupling stiffness krθ and radial damping cr :
krθ
w = (10.89)
cr
where the cross-coupling stiffness and radial damping are dependent on the eccen-
tricity ratio ε0 , which measures where in the journal housing the journal is
operating. A zero value denotes a centered position, whereas a unit value denotes
contact of the journal with the wall. Data on the variations of stiffness and damping
characteristics with eccentricity ratio are usually provided by the bearing manu-
facturers. For typical plain journal bearings the whirl ratio typically comes out to
approximately one-half. The whirl ratio is then used to approximate the onset of
instability (OSI):
ωn
OSI = (10.90)
w
where ωn is the rotor shaft critical speed. Thus, instability would be expected
to occur at approximately twice the lowest shaft critical speed. As discussed in
detail by Ehrich, a number of devices have been developed to alleviate this type
of instability.
10.5.3 Aerodynamic Cross-Coupling—Tip-Clearance Excitation

One form of instability typically found in turbomachinery relates to the genera-
tion of local pockets of increased and decreased aerodynamic efficiency caused by
tip clearances. As shown in Fig. 10.30, a laterally deflected turbomachinery stage
will cause the side nearest to the wall to close, thereby increasing its aerodynamic
efficiency, while the side opposite will have an increased tip clearance rendering
a decrease in efficiency.
The dissymmetry in local aerodynamic efficiency produces an increased turbine
force for the closer side and a reduced force on the opposite side. For a forward
whirl, a net force in the whirl direction is produced, thereby producing the desta-
bilizing force in phase with the whirl direction. Empirical experience suggests a
dependence of stability on the pressure parameter P2 P, where P2 is the inlet
pressure and P is the pressure change across the turbine stage. Other forms of
stabilization would include special treatments to the wear-ring seal.
10.5.4 Dry Friction Rubbing

Although the operation of a rotating shaft in its housing without or loss of lubri-
cation would reasonably be a prime facie violation of good operation procedure,
the situation bears examination because it is a clear example of the physics of
instabilities. Figure 10.31 shows a rotating shaft in a static housing with a finite
clearance between the two.
As noted earlier, the shaft can experience whirling motion in either the forward
or retrograde directions. As shown, if the whirl is in the retrograde direction and
contact is made between the shaft and its housing, coulomb friction will generate
“chapter_10(new)” — 2005/11/16 — page 382 — #48

Fig. 10.30 Tip-clearance excitation (after Ehrich and Childs).
Fig. 10.31 Dry friction whip (after Ehrich).
“chapter_10(new)” — 2005/11/16 — page 383 — #49

a contact force opposite to the shaft’s rotation, and hence, in the direction of
the whirling motion. Furthermore, the coulomb friction force can be expected to
increase with increased deflection. Thus, both conditions are met for establishing
the existence of an instability. The obvious remedy for this type of instability is to
provide lubrication. If that is not feasible, abradability of either of the two elements
should be provided. Also, whirl of the shaft with its housing can be controlled by
maintaining the natural frequencies of the two elements to as dissimilar as possible.
References
Section
10.1 Brosens, P. J., and Crandall, S. H., “Whirling of Unsymmetri-
cal Rotors,” Journal of Applied Mechanics, Vol. 83, Sept. 1961,
pp. 355–362.
Crandall, S. H., and Brosens, P. J., “On the Stability of a Rotor with
Rotationally Unsymmetric Inertia and Stiffness Properties,” Journal
of Applied Mechanics, Vol. 83, Dec. 1961, pp. 567–570.
Gladwell, G. M., and Stammers, C. W., “On the Stability of an
Unsymmetrical Rigid Rotor Supported in Unsymmetrical Bearings,”
Journal of Sound and Vibration, Vol. 3, No. 3, 1966, pp. 221–232.
Wilkerson, J. B., “An Analytical Study of the Mechanical Stability
of Two-Bladed Rigid Rotor Systems,” Naval Ship Research and
Development Center Report 3351, Jan. 1970.
10.2 Dowell, E. H., Curtiss, H. C., Scanlan, R. H., and Sisto, F., A Mod-
ern Course in Aeroelasticity, Sijthoff & Noordhoff, Rockville, MD,
1978.
Peters, D. A., “Flap-Lag Stability of Helicopter Rotor Blades in For-
ward Flight,” Journal of the American Helicopter Society, Vol. 20,
No. 4, 1975, pp. 2–13.
10.4 Ormiston, R. A., and Hodges, D. H., “Linear Flap-Lag Dynamics
of Hingeless Helicopter Rotor Blades in Hover,” Journal of the
Peters, D. A., “Flap-Lag Stability of Helicopter Rotor Blades in For-
ward Flight,” Journal of the American Helicopter Society, Vol. 20,
No. 4, 1975, pp. 2–13.
10.5 Childs, D. W., “Identification and Avoidance of Instabilities in High-
Performance Turbomachinery,” Von Karman Institute for Fluid
Dynamics, lecture, Rhode-Saint-Genese, Belgium, 1987.
Ehrich, F. F., Handbook of Rotordynamics, McGraw–Hill, New York,
1992.
Ehrich, F. F., “Self-Excited Vibration,” Shock and Vibration Handbook,
Third Edition, edited by C. M. Harris, McGraw–Hill, New York,
1988, Chap. 5.
“chapter_10(new)” — 2005/11/16 — page 384 — #50

Problems
10.1 For the unsymmetrical rotor instability phenomenon consider the com-
pletely isotropic case ( i = s = 0), but with both internal and external
viscous damping (both isotropic) ci and ce , respectively. Denote nondi-
mensional forms of these damping sources as
σi = ci /2I; σe = ce /2I; where I = Iξ = Iη (10.91)
(a) Derive the additional (nondimensional) terms in the equations of

motion for the inclusion of both internal and external sources of damping.
(b) Examine the stability condition in terms of the nondimensional
natural frequency:

ω= k/I; (k = ka = kb ) (10.92)
the axial inertia term A and the nondimensional damping values σi and σe .
(c) Relate the result of part b to Bishop’s findings for stable operation
of supercritical shafting.
10.2 The effects of high-speed compressibility are generally confined to the

blade-tip region on the advancing side wherein the local Mach numbers
can reach values in excess of the drag-divergence values Mdd for the tip
blade sections. Two effects of these high blade-tip Mach numbers are, first,
that the lift-curve slope a is increased by the Prandtl–Glauert correction:

[a]M>0 = [a]M=0 / (1 − M 2 ) (10.93)
and, second, that a substantial drag rise is incurred. Using the ratio of the
drag-divergence Mach number to the advancing blade-tip Mach number
(= Mdd /MR ), formulate a Fourier-series representation for the value of
the lift-curve slope for radii that experience Mach numbers greater than or
equal to the drag-divergence value.
10.3 Derive an expression for the (periodic) flap damping of an articulated rotor
blade (i.e., δUp = −xRδ β̇) taking into account reversed-flow effects. [Note:
Group the linear aerodynamic descriptor constants together into a single
aerodynamic effectivity constant Aa (= ρ/2 acR3 R).]
10.4 Derive the inertial loading terms in the equations of motion for rotor
weaving using a Newtonian approach and the theorem of Coriolis.
10.5 A not uncommon feature of teetering rotors is the existence of built-in

geometric pitch-flap coupling or the “δ3 effect.” This effect consists of a
direct mechanical gearing between flapping and pitching so that the blade
“chapter_10(new)” — 2005/11/16 — page 385 — #51

is pitched incrementally in proportion to flapping:

∂θ
θ = β = β β (10.94)
∂β
where β = − tan δ3 .
Derive the additional terms in the equations of motion for rotor weaving to
account for pitch-flap coupling.
Hint: The basic equation set can be represented in abbreviated matrix
form as

β Lβ
[L( p̃)] = = {0} (10.95)
θ Lθ
The action of the pitch-flap coupling can be represented as
βnew = βold ; θnew = θold + β βold (10.96)
or, in matrix form,

β 1 0 β β
= = [J] (10.97)
θ new β 1 θ old θ old
Finally, it can be shown that a consistent statement of the virtual work

caused by virtual displacements of the variables requires that

T Lβ
[J] = {0} (10.98)
Lθ
10.6 Show that for equal values of the structural coupling factors Rβ and Rζ ,
they are also equal to the structural coupling factor given by Peters:
1/KβB − 1/KζB
R≡ (10.99)
[1/KβB + 1/KβH ] − [1/KζB + 1/KζH ]
“chapter_10(new)” — 2005/11/16 — page 386 — #52

11
Mechanical and Aeromechanical Instabilities
of Rotor-Pylon Systems
11.1 Multiblade Coordinates and Rotor Modes

The stability phenomena described in this chapter are characterized by signif-
icant motion of the rotor hub with or without motion of the several rotor blades
relative to the (rotating) hub. A further characterization of these phenomena is
that the system motion (the vector of total system degrees of freedom) consists
of components that couple with each other and are measurable in both the rotat-
ing coordinate system (the rotor blades) and in the nonrotating coordinate system
(the rotor pylon). If the system modeling were to retain this combination of both
rotating and nonrotating degrees of freedom, three resulting difficulties would
emerge:
1) The total number of system degrees of freedom would equal the maximum
possible needed to describe the dynamics of the phenomenon. For example, for
a four-bladed (elastic) rotor involving six modes (e.g., three flatwise bending,
two edgewise bending, and one torsion) along with five pylon degrees of freedom
(e.g., three translations and the roll and pitch rotations), the total number of system
degrees of freedom would be 29. Thus, the attendant arrays (mass, damping, and
stiffness matrices) needed to define the mathematics would be relatively large and
would thereby maximize the required computer resources and the attendant costs
for obtaining solutions.
2) The response solutions obtained (either by an eigenvalue solution or by a
direct time-history solution) would inherently contain all responses whether they
represent responses that are coupled or responses that are uncoupled. Thus, effort
would have been spent obtaining some response information that is either not of
direct interest or is obtainable from a simpler analysis that does not contain one or
the other of the two subsystems.
3) The coupling terms in the equations of motion (the forces in the rotating
coordinate system accruing from motion in the nonrotating coordinate sys-
tem, and vice versa) would necessarily contain periodic terms. Such a system
would represent a “pseudo-periodic coefficient system” and would not necessarily
exhibit the characteristics distinctive of systems with true periodic coefficients:
aliasing and the presence of parametric instabilities (i.e., those involving char-
acteristic frequencies that are integral harmonics and/or pure fractions of a
harmonic).
387
“chapter_11(new)” — 2005/11/24 — page 387 — #1

11.1.1 Collective and Cyclic Rotor Modes

An alternative to the concurrent use of both rotating and nonrotating coordinates
is to use a description of the rotating blade motion that 1) incorporates variables
which can be defined in the nonrotating coordinate system, 2) isolates those blade
motions that actually couple with pylon motion, and 3) is relatively independent of
the number of rotor blades. Such a blade motion description can be achieved with
the use of so-called rotor modes or, alternatively, multiblade coordinates. In its
simplest form the idea is to represent each of the motions of any individual blade
as the sum of two basic components: One component defines the generally low-
frequency motion that all of the blades share; this motion we denote the collective
responses. The second component can be compared to an amplitude modulated
“signal” whose “carrier frequency” is the rotor frequency and whose Fourier
component “amplitudes” are the rotor modes. This motion we denote as the cyclic
responses. As an example, consider the flapping motion of the kth rotor blade:
βk (ψ) = β0 (ψ) + β1c (ψ) cos [ψ + 2π(k/b)]
+ β1s (ψ) sin [ψ + 2π(k/b)] k = 1, 2, . . . , b (11.1)
If the rotor were to be viewed as rotating inside a thin elastic membranous (flat)
shell so that the shell did not rotate with the rotor but assumed the volume in space
of the rotor, then the quantities β0 , β1c , and β1s would respectively represent the
umbrella-like motion and the longitudinal and lateral tilt angles of the membrane.
This concept can be generalized to any of the rotating degrees of freedom used to
define the motion of the blade. Thus, the mth elastic flatwise bending mode gener-
alized coordinate of the kth rotor blade qw(k)m (ψ) can be similarly expressed using
q0 (ψ), q1c (ψ), and q1s (ψ), where these Fourier amplitudes respectively represent
circumferential “distortions” of the membrane in the vertical, longitudinal, and
lateral directions. The spanwise variation of the distortion would be defined by the
mth spanwise mode shape γwm (x).
One further example that will prove useful in the developments to follow is the
application of this concept to the case of in-plane (or, alternatively, edgewise) blade
elastic deformations . Again, we assume the existence of a nonrotating membrane
surrounding the rotor and occupying the same volume. For blade motion defined
by the in-plane bending mode generalized coordinate, the membrane will again
distort, but this time in the plane of rotation. Thus, let the in-plane motion of the
kth blade be defined by the following rotor mode description:
k (ψ) = 0 (ψ) + y (ψ) cos [ψ + 2π(k/b)]
− x (ψ) sin [ψ + 2π(k/b)] k = 1, 2, . . . , b (11.2)
The 0 (collective) motion represents a “winding-up” distortion of the membrane
about the axis of rotation. The x and y (cyclic) motions represent distortions that
can be visualized by the depiction given in Fig. 11.1.
11.1.2 Rotor Center of Gravity

The use of rotor modes greatly facilitates the definition of the position of the
center of gravity of the rotating rotor in the nonrotating coordinate system. For
“chapter_11(new)” — 2005/11/24 — page 388 — #2

INSTABILITIES OF ROTOR-PYLON SYSTEMS 389
Fig. 11.1 Depiction of the in-plane cyclic rotor modes.
flapping and flatwise bending the collective rotor modes represent a measurement
of the out-of-plane displacement of the rotor center of gravity from the undeflected
rotor plane. Likewise the in-plane lead-lag and edgewise bending cyclic rotor mode
descriptions provide a corresponding measurement of the in-plane position of the
rotor center of gravity position. Similarly, the cyclic flatwise bending rotor modes
define the pitching and rolling rotational motion of the rotor, and the collective
edgewise bending rotor modes define the yawing rotational motion of the rotor
from which the effective torsional inertia effects can be defined.
The importance of the collective and cyclic components of the rotor mode
description arises from the fact that it is only these modes that elastomechanically
couple with the pylon. A complete rotor mode description of the rotor could include
a Fourier-series expansion that would contain second (and higher) harmonic com-
ponents. However, these portions of the expansion involve neither translations
nor rotations of the rotor center of gravity and, hence, do not produce hub loads
(either forces or moments); these rotor modes are essentially “reactionless.” For
this reason further consideration of them will be omitted herein.
11.2 Rotor-Nacelle Whirl Flutter

Rotor- (propeller) nacelle whirl flutter was first identified as a dynamic in-
stability phenomenon of propellers at a time (around 1938) when aircraft were
configured with reciprocating engines that were relatively massive compared to the
propellers they powered. Consequently, typical engine-propeller installations then
defined ranges of parameters pertinent to this phenomenon that could not produce
the instability. The phenomenon was duly forgotten until its “rediscovery” in 1960,
when two turboprop aircraft were fatally lost. Subsequent investigations showed
“chapter_11(new)” — 2005/11/24 — page 389 — #3

that the propeller-nacelle whirl flutter phenomenon was the most probable cause
of the loss of the aircraft. Retrofit stiffenings were made to the remaining aircraft
of this type, based on knowledge of the stabilizing trends for this phenomenon,
and thereafter no more aircraft were lost.
The study of this phenomenon forms a convenient transition from the material
of the preceding chapter to that of the present one. The basic characteristics of
the more generally defined rotor-nacelle whirl flutter are easily defined using a
relatively simple dynamic representation; at the heart of the formulation is the
dynamics of a relatively simple gyroscopic element. As such, the theory builds not
only on the gyroscopic theory tools covered in an earlier chapter but also on the
material relating to the unsymmetrical rotor of the preceding chapter.
11.2.1 Elastomechanical Description

As shown in Fig. 11.2, the basic dynamic system consists of the rotor (propeller),
forming the gyroscopic and aerodynamic element, supported horizontally by a
pylon that is pivoted at some wing attachment point. The spin axis of the propeller
is oriented more or less into the direction of the freestream flight velocity V with
angular perturbations from this vector in pitch θ and yaw ψ.
Let us denote the angular moment of inertia of the pylon (including the diametral
moment of inertia of the rotor and nacelle inertia) about the pivot point as In and
the polar moment of inertia of the rotor as Ix ; thus, the following basic equations
of motion can be written for this dynamic system:
In θ̈ + Cθ θ̇ + Kθ θ − Ix ψ̇ = Mθ (11.3)
Ix θ̇ + In ψ̈ + Cψ ψ̇ − Kψ ψ = Mψ (11.4)
Fig. 11.2 Schematic of basic elastomechanical features of the rotor-nacelle dynamic

system.
“chapter_11(new)” — 2005/11/24 — page 390 — #4

where Mθ and Mψ are, respectively, the applied moments in pitch and yaw, taken
about the pivot point, arising from the aerodynamic loads.
11.2.2 Aerodynamic Description Using Quasisteady Theory

Using the techniques of the preceding chapter, the perturbational aerodynamic
loads can be formulated using strip theory, wherein only the vector diagram of
the air velocities at a typical section need be considered. The principal quantities
that must be mathematically modeled are the total air velocity at the section and
its components in and out of the plane of the rotor disk; the geometric angle of
attack change caused by pitch and yaw α1 ; and the change in the inflow angle
caused by the perturbational velocities caused by pitch and yaw rates. As viewed
from the front of the rotor, in Fig. 11.3, a typical section is located at an arbitrary
spanwise location r of a blade positioned at some azimuth angle t and with the
hub displaced at an arbitrary position in pitch and yaw.
The pertinent geometrical features of the typical airfoil section are then shown
in Fig. 11.4. The tangent and perpendicular components of velocity at the blade
section are given by
UT = r + ṡ + V sin α1 (11.5)
UP = ẇ + V cos α1 (11.6)
where
α1 = ψ sin t − θ cos t (11.7)

ṡ = aR(θ̇ cos t − ψ̇ sin t) (11.8)
ẇ = −r(ψ̇ cos t + θ̇ sin t) (11.9)
Fig. 11.3 Front view geometry of deflected hub and radially located typical section.
“chapter_11(new)” — 2005/11/24 — page 391 — #5

Fig. 11.4 Air velocity components and airload components at the typical blade
section.
The square of the effective resulting velocity (that producing dynamic pressure) is
then approximated by
Ue2 = (2 r 2 + V 2 ) + 2rṡ + 2V ẇ + 2rV α1 + (higher-order terms) (11.10)

U2
The lift at the airfoil section can then be expressed in terms of the effective
dynamic pressure and the angle of attack of the section:

dL(r, t) 1
= = ρUe2 · [clα α] ·
c
dr 2
cl chord
dynamic
pressure
1
= ρclα c (U 2 + 2V ẇ + 2rṡ + 2rV α1 ) · (β − γ ) (11.11)
2
where
β − γ = β − tan−1 (UP /UT )
r V V2
= β − tan−1 (V / r) − 2 ẇ + 2 ṡ + 2 α1 (11.12)
U U U
≡ α0
After the various formulations are combined, the resulting expression can be
expanded wherein only the linear terms are retained. The following expression
for the lifting load distribution is then obtained:

dL(r, t) 1 r 2V
= ρclα cU α0 − 2 1 −
2
α0 ẇ
dr 2 U r

V 2r V2 2r
+ 2 1+ α0 ṡ + 2 1 + α0 α1 (11.13)
U V U V
“chapter_11(new)” — 2005/11/24 — page 392 — #6

or, after including the expressions for ẇ, ṡ, and α1 :

dL(r, t) 1 r 2V
= ρclα cU α0 + 2 1 −
2
α0 r(ψ̇ cos t + θ̇ sin t)
dr 2 U r
V 2r
+ 1+ α0 aR(θ̇ cos t − ψ̇ sin t)
U2 V

V2 2r
+ 1+ α0 (ψ sin t − θ cos t) (11.14)
U2 V
This lift distribution can then be used to calculate the total perturbational in-
plane forces and pitching and rolling moments at the hub accruing from all of the
blades. Within the context of the simplicity of the quasi-steady assumption, the
terms involving α0 can be neglected resulting in the following expressions for
the perturbational forces and moments at the hub:
b ψ̇ θ̇
Ly = Ka A1 ψ − aA1 + A2 (11.15)
2
b θ̇ ψ̇
Lz = Ka A1 θ − aA1 − A2 (11.16)
2
b ψ̇ θ̇
My = Ka R A2 ψ − aA2 + A3 (11.17)
2
b θ̇ ψ̇
Mz = Ka R A2 θ − aA2 − A3 (11.18)
2
where b is the number of blades, and
Ka = 21 ρclα R4 2 (11.19)
The various integrals given in the preceding expressions are defined as follows:
1 c (J/π )2
A1 = dη; A1 = (J/π )A1 (11.20)
0 R (J/π )2 + η2
1 c (J/π )η2
A2 = dη; A2 = (J/π )A2 (11.21)
0 R (J/π )2 + η2
1 c η4
A3 = dη; A3 = (J/π )A3 (11.22)
0 R (J/π )2 + η2
where J = πV / R ≡ the propeller advance ratio.
The moments about the pivot are then given by
Mθ = −My + aRLz (11.23)
Mψ = Mz + aRLy (11.24)
“chapter_11(new)” — 2005/11/24 — page 393 — #7

which result in the following final useful forms:

b θ̇
Mθ = Ka R −(A3 + a A1 ) − A2 ψ + aA1 θ
2
(11.25)
2

b ψ̇
Mψ = Ka R −(A3 + a2 A1 ) + A2 θ + aA1 ψ (11.26)
2
11.2.3 Basic Stability Characteristics

With the finalized expressions for the aerodynamic moments about the (wing)
pivot point the dynamic equations can be used to assess the stability characteristics
of the system. Because the basic equations of motion are linear and of constant
coefficient type, use of the Routh–Hurwitz stability criterion will yield stability
boundaries for flutter and divergence. Within the scope of this simple description
of the phenomenon, the stability characteristics are found to be functions of the
following parameters: 1) rotor speed ; 2) polar moment of inertia Ix ; 3) pylon
(nacelle) inertia In ; 4) stiffnesses in the two directions Kψ and Kθ ; 5) damping in
the two directions Cψ and Cθ ; and 6) pivot to hub distance to radius ratio a.
The flutter instability characteristics are directly related to the gyroscopic pre-
cessional behavior of the rotor. For the rigid-blade rotors considered herein, where
all of the system flexibility resides in the two wing attachment pivot springs, it is
found that the rotor can experience flutter instability only in the backward-whirl
mode. This type of whirl motion consists of the rotor hub tracing a circular motion
in the rotor plane in a direction opposite to that of the rotor rotation . As demon-
strated in a later section, the backward-whirl mode aspect of the instability can be
shown to produce a condition of negative damping by virtue of the aerodynamic
hub loads. Consequently, because the pertinent aerodynamic terms are directly
proportional to rotor speed squared, the instability characteristics monotonically
increase with rotor speed . Also, because the backward-whirl mode is determined
in large measure by the precessional characteristics of the rotor (as determined by
the gyroscopic terms), this instability condition would again be aggravated by the
rotor speed, as well as the polar moment of inertia Ix . Note that, because the driver
for the unstable motion is of aerodynamic origin and a positive-definite damping
matrix exists, the system cannot be stabilized by the gyroscopic terms.
Apart from these preliminary considerations, the stability characteristics (both
in flutter and divergence) depend on the remaining parameters. A qualitative indi-
cation of how the stability characteristics depend on these basic parameters is given
in Fig. 11.5 (as extracted from Reed).
These basic trends can be interpreted from the expressions for the aerodynamic
moments about the pivot point given in the preceding section. The condi-
tions of divergence instability are indicated in Fig. 11.5 by the lower limiting
values on the yaw and pitch spring rates. These divergence conditions origi-
nate from the (+)(aA1 θ ) and (+)(aA1 ψ) terms in the expressions just given for
the aerodynamic moments. These terms effectively define negative springs [i.e.,
(+)moment/(+)angle conditions] and are the sources of the aperiodic divergence
instabilities.
“chapter_11(new)” — 2005/11/24 — page 394 — #8

Fig. 11.5 Summary of key results of rotor whirl trend studies of propellers with rigid
blades (after Reed).
For conditions of backward whirl, θ and ψ̇ are in phase with each other, and
ψ and θ̇ are out of phase. For these conditions the expressions just given for
aerodynamic moment show that the second terms (−)A2 ψ and (+)A2 θ in each of
these expressions are out of phase with their respective rate-dependent terms and
therefore act as negative dampers. These terms are then the source of the flutter
instability.
The following detailed interpretations can be drawn from the trends depicted
in Fig. 11.5:
1) The extent of the susceptibility of the rotor to the instability is directly related
to the isotropy of the pylon structure. The more nearly isotropic the structure,
the more likely the characteristic response of the hub will be a circular whirl.
Increasing amounts of anisotropy serve to distort this response into an elliptic form
“chapter_11(new)” — 2005/11/24 — page 395 — #9

with commensurately less ability of the motion to generate the already identified
destabilizing driver forces.
2) For small amounts of damping, the effect of incremental damping is sig-
nificantly greater than that for moderate amounts. Thus, it is quite important to
provide the pylon structure with at least a modicum of damping. However, beyond
a certain point, incremental damping does not offer a viable means for stabilizing
the rotor.
3) The stabilizing effect of the pivot to hub distance aR is attributed to the
additional aerodynamic damping afforded by the motion of the rotor in its plane
(Ly and Lz caused, respectively, by ẏ and ż motion).
11.2.4 Effects of Wing Flexibility

The basic formulation of the rotor-nacelle whirl flutter instability given earlier
assumes that the pivot point is located on a rigid wing. The resulting analysis
is relatively simple and adequately serves to identify the principal mechanisms
involved in the phenomenon. However, real-world applications of the theory must
include the effects of wing flexibility. For cases wherein the wing has a high
degree of stiffness, so that the wing natural frequencies are all much higher than
the uncoupled pylon modes (as defined by the nacelle inertia In and the stiffnesses
Kψ and Kθ ), the rigid pivot attachment point assumption is reasonably valid, and
the resulting trends are valid. For cases wherein the lowest wing natural frequency
(typically that associated with bending ωh ) is close to the whirl flutter frequency, the
effects of the wing aeroelasticity are generally stabilizing because of the additional
source of damping afforded by the wing in bending. This and other trends are
important to the understanding of the phenomenon and are summarized in Fig. 11.6
(again, as extracted from Reed).
Fig. 11.6 Influence of wing flexibility on whirl flutter (after Reed).
“chapter_11(new)” — 2005/11/24 — page 396 — #10

Figure 11.6 shows three different types of stability boundaries, depending on

the degree and type of wing flexibility. The basic rigid-wing result is typified by the
point of maximum instability, point A, which corresponds to the case of isotropic
stiffness. With the introduction of only wing bending flexibility, the region of
instability shrinks in accordance with the fact that the total dynamic system has
an additional source of damping. The points of maximum instability reduction,
relative to the rigid-wing results, are denoted by the points indicated by B. At
these points the whirl frequency ω is equal to the wing bending frequency. In
effect, the wing is acting as a type of tuned damper that operates to maximum
advantage when the system frequency is close to the frequency of the damper. The
third result given in Fig. 11.6 relates to the case wherein the first coupled wing mode
frequency ω1 , which now involves both flatwise bending and torsion motion (h and
α, respectively), is approximately the same as the uncoupled bending frequency
ω considered earlier. The trend shown by this result represents an exception to
the general stabilizing trend of wing flexibility discussed earlier and is denoted
point C. This point corresponds to the case wherein the coupled frequency ω1 is
approximately equal to the frequency in yaw ωψ . For this case the effective nacelle
impedance in pitch, as seen by the rotor, is approximately equal to that in yaw,
and the rotor responds as if it were at point A. In either case, points A or C, the
region of instability is at its maximum extent, in accordance with the conditions
favorable to circular whirling motion.
11.2.5 Stability Characteristics with Blade Flapwise Flexibility

In the preceding sections the formulations for rotor-nacelle whirl instability
have concentrated on the essentially rigid rotor case. This case corresponds to the
original work that was done in response to the catastrophic problems encountered
with “rigid” propellers. For this relatively simple system the instability, as shown
earlier, always involves a regressive or backward whirl. However, in the course
of investigating the effects of various parametric variations about the rigid-blade
rotor case, it was (experimentally) found that, for rotors with flapwise flexibility
or, alternatively, with flapping hinges, the instability could take the form of a limit
cycle whirl in the advancing or forward-whirl direction. Furthermore, it was found
that the forward whirl could occur only for flapping frequency to rotor speed ratios
(ωβ / ) that do not greatly exceed unity (≈1.06) and only in conditions of low
disk loading. For frequency ratios above this value, the instability was found to
revert to the more usual backward-whirl form.
The early investigators of the basic instability problem had concentrated on
the (rigid) propeller case. Consequently, the impetus to explore the forward-whirl
problem in depth came, in part, from the requirements of the vertically lifting tilt-
rotor configurations evolving from within the helicopter technology community.
Such configurations typically consist of rotor-bearing nacelles mounted on the
ends of wings; the nacelles are then provided with a swiveling capability so that the
rotors can act both as helicopter-like lifting rotors (with the rotor axis in the vertical
position) and as thrusting propellers (with the axis in the horizontal direction). For
such designs the occurrence of the forward-whirl instability (or, indeed, other
more exotic variants of the basic instability) was a very real possibility and hence
had to be addressed and understood. The experimentally observed features of the
“chapter_11(new)” — 2005/11/24 — page 397 — #11

soft flapwise stiffness rotor case were successfully predicted in theory by Young
and Lytwyn. Their formulation uses a straightforward Lagrangian procedure and
includes a fairly standard application of (incompressible) quasi-steady lift and drag
airloads. The reader is referred to the literature for a detailed description of the
equations of motion used to model the instability. Alternatively, the generalized
equations of motion for ground and air resonance given in Appendix D can be
used as a basis for defining an approximate formulation to the problem. The key
findings of the work of Young and Lytwyn are summarized in Figs. 11.7 and 11.8,
which were extracted from their work.
Figure 11.7 shows that for blade flapping frequency ratios above 1.1 the effect of
flapping flexibility is to increase the amount of nacelle stiffness required to stabilize
the usual (backward-) whirl mode of instability. Figure 11.8 shows the presence
of the forward-whirl instability for a range of flapping frequency ratios less than
or equal to approximately 1.07. The figure also indicates that for frequency ratios
in this range the prop rotor can experience instability in both the forward- and
backward-whirl modes.
The analysis of actual real-world prop rotors of the tilt-rotor variety must nec-
essarily require a more sophisticated analysis than that provided by either the work
of Young and Lytwyn or a modification of Appendix D. Not only are the (omitted)
effects of unsteady aerodynamics significant to the stability boundaries, but the
details of specific couplings (peculiar to any given prop-rotor configuration) can
greatly influence or even stabilize an existing instability (see Wernicke and Gaffey).
11.2.6 Approximate Corrections to Aerodynamics for

Compressibility and Finite Blade Span Effects
Houbolt and Reed give approximate corrections to the (incompressible) quasi-
steady airloads formulations given earlier. These corrections consist of applying a
Fig. 11.7 Influence of blade flapping frequency ratios greater than 1.1 on the
(backward-whirl) stability of propeller rotors (after Young and Lytwyn).
“chapter_11(new)” — 2005/11/24 — page 398 — #12

Fig. 11.8 Influence of blade frequency ratios from 0.90 to 1.12 on the (forward-whirl)
stability of propeller rotors (after Young and Lytwyn).
combination of two factors to the incompressible lift-curve slope clα (F1 and F2 ,
(=c · F · F ), where
respectively) to define a “corrected” lift-curve slope clα lα 1 2
(AR)
F1 = 1/ 1 − M 2 (r); F2 = (11.27a)
(AR) + 2
and where
M(r) ≡ Mach number at station r
(AR) ≡ Mach number corrected aspect ratio of the blade section,
= (AR) [1 − M 2 (r)] (11.27b)
(AR) ≡ aspect ratio of the entire blade

Accordingly,
(AR)
F1 · F2 = (11.28)
2 + (AR) 1 − M∞
2 [1 + (π 2 /J 2 )η2 ]
where M∞ is the forward-flight Mach number and η = r/R. This combined factor
is then inserted under the integral in the expressions given earlier for the coefficients
A1 , . . . , A3 , so that it forms part of each of the integrands.
When available, wind-tunnel measured propeller derivatives should be used to
check the preceding results. Furthermore, in addition to the approximations just
given to the compressibility and finite-span effects, additional modifications can
be added to the analysis to approximate unsteady aerodynamic effects. Methods
for the inclusion of these effects are covered in a subsequent chapter and therefore
will not be treated in any further detail at this point.
“chapter_11(new)” — 2005/11/24 — page 399 — #13

11.2.7 Comprehensive Analyses

Laplace transform space eigenvalue analyses. With the evolving interest
in tilt-rotor and tilt-wing rotorcraft, the need for accurate analyses has been met
by the emergence of a number of governmental and proprietary comprehensive
computer codes. Another driver for the development of these codes has been the
increasingly more powerful and yet less expensive computer resources available.
Three major analyses have emerged, respectively, from those Kvaternik, Johnson,
and Popelka et al. A representative example of the degree of modeling undertaken
for the analysis of tilt-rotor aircraft is presented in Fig. 11.9.
Application of the multi-degrees-of-freedom frequency-domain method-

ology. Part of the difficulty in analyzing the rotor-nacelle aeroelasticity is that
there are three interacting dynamic subsystems, each with its own modeling
difficulties:
1) The rotor has all of the complexities of a normal helicopter rotor as far as
structural and aerodynamic modeling is concerned. Indeed, the requirements are
exacerbated by the relatively high twist in tilt-rotor and tilt-wing aircraft.
2) The wing and nacelle comprise a flexible structure providing rotor attach-
ment to the fuselage. It has a number of internal structural and nonstructural
components that all contribute to the elastomechanical characteristics of the
structure.
Fig. 11.9 Schematic presentation of the modeling detail used in the DYN4 analysis
(Popelka et al.).
“chapter_11(new)” — 2005/11/24 — page 400 — #14

3) The wing unsteady aerodynamic loading comprises a separate dynamic

subsystem that interacts with the wing elastomechanics, which in the absence
of the rotor, would define a classic wing flutter problem. The wing unsteady
aerodynamics is usually most accurately defined in the frequency domain.
Figure 11.10 exemplifies the latter two of the subsystems just identified (with
the rotor included as a concentrated mass).
Figure 11.11 presents a conceptualization of how the three major dynamic
subsystems interact. The purpose of defining the problem in this particular way is
to isolate the mathematical descriptions needed for each of the subsystems. For
example, the unsteady aerodynamics of the wing can be handled in a variety of
ways, as is described in Chapter 12. In particular, the most accurate analytical
treatment of the airloads is in the frequency domain. Furthermore, the wing and
nacelle elastomechanics can be addressed with a suitable finite element analysis or
even with experimental shake-test data. Lastly, the rotor dynamics can be modeled
with any of the major available rotor aeroelastic analyses that include hub motion.
The use of this methodology begins by considering the abbreviated form of
the total equations of motion, as would be used by a Laplace transform space
eigenvalue analysis. The various degrees of freedom are grouped into three
main partitions of the state vector, and the matrices that couple them are appro-
priate partitions of the total matrix dynamic equation. {Xhub }, {Xwing }, and {Xrotor }
are, respectively, the six spatial degrees of freedom of the hub (three translations
and three rotations), the spanwise distribution of discrete flatwise, edgewise, and
torsion deflections of the wing, and the various degrees of freedom selected to
describe the rotor aeroelastic response. Except for the hub degrees of freedom, the
actual selection of degrees of freedom for the wing and rotor is arbitrary, depend-
ing on engineering judgement and available computer resources. The abbreviated
Fig. 11.10 Schematic of the basic dynamic subsystems defining the rotor-nacelle
aeroelastic problem.
“chapter_11(new)” — 2005/11/24 — page 401 — #15

Fig. 11.11 Conceptualization of the interaction of the dynamic components defining

rotor-nacelle flutter.
form of the total coupled equations of motion can be written as

  
H11 (s) H12 (s) H13 (s)  Xhub 
H21 (s) H22 (s) 0  Xwing = {0} (11.29)
 
H31 (s) 0 H33 (s) Xrotor
The H11 , H13 , H31 , and H33 matrix partitions constitute the modeling that would
be present in a comprehensive rotor aeroelastic analysis that included hub motion
dynamics. The supplementary matrix partitions H12 , H21 , and H22 , relate to the
coupling of the hub with the wing. Of all of these the H11 matrix represents the
dynamics of the hub taken as a point.
As far as the rotor dynamics part of the problem is concerned, it represents
a form of attachment impedance. The H13 and H31 matrices form the essential
coupling of the rotor dynamics with the hub. Equation (11.30) represents the form
of a typical comprehensive rotor aeroelastic response analysis:

H11 (s) | H13 (s) X
hub
−−−−−−− = {0} (11.30)
H (s) | H (s) X rotor
31 33
Consider now the third row of the partitioned Eq. (11.29):

 
 Xhub 
−−
−−
[H31 (s) 0 H33 (s)] Xwing = {0} (11.31)

 
Xrotor
This equation allows the elimination of the rotor degrees of freedom from the
active state vector:
{Xrotor } = −[H33 (s)]−1 [H31 (s)]{Xhub } (11.32)
“chapter_11(new)” — 2005/11/24 — page 402 — #16

The resulting total equation set then becomes

 −1 
H11 (s) − H13 (s)H33 (s)H31 (s) | H12 (s) Xhub
− − − − − − − − − − − − − − − − − = {0} (11.33)
H21 (s) | H22 (s) Xwing
This equation can be written in the following alternative form:

G2 (s) | 0

Xhub Xhub
[Ĥ(s)]elasto−mech. + [Ĥ(s)]aero. − − − − − − = {0}
Xwing 0 | 0 Xwing

G2 (s) | 0 Xhub
[Ĥ(s)] − − − − − − = {0} (11.34)
0 | 0 Xwing
where [G2 ] = [H13 ][H33 ]−1 [H31 ] and is composed of an elastomechanical part
and a part caused by the unsteady airloads:

H11 | H12
[Ĥ(s)][Ĥ(s)] = − − − − − = [Ĥ(s)]elasto−mechan. + [Ĥ(s)]aero. (11.35)
H21 | H22
Equation (11.34) can then be again rewritten as

G2 (s) | 0

Xhub Xhub
[Ĥ(s)]elasto−mech. + [Ĥ(s)]aero. − − − − − − = {0}
Xwing 0 | 0 Xwing
(11.36)
From this point on, we assume sinusoidal motion and continue the formulation
in the frequency domain. This amounts to just substituting iω for the Laplace
transform variable s and recognizing that the equality must be modified with a
multiplier.

Xhub G2 (ω) | 0 Xhub
−1
= [Ĥ(ω)]elasto−mech. −[Ĥ(ω)]aero. + − − − − −
Xwing 0 | 0 Xwing
(11.37)
We then note that the inverse of the [Ĥ(ω)]elasto−mech. matrix (as taken in the
frequency domain) is the frequency-response function of the hub (and wing) as
a result of hub and wing excitation loads, that is, the G1 (FRF) matrix. Thus, in
terms of the quantities given in Fig. 11.11 the final frequency-domain formulation
becomes

Xhub 1 G2 (ω) | 0
Xhub
= [G1 (ω)] ρU [Qa (k, · · · )] + − − − − −
2
Xwing 2 Xwing
0 | 0
(11.38)
“chapter_11(new)” — 2005/11/24 — page 403 — #17

where
[G2 (ω)] = [H13 (ω)][H33 (ω)]−1 [H31 (ω)] (11.39)
It should be stressed that each of the three component matrices in Eq. (11.38)
is obtainable from separate sources. Also, the matrices comprising Eq. (11.39)
(which are obtained from the selected rotor aeroelastic analysis) can be alterna-
tively obtained by using a time-history solution of the computer code to establish
responses that have converged to a purely sinusoidal state. This is especially useful
for treating nonlinearities in the rotor equations. Finally, Eq. (11.38) is now in the
proper form for application of the multiple-degree-of-freedom frequency-domain
methodology described in Chapter 9.
11.3 Ground Resonance Instability

Of the several dynamics-related potential problem areas in rotorcraft, the phe-
nomenon of ground resonance represents the first instance where the coupling
between the rotor and pylon (fuselage) was considered and successfully analyzed.
The high degree of success attained was most likely caused in part by the pio-
neering analytic investigation by Coleman and Feingold in the 1940s and to the
fact that this instability is relatively free of aerodynamic effects and is a true elas-
tomechanical instability. The basic equations of motion are somewhat different
depending on whether the rotor has isotropy (three-or-more-bladed case) or not,
that is, anisotropy (one- or two-bladed case). In the material to follow, these two
cases are developed separately.
11.3.1 Dynamic Description for the Isotropic Rotor Case

As described in an earlier section, a major analysis problem to be addressed
in modeling rotor-fuselage coupled systems is that a common coordinate system
must be defined. Generally, the most convenient coordinate frame is the one that
has the anisotropy. This avoids the problem of dealing with equations with periodic
coefficients. For the three-or-more-bladed case, with all blades elastomechanically
the same, the rotor is isotropic, and it is appropriate to write the equations of motion
in the fixed or nonrotating coordinate system. The most simple ground resonance
analysis also happens to be remarkably effective. This simple approach considers
the fuselage to be represented by equivalent masses and springs in the longitudinal
and lateral directions, respectively. As such, the elastomechanical description of
the fuselage consists of equivalent masses and dampers together with characteristic
natural frequencies in each of the two in-plane directions, as shown in Fig. 11.12.
Note that the longitudinal and lateral motions of the hub, x and y, respectively,
arise from pitching and rolling motions of the fuselage about some effective rota-
tion centers that are generally not located in the ground plane but typically below it
for the lowest-frequency modes of the fuselage. Using the distances of the effective
rotation centers from the rotor hub hθ and hφ , effective masses in the x and y direc-
tions, mFx and mFy , respectively can be defined. The x- and y-direction “effective”
springs and dampers, kFx , kFy , and cFx , cFy , respectively, are obtained from the
effective masses together with the frequency and damping ratios appropriate to the
pitch and roll modes of fuselage motion on the landing gear (ωθ , ζθ , ωφ , and ζφ ).
“chapter_11(new)” — 2005/11/24 — page 404 — #18

Fig. 11.12 Simplified dynamic description of fuselage for ground resonance analysis.
The dynamic description of the flexible rotor blades is achieved using cyclic rotor
modes as described in an earlier section (see Fig. 11.1).
Equations of motion. A presentation of a basic unified set of equations for

both ground resonance and air resonance is given in Appendix D. This appendix
is meant to provide a somewhat more detailed version of the simpler material pre-
sented herein, which we need for investigating the major features of the instability.
From the basic equations in Appendix D, we can extract a simplified set of equa-
tions of motion for the three- or more-bladed case; these are given in the following
“chapter_11(new)” — 2005/11/24 — page 405 — #19

without derivation. Note that the first two equations define the force equilibrium of
the nonrotating hub, and the second two define the force equilibrium of the cyclic
in-plane rotor modes.
Hub longitudinal force Fx :
Mx ẍ + cFx ẋ + kFx x + (b/2)S48 ¨x = 0 (11.40)
Hub lateral force Fy :
My ÿ + cFy ẏ + kFy y + (b/2)S48 ÿ = 0 (11.41)
Rotor longitudinal edgewise excitation x :
S48 ẍ + S49 [¨x + c ˙x + (ωv2 − 2 ) x ] + S49 (2˙y + c y ) = 0 (11.42)
Rotor lateral edgewise excitation y :
S48 ÿ + S49 [ÿ + c ˙y + (ωv2 − 2 ) y ] − S49 (2˙x + c x ) = 0 (11.43)
where the effective masses in the x and y directions, Mx and My , are respectively
given by
Mx = mFx + mR ; My = mFy + mR (11.44)
The blade in-plane damping c can be alternatively expressed in terms of a critical

damping ratio and natural frequency at zero rotor speed ωv0 :
c = 2ζv ωv0 (11.45)
and the various inertia related integrals Sn are defined in Appendix D.
Stability solution. The solution for the stability descriptors is standard and
makes use of the solution of either an expanded characteristic polynomial (in this
case an eighth-degree polynomial) or a matrix eigenvalue solution. The roots λj
will have the usual form
λj = σj + iωj (11.46)
and stability is ensured if all σj < 0.
Frequency characteristics: Coleman diagram. The coupled frequencies

of the system are just the imaginary parts of the eigenvalues ωj . If all damping is
removed from the system (cFx = cFy = ζv = 0), then the roots will occur either
as imaginary pairs (pure undamped frequencies) or as quartets of roots:
λj = ± σj ± iωj (11.47)
For this latter case the eigensolution clearly indicates an instability. These results
are conveniently shown on a Coleman diagram shown in Fig. 11.13; this diagram
“chapter_11(new)” — 2005/11/24 — page 406 — #20

Fig. 11.13 Typical Coleman diagram for ground-resonance characteristics of an

isotropic rotor and anistropic fuselage.
shows the variation with rotor rotational speed of the resulting coupled frequencies
(both blade and fuselage modes) as seen in the nonrotating coordinate system.
Note that the regions of instability always occur at the intersections of the degen-
erate “regressive” lead-lag mode with one or more of the fuselage modes. For these
conditions the regressive lead-lag mode is denoted as “degenerate” because the
rotor has become supercritical; the blade frequency has become less than the rotor
rotational frequency. At these conditions the normally regressive mode has changed
direction and become a slower moving progressive mode. More specifically, the
following characteristics can be noted from this figure:
1) The basic structure of this diagram is similar to what the various modes would
look like with zero coupling (i.e., if the ratios of S48 to each of the hub masses,
Mx and My → 0). For zero coupling the intersections between the blade modes
and the pylon modes would show no deviations. However, with finite coupling the
various intersections become coalescences. For coalescences that occur above the
1/rev (lP) line (ω = ), the coupling is smooth, and only pure frequencies result.
Only for coalescences that occur below the 1/rev line can the coupling result in
instabilities.
2) The degree of inertia coupling as defined in terms of the earlier identified
terms is a major measure of the susceptibility of the system to the instability.
3) The shaft critical speeds are indicated as the points where the 1/rev line
intersects the pylon frequencies.
4) At the point where the coupled blade mode crosses the horizontal axis, that
frequency has a value of zero and represents a condition wherein a steady force
can excite the system. Note that this condition is also the point where the rotating
blade mode frequency is exactly equal to the rotor rotational frequency.
5) The figure implies one effective means of avoiding the instability: Do
not operate the rotor near the instability regions that are closely approxi-
mated by conditions wherein [the blade lead-lag (rotating) frequency] + [one of
the pylon mode frequencies] = [the rotor rotational frequency].
“chapter_11(new)” — 2005/11/24 — page 407 — #21

Mass coupling parameter. In the original work by Coleman and

Feingold, the use of graphical methods is stressed for analyzing ground-resonance
problems; more specifically, selected nondimensional parameters are used to
define the stability characteristics. The principal parameters defined and used in
this classic work are the 1 , 2 , µ, and 3 parameters. It turns out that the first
two parameters, 1 and 2 , define the lead-lag frequency characteristics of an
articulated blade. The remaining two parameters, µ and 3 , are, respectively, the
mass ratio (effective rotor mass to the total mass at the hub) and the mass coupling
parameter. The mass coupling parameter 3 is defined in Coleman and Feingold
in terms of µ. Herein, however, this parameter is defined more broadly in terms of
the generalized first edgewise mass moment S48 , the generalized second edgewise
mass moment or generalized (modal) mass S49 , and the effective in-plane mass at
the hub Meff (= Mx or My ), as follows:
2
bS48
3 = (11.48)
2Meff S49
This mass coupling parameter is a convenient factor for defining the degree of
interaction between the rotor and the fuselage. The degree of instability varies
monotonically with this parameter.
11.3.2 Damping Characteristics

To stabilize the ground resonance instability when frequency coalescence is
unavoidable, recourse must be made to the use of dampers (in both the blades and
in the fuselage undercarriage). Because damping in these components is needed
for other reasons as well, the use of dampers is virtually universal with three-or-
more-bladed rotor systems. The exception is the case of hingeless rotors where
the incorporation of a damper becomes problematic. In such cases, the inherent
structural damping in the blade and blade retention structure must be enhanced and
made to do the job. Now for helicopter systems that are simple enough, recourse can
be made to the use of the Routh stability criterion to establish stability boundaries
rather than actually solving for the roots of the characteristic equation. Figure 11.14
presents typical stability boundary information. One important feature of these
plots is that, for any fixed value of blade damping, the width of the instability
range of rotor speed actually increases with increased fuselage damping before
the instability region is eventually closed.
Product of damping criteria. Let us return to considerations of the char-

acteristic equation itself in order to investigate the role of the damping values
of the blade vis-à-vis the pylon. Let us further consider that the three possible
extremes existing are that the pylon be either isotropic, or have either of two
infinitely anisotropic conditions (i.e., the x direction or y direction with all deflec-
tions and derivatives identically zero). All other cases lie somewhere in between
these three cases. Observations of the Coleman diagrams show that the ground-
resonance instabilities tend to cluster symmetrically about the intersection points
of the blade mode and one or the other of the pylon modes.
“chapter_11(new)” — 2005/11/24 — page 408 — #22

Isotropic case. Here we set the hub and rotor center of gravity displacements
to complex forms:
z̄ = x − iy; ¯ = x − iy (11.49)
Use is also made of the (nondimensional) coupling factor already described:

2
bS48
˜3 = (11.50)
2M̃S49
where
M̃ = (mFx + mR ) = (mFy + mR ) (11.51)
For the product of damping to function properly, the blade and fuselage damping
must be expressed in a similar fashion:
c = 2ζv ωv0 ; c̃ = cFx /M̃ = cFy /M̃ (11.52)
Next, let us assume a neutral stability point right on the stability boundary:
λ = iω, where ω̂ is the frequency on the Coleman plot taken at the center of
instability point, = ω̂ = ωx = ωy = − ω . (Note that here ω is an alternate
expression of the edgewise natural frequency ωv .) With these simplifying assump-
tions the characteristic equation is expressible as a real part and an imaginary part,
both equal to zero. Expressing the damping in the pylon to be some isotropic value
c̃ then permits the real part of the characteristic equation (after some algebra) to
be written in the following compact form:
˜ 3 ω̂3
c̃c = (11.53)
ω
Fig. 11.14 Typical stability boundaries for variations in damping.
“chapter_11(new)” — 2005/11/24 — page 409 — #23

Because the preceding expression denotes the approximate conditions at the

stability boundary, the product of damping criterion states that, for stability,
˜ 3 ω̂3
c̃c ≥ (11.54)
ω
Infinitely anisotropic case. For the case of an infinitely anisotropic pylon,

the pylon frequency will be either ωx or ωy . For present purposes we assume
that the helicopter is infinitely anisotropic, with all motion occurring only in roll.
Therefore, the appropriate frequency and mass items are ωy and (mFy + mR ). For
this case we define the coupling factor in terms of the “noninfinitely rigid” or
flexible degree-of-freedom properties ( )y :
2
bS48
3v = (11.55)
2My S49
where
My = (mFy + mR ) (11.56a)
Also,
c̃y = cFy /My (11.56b)
Using similar algebraic manipulation, one can then show that for the infinitely
anisotropic case,
3
3v ωy
c̃y c ≥ (11.57)
2ω
which is the product of damping criterion for the infinitely anisotropic case, as
written here for roll flexibility. The result is the same for pitch flexibility only
infinitely anisotropic case, except that ( )y → ( )x . Note that this result is identical
to the isotropic case except for the factor of two in the denominator.
Limitations on criterion. The use of the product of damping criterion is

subject to the following limitations:
1) The product of damping is reasonably accurate only when the pylon and
blade damping levels are of comparable orders of magnitude, that is,

(c̃/ω̂) (c /ω )
O =O (11.58)
(c /ω ) (c̃/ω̂)
and (c̃/ω̂) and (c /ω ) are both of O(0.10).

2) When the preceding requirements are not met, then the product of damping
underestimates the damping required.
3) The product of damping is proportional to 3 (i.e., the mass coupling) in
the system. For more mass coupling more energy must be dissipated.
“chapter_11(new)” — 2005/11/24 — page 410 — #24

Exact stability boundaries for single-degree-of-freedom pylons with

isotropic rotors. The essence of the ground resonance instability is the coupling
of the regressive lead-lag blade motion with one or more of the pylon in-plane
translational degrees of freedom (expressed as either discrete linear motion coor-
dinates or as modes). However, despite the opportunity of this blade lead-lag
degrees of freedom to couple with any of the various pylon translational degrees
of freedom the coupling process takes place one pylon degree of freedom at a
time, depending on the frequency coalescence characteristics. Hence, the ground
resonance phenomenon can be approximated for many cases using one pylon
degree of freedom at a time. This simplification reduces the basic formulation
to a three-degree-of-freedom system and consequently leads to a sixth-degree
characteristic equation.
Gabel and Capurso have examined this simplified system and, using the Routh
array approach, have calculated useful charts giving exact stability boundaries for
an articulated rotor for various values of rotor speed mass coupling parameter
3 and blade and pylon damping parameters λζ and λy , respectively. The results
obtained in this study are similar to the stability boundaries shown qualitatively in
Fig. 11.11. A typical example of the results obtained is shown in Fig. 11.13.
The parameters shown in Fig. 11.15 utilize nondimensionalizations based on the
lateral ( y motion) equivalent mass My and natural frequency ωy of the hub and the
blade mass moment of inertia about the lag hinge Iζ :
= /ωy (11.59a)
λζ = cζ /Iζ ωζ ; cζ = blade lag damper rate (11.59b)
λy = cy /My ωy ; cy = equivalent pylon damper rate (11.59c)
in the y direction
1 = ωζ2 / 2 ; ωζ = blade lead-lag frequency (11.59d)
(=ω = ωv )
Fig. 11.15 Exact ground-resonance instability regions, single-degree-of-freedom

pylon, 1 = 0.10, 3 = 0.025 (after Gabel and Capurso).
“chapter_11(new)” — 2005/11/24 — page 411 — #25

Note that in Fig. 11.15 the values of hub (pylon) damping required for stability, as
determined by the product of damping criterion, are included for comparison. As
developed therein for rigid lead-lagging blades and an infinitely anisotropic pylon,
the product of damping criterion is expressed as
1 1
λy λζ ≥ 3 √ −1 (11.60)
2 1
As shown in Fig. 11.15, the inadequacy of the criterion is evident at both the
high and low values of blade-blade damping wherein the hub and blade damping
values are most dissimilar. The agreement is also quite apparent, however, when
the damping values are approximately equal.
The usefulness of the Gabel and Capurso results (as well as other such graphic
results) lies in their ability to provide quick estimates of the ground-resonance
characteristics for each pylon mode separately, at least for cases wherein the modes
are reasonably well separated in frequency. By combining the results obtained
separately for each pylon mode, the combined Coleman plot (such as that depicted
in Fig. 11.13) can be approximated.
Use of interblade dampers. Sela and Rosen have proposed an alternate

lead-lag damper configuration wherein the dampers are attached blade to blade
rather than blade to hub. Based on the assumption of angular dampers between
the blades, an effectiveness ratio Ef was formulated to express the gain of the
blade-to-blade attachment over that for the blade-to-hub attachment:
2π
Ef = 2 1 − cos (11.61)
N
where N is the number of rotor blades. It can be clearly seen that this expres-
sion for effectiveness ratio will yield an improvement only for six blades or less.
Figure 11.16 puts this effectiveness ratio in perspective.
The figure begins with a reference value of blade lead-lag damper cref and
then shows the theoretical damping achieved using Eq. (11.61) (cSR = Ef cref ).
However, any practical damper installation will use linear dampers, as also shown
in Fig. 11.16. Because of the skewed attachments of the damper to the successive
blades, the effectiveness of the damper in creating damper moments about the
blade lead-lag hinges is degraded. The factor of effectiveness caused by skewing
of the damper κN is expressed in terms of the angle γ shown in Fig. 11.16 and
given by:
M = Fdamper l cos γ
Fdamper = clinear dżdamper = clinear l cos γ dθ̇ (11.62)
M = clinear l2 cos2 γ dθ̇ = (cos2 γ )Ef cref dθ̇
Thus, the skewing degradation factor κN is equal to cos2 γ , where γ is equal to half
the angle between the successive blades 2π/N. The results of the two effectiveness
factors are shown in Table 11.1.
“chapter_11(new)” — 2005/11/24 — page 412 — #26

Fig. 11.16 Pictorial representation of blade-to-blade lead-lag damper installation.
The results shown in Table 11.1 would indicate that there is no payoff for
using a blade-to-blade damper configuration. Yet, as shown in Fig. 11.17, such a
configuration has been incorporated in the hub design of a (four-bladed) production
helicopter.
Although there is no payoff in terms of increased effectiveness, this configura-
tion is practical for two reasons: first, a longer moment arm, l, can be used while
still keeping the hub relatively compact, and second, the design results in a simpler
and, therefore, lighter hub. Furthermore, because the dampers are attached further
outboard they will be easier to inspect and service.
11.3.3 Examination of the Mechanics of the Instability

The use of the eigenvalue analysis of the ground resonance instability yields
information as to whether the system of equations represents a stable or unstable
configuration and, by examining the eigenvalues, by how much. All too often
Table 11.1 Blade-to-blade effectiveness ratios as functions

of number of blades N
N 3 4 5 6
Ef 3 2 1.38 1
κN 0.25 0.5 0.66 0.75
κN ·Ef 0.75 1 0.90 0.75
“chapter_11(new)” — 2005/11/24 — page 413 — #27

Fig. 11.17 Installation of blade-to-blade lead-lag dampers on the NH Industries

NH-90 helicopter (photo courtesy of Burkhard Domke).
the eigenvector information is underutilized in the process. A simple unstable

ground resonance example is now examined from the standpoint of gleaning some
understanding of the physics involved. Table 11.2 presents the data defining the
example.
The eigenvector is shown in planform view in Fig. 11.18. The arrangement
depicted shows the relationship of the hub and the center of gravity of the hub, as
defined by the two components of cyclic inplane deflection εx and εy . The whirl
speed is in the same direction as, but at a fraction of, the rotor speed. It is clear
that the rotor center of gravity is leading the fuselage mass (the hub) and that it
Table 11.2 Data defining an unstable ground-resonance configuration
Fuselage:
Mx = 131.479 lb-s2 /ft, My = 28.687 lb-s2 /ft,
ζx = 0.05, ζy = 0.05,
ωx = 35.0 rad/s, ωy = 82.1 rad/s
Rotor:
R = 28 ft, ref = 221.68 rpm,
b = 4, ζv = 0.005,
Mb = 3.8801 lb-s2 /ft, (ωv )ref = 26.979 rad/s,
S48 = 60.214 lb-s2 , Kv 0.256,
S49 = 1144.0 lb-s2 2-ft
Operational condition:
= 68.068 rad/s (R = 650 ft/s)
Ground-resonance eigensolution results:
λGR = 1.31428 + i31.8026
eigenvector:
φGR = 1 + i0, 0.1712 − i0.1061, 0.0900 + i0.2289, 0.2275 − i0.0954
“chapter_11(new)” — 2005/11/24 — page 414 — #28

Fig. 11.18 Planform depiction of the ground-resonance eigenvector.
is driving the fuselage mass by virtue of the centrifugal force arising from the
whirl motion.
Figure 11.19 presents the results of the force-phasing matrix calculation made
using the eigenvector information along with the mass, damping, and stiffness
matrices for the system. This figure shows that the only energy flow path and critical
drivers are those that result from the cyclic in-plane rotor modes. The fuselage
degrees of freedom do not directly take part in the energy flow path, but are instead
driven by the rotor modes through the mass matrix coupling terms m13 and m24 .
Fig. 11.19 Energy-flow diagram for the example ground-resonance instability

condition.
“chapter_11(new)” — 2005/11/24 — page 415 — #29

The terms defining the energy flow path are blade inertia and damping related that
arise from the rotor mode transformation defined by Eq. (11.2). Also, although not
shown, the diagonal stiffness terms for the cyclic in-plane rotor modes are negative,
commensurate with the rotor being supercritical. The role that fuselage dynamics
play in the physics is thus seen to be that of an “enabler” in that the fuselage provides
a low impedance support for the rotor center-of-gravity’s motion at the coincident
frequency of − ωe (=ωFx ). These results are not particularly significant as far as
defining any new methods for stabilization, rather they underscore well-established
knowledge gleaned pragmatically from looking at trends of stability eigenvalues.
The results also underscore the basic well-established method for stabilization:
avoid any frequency coalescence between the rotor and fuselage and increase the
impedance of the fuselage (by adding damping).
11.3.4 Application of the Multiple-Degree-of-Freedom

Frequency-Domain Methodology
Mathematical formulation. The application of this methodology to the
ground resonance problem poses a simpler formulation than was required for
the rotor-nacelle flutter problem owing to the fact that ground resonance is prin-
cipally an elastomechanical phenomenon not requiring aerodynamic interactions.
Figure 11.20 presents the conceptualization of the rotor-airframe interaction in the
form required for the methodology.
The G1 and G2 transfer function matrices must be gleaned from the standard
ground-resonance equations:

H11 (s) | H12 (s)
XF
−−−−−−− = {0} (11.63)
XR
H21 (s) | H22 (s)
where X F represents the state variables of the fuselage plus the rotor as a point
mass and X R represents the variables defining the blade lead-lag rotor modes. The
partitions of Eq. (11.63) can be rewritten as two matrix equations, which then
Fig. 11.20 Conceptualization of the rotor-airframe interaction in ground resonance.
“chapter_11(new)” — 2005/11/24 — page 416 — #30

allows for isolation of the separate vectors:
{XF } = −[H11 ]−1 [H12 ]{XR } (11.64)

{XR } = −[H22 ]−1 [H21 ]{XF } (11.65)
The conversion to the frequency domain (i.e., substituting iω for s) requires

closing the loop defined in Fig. 11.20 by relating the X F2 vector to X F1 with the
characteristic multiplier . The final required form of the equation then becomes
(ω) {XF } = [H11 (ω)]−1 [H12 (ω)] [H22 (ω)]−1 [H21 (ω)] {XF } (11.66)
or, more compactly
(ω) {XF } = [G1 (ω)] [G2 (ω)] {XF } (11.67)
where
[G1 (ω)] = [H11 (ω)]−1 (11.68a)

[G2 (ω)] = [H12 (ω)] [H22 (ω)]−1 [H21 (ω)] (11.68b)
Note that the methodology works equally well if the lead-lag vector X R is used
in Eq. (11.66) and when
[G1 (ω)] = [H22 (ω)]−1 (11.69a)

[G2 (ω)] = [H21 (ω)] [H11 (ω)]−1 [H12 (ω)] (11.69b)
The choice of decomposition [i.e., using Eqs. (11.68a), (11.68b), or (11.69a),

(11.69b)] is arbitrary, but some computation time would be saved by using the
resulting smaller-sized [G1 · G2 ] matrix. For instance, if the fuselage dynamics
involved a full complement of six degrees of freedom, Eqs. (11.69a) and (11.69b)
would be preferable because the size of the [H22 ]−1 matrix is only 2 × 2, whereas
the [H11 ]−1 matrix would be 6 × 6.
Analytically, either G1 and/or G2 can be calculated using a straightforward
numerical integration of the nonlinear fuselage (or rotor) equations of motion.
A standard high-performance numerical integration scheme would be used, such
as the Runge–Kutta method (fourth order with Gill coefficients, as described in
Chapter 9), and then the time histories are Fourier analyzed for their fundamental
harmonic cosine and sine components. (Note that these calculated quantities are the
exact dynamic quantities that would be directly measured in subsequent ground-
vibration tests!) One practicality of this methodology is that, although either G1
and/or G2 , can always be calculated in the time domain, for many conditions
G2 , the rotor impedance, is defined by linear terms. The required matrix inversion
can then be performed quite simply and accurately without the integration of the
equations of motion. The fuselage mobility requires the calculation of the matrix
[H11 ]−1 , which actually constitutes identically the fuselage x- and y-direction
sinusoidal responses caused by sinusoidal (test) forces at the hub in the x- and
y-directions, that is, FRFs ( frequency response functions). For weakly nonlinear
“chapter_11(new)” — 2005/11/24 — page 417 — #31

systems the choice of amplitude for the sinusoidal test forces would be of some
concern. It would be prudent to keep the amplitude small enough to maintain
linearity and yet large enough to keep the results out of the “round-off noise.”
The issue of the presence of nonlinearities in the system is implicitly addressed
in this methodology, in that the fuselage mobilities thus calculated are linear-
effective mobilities. (Note that these would consequently vary with response
amplitude for nonlinear systems). Although this procedure forms a “neat” way
of addressing the nonlinearities, it needs to be stated that this form of linearization
is no less valid than (and is actually equivalent to) the use of describing func-
tions, which are routinely used as a standard basis for linearization, as described
by Savant. Furthermore, as a result the new methodology easily enables the pre-
diction of limit-cycle conditions resulting from nonlinearities. The analysis of
limit-cycle conditions requires that suitable variation of the sinusoidal test loads
be made. Lastly, although not a nonlinearity per se, the condition of a failed or par-
tially failed lead-lag damper (the doubly anisotropic problem) is also addressable
by this methodology without recourse to Floquet theory (Bielawa).
Practical implementation issues. As just formulated, the basic equation,

Eq. (11.66), is a matrix eigenvalue equation whose solution yields discrete values
of the characteristic multipliers j as functions of discrete values of the scan
frequency ωj . The search for the point of crossing of the real axis (in order to
assess stability) must then be accomplished using supplementary techniques. First
off, a suitable range of scan frequencies must be established. A convenient method
is to identify a median frequency ωM and then define the scan range using selected
frequencies below and above that frequency. For the ground resonance problem the
appropriate median frequency would be the center-of-instability frequency, defined
by the slower blade progressive mode (i.e., the degenerate regressive mode after the
rotor has gone supercritical). This frequency, as is well identified in the literature,
is then given by
ωM = − ωe (11.70)
where ωe is the blade lead-lag natural frequency (at rotor speed). It is then a
straightforward programming exercise to evaluate the phases (or alternatively, the
imaginary parts) of the characteristic multipliers to identify the real-axis crossing
value, as is described in Chapter 9.
Consistency of eigenvalue results with frequency-domain results. The

two basic methods for determining stability (i.e., Laplace transform space eigen-
value vs frequency domain) use the basic matrix eigenvalue solution in distinctly
different ways. To show mathematical consistency between the two methods, they
were separately used on exactly the same (linear) ground-resonance equations of
motion for the example configuration presented in the preceding subsection. To
this end, the linear lead-lag damping value was artificially adjusted to produce
three conditions: stable, exactly neutrally stable, and unstable. The results of this
exercise are shown in Figs. 11.21a, 11.21b, and 11.21c, respectively: 1, stable
case; 2, neutrally stable case; and 3, unstable case.
“chapter_11(new)” — 2005/11/24 — page 418 — #32

Fig. 11.21 Characteristic multiplier vs phase angle (representative results from

example ground resonance configuration).
“chapter_11(new)” — 2005/11/24 — page 419 — #33

These figures show the parametric variations of the respective characteristic

multiplier magnitudes and phase angles with scan frequency ω. Included in each
of these figures are the values of the stability critical eigenvalues, as separately
determined by the conventional (Laplace transform space) eigenvalue solution.
The figures clearly show that the methods do, indeed, give consistent qualitative
indications of the stability of the system.
Implications for safer ground-resonance testing. As has been verified by

numerous ground test programs, the testing of full-scale helicopters for ground
resonance is extremely risky. The established procedure for exiting an incursion
into ground resonance is to get the helicopter airborne quickly so as to signifi-
cantly change the fuselage impedance. For ground tests, this is not an option, and
there are few other practical options for quenching an instability once it “locks
in.” The concept of reducing the instability to an examination of the component
dynamic subsystems can be readily applied to the full-scale testing situation to
create a considerably safer exercise. Figure 11.22 demonstrates how this can be
accomplished.
The first step is to conduct a shake test on the fuselage with the rotor blades
removed, but with a concentrated mass attached to the hub to represent the rotor
mass. The purpose of the shake test is to measure the frequency response function
(matrix) [G1 (ω)]. The FRF measurements should be made with the shaker exciting
Fig. 11.22 Schematic for applying the frequency-domain stability methodology to

testing full-scale rotorcraft for ground resonance.
“chapter_11(new)” — 2005/11/24 — page 420 — #34

the hub separately in both the fore and aft and lateral directions, and, for each
direction of excitation, responses should be measured in both the fore and aft and
lateral directions. The rotor impedance matrix [G2 (ω)] on the other hand can be
calculated with considerable accuracy [using well-established ground resonance
equations and Eq. (11.68b)]. With the two matrices available, the characteristic
multiplier calculation procedure can then be made to indicate the presence of
instability, as well as a measure of the stability level using the techniques given in
Chapter 9.
11.3.5 Two-Bladed Case

Because of the dynamic isotropic symmetry of the three-or-more-bladed rotor,
the dynamic equations can be developed in the nonrotating coordinate frame, and
the different dynamic properties of the rotor support structure (fuselage) in the x-
and y-directions could be accommodated without undue difficulty. However, in the
case of the two-bladed rotor the rotor itself has anisotropic dynamic properties.
Consequently, the dynamic equations must be developed in the rotating coor-
dinate system, and the convenient (constant coefficient) stability solutions can
be obtained only for an isotropic support structure. For systems with significant
fuselage anisotropy, recourse must then be made to a Floquet-type solution.
Derivation and general presentation of the actual equations of motion for the
two-bladed case are omitted in the interest of clarity and to permit us to focus
on the engineering results pertinent to rotor stability. In practice, a full-fledged
Floquet theory solution is seldom implemented for the two-bladed case. Instead,
the analysis is typically conducted twice assuming an isotropic pylon: once with
the isotropic properties set to the longitudinal properties and then once using the
lateral properties. Although the two-bladed case equations are set up in the rotating
coordinate system, most of the same generalizations used for the three-or-more-
bladed case are still valid for the two-bladed case.
Equations of motion: isotropic pylon case. For the case of an isotropic

pylon, the rotating coordinate system now contains the system anisotropy, and it
is appropriate to derive the equations of motion in the rotating coordinate system.
Figure 11.23 shows the basic modeling of the asymmetrically deflected rotor blades
in the rotating coordinate system.
The dynamic equations for two-bladed rotor ground resonance, as are those for
the three-or-more-bladed case, are typically derived using a Lagrangian approach,
the details of which are omitted herein. When the assumption of an isotropic pylon
is made, we can represent the pylon properties, which we formerly ascribed to the
x- or y-directions, as averaged quantities and denote them as barred quantities (− ).
With this point in mind, the appropriate equations of motion can then be written for
the (rotating) xR and yR motions, and the cyclic in-plane blade bending degrees
of freedom as follows:
xR -direction force:
M ẍR + c̄ẋR + M(ω2 − 2 )xR + 2M ẏR + c̄yR + 4S48 ˙ = 0 (11.71)
“chapter_11(new)” — 2005/11/24 — page 421 — #35

Fig. 11.23 Pictorial description of the asymmetrically deflected two-bladed rotor in

the rotating coordinate system.
yR -direction force:
− 2M ẋR − c̄xR + M ÿR + c̄ẏR + M(ω2 − 2 )yR

+ 2S48 (¨ − 2 ) = 0 (11.72)
Cyclic in-plane excitation:
−2S48 ẋR + S48 ( ÿR − 2 yR ) + S49 (¨ + 2ζv ωv0 ˙ + ωv2 ) = 0 (11.73)
where Mω2 = K, the pylon stiffness in either the x- or y-direction.
Frequency characteristics: Coleman diagram. The preceding equations

of motion are then seen to be of constant coefficient form and quite amenable to
a straightforward eigenvalue solution. If, for a case with zero damping (in both
the blades and the pylon), the imaginary parts of the resulting eigenvalues are
plotted vs rotor rotational speed, the results would typically look like those shown
in Fig. 11.24.
In addition to the points discussed earlier for the three-or-more-bladed case, the
following additional observations can be made:
1) The horizontal axis represents the shaft critical condition, and the intersec-
tions of the now-diagonal lines (representing the fuselage mode) with the horizontal
axis define the shaft critical speed conditions.
2) Because the rotor has different mechanical properties relative to the shaft
depending on whether the hub has deflected in a direction parallel or perpendicular
to the blade axis, two different shaft critical speed conditions exist. These two
“chapter_11(new)” — 2005/11/24 — page 422 — #36

Fig. 11.24 Typical Coleman diagram for ground resonance of two-bladed anisotropic
rotor and isotropic fuselage.
shaft critical speeds are the actual intersection points with the horizontal axis
as shown. However, the rotor speed range between these two shaft critical speeds
defines (in the rotating coordinate system) an additional aperiodic instability range
characteristic of two-bladed rotors, not present with the isotropic rotors. These
two shaft critical speeds are seen in the rotating coordinate system as a divergence.
Therefore, the values of shaft critical speed are then obtained by omitting the time
derivative and damping terms in the characteristic equation and solving for the two
rotor speeds C1 and C2 :
(ω2 − 2C )[(ω2 − 2C )ωv2 − 2 3 C ]

4
=0 (11.74)
The first shaft critical speed is given simply as
C1 = ω (11.75)
This critical frequency is seen to exclude any of the blade bending descriptors.
It is, in fact, the shaft critical speed that results when the shaft is deflecting in a
direction parallel to the blade axis (with zero blade bending). The second shaft
critical speed can be obtained in a compact form by making use of the simplified
relationship for the blade rotating natural frequency in terms of the nonrotating
frequency ωvNR and the rise factor Kv :
ωv2 = ωv2NR + Kv 2 (11.76)
“chapter_11(new)” — 2005/11/24 — page 423 — #37

Thus,
 1/2
 Kv ω − ωvNR + (Kv ω2 − ωv2NR )2 + 4(2 + Kv )ω2 ωv2NR 
2 2
3
c2 = 
2(2 3 + Kv )
(11.77)
This shaft critical speed does involve blade bending descriptors and corresponds
to the case wherein the shaft is deflecting normal to the blade axis and thereby
involves blade bending as well.
3) For the conditions wherein λ = i, a steady force in the fixed coordinate
system can excite a 1/rev (1P) response in the rotating coordinate system. The ram-
ification of this 1P response condition is that a 2P response will also be manifested
in the fixed coordinate system.
4) The implied methods for avoiding ground response with two-bladed rotors
are the same as those for the three-or-more-bladed case: Keep the blade natural
frequency subcritical (at a value above 1P), avoid values of rotor speed that result
in frequency coalescences with the pylon (fuselage) natural frequencies, reduce
the mass ratio, and provide damping in both the blade and pylon systems.
Effects of damping. The stability boundaries resulting from variations in

each of the damping coefficients follow the same trends as those found with the
three-or-more-bladed case. Also, a product of damping criterion can be formulated
for the two-bladed case as well, subject to the same limitations.
11.3.6 Aeroelastic Considerations

When ground resonance was first encountered, it was with conventional artic-
ulated rotor systems, and virtually no coupling occurred with the flapping degree
of freedom. Consequently, the instability could be considered to be a mechani-
cal instability governed principally by the elastomechanics of the blade lead-lag
motion and the in-plane impedance characteristics of the fuselage at the hub.
Rotor aerodynamics were consequently not needed for practical calculations of
the ground-resonance characteristics. However, with the advent of hingeless and
bearingless rotor systems the coupling between the flapping and in-plane blade
degrees of freedom has markedly increased. Consequently, for accurate calcula-
tions there are findings which indicate that the inclusion of quasi-steady airloads
and, in particular, of dynamic inflow should be included. This is because of the
ability of these newer rotor types to impart greater moments into the air mass than
was possible with articulated rotor types. Although some analytic results confirm
the increased correlational accuracy obtained by using dynamic inflow, the physics
underlying this increased accuracy is not yet well understood.
11.4 Air Resonance

11.4.1 Basic Considerations and Dynamic Equations
Of all rotorcraft aeroelastic phenomena presently known, air resonance poses
perhaps the greatest challenge to the aeroelastician by virtue of the richness and
“chapter_11(new)” — 2005/11/24 — page 424 — #38

variety of the interacting forces (inertial, aerodynamic, elastic, and gravitational)

needed to define its characteristics. Air resonance, as a rotorcraft instability phe-
nomenon, has continued to receive considerable attention in the literature, and
much is known both qualitatively and quantitatively of its physics. Stated briefly,
air resonance can be thought of as a combination of three forms of more ele-
mentary instability phenomena: the mechanical instability of rotors known as
ground resonance (but one that occurs in flight), propeller-nacelle whirl flutter,
and flap-lag instability. It is considered to be most related to the ground-resonance
phenomenon. The principal features of both ground resonance and air resonance
are given as follows: 1) the presence of a blade in-plane response degree of freedom
(bending mode or rigid lead-lag motion) characterized by being supercritical (i.e.,
one whose frequency in the rotating system is less than rotor rotation frequency );
2) a coincidence or close proximity of the frequency of the regressive, nonrotating
coordinate system manifestation of the supercritical blade in-plane mode with that
of a nonrotating fuselage mode having in-plane hub response components; and 3)
marginal to insufficient internal damping in either or both the blade in-plane mode
or the related fuselage mode.
The principal differences between ground and air resonance arise from their
respective sources of fuselage stiffness and damping. In ground resonance the sole
source of both stiffness and damping is typically the landing-gear assembly; in air
resonance the sources are multiple: gravity (or, alternatively, rotor thrust) and the
flapwise bending and aerodynamic damping of the rotor blades.
Pertinent degrees of freedom. Appendix D presents a simplified set of 11

equations used for examining the physics of the air-resonance phenomenon. The
minimum description selected for this simplified analysis includes nine degrees
of freedom: longitudinal, lateral, and vertical hub translations; hub roll and pitch
rotations; blade longitudinal and lateral in-plane bending; blade rolling and pitch-
ing flatwise bending and cosine and sine components of cyclic blade torsion. The
resulting state vector is given by
{Z} = [ẋ, ẏ, ż, θxF , θyF , x , y , θxR , θyR , φx , φy ]T (11.78)
These degrees of freedom are defined in Fig. 11.25.

The increased complexity of the equations for air resonance (over those for
ground resonance) is commensurate with the need to include aerodynamic as well
as gravitational effects. Consequently, coupling terms abound in the air resonance
equations.
Blade frequency characteristics. The behavior of the edgewise bending

mode frequency variation with rotor speed, as seen in the nonrotating coordi-
nate system, is the same for air resonance as it is for ground resonance. Because the
edgewise mode frequency is little affected by aerodynamics, the nonrotating coor-
dinate system perceived frequencies will be (ωv + ) and |ωv − |, respectively.
Because the flatwise bending mode is much more affected by the aerodynamics,
the damped frequency of the blade in flapping will be heavily affected by rotor
speed. Let the uncoupled flapping dynamics of the blade be given by the following
“chapter_11(new)” — 2005/11/24 — page 425 — #39

Fig. 11.25 (Eleven)-degrees-of-freedom basis for mathematical modeling of

air-resonance instability.
expression:
∗∗ γ ∗
β + β +ω2β β = 0 (11.79)
8
The (nondimensional) characteristic roots for this simple second-order system then
become
γ ! γ "2
λ=− ± i ω2β − (11.80)
16 16
where, using the rise factor representation for the rotating natural frequency.
1
ω2β = (ω2 + Kβ 2 ) (11.81)
2 βNR
A typical value for the rise factor Kβ of the first (flatwise) bending mode is 1.05.
These relationships can be combined to give the following expression for the
dimensional damped characteristic root:
γ
λ = λ = − ± i ωβ2NR + [Kβ − (γ /16)2 ]2 (11.82)
16
The behavior of the damped frequency ωβD [= Im(λ)] is shown in Fig. 11.26.
As this figure indicates, there is a certain value of Lock number for which the
flatwise frequency remains invariant. For Lock numbers greater than this value,
the damped frequency will eventually go to zero (at which point the characteristic
root pair becomes two aperiodic roots) for a sufficiently high value of rotor speed.
“chapter_11(new)” — 2005/11/24 — page 426 — #40

Fig. 11.26 Behavior of uncoupled flatwise bending frequency with rotor speed and
Lock number.
11.4.2 Extended Coleman Diagram

The Coleman diagram shown in Fig. 11.13 for the ground-resonance case can
be extended to the air-resonance case. For this case Fig. 11.27 identifies some
of the key features as they relate to the 11-degrees-of-freedom system defined in
Fig. 11.25.
Pertinent features of this extended diagram are given in the following:
1) There is comparatively more information on this Coleman diagram than for
the ground resonance case because of the increased number of pertinent degrees
of freedom.
2) The pylon degree-of-freedom modes are more sensitive to rotor speed since
a principal part of their stiffening derives from blade bending.
Fig. 11.27 Extended Coleman diagram for air-resonance instability modeling.
“chapter_11(new)” — 2005/11/24 — page 427 — #41

3) The coalescence of the edgewise modes with the pylon modes as a prime fea-
ture of the instability is not a distinguishing feature of air resonance. An instability
can be established without a clear-cut coalescence.
4) The air resonance phenomenon involves substantial coupling of the modes,
several of which are of relatively low frequency. Indeed, the modeling is capable
of simulating some of the low-frequency dynamics appropriate to evaluations of
vehicle dynamic stability and handling qualities. Although these roots are unstable,
they are not generally considered a structural dynamics problem because of the
pilot’s and/or autopilot’s ability to interact and stabilize them.
11.4.3 Pertinent Parameters and Scaling Considerations

As described in the preceding section, the ground resonance phenomenon was
found to depend on a small number of nondimensional parameters. The use of such
parameters is quite useful not only in describing the stability characteristics of a
wide range of configurations but in defining scaling considerations whereby the
phenomenon can be explored at model scale. Because the air resonance instability
depends on more physical effects than does ground resonance, a discussion of the
nondimensional parameters or scaling considerations is in order.
Nondimensional scale factors of the rotor. For a complete aeroelastic

modeling of the rotor, four basic scaling considerations must be met relating to
the proper interactions of the aerodynamic, elastic, inertial, and gravity forces
(see Hunt). Assuming complete geometric modeling, these interactions can be
stated mathematically in terms of the following nondimensional parameters, which
should be maintained invariant:
Frequency scaling:
E E
λf = 2
= (11.83)
m ρ(R)2
Lock number:
ρacR4
γ = (11.84)
Ib
Advance ratio:
V
µ= (11.85)
R
Froude number:
2 R
F= (11.86)
g
The frequency scaling parameter ensures that the rotor blades have the correct
natural frequencies in bending in relationship to rotor frequency. It is therefore
a measure of the ratio of elastic stiffening to centrifugal stiffening. The Lock
“chapter_11(new)” — 2005/11/24 — page 428 — #42

number ensures that the rotor has the correct aerodynamic damping and aerody-
namic coupling characteristics, and the advance ratio ensures that the scaling of
forward flight speed is correct in relationship to rotor rotational speed. The Froude
number ensures that the gravity effects, in terms of gravity springs and the rotor
thrust, are properly scaled in relation to the other three basic forces. The Froude
number is typically in the order of 500–700 and becomes increasingly important
with rotor size. Because the Froude number is relatively large compared with
the other nondimensional parameters, strict scaling of the gravitational terms can
sometimes be relaxed if their effects (as they relate to the phenomenon at hand)
can be approximated.
Scale factors of the pylon. For the rotor pylon to be properly scaled relative
to the rotor itself, it must present to the rotor a properly scaled impedance. This
can be achieved by matching 1) the mass ratios between the rotor mass and that
of the pylon, 2) any couplings existing between the hub degrees of freedom, and
3) the pylon natural frequencies (as nondimensionalized by the rotor frequency).
Mass ratios. Inspection of the air resonance equations shows that two mass
ratios of importance to the air-resonance instability phenomenon are 1) the already
discussed mass coupling parameter (involving rotor in-plane blade mode general-
ized mass and in-plane pylon effective mass) 3 and 2) a new parameter 4 . This
new parameter involves the rotor flapping blade mode generalized mass and pylon
rotational inertia, as taken about an effective focal point heff (see Fig. 11.12):
bS122
4 = (11.87)
2Ieff S10
where
Ieff = Meff heff

2
(11.88)
and where the integrals S10 and S12 are respectively given by
R R
S10 = R2 m γw2 dr; S12 = R m rγw dr (11.89)
0 0
Couplings. Couplings of the pylon degrees of freedom can occur because of

longitudinal center-of-gravity locations off the rotor rotation axis and because of
the focusing of the roll and pitch rotations about some position below the rotor
plane. A reasonable approximation to the free-flight condition is to take the focal
point to be the point on the rotor rotation axis that intersects the horizontal plane
containing the total aircraft gravity center. Indeed, it can be seen that only for a
scaling of the focal point at the aircraft center of gravity do the already described
mass ratios 3 and 4 scale commensurately. The choice of the aircraft c.g. as the
focal point is convenient because it eliminates the gravity forces as contributors to
the roll and pitch spring rates and thereby minimizes the effects of any non-Froude
scaling.
“chapter_11(new)” — 2005/11/24 — page 429 — #43

Pylon natural frequencies. The requirements for scaling the pylon impedance
also require that the natural frequencies (with respect to rotor rotational frequency)
of the pylon (with the constraint of it being focused at a point heff below the hub)
must be maintained. In the present context this scaling principle becomes important
when we might wish to alter the properties of an unscaled rotor also having an
unscaled focal point. Thus, for two configurations, ( )1 and ( )2 , that have the same
effective mass Meff , but are operating at different rotor speeds, 1 and 2 , the
frequency criterion then becomes that of maintaining the same effective rotational
stiffness.
The pylon stiffnesses in pitch and roll can each be approximated as a sum of
an explicit spring rate (around the focal point) Kp and an implicit one caused by
rotor flapping flexibility Kr . Furthermore, this rotor flapping spring rate Kr can
be conveniently expressed as a factor K̄r multiplying the rotor speed squared.
This factor is frequency dependent and proportional to the number of blades
and the already defined blade integration constant S12 . The K̄r factor depends
principally on Lock number γ , the nondimensional frequency of vibration ω,
and the blade flapping natural frequency ωw , again nondimensionalized by rotor
speed :
b
Kr = 2 K r S12 , γ , ω, ωw (11.90)
2
Then, the invariancy of pylon frequency criterion can be written as
Kp + K r 2
= invariant (11.91)
Ieff 2
2 , and M
But, Ieff equals Meff heff eff itself must also be invariant. Therefore, for
two different configurations that must both present the same impedance to the
rotor,

Kp / 2 + K r Kp / 2 + K r
2
= 2
(11.92)
heff heff
1 2
Noting that K̄r must be the same for both configurations, we can then
rewrite the expression to separate out the explicit stiffness rate for the second
configuration:

2 2
h eff K p heff
Kp2 = 22 2 1
+ K̄r 2
−1 (11.93)
heff1 21 heff1
It is readily apparent that, for 1 = 2 and heff1 = heff2 , the two explicit stiffnesses
are equal.
Effects of torsional dynamics and precone angle. An extensive study

(both analytical and experimental) of the effects on air resonance of a wide range of
parameters was made by Burkam and Miao. The experimental tests were conducted
“chapter_11(new)” — 2005/11/24 — page 430 — #44

Fig. 11.28 Analytical predictions of air resonance stability variation for the BO-105
model with feathering axis precone (after Burkam and Miao).
on a 1/13.8 Froude-scaled BO-105 model having nondimensional blade first flap,

lag, and pitch frequencies equal to those of the full-scale aircraft. This study
identified the importance of the proper selection of precone angle for those blade
designs that are relatively “soft” in torsion. The nominal torsion natural frequency
for this aircraft was given as 3.6. As is discussed in Chapter 3, a significant form
of nonlinear coupling exists giving a torsion moment proportional to the product of
flatwise(flapwise) bending and edgewise (lead-lag) bending. The amount of built-
in feathering axis precone angle in the rotor configuration β0 determines the degree
of steady flapwise bending the rotor will maintain. Through the mechanism of the
aforementioned nonlinear coupling, this steady flapwise bending creates a form
of pitch-lag coupling that can impact on the air resonance stability. Figure 11.28
presents analytical results for the BO-105 model showing the effect of feathering
axis precone angle on the air stability.
The figure clearly shows the beneficial effects of zero to even negative values
of precone in stabilizing the instability. Also in evidence is the destabilizing effect
of load factor. The stability degradation at the higher load factors is related to the
higher commensurate collective angles and the stiffening of the aircraft roll mode so
that this mode can then coalesce with the air resonance critical regressive lead-lag
mode (nonrotating frequency = − ωv ). The same trends were experimentally
confirmed with tests with the model, as shown in Fig. 11.29.
This figure not only confirms the analytical predictions for the effects of precone,
but also confirms the general presence of the higher collective angle instability
region. The point is made in the Burkam and Miao study that the instabilities were
dependent on the torsional stiffness of the blades and that stiffening the blades
generally improved the stability.
“chapter_11(new)” — 2005/11/24 — page 431 — #45

Fig. 11.29 Experimental stability boundaries for the BO-105 model, showing
variations in collective angle and precone angle (after Burkam and Miao).
11.4.4 General Guidelines for Stabilization

Because of the similarity of the air resonance phenomenon with three of the
instabilities already studied (rotor flap lag, rotor-nacelle whirl flutter, and ground
resonance), stabilization of this instability must draw upon knowledge of these
former phenomena. General guidelines that can be followed for stabilization are
as follows:
1) Frequency placement: A crucial item is the coalescence of the regressive
in-plane frequency with one of the fuselage response modes (usually rigid-body
pitch and/or roll). These coalescences should therefore be avoided or minimized.
2) Blade in-plane damping: The fuselage pitch and roll responses are sub-
stantially damped because of the aerodynamics of the main rotor. The in-plane
damping on rotors that would be subject to air resonance instability (hingeless and
bearingless rotors) is typically low because of the absence of lead-lag dampers.
Some damping can be introduced into the rotor by a judicious spar design and
selection of a material and/or fabrication technique that results in a (relatively)
high degree of structural damping. Additionally, a discrete damper (usually of
elastomeric design) can be installed.
3) Elastomechanical couplings: Some form of pitch-lag coupling can be
included that would otherwise stabilize an unstable flap-lag instability. Such cou-
plings can be incorporated in the detailed design of the torque-tube component of
bearingless rotor configuration, for example.
4) Precone of feathering axis: Depending on the torsional stiffness of the rotor
blade, the rotor blades should be designed with minimal precone angle. Generally,
keeping the blade torsion frequency relatively high will minimize the effects of
the nonlinear torsion coupling of flatwise and edgewise bending.
“chapter_11(new)” — 2005/11/24 — page 432 — #46

5) Active pitch control: The already present active stability augmentation sys-
tems in rotor control systems form the basic technology upon which an active air
resonance stabilization system can be configured. The frequency response required
of such a controller is well within the state of the art, and suitable feedback signals
can be obtained by accelerometer technology instrumentation. However, attention
must be paid to ensuring that such a stabilization scheme does not degrade existing
control system characteristics.
11.5 Air Mass Dynamics

A crucial aspect of instabilities addressed in this chapter is the ability of the
rotor blades to interact with each other through a suitable mechanism (primarily
the compliant motion of the pylon and, as a result, of the hub, i.e., hub mobil-
ity), in contrast to the case covered in the preceding chapter, wherein each blade’s
dynamic description is essentially uncoupled from that of all the others. However,
an additional source of coupling between the blades that typically operates in con-
junction with the coupling afforded by the hub mobility is that accruing from rotor
air mass dynamics. Air mass dynamics of the rotor is an idealization that attempts
to account for the gross effects of the unsteadiness of the rotor aerodynamic envi-
ronment when considered as a large rotor-sized air mass that operates on the rotor
as a whole and, in particular, only as a virtual mass and a virtual inertia.
11.5.1 Basic Concepts and Assumptions

The primary basic concept of the air mass dynamics formulation is that the mass
of air flowing through the rotor disk reacts as a virtual mass and inertia, that is,
accelerates, in response to the dynamic lifting force and hub moments generated
by the rotor, as a whole. As such, some simplifying assumptions must be made to
define the form of the accelerations in a suitably tractable, yet realistic, manner.
The formulations derive from the following basic assumptions:
1) The rotor operates as an actuator disk, wherein the air mass “sees” the rotor
to consist of an infinite number of blades.
2) The momentum equation holds on infinitesimal elements of rotor area, with
each such area uncoupled from all others.
3) The flowfield is void of any flow singularities (as would be associated with
vortical flow).
4) The radial and azimuthal variations of the air velocities induced by the rotor
lift and hub moments are given by the following simple function:
νi = R[ν̄0 + x(ν̄c cos ξ + ν̄s sin ξ )] (11.94)
5) The total rotor inflow (ratio) consists of the induced velocity and the ram
part caused by forward, axial, and/or rotational motion of the rotor through the
air:
λ(x, ξ ) = (V̄z + ν̄0 ) + x[(ν̄c + κν ν̄0 − ωy ) cos ξ + (ν̄s + ωx ) sin ξ ] (11.95)
where (− ) denotes nondimensionalization with respect to R and/or ; V̄z is the

axial flight speed, and ωy and ωx are, respectively, the rolling and pitching angular
“chapter_11(new)” — 2005/11/24 — page 433 — #47

rates of the hub. The preceding expression defines an obliquely truncated cylin-
drical function over the rotor disk; the spanwise variable x defines the “tilt” of the
truncation plane.
The κν term represents an empirically formed function to account for the fore
and aft tilt of the truncation plane in forward flight caused by rotor lift (even in the
absence of rotor hub moments). Stated in physical terms, rotor lift in forward flight
produces an “upwash” in the forward part of the rotor disk and a “down-wash” in
the aft part of the disk. This fore and aft inflow distribution would be typical of any
lifting surface in forward flight. The empirical representation for κν , as formulated
by Payne, is as follows:
4µ
κν = (11.96)
3.6|λ0 | + 3µ
where λ0 represents only the uniform part of the total inflow.

6) The cosine and sine components of the induced velocities ν̄c and ν̄s ,
respectively, are small relative to the uniform component ν̄0 .
11.5.2 Mathematical Formulation Using Momentum Considerations

Using the basic assumptions described earlier, we can construct a simple flow
model to relate the increment in pressure across an element of the rotor disk area
to the change in momentum experienced within a streamtube of air going through
the element of area. This modeling is illustrated in Fig. 11.30.
Incremental pressure. Using a standard implementation of rotor actuator

disk theory, but with the inclusion of nonstationary effects, we can express the
incremental element of pressure as
p
= 2 Vx2 + νz2 νi + he ν̇i (11.97)
ρ
where he is the effective height of a column of air that is being directly accelerated
(∂/∂t part of D/Dt), in contrast to the air that is being accelerated by convection
(U∂/∂x part of D/Dt). The actual value of he to be used depends on the type
of motion, but this issue will be deferred for the present. Invocation of the sixth
assumption given earlier allows the magnitude of the particle transport velocity
through the disk element to be expressed in the following manner:
 

 λ0 ν̃i 
Vx2 + νz2 ∼
= R  µ2 + λ20 +  (11.98)
µ2 + λ20
where
Vz + ν0
λ0 = (11.99)
R
“chapter_11(new)” — 2005/11/24 — page 434 — #48

Fig. 11.30 Idealized stream tube air flow through the rotor disk.
Here ν̃i , represents the part of the inflow velocity that can be assumed to be “small”:
ν̃i = x[(ν̄c + κν ν̄0 − ω̄y ) cos ξ + (ν̄s + ω̄x ) sin ξ ] (11.100)
Now let us expand the momentum equation, retaining linear terms where feasible
and nondimensionalizing by the rotor tip speed R. Also, from this point on the
nondimensionalization symbology is discarded:

p ∗
= 2 µ2 + λ20 ν0 + he0 ν 0
ρ(R)2
 
 µ2 + λ20 + λ0 ν0 ∗
+ x 2 νc + he1 νc  cos ξ
µ2 + λ20
 
 µ2 + λ20 + λ0 ν0 ∗
+ x 2 νs + he1 νs  sin ξ (11.101)
µ2 + λ20
Integrated hub loads. The integrated hub loads (rotor thrust, pitching, and
rolling moments) are simply formed from the incremental pressures. A simple
(very approximate) way of dealing with three-dimensional effects at the blade tips
is to delimit the extent of the integration at the blade tip means of the tip loss factor
B (typical value = 0.97). Also, at the same time the hub loads can be put into the
usual appropriate nondimensional rotor load coefficient form:
“chapter_11(new)” — 2005/11/24 — page 435 — #49

Rotor thrust:
1 2π B
CT = px dx dξ (11.102)
πρ(R)2 0 0
Rotor pitching moment:
−1 2π B
CM = px 2 cos ξ dx dξ (11.103)
πρ(R)2 0 0
Rotor rolling moment:
1 2π B
CL = px 2 sin ξ dx dξ (11.104)
πρ(R)2 0 0
Basic working form of equations. The preceding expressions can then be

combined to obtain working differential equations for the determination of the
induced velocities. In the implementation of the integrals, it can be recognized
that the integrations of the effective heights (he0 and he1 ) define classic problems
in incompressible fluid dynamics relating to the acceleration of ellipsoidal solids
(see Lamb). Thus, the integrations of these two quantities in the force and moment
equations given earlier results in different values. The final resulting equations are
then as follows:

CT 8 ∗
= 2 µ2 + λ20 ν0 + ν0 (11.105)
B2 3π
CM µ2 + λ20 + λ0 ν0 16 ∗
− 3 = νc + νc (11.106)
B 2 µ2 + λ2 45π
0
CL µ2 + λ20
+ λ0 ν0 16 ∗
= νs + νs (11.107)
B3 2 µ2 + λ20 45π
11.5.3 Modifications to the Basic Equations

The preceding equations thus constitute a set of first-order differential equations
that are weakly nonlinear and can thus either be used as is or linearized about some
steady value of ν0 . Before we discuss how these equations can be used, a few of
the equation set’s basic properties should be explored. First, the entire formulation
depends on the validity of using the momentum equation in such a simplistic type
of application. For certain rotor conditions the flow details are just too complicated
to allow for a gross modeling of the characteristics. Second, it must be realized that
the concept of air mass dynamics is really a mathematical construct and, as such, is
subject to a variety of interpretations and implementations. Indeed, the literature is
rife with alternate assumed forms that have been “validated” on the alternate bases
of mathematical self-consistency and/or correlation with experimental results. It
should be kept in mind that the concept is really an engineering tool that should
be used carefully.
“chapter_11(new)” — 2005/11/24 — page 436 — #50

Breakdown of momentum equation. Examination of the √ preceding equa-

tions shows that, as the total through flow of the rotor [the (µ2 + λ20 ) term]
approaches zero, 1) the thrust equation requires massive induced velocity ν0 in
order to produce a finite thrust, and 2) terms in the moment equations appear to
become singular. This state of affairs is clearly a breakdown of the momentum
assumption. Furthermore, a well-established feature of real rotor flow conditions
is the existence of the vortex-ring state. In this condition the rotor operates with
an inflow in a direction in opposition to the vertical translational motion of the
rotor (i.e., rotor in a vertical descent with inflow directed downward through the
rotor). This condition is known to be characterized by substantial recirculation of
the air through the rotor forming a ringlike flow structure. This flow condition
is quite unstable, and the rotor typically experiences significant loss of moment
controllability. It is no coincidence that the breakdown of the momentum equation
assumption correlates closely with the onset of the vortex-ring state.
The form of the momentum equations already derived is an attractive one from
the dynamicist’s point of view in that the form of the equations lends itself to a
variety of useful and easily implemented applications. Hence, it is advantageous
to modify the equation set to preserve its usefulness. To this end, the following
development is presented.
First, we rewrite the thrust equation in the following abbreviated form:
CT
= | | ν0 + A (11.108)
2B2
where A is the nonsingular time-derivative part, and where

| |= µ2 + λ20 (11.109)
Then this equation can be further rewritten as
A
1 = sgn(CT )| ˆ | ν̂0 + (11.110)
CT /2B2
where
(· · · )
(· ˆ· ·) = (11.111)
|CT |/2B2
The basic thrust equation can then be rewritten to isolate the offending singularity as
1 A
ν̂0 = sgn(CT ) 1− (11.112)
|ˆ| CT /2B2
which can then be generalized to the following form:
A
ν̂0 = sgn(CT )F ( ˆ ) 1 − (11.113)
CT /2B2
“chapter_11(new)” — 2005/11/24 — page 437 — #51

where the F ( ˆ ) is denoted the rotor-induced velocity function and for steady con-
ditions is the value of the equivalent induced velocity. In practice, the momentum
equation is reasonably valid in two operating conditions wherein there is substan-
tial generally uniform flow through the rotor disk. One is the thrusting propeller
state, wherein the flow is in a direction opposite to the thrust. The other valid case
is the windmill brake state, wherein the thrust is in the same direction as the flow.
These two cases define the two valid branches of the momentum equation that are
separated by the singular zero through-flow case and that must be joined by some
suitable empirical relationship.
Thus, the construction strategy is to join these two branches of the momen-
tum equation by some smooth nonsingular functionality. One way of achieving
the required approximation to this function is to use steady induced velocity test
data for rotors operating at or near the vortex ring state. Two sets of data for
rotors in vertical flight conditions are available for this purpose. A set of data col-
lected and presented in the work of Gessow and Myers and another set by Castles
and Gray (available in report form, see Bibliography for both works) provide the
basis for devising empirical fits to achieve a smooth transition between the two
branches of the momentum equation. The resulting complete functionality is given
in Fig. 11.31.
Observations that can be made from Fig. 11.31 are as follows:
1) Of the two experimental data sets available, the data set of Castles and Gray
shows the most marked deviation from the momentum equation. This curve is
Fig. 11.31 Rotor-induced velocity function.
“chapter_11(new)” — 2005/11/24 — page 438 — #52

probably more reliable in that the details of the test results are more accessible
and show features that are more consistent with actual identifiable flow details. All
subsequent observations are made relative to these data.
2) By normalizing the total through flow by the square root of the rotor thrust
coefficient (to give ˆ ), a measure of the relative magnitude of the thrust of the
rotor can be made. It is in the region of low absolute value of ˆ that the thrust can
be deemed to be “high,” in this case high relative to the valid momentum analysis
values.
3) The curve is strictly valid only in the vertical flight regime. However, the
empirical part of the curve deviates from the momentum equation only when the
nondimensional inflow is small, which occurs only in vertical flight. Indeed, for
both forward translational flight (helicopter operation) or forward axial flight (tilt-
rotor operation), the inflow parameter will assume high (negative) values, which
is seen to reduce the curve to the momentum analysis values.
4) The experimental bridge between the two momentum equation branches
at inflow parameter values near +1 is seen not only to have an infinite slope
but to be actually multivalued. This type of functionality, as described in a later
chapter, typically gives rise to hysteretic behavior in the dynamics. Indeed, such
behavior could well provide a mathematical vehicle for describing the unstable
flow conditions in the vortex-ring state.
5) The composite function assumes the role of a universal function servicing a
variety of rotor-related machines, both flight and nonflight (wind turbine) related.
6) The F = ˆ line corresponds to the hovering flight condition, and it is seen
that the fit of the data produces modification to the valid momentum value even at
hover.
7) In rotorcraft it is difficult to achieve “high” thrust conditions in the wind-
mill brake state (small positive values of ˆ ) because the actual inflow is typically
substantial, and the thrust is limited to the weight of the vehicle. However, in the
case of a stationary wind turbine neither of these limitations is ensured; indeed, the
thrust of a wind turbine is limited only by its stall conditions. Thus it is quite possi-
ble for a wind turbine to generate high thrust conditions with attendant deviations
from the momentum validity values of induced velocity.
8) For high positive values of the inflow parameter (representative of wind
turbine operation), the deviation of the experimental values from the momen-
tum validity values is consistent with the turbulent windmill state. In this
state the momentum equation requires a smooth deceleration and significant
streamtube expansion in the flow that is not attainable as a result of real flow
effects.
Reformulation using the rotor-induced velocity function. The develop-

ment already described can then be continued by first solving for the time
derivative part:
CT |CT | v0
A= − (11.114)
2B2 2B2 F ( ˆ )
“chapter_11(new)” — 2005/11/24 — page 439 — #53

Then the final (still nonlinear) working form of the thrust equation is
4 ∗ |CT | v0 CT
v0 + − 2 =0 (11.115)
3π 2B F ( ˆ ) 2B
2
This equation is now well behaved in that the F ( ˆ ) function never equals zero.
Also, examination of the moment equations shows that, for the critical vertical
flight condition where λ0 goes to zero (and concurrently, µ = 0), the similar terms
for vc and vs are seen to be proportional to the parameter V (a combination of v0
and λ0 , defined in the following section as the inflow mass flow parameter), which
is at least mathematically well behaved. However, more needs to be learned about
the behavior of the actual rotor moment characteristics in this critical condition.
11.5.4 Alternate Forms of the Air Mass Equations

Because the air mass dynamics equations are formed on the basis of rather
restrictive assumptions, the concept is one based more on mathematical grounds
than on real-world physics. Consequently, a variety of implementations of the basic
equations has arisen, mostly because any one implementation might or might not be
more practical than another in a given mathematical solution technique. Although
two basic general forms of the equations appear in the literature, both share the same
characteristic of defining a first-order differential equation set and are often seen
in linearized form wherein the vector quantities are “perturbational” quantities.
Form 1:
∗ 

  v 0   
v 0 
    CT 
∗
[τ ] v c + vc = [L] −CM (11.116)

     
∗   vs CL
vs
Form 2:
∗ 
     
v 0 
  v0   CT 
∗
[Ma ] v c + [L]−1 vc = −CM (11.117)

     
∗   vs CL
vs
where, based on the preceding development, the various matrices are given by
 8 
0
 3π 
 
 16 
[Ma ] =   (11.118)
 45π 
 16 
0
45π
“chapter_11(new)” — 2005/11/24 — page 440 — #54

 
2 0
 1 
 
[L]−1 = V  2  (11.119)
 
1
0
2
where V is typically referred to as the inflow mass flow parameter:
µ2 + λ20 + λ0 v0
V= (11.120)
µ2 + λ20
Pitt model. One air mass dynamics model that is often used because of its
high degree of mathematical consistency is that devised by Pitt and Peters. In this
model the [L] matrix is written in a form that smoothly transitions from hovering
flight to forward flight:
 $ 
1 15π 1 − sin χ
 − 0 
 2 64 1 + sin χ 
 
 $ 

1 15π 1 − sin χ 
[L] =  
4 sin χ
0

 (11.121)
V  64 1 + sin χ 1 + sin χ 
 
 
 4 
0 0
1 + sin χ
where
χ ≡ wake skew angle = tan−1 (λ0 /µ)

π
χ= (for hover) (11.122)
2
11.5.5 Use of Air Mass Dynamics in Aeroelastic Analyses

An important property of the formulation given for air mass dynamics is that,
in order to maintain the assumed character of the flow (that the induced velocity
structure varies spatially only to first harmonic in azimuth and linearly with span),
the air mass is responding to integrated rotor forces and moments and not individual
blade loadings. As a result, the air mass dynamics formulation is inherently a
low-frequency concept and, hence, must be deemed a low-frequency unsteady
aerodynamic analysis tool only!
Therefore, the air mass dynamics concept typically finds use in eigenvalue
analyses of low-frequency rotor aeromechanical and aeroelastic phenomena such
as total aircraft flight stability and handling qualities and flap-lag stability. A
typical implementation of linearized air mass dynamics would include 1) cou-
pling terms relating the perturbational forces (as defined by the aeroelastic
equilibrium equations) to the (also perturbational) air mass velocity components
“chapter_11(new)” — 2005/11/24 — page 441 — #55

(δvj = [δv0 , δvc , δvs ] and 2) a perturbational form of the air mass equations with
coupling terms relating the perturbations of the rotor hub loads accruing from per-
turbations of the vector of the selected rotor aeroelastic variables (δqR ). In matrix
form this implementation would have the following form:
 
..
 A 11 (λ) . A 12 (λ)  δqR
 . 
· · · · · · · · · · · · .. · · · · · · · · · · · · . . . . . . = 0 (11.123)
.. δvj
A (λ)
21 .
22 A (λ)
where
A11 (λ) ≡ basic aeroelastic equations for perturbational blade (modal) responses

= [M]λ2 + [C]λ + [K] (11.124)
A12 (λ) ≡ additional excitations to modal responses from perturbational air mass
dynamics velocities (arising from perturbational quasisteady blade air-
load distributions)
A21 (λ) ≡ perturbations in the integrated hub loads caused by perturbations in the
blade modal responses

∂(CT , −CM , CL )
=− (11.125)
∂(qR )

∂(CT , −CM , CL )
A22 (λ) = [Ma ]λ + [L]−1 − (11.126)
∂(vj )
References
Section
11.1 Hohenemser, K. H., and Yin, S.-K., “Some Applications of the
Method of Multiblade Coordinates,” Journal of the American
11.2 Houbolt, J. C., and Reed, W. H., “Propeller-Nacelle Whirl Flut-
ter,” Journal of the Aeronautical Sciences, Vol. 29, No. 3,
1962, pp. 333–346.
Johnson, W., “A Comprehensive Analytical Model of Rotorcraft
Aerodynamics and Dynamics, Part I: Analysis Development,”
NASA TM-81182, 1980.
Kvaternik, R.G., “Studies in Tilt Rotor VTOL Aircraft Aeroe-
lasticity,” Ph.d. Dissertation, Dept. of Solid Mechanics,
Structures and Mechanical Design, Case Western Reserve
Univ., Cleveland, OH, 1973.
“chapter_11(new)” — 2005/11/24 — page 442 — #56

Popelka, D. A., Sheffler, M. W., and Bilger, J., “Correlation of
Test and Analysis for the 1/5-Scale V-22 Aeroelastic Model,”
Journal of the American Helicopter Society, Vol. 32, No. 2,
1987, pp. 21–33.
Reed, W. H., “Propeller-Rotor Whirl Flutter: A State-of-the-Art
Review,” Journal of Sound and Vibration, Vol. 4, No. 3, 1966,
pp. 526–544.
Wernike, K. G., and Gaffey, T. M., “Review and Discussion of
‘The Influence of Blade Flapping Restraint on the Dynamic
Stability of Low Disk Loading Propeller-Rotors,’” Journal
of the American Helicopter Society, Vol. 12, No. 4, 1967,
pp. 55–57.
Young, M. I., and Lytwyn, R. T., “The Influence of Blade Flap-
ping Restraint on the Dynamic Stability of Low Disk Load-
ing Propeller-Rotors,” Journal of the American Helicopter
Society, Vol. 12, No. 4, 1967, pp. 38–54.
11.3 Bramwell, A. R. S., Helicopter Dynamics, Wiley, New York,

1976.
Coleman, R. P., and Feingold, A. M., “Theory of Self-Excited
Mechanical Oscillations of Helicopter Rotors with Hinged
Blades,” NACA Rept. 1351, 1958.
Gabel, R., and Capurso, V., “Exact Mechanical Instability
Boundaries as Determined from the Coleman Equation,”
1962, pp. 17–23.
Savant, C. J., Jr., Basic Feedback Control System Design,
McGraw–Hill, New York, 1958.
Sela, N. M., and Rosen, A., “The Influence of Alternate Inter-
Blade Connections on Ground Resonance,” Journal of the
11.4 Bielawa, R. L., “Validation of a Method for Air Resonance

Testing of Helicopters at Model Scale Using Active Control
of Pylon Dynamic Characteristics,” Journal of the American
Bielawa, R. L., “Simplified Dynamic Equations for Ground
and Air Resonance,” Rensselaer Polytechnic Inst., Rotorcraft
Technology Center, Troy, NY, Rept. D-90-1, 1990.
Bielawa, R. L., “An Experimental and Analytical Study
of the Mechanical Instability of Rotors on Multiple-Degree-
of-Freedom Supports,” Princeton Univ., Princeton, NJ, Dept.
of Aeronautical Engineering Rept. 612, 1962.
“chapter_11(new)” — 2005/11/24 — page 443 — #57

Burkam, J. E., and Miao, W., “Exploration of Aeroelastic Stabil-
ity Boundaries with a Soft-in-Plane Hingeless Rotor Model,”
1972, pp. 27–35.
Donham, R. E., Cardinale, S. V., and Sachs, I. B., “Ground and
Air Resonance Characteristics of a Soft In-Plane Rigid Rotor
System,” Journal of the American Helicopter Society, Vol. 14,
No. 4, 1969, pp. 33–41.
Hunt, G. K., “Similarity Requirements forAeroelastic Models of
Helicopter Rotors,” Aeronautical Research Council, London,
CP-1245, 1972.
Lytwyn, R. T., Miao, W.-L., and Woitsch, W., “Airborne and
Ground Resonance of Helicopter Rotors,” Journal of the
Lytwyn, R. T., “Aeroelastic Stability Analysis of Hingeless
Rotor Helicopters in Forward Flight Using Blade and Air-
frame Normal Modes,” Proceedings of the 36th Annual
National Forum of the American Helicopter Society, Paper
79-23, 1979.
Miao, W.-L., and Huber, H. B., “Rotor Aeroelasticity Coupled
with Helicopter Body Motion,” Proceedings of Specialists’
Meeting on Rotorcraft Dynamics, NASA Ames Research
Center, 1974.
11.5 Castles, W. Jr., and Gray, R. B., “Empirical Relation Between
Induced Velocity, Thrust and Rate of Descent of a Helicopter
Rotor as Determined by Wind-Tunnel Tests on Four Model
Rotors,” NACA TN 2474, 1951.
Gaonkar, G. H., Sastry, V. V. S. S., Reddy, T. S. R., Nagab-
hushanam, J., and Peters, D. A., “The Use of Actuator-Disc
Dynamic Inflow for Helicopter Flap-Lag Stability,” Journal
of the American Helicopter Society, Vol. 28, No. 3, 1983,
pp. 79–88.
Gessow, A., and Myers, G. C., Aerodynamics of the Helicopter,
Ungar, New York, 1967.
Lamb, H., Hydrodynamics, Dover, New York, 1932.
Pitt, D. M., and Peters, D. A., “Theoretical Prediction of
Dynamic-Inflow Derivatives,” Vertica, Vol. 5, No. 1, 1981,
pp. 21–34.
Rao, B. M., and Jones, W. P., “Application to Rotary Wings of a
Simplified Aerodynamic Lifting Surface Theory for Unsteady
Compressible Flow,” Proceedings of Specialists’ Meeting on
Rotorcraft Dynamics, NASA Ames Research Center, 1974.
“chapter_11(new)” — 2005/11/24 — page 444 — #58

Problems
11.1 A simple rotor-nacelle propulsion system has a four-bladed rotor with
constant chord blades and the following specifications:
R = 3 ft aR = 4 ft c/R = 0.10 b=4

= 3000 rpm WR = 200 lb
I nacelle = 40 lb-ft-s 2
airfoil section: c/α = 0.1/deg
(less rotor)
The propulsion system is to be stable (with 10% speed margin) at a flight

speed of 300 mph at standard sea-level conditions.
a) Determine the static divergence boundaries for the design at the crit-
ical operating condition in the Kψ − Kθ plane (in terms of axis intercepts
and asymptotes).
b) Assuming isotropic stiffnesses (K = Kψ = Kθ ) and critical damping
ratios in both the ψ and θ directions of 0.10, determine the spring rate K
needed for flutter stability.
11.2 Using the basic (rigid-blade) rotor-nacelle whirl flutter equations of motion
(as simplified to the isotropic case), show that the whirl instability must be
in the retrograde direction. (Hint: Combine the two degrees of freedom into
a single complex variable, Z = ψ + iθ , and assume sinusoidal motion.)
11.3 Using the results of Problem 11.2,

(a) Derive an expression for the (isotropic) spring rate at the flutter
boundary.
(b) For the specifications given in Problem 11.1, evaluate this spring
rate.
11.4 Verify that the product of damping stability criterion given in Sec. 11.3.2
reduces to the criterion used by Gabel and Capurso given in Sec. 11.3.3.
11.5 Derive the product of damping stability criterion given for the two-bladed
rotor case.
11.6 Using the momentum equation for hub (pitching) moment in hover, derive an
expression for the lift deficiency function [≡(CM )unsteady /(CM )QS ]. Note:
use the following assumptions and considerations:
1) A separate expression for pitching moment as a result of pitch rate
(using strip theory) is required:
(CM ) strip = F (ωy , vc ) = (CM )unsteady (11.127a)

theory
2) (CM )QS = F (ωy , 0) (11.127b)
“chapter_11(new)” — 2005/11/24 — page 445 — #59

3) Airfoil section: clα = 2π , B=1 (11.127c)
4) ∗
vc = 0 (11.127d)
11.7 Consider the tail rotor of a helicopter in yawing (hovering) flight at stan-
dard sea-level conditions. The (two-bladed) tail rotor specifications are as
follows:
R = 3.25 ft thrust T = 275 lb chord c = 0.65 ft

R = 712 ft/s tip loss B = 0.95
tail-rotor position aft of aircraft c.g., = 20.5 ft
Find the percent loss in first harmonic aerodynamic effectivity (lift deficiency
function) for respective helicopter yaw rates of ±100 deg/s. Note: rotor
inflow for the rotor ascending (i.e., moving in the same direction as the
thrust) at an ascent speed Vz is given by
CT
λ= (11.128)
2B2 (λ + Vz / R)
“chapter_11(new)” — 2005/11/24 — page 446 — #60

12
Unsteady Aerodynamics and Flutter of Rotors
12.1 Introduction and Classification

As our understanding of the dynamics and aeroelasticity of rotors increases,
the single most important item still needed for successful analysis continues to be
an accurate, universal modeling of the blade unsteady airloads. The paramount
importance of the airloads becomes quite apparent when one considers that a
helicopter rotor blade, unlike a conventional airplane wing, is quite flexible and
obtains its rigidity, for the most part, from centrifugal forces. Consequently, the
major forces acting on an element of a rotor blade are the inertia loads and the
airloads. Furthermore, as will become more apparent, the airloads even in steady
trimmed flight are characteristically quite unsteady and cause the rotor blade to
respond, to a marked degree, in such a way as to produce other significant unsteady
airloads.
12.1.1 Characteristics of the Rotary-Wing Aerodynamic

Environment
The importance of predicting these unsteady airloads to the helicopter rotor-
blade dynamics is equalled if not exceeded by the complexity in trying to make
a correct prediction of them. Some of the key features of the rotary-wing aero-
dynamic environment contributing to the complexity of the unsteady airloads are
as follows:
1) In forward-flight conditions the tangential velocity component at any arbi-
trary blade section UT has a relatively large first-harmonic content in rotor azimuth
angle ψ(=t), thereby producing zeroth-, first-, and second-harmonic (Fourier)
components of dynamic pressure.
2) The local section angle of attack α is composed of two parts: the geometric
part θ and the inflow part φ. Both of these portions of the angle of attack generally
have significant harmonic content because of the requirements for rotor control
and the physics of the induced airflow through the rotor disk.
3) At high forward-flight speeds the outer portions of the blade of the advancing
side of the disk encounter compressibility effects (i.e., elevated Mach numbers
arising from the summing of the rotational and forward-flight components of air
velocity). However, on the retreating side these same blade sections encounter
very low subsonic flow conditions (arising from the differencing of these same
air velocity components) together with high angles of attack. Such a combination
447
“chapter12(new)” — 2005/11/23 — page 447 — #1

generally leads to stalled airfoil conditions. Although each of these characteristics

is governed by nonlinear processes, their respective mathematical descriptions are
completely dissimilar. These two aerodynamic phenomena are the major sources
of nonlinear aerodynamic excitation to the blade.
4) The lifting blade leaves the disturbances (such as those arising from blade
motion, as well as from the strong trailing tip vortices) in its wake. Because the
mean velocity of the airflow normal through the rotor disk is generally small
compared to the tangential velocity, the rotor blade creating the disturbances, as
well as all of the other blades, can “sense” these disturbances at subsequent points
in time during several rotor revolutions later.
5) In response to the primary harmonic loads resulting from the generally
harmonic nature of the dynamic pressure in forward flight and the trim control
angles, the rotor blade will deflect elastically in bending and torsion in such a
manner that it generates significant secondary airloads that are also of a harmonic
nature.
6) In forward flight the rotor wake containing the disturbances already men-
tioned becomes distorted principally in the out-of-plane direction as it is being
swept rearward. It is this distortion, more than any other effect, that gives the
blade airloads their large higher harmonic content.
12.1.2 Basic Assumptions Defining the

Analytical Development
Historically, the development of rotary-wing unsteady aerodynamic theory has
been one of improvisation using elements gleaned from theories developed prin-
cipally for fixed-wing applications. The “fit” of such theories to the rotary-wing
case has been generally spotty. There does not yet exist a single, universal compre-
hensive theory that can be accurately applied, with ranges of parameters suitable
to most rotor aeroelastic problems. The use of aerodynamic theories of fixed-wing
origin is based on the concurrence of assumed conditions for the fixed-wing and
rotary-wing cases. The following list of basic assumptions helps to define the lim-
its to which fixed-wing formulations have proven useful in obtaining the albeit
limited present analytical tools for aeroelastic formulations:
1) Two dimensionality: An enduring assumption that has facilitated develop-
ment is as follows: The unsteady airloads are, except for local effects near the blade
tips, spanwise independent and can be treated on a two-dimensional basis. This
assumption follows from the fact that helicopters and other related rotary-wing
aircraft use rotor blades with relatively high aspect ratios, (AR) = R/c ≈ 10. The
use of two-dimensionality is further justified by the trend observed in fixed-wing
formulations, which show that with increased unsteadiness the relative effects of
spanwise propagations of disturbances (three dimensionality) are commensurately
diminished.
2) Reduced frequency: In accordance with the applicability of two-dimensional
formulations, the extent of aerodynamic unsteadiness can be assessed using the
basic parameter formulated for fixed-wing theory, the reduced frequency k, as
“chapter12(new)” — 2005/11/23 — page 448 — #2

UNSTEADY AERODYNAMICS AND FLUTTER 449
given in the following:

cω
k≡ (12.1)
2U
where, for rotary-wing applications, typical values can be established using the
tangential velocity caused by rotation (as measured at the 75% span location):
c ω 2cω
k≈ = (12.2)
2 (0.75 R) 3R
By virtue of the typical values for rotor-blade aspect ratio, the reduced frequency
is seen to be approximated by 0.0667(ω/ ). The value of the resulting reduced
frequency is then determined by the nondimensional frequency of the aeroelastic
phenomenon. Table 12.1 summarizes the resulting range of frequency ratios and
reduced frequencies (as seen in the rotating coordinate system) that exist for various
types of rotary-wing aeromechanical and aeroelastic phenomena.
As subsequent material will show, a measure of the significance of unsteadiness
is whether the reduced frequency is greater than approximately 0.06. For reduced
frequencies greater than this value, the unsteady airloads are reduced in magnitude
by approximately 10%. In addition to attenuations of the airloads in magnitude, and
probably more important, are the phase lags that unsteady effects introduce. Again,
as subsequent material will show, the maximum phase-lag effects occur at reduced
frequencies of approximately 0.25. The results of Table 12.1 indicate the follow-
ing: unsteady airloads become most important with regard to forced (vibratory)
response rotary-wing aeroelasticity problems and to higher-mode flutter problems.
3) Variable flow and airfoil motion: In contrast to the typical problems in fixed-
wing aeroelasticity, the aeroelastic problems of rotary wing (in forward flight) are
characterized by relatively large fluctuations in both the flow components and air-
foil section motion, simultaneously in both the longitudinal and vertical directions.
The full ramifications of this assumption have not yet been integrated into present
rotary-wing unsteady airloads theory. At best, pieces of the problem solution that
Table 12.1 Comparison of aeroelastic phenomena by reduced frequency
Nondimensional Reduced
Phenomenon Rotor type frequency ω/ frequency k
Pitch-lag instability Artic. 0.3 0.020

Ground resonance Artic. 0.3 0.020
Air resonance Nonartic. 0.8 0.053
1P flapping All 1.0 0.067
Flap-lag instability Nonartic. 1.2 0.080
2P flapping All 2.0 0.133
4P vibratory resps. All 4.0 0.267
5P vibratory resps. All 5.0 0.334
Torsional flutter All 4.0–6.0 0.267–0.400
“chapter12(new)” — 2005/11/23 — page 449 — #3

relate to the variability of flow in the two directions have been assimilated from the
fixed-wing case. However, the effects of large response amplitudes of the airfoil
section in the rotary-wing case have not as yet been fully addressed, and the fixed-
wing solutions, which have almost universally assumed infinitesimal motion, are
suspect and of limited usefulness.
4) Near- and far-field effects: Because of the rotor-blade’s ability to be influ-
enced by past disturbances from both itself and other blades, the rotor-blade wake
is composed of a near-field wake that consists of the trailing and shed vortex system
extending a few chord lengths behind the trailing edge of the blade and a far-field
wake that is incorporated with those from the other blades into a total rotor wake.
The modeling of this rotor wake has been approached from two directions: 1) the
concept of air mass dynamics, as treated in an earlier section, and 2) a detailed
flow calculation built up using vortex elements.
12.1.3 Classification of Rotor Unsteady Airloads Theories

Rotor unsteady airloads theories can generally be classified according to five
categories: purpose, complexity, elemental mathematical building block, treatment
of rotor wake, and method of implementation. Within this framework the optional
paths for each of these categories are as follows:
A. Purpose
1. Steady-state harmonic loads computation (both hover and forward flight)
2. Nonharmonic airloads for (higher mode) aeroelastic response and flutter
calculations (in hover)
3. Unsteady harmonic airloads for flutter and aeromechanical stability calcu-
lations in forward flight
B. Complexity
1. Lifting-line theory (strictly two-dimensional)
2. Lifting-surface theory (for low-aspect-ratio rotor blades and for conditions
at the tip)
C. Elemental mathematical principle (building block)
1. Two-dimensional airfoil
2. Elemental actuator disk (momentum equation)
3. Vortical elements
4. Pressure doublets
D. Treatment of the blade and rotor wakes
1. No blade wake and constant inflow only
2. Addition of one trailing sheet
3. Addition of many returning shed and trailing blade wakes
4. Nonuniform, time-varying rotor wake, creating a variable-inflow condition
at every instant of time at the airfoil
E. Method of (mathematical) implementation
1. Operations in the frequency domain
2. Use of complex frequency domain (Laplace operators)
3. Time domain using step-by-step time-history integration
It is within the context of this classification that the descriptions of rotary-wing
unsteady airloads to follow are presented. These descriptions constitute the first
“chapter12(new)” — 2005/11/23 — page 450 — #4

portion of this chapter; later portions of the chapter deal with two aeroelastic
instabilities, each of which require an appropriate form of unsteady aerodynamics
for an accurate analysis: bending-torsion flutter and stall flutter.
12.2 Two-Dimensional Frequency-Domain Theories

This and subsequent sections present four related aerodynamic theories that
were originally formulated either for fixed-wing applications and subsequently
made applicable to the rotary-wing case or derived for the rotary-wing case directly
using established fixed-wing theoretical methods. The four theories all make the
small-perturbation assumption (including the assumption of a “thin” airfoil). In
addition, the following specific assumptions are made consistent with linearized
theory:
1) The airfoil section consists of a flat plate with “small” initial angle of attack,
with a wake shed from the trailing edge of the airfoil rearward to infinity without
any downward velocity or vertical distortion.
2) The flow is inviscid and incompressible (M = 0).
3) The flow is governed by linear potential flow; the nonlinearities arising from
stall are omitted.
4) The disturbances, whether in the motion of the airfoil or in the flow, are
simple harmonic.
The four theories deal, respectively, with motion of the airfoil in pitch and
plunge (the Theodorsen problem), vertical sinusoidal gustiness imbedded in the
flow into which the airfoil is moving with constant speed (the Sears problem),
an expansion of the Theodorsen problem with an accounting of the returning
wakes from the same and other rotor blades (the Loewy problem), and sinu-
soidal variations of the flow in the longitudinal (streamwise) direction (the Isaacs
problem).
12.2.1 Theodorsen Function

The two-dimensional theory formulated by Theodorsen is one of the most endur-
ing (and probably overused) theories transcribed from fixed-wing aerodynamics.
This theory relates to the perturbational plunging (ḣ) and pitching (alternatively,
θ or α) motions of an airfoil in an air mass that is initially at rest (see Fig. 12.1).
Fig. 12.1 Schematic of the Theodorsen problem.
“chapter12(new)” — 2005/11/23 — page 451 — #5

As derived in numerous texts, the result of the formulation is the following

expression for the lift and pitching moment on an airfoil (having a unit span):
L = πρb2 [ḧ + U θ̇ − baθ̈ ] + 2πρUbC(k)[ḣ + Uθ + b(1/2 − a)θ̇] (12.3)

Mθ = πρb3 [aḧ − U(1/2 − a)θ̇ − b(1/8 + a2 )θ̈ ]
+ 2πρUb2 (1/2 + a)C(k)[ḣ + Uθ + b(1/2 − a)θ̇ ] (12.4)
where C(k) is the Theodorsen function, or, alternatively, the lift-deficiency

function. This function is complex valued and is defined by the following
expressions:
(2)
H1 (k)
C(k) = F(k) + iG(k) = (2) (2)
(12.5)
H1 (k) + iH0 (k)
(2)
where Hn are Hankel functions of the second kind, which are, in turn, composed
of Bessel functions of the first and second kinds. An extensive tabulation of the
Theodorsen function is given in the work of Luke and Dengler.
Approximate expression for C(k). For most applications, any possible

additional accuracy afforded by the rigorous use of the expression already given for
the Theodorsen function is unwarranted in light of the inherent restrictive assump-
tions in its formulation relative to the rotary-wing case (e.g., lack of small motions,
compressibility effects, nonstationary air mass, etc.). Furthermore, although the
function is comprised of analytical functions, it is still relatively cumbersome to
work with because the Hankel functions are transcendental functions and therefore
do not have convenient Laplace or (inverse) Fourier transforms. For these reasons
the unsteady airloads descriptions given earlier are often used with an approximate
Theodorsen function as given by
0.165 0.335
C(k) ≈ 1 − − (12.6)
1 − (0.0455/k)i 1 − (0.3/k)i
Note that this approximation now has well-defined transforms and has proven quite
useful, as we will see later, in providing aeroelastic operators that can operate in
the complex-frequency domain.
Basic characteristics of Theodorsen function. Figure 12.2 shows the

variations of the real and imaginary parts of both the actual and approximate
Theodorsen functions with reduced frequency. Note that the vertical scales in
the two parts of the figure are different. The real and imaginary parts are
asymptotic to 0.5 and 0, respectively.
Noncirculatory and circulatory components of airloads. Examination of

the preceding expressions for the unsteady lift and pitching moment shows that
each is comprised of two distinct types of terms. One part contains virtual mass
terms and acts as an inertia (involving substantial second derivatives in time). The
“chapter12(new)” — 2005/11/23 — page 452 — #6

Fig. 12.2 Real and imaginary parts of the Theodorsen function F(k) and G(k),
respectively.
second part contains circulatory terms and looks like a static airfoil coefficient
except with a multiplication by the Theodorsen function C(k). This term in these
expressions has the form of an attenuation factor on what would otherwise be
static (circulation-dependent) airfoil coefficients. Hence, the Theodorsen function
is generalized to represent a lift-deficiency function, with the understanding that
alternate lift-deficiency functions from other sources might be used in the same
manner as that stated in these equations.
Virtual mass terms. The inertia-like parts of the expressions given earlier
for lift and pitching moment are usually referred to as the noncirculatory or virtual
mass terms and represent the reactive forces exerted by the “equivalent” mass
and inertia of the air mass surrounding the airfoil. For rotary-wing applications
the doubly differentiated acceleration-like terms (¨) are typically small relative to
“chapter12(new)” — 2005/11/23 — page 453 — #7

the blade’s mass properties and are thereby neglected. The single-differentiated
velocity-like terms (˙) are relatively small in the expression for lift, but are large in
the expression for the pitching moment. In fact, the principal source of damping
in pitch comes from this virtual mass term:
1
Mθ = · · · − Ub 2 − a θ̇ − · · · (12.7)
Circulatory terms and quarter-chord points. The respective (bracketed)

terms in the expressions for the unsteady lift and pitching moment of the airfoil
are each in the form of the lift-deficiency function C(k) multiplying other aero-
dynamic descriptors. These descriptors include the freestream velocity U and the
quantity in brackets [ḣ + · · · ]. This bracketed quantity is merely the quasi-steady
angle of attack αQS multiplied by the freestream velocity, again U. This bracketed
quantity represents the air velocity normal to the airfoil as measured at the three-
quarter-chord point. Thus, the unsteady formulation can be reduced to a quasi-
steady formulation if 1) the angle of attack is defined using the normal air velocity
as measured at the quarter-chord point and 2) if the resulting modified angle of
attack is then multiplied by the lift deficiency function C(k). Thus, the unsteady
lift coefficient is denoted by
(cl )unst = cl (αequiv )static (12.8)
where the equivalent angle of attack is given by

αequiv = C(k) ḣ + Uθ + b 21 − a θ̇ /U (12.9)
The second chordwise point of importance is the one-quarter-chord point, which

defines where the lift acts to produce the pitching moment caused by the lift.
Thus, it can be seen that these two chordwise points, the one- and three-quarter-
chord points, define the role of the pitching motion and pitching moment in the
unsteady aerodynamic formulation: The unsteady airfoil behaves in a quasi-steady
manner with regard to the circulatory effects if the angle of attack is defined at the
three-quarter-chord point and the lift is taken at the one-quarter-chord point. This
principle is illustrated in the following sketch:
“chapter12(new)” — 2005/11/23 — page 454 — #8

Modification of the basic Theodorsen formulation to the rotary-wing

case. With the aforementioned interpretations in mind, the Theodorsen formu-
lation can be put in a form more appropriate to rotary-wing applications by means
of the following substitutions:
2π (lift-curve slope) = clα ; ba = xa − c/4

b (semichord) = c/2; U = r
ḣ (plunge motion) = (−)ż (flapping motion)
2 c ω
k= (12.10)
3R
where xa is the distance (forward) from the pivot point to the aerodynamic center.
With these substitutions the following general expressions for the lift and pitching
motion distributions (same as lift and pitching moment per unit span just given)
are obtained:
dL π c
= − ρc2 2 z̈/ 2 − r θ̇/ + xa − θ̈/ 2
dr 4 4
1 c
+ ρclα cr2 C(k) −ż/ + rθ + − xa θ̇/ (12.11)
2 2
dMθ π c c
= − ρc2 2 xa − z̈/ 2 + r − xa θ̇/
dr 4 4 2
3 2 c
+ c − xa + xa2 θ̈/ 2
32 4
1 c
+ ρclα cxa r2 C(k) −ż/ + rθ + − xa θ̇/ (12.12)
2 2
Generalization of the modified Theodorsen formulation. Although the
use of the formulation already given would appear quite limited in that it is based on
the relatively restrictive Theodorsen problem, it can be easily generalized by noting
that the unsteadiness is essentially modeled by the lift-deficiency function (in this
case, the Theodorsen function). By substituting different lift-deficiency functions
in place of the Theodorsen function, as appropriate, the unsteady airloads for other
assumed variable flow/airfoil motion conditions can be easily expressed. In the
following sections three different types of lift-deficiency functions are considered.
All of these functions can be used with the preceding equations to model alternate
identified characteristics of the rotary-wing aerodynamic environment.
12.2.2 Sears Function

The formulation of Sears is similar to that of Theodorsen: The airfoil is at zero
incidence and moving through the air mass with a uniform velocity U. The Sears
formulation differs from that of Theodorsen in that the airfoil is now assumed to
“chapter12(new)” — 2005/11/23 — page 455 — #9

Fig. 12.3 Schematic of the Sears problem.
be motionless in both pitch and plunge. The unsteadiness of the problem is instead
afforded by the sinusoidal vertical gust pattern imbedded in the air mass into which
the airfoil is moving, as shown in Fig. 12.3.
Thus, if the coordinate axes are fixed to the airfoil, the vertical gust can be
represented by a (vertical) velocity distribution flight U of the following form:
w(x, t) = Weiω(t−x/U) (12.13)
The waviness of the sinusoidal gust can alternatively be expressed using the wave-
length of the imbedded gust pattern or the frequency ω, with which the wave
passes any point of the airfoil:
ω = 2π U/ (12.14)
The results of Sears’ formulation give expressions again for unsteady lift and
pitching moment about the midchord. Sears’ formulation is unlike the Theodorsen
formulation, however, in that it is devoid of virtual mass terms (which is reasonable
because the airfoil is not moving) and similar to it in the appearance of a lift-
deficiency functionality to describe the unsteady features of the problem:
L = πρcUWeiωt φ(k); Mc/2 = L · (c/4) (12.15)
where the Sears function is again defined in terms of Bessel functions,

or, alternatively, by using the Theodorsen function in combination with Bessel
functions:
φ(k) = [J0 (k) − iJ1 (k)]C(k) + iJ1 (k) (12.16)
Figure 12.4 shows the variation of the Sears function in the complex plane together
with the Theodorsen function for comparison. Note that the asymptotic values for
the Sears and Theodorsen functions are, respectively, (0 + 0i) and (0.5 + 0i) as
k approaches ∞. As can be seen by this comparison, for frequencies typical of
helicopter aeroelastic problems there is little difference between these two lift-
deficiency functions. However, the figure does serve to highlight the difference
between the effects of unsteady angles of attack caused by airfoil motion (θ̇ ,
θ, and ḣ/U, i.e., the Theodorsen problem), compared with those of an angle
of attack caused by air mass motion (inflow angle φ, i.e., the Sears problem),
“chapter12(new)” — 2005/11/23 — page 456 — #10

Fig. 12.4 Vector diagram showing the real and imaginary parts of Sears’ φ(k) and
Theodorsen’s C(k) functions as functions of the reduced frequency k.
where the total angle of attack α is generally a combination of both basic types of
effects.
12.2.3 Physical Interpretation of Lift Deficiency Function

In the preceding sections the sinusoidally varying unsteady airloads are
presented in the quite useful form of a quasi-steady description modified by
a complex-valued (lift-deficiency) function. Generally, the effect of the lift-
deficiency function is to attenuate the airloads and to introduce a phase (lag)
angle. Herein, we will denote C(k) “generically” so that it will refer not only
to the Theodorsen function, but also to any function that mathematically modi-
fies the quasi-steady circulatory lift and pitching moment in the manner described
earlier and as given in the following abbreviated form:
Lcirc = LQS · C(k) (12.17a)

Mcirc = MQS · C(k) (12.17b)
where k = bω/U, and
C(k) = F(k) + iG(k) = C̄(k) · eiη(k) (12.18)

C̄(k) = F 2 (k) + G 2 (k) (12.19a)
−1
η(k) = tan [G(k)/F(k)] (12.19b)
Note that the Theodorsen function η(k) is always ≤0 because G ≤ 0

and F > 0. For this basic case the negative sign on η implies an aerodynamic
phase lag.
“chapter12(new)” — 2005/11/23 — page 457 — #11

Example 12.1
Consider a simple two-dimensional airfoil taken as a spring mass damper con-
strained to pivot about a (variable) point located at a distance xa behind the
aerodynamic center (quarter-chord). Thus, the airfoil is constrained to have zero
plunge (ḧ = ḣ = 0), as shown in the following sketch:
When the doubly time differentiated virtual mass terms are omitted, the equation
of motion for this system can be stated very simply as
Iθ θ̈ + cθ θ̇ + kθ θ = Mθ (12.20)
where the (sinusoidal) aerodynamic moment Mθ resulting from the θ motion (of
amplitude θ̄ ) is given by
 

 2x 

a 2xa ik 1 − 2xa
Mθ = A 1 − ik 1 − C̄(k)eiη(k) − θ̄ (12.21)

 c c 2 c 

(= F+iG)
and where A ≡ 2πρU 2 b2 .

It is instructive to investigate the aerodynamic pitch damping for rough-cut
variations in the pivot to quarter-chord distance xa . Denote the quantity (2xa /c) as
X, and rewrite the preceding equation in the following more compact form:

ik G
Mθ = A X[F − Gk(1 − x)] − 1 − X − 2X + F(1 − X) θ̄ (12.22)
2 k
or

ik
Mθ = A XR − S θ̄ (12.23)
2
“chapter12(new)” — 2005/11/23 — page 458 — #12

where
R = F − Gk(1 − X) (12.24a)
S = (1 − X)(1 − 2XF) + X(−2G/k) (12.24b)
The AXR quantity is then the effective aerodynamic spring, and the S factor is a
measure of the pitch-damping effectivity.
Because the possible range of X is from −0.5 (leading-edge pivot) to +1.5
(trailing-edge pivot), the R factor is always positive. Therefore, as expected, the
aerodynamic spring stiffness is proportional to the negative of X (i.e., a pivot point
ahead of the quarter-chord yields a stable “restoring” spring). Let us then examine
the behavior of the S factor for four principal pivot axis locations:
X = − 21 (corresponds to pivoting about the leading edge): For this value of X,
the S factor is given by
S = 23 (1 + F) − 21 (−2G/k) ≈ 21 [6 − (−2G/k)] (12.25)
Fung presents a handy tabulation of the Theodorsen function as well as the function
(−2G/k); from this tabulation it can be shown that for reduced frequencies less
than approximately 0.03, the quantity (−2G/k) is greater than 3(1 + F) (i.e., ≈6).
Therefore, for configurations with negligible mechanical damping, the airfoil has
negative pitch damping and is therefore unstable for these low values of reduced
frequency.
X = 0 (pivot location at the quarter-chord): For this value of X, the aerodynamic
spring rate goes to zero, and the pitch-damping effectivity S is equal to 1. Thus,
the system is unconditionally stable.
X = + 21 (pivoting about the midchord): For this case the aerodynamic spring
is unstable, but the pitch-damping effectivity is given by
S = 1/2[1 − F + (−2G/k)] (12.26)
which is positive for all values of k. Thus, for an airfoil that has sufficient mecha-
nical stiffness, the airfoil would be stable.
X = +1 (pivoting about the three-quarter-chord point): This case corresponds
to the case wherein the principal part of the pitch damping (accruing from the
virtual mass terms only) would be zero. In this case, however, the pitch-damping
effectivity is equal to (−2G/k), which is always positive, and the airfoil is still
stable. The variations already given serve to demonstrate how the phase lag of
the unsteady aerodynamics can destabilize an otherwise stable configuration and
stabilize an otherwise neutrally stable one.
12.2.4 Loewy Function

Unsteady effects including the returning wake. By way of introduction to
this topic, it is useful to consider the physics of the flowfield in the vicinity of a rotor
in hover or vertical flight. Such a rotor has an axial component of velocity through
the disk plane u accruing from the thrust that the rotor is developing and from the
“chapter12(new)” — 2005/11/23 — page 459 — #13

axial translational speed Vz . If this inflow velocity is assumed constant over the
rotor disk, standard momentum considerations give the following expression for
the inflow velocity u of the rotor:

Vz Vz2 T
u= + + (12.27)
2 4 2Aρ
or, in nondimensional form,

 
V̄z V̄z2 CT 
u = Rλ = R  + + (12.28)
2 4 2
where V̄z is the nondimensionalized vertical velocity = Vz / R, and CT is the

rotor thrust coefficient,
T
CT = (12.29)
π R2 ρ(R)2
The fully three-dimensional representation of the returning-wake structure can be

appreciated by building up the wake with increasing levels of complexity to the
problem as indicated in Fig. 12.5.
The predominant feature of the rotor flowfield is the strong trailing tip vortex,
which is blown downward at the velocity u to form a contracting helix, as shown in
Fig. 12.5a. Figure 12.5b shows the loci of the local airstreams leaving the trailing
edge of the rotor blade, thereby forming a helical surface. In general, the rotor
blade will develop oscillating loads that, from basic theory, will cause a radial
distribution of vorticity to be “shed” from the trailing edge. This shed vorticity
will lie on the helical surface, as depicted in Fig. 12.5c. Finally, because the
field is truly three dimensional, the flow description is further complicated by the
presence of unsteady trailing vorticity streaming from the trailing edge, as shown
in Fig. 12.5d.
These schematic drawings are meant to indicate, in general, the complexities
involved. To analyze the flow, some simplifying assumptions (where admissible)
must be invoked. One means of simplification is to consider extreme values of
inflow velocity so that the spacing between the surfaces (2π u/Q), where Q is
the number of blades is great. When this condition is fulfilled, the effect of the
vorticity in the helical surface is negligible after a fraction of “screw rotation,” or
distance down the helical surface. This case, which corresponds more nearly to
the Theodorsen problem considered earlier, is shown in Fig. 12.6a.
However, when the spacing is small, the helical surfaces “stack up” one on
another, as shown in Fig. 12.6b. Then, relative to the rotor blade, the vorticity in
the various surfaces, after integral multiples of screw rotation, can become more
important than vorticity in the helical surface only a fraction of screw rotation
behind the trailing edge. In a very real sense the aerodynamic influence of the
returning wakes defines the terms high and low inflow.
“chapter12(new)” — 2005/11/23 — page 460 — #14

Fig. 12.5 Elements of the hovering rotor flowfield (after Loewy).
“chapter12(new)” — 2005/11/23 — page 461 — #15

Fig. 12.6 Schematic representations of unsteady rotor fields (after Loewy).
Review of the Loewy theory. The returning wake concepts developed ear-
lier are incorporated in an approximate two-dimensional manner in the theory of
Loewy. Historically, this analysis was the first successful attempt to incorporate
the returning wake into the unsteady aerodynamics problem. Fortunately, because
only the hovering/vertical flight condition was considered, a closed-form solution
was possible. In addition to the assumptions inherent in the Theodorsen problem, it
is assumed here that the returning wake can be represented by a countably infinite
number of infinite (in length) sheets of vorticity parallel to the rotor plane as shown
in Fig. 12.7.
Fig. 12.7 Schematic of the Loewy problem.
“chapter12(new)” — 2005/11/23 — page 462 — #16

By use of two-dimensional vortex sheets, the following integral equation

relating vortex strengths and airfoil velocity is obtained:
! +b ! ∞
1 γa (ξ , t) dξ γ00 (ξ , t) dξ
va (x, t) = − C +
2π −b x − ξ b x − ξ
""
Q−1 ∞ ! ∞
γnq (ξ , t)(x − ξ ) dξ
+ 2
−∞ (x − ξ ) + (nQ + q) h
2 2
q=0 n=0
∞ !
#
" ∞ γn0 (ξ , t)(x − ξ ) dξ
+ (12.30)
(x − ξ )2 + n2 Q2 h2
n=1 b
Neglecting the last two terms, which are a result of the returning wakes, reduces the
formulation to that of Sec. 12.2.1 (the Theodorsen problem). The usual unsteady
aerodynamics assumption of sinusoidal motion is made at this point:
va = v̄a eiωt , γa = γ̄a eiωt (12.31)
The strengths of the trailing and returning wakes are then related to the strength
of the bound vorticity by means of 1) the phasing of the motion between the rotor
blades ψq ; 2) the spacing of the returning wakes h; 3) the number of rotor blades Q;
and 4) the reduced frequency k. After a nondimensionalization of the lengths by
the semichord b, the preceding integral equation then reduces to
! +b ! ∞ −ikx #
1 γ̄a (ξ ) dξ e dξ
v̄a (x) = − C − ik ¯ ¯
+ π k e −ikx
W (12.32)
2π −b x − ξ 1 x−ξ
where the amplitude to the total vorticity over the blade section ¯ is given by
! +1
¯ = eik γ̄a (x) dx (12.33)
−1
and W , which accounts for the total effect of the returning wake, is given by
$ %
Q−1 &(Q−q)/Q
1+ ekhQ ei2πm eiψq
q=1
W= (12.34)
ekhQ ei2πm − 1
In a like manner, as was used for the fixed-wing case (Theodorsen problem),
the preceding integral equation for ¯ is solved using Söhngen’s inversion formula.
The total circulation over the airfoil then becomes
! +1
2 (1 + ξ /1 − ξ )v̄a (ξ ) dξ
¯ −1
= ' ( (12.35)
(2) (2)
iπ 1/2 H1 (k) + iH0 (k) + [J1 (k) + iJ0 (k)] W
“chapter12(new)” — 2005/11/23 — page 463 — #17

Bernoulli’s equation, which relates the solution variable (vorticity) to the desired
working quantity (pressure), for sinusoidal motion is given by
! x
p̄(x)
= −γ̄ (x) − ik γ̄a (ξ ) dξ (12.36)
ρV −1
Use of this equation then yields the following expression, which represents the
desired solution as a function of the known (or prescribed) velocity at the airfoil
section:
(2)
−p(x) 2i/π [H0 (k) + 2J0 (k)W ]
= (2)
ρr (2)
H1 (k) + iH0 (k) + 2[J1 (k) + iJ0 (k)]W
! +1 )
1−x 1+ξ
× v̄a (ξ ) dξ
−1 1+x 1−ξ
! ) #
2 +1 1−x 1+ξ 1
+ C − ik1 (x, ξ ) v̄a (ξ ) dξ (12.37)
π −1 1+x 1−ξ x−ξ
√ #
1 1 − xξ + 1 − ξ 2 1 − x 2
1 (x, ξ ) = n √ (12.38)
2 1 − xξ − 1 − ξ 2 1 − x 2
The preceding equation for p̄ can be shown to be of the same form as a cor-
responding equation for the two-dimensional fixed-wing problem discussed in a
preceding section. Therefore, a modified lift-deficiency function exists and can be
expressed in the following general form:
C (k, m, h, Q, ψq ) = F (k, m, h, Q, ψq ) + iG (k, m, h, Q, ψq ) (12.39)
or, more specifically,

(2)
H1 (k) + 2J1 (k)W (k, m, h, Q, ψq )
C (k, m, h, Q, ψq ) = (2) (2)
(12.40)
H1 (k) + iH0 (k) + 2[J1 (k) + iJ0 (k)]W
It is readily apparent that the preceding equation degenerates to the Theodorsen

lift-deficiency function C(k) given in an earlier section, when W , the measure of
the influence of the returning wake effect, approaches zero.
Figures 12.8–12.10, reproduced from Loewy’s work, show the effects of varying
the newly defined parameters. Also shown are the deviations of the modified lift
deficiency from the Theodorsen function, which corresponds to the h = ∞ case
shown in the figures.
h ≡ h /b = 2π u/bQ = 4|λ|/σ ; (σ ≡ rotor solidity)

r ≡ r /b; m ≡ ω/ ; Q ≡ number of blades
ψq (for cyclic disturbances) = 2π(q/Q)(ω/ ); q ≡ blade index (12.41)
“chapter12(new)” — 2005/11/23 — page 464 — #18

Fig. 12.8 Real and imaginary parts of the Loewy function vs k as functions of inflow
parameter (m = 0) (after Loewy).
A significant and useful result of this formulation is that the modified (Loewy) lift-
deficiency function C can be directly inserted into the simple expressions for rotor-
blade lift and moment distributions given in Section 12.2.1 to give a more accurate
flutter/aeromechanical stability calculation for hovering or vertically ascending
rotors. Furthermore, the Loewy analysis represents a base upon which even more
accurate formulations can be formulated.
Damping characteristics. The impact of the returning wake can be appreci-

ated by examining the effects of the wake on the effective pitch and flap damping
coefficients. These effects are shown in Figs. 12.11 and 12.12. These figures can
be interpreted in the following ways:
1) The possibility of negative pitch damping exists when the frequency of
oscillation is close to a harmonic value. This implies the possibility of single-
degree-of-freedom flutter in blade torsion. Other figures given in the work of
Loewy (but not shown here) show that this negative damping is a strong function
of (pitch) rotation center position; the effect of the wake is found to be zero
when the rotation center is at the quarter-chord position, consistent with the other
two-dimensional theories given earlier in the chapter.
“chapter12(new)” — 2005/11/23 — page 465 — #19

parameter (m = 1/4) (after Loewy).
2) At integral harmonics of frequency, the flap damping is greatly attenuated

by the returning-wake effects. Two important simplifications to the modified lift-
deficiency function result in the limit as either h or k approaches zero:
J1 (k)
C = (12.42)
lim J1 (k) + iJ0 (k)
m→integer
h→0
h 1
C = = (12.43)
lim h+π 1 + (π σ/4|λ|)
m→integer
k→0
“chapter12(new)” — 2005/11/23 — page 466 — #20

parameter (m = 1/2) (after Loewy).
It can be verified that the minimum (inverted) peaks shown in Fig. 12.12 for flap
damping, when computed for nonzero values of k, are closely approximated by
this formula.
3) At nonintegral harmonics of the oscillatory frequency, as the spacing h
approaches zero, and as the reduced frequency k approaches zero (by virtue of
r → ∞), the lift-deficiency function approaches unity:
C =1 (12.44)
lim
m→integer
h→0
k→0
As a practical aeroelastic tool for rotors, the Loewy theory has four inherent
limitations.
First, it does not account for oscillations in the streamwise air velocity and is
therefore not valid for forward-flight conditions. Second, being a two-dimensional
“chapter12(new)” — 2005/11/23 — page 467 — #21

Fig. 12.11 Pitch-damping coefficient vs frequency ratio as a function of inflow

parameter (after Loewy).
theory, it cannot account for the interactional characteristics of spanwise

phenomena, such as the induced velocities of the concentrated tip vortex. Third,
because the theory makes the same assumption on the infinitesimal magnitudes
of the blade motion as the Theodorsen theory, it cannot account for the rela-
tively large motions of the blade sections of actual rotors. And fourth, because
Fig. 12.12 Flap-damping coefficient vs frequency ratio as a function of inflow

parameter (after Loewy).
“chapter12(new)” — 2005/11/23 — page 468 — #22

the theory is again based on the Theodorsen formulation approach, it must treat
aeroelastic problems only in the frequency domain and cannot easily treat prob-
lems of arbitrary motion in the time domain. The theory nevertheless represents
a substantial achievement in unsteady aerodynamics and is a pioneering effort in
the development of a truly rotorcraft-oriented theory. Moreover, this theory has
become a touchstone for newly emerging theories that continue to relax the basic
assumptions.
The crucial assumption of sinusoidal motion of the aerodynamic environment
is a basic impediment to an expanded, more generally applicable unsteady aerody-
namic formulation. In the following material concepts are presented that provide
basic arbitrary motion aerodynamic formulations. It is hoped that these concepts
will eventually lead to such a general unsteady aerodynamic formulation.
12.3 Two-Dimensional Arbitrary Motion Theories

The development of unsteady aerodynamic theories, suitable for predicting time
histories of arbitrary motions of the aeroelastic responses, has in one sense par-
alleled the early work on aerodynamic theory for sinusoidal motion, as described
earlier. The Theodorsen and Sears problems, which relate respectively, to sinu-
soidal airfoil motion in an otherwise motionless air mass and to a motionless airfoil
entering an air mass in sinusoidal motion, have counterparts in the time-history
domain in the Wagner and Küssner problems. In these latter (arbitrary motion)
theories the problem statements are the same (as for the Theodorsen and Sears
problems), except that the motions are defined in terms of unit step functions.
Development of more general rotary-wing formulations in the time-history
domain is a current direction being actively pursued. Much work is in progress,
and much yet remains to be accomplished. As developed in the ensuing sections,
the relationship between the Theodorsen and Wagner functions is that of a Fourier
transform pair. The characteristics of the Wagner problem have pointed the way
to a more general formulation of the rotor unsteady aerodynamic problem. The
Wagner problem and a more general implementation of it are presented in the
following material.
12.3.1 Classical Methods Using Theodorsen and

Wagner Functions
Recall from earlier sections that the lift and moment on a two-dimensional
airfoil can be characterized by the presence of noncirculatory (virtual mass) and
circulatory terms that have the following features:
1) The noncirculatory terms are already defined in the time-history domain and
hence are appropriate for descriptions of arbitrary motion θ(t) and h(t).
2) The circulatory terms are composed of a product of two terms: C(k) and the
vertical velocity at the three-quarter-chord point on the airfoil. The latter quantity
is also defined in the time-history domain:
w3c/4 (t) = [ḣ + Uθ + b(1/2 − a)θ̇ ] (12.45)
“chapter12(new)” — 2005/11/23 — page 469 — #23

3) The lift-deficiency function C(k) is defined only for simple harmonic motion,
that is, in the frequency domain.
We require an unsteady aerodynamic formulation that is described entirely for
arbitrary airfoil motion. To this end, four possible approaches present themselves
and are respectively described in the subsequent material: 1) use of the Theodorsen
function together with a rigorous application of Fourier transform (integral) tech-
niques; 2) use of the Wagner function (derivable from the Theodorsen function)
together with the Duhamel integral; 3) direct use of the generalized Theodorsen
function C( p), with a (complex) Laplace transform argument; and 4) use of an
approximate Wagner function, again using Laplace transform techniques.
Fourier transform method. The complete transformation from the freq-

uency domain to the time domain and back again is ensured by a straightforward
usage of the Fourier integral transform. Let
! +∞
f (ω) = w3c/4 (t)e−iωt dt (12.46)
−∞
Then the expressions for lift and moment become
L = πρb2 [ḧ + U θ̇ − baθ̈]

! +∞
+ ρUb C(ωb/U) f (ω)eiωt dω (12.47)
−∞
Mθ = πρb2 [baḧ − Ub(1/2 − a)θ̇ − b2 (1/8 + a2 )θ̈]

! +∞
+ ρUb (1/2 + a)
2
C(ωb/U) f (ω)eiωt dω (12.48)
−∞
The principal disadvantage with this method is that the various integrals, although
completely valid for any arbitrary motion, are quite cumbersome to evaluate for any
general arbitrary vertical velocity w3c/4 (t). The method described in the following
section circumvents this disadvantage by performing the required Fourier integral
transform once for a basic type of transient motion, the unit step function, and then
by building up arbitrary motion using superposition.
Wagner function and Duhamel integral. Wagner used the technique of the
previous Fourier integral method for a step change in angle of attack:

0, t<0
w3c/4 (t) = (12.49)
Uα0 , t ≥ 0
Then, the Fourier transformed (circulatory) lift is given by

! +∞
C(k) iks
Lc (s) = ρbU 2 α0 e dk = 2πρbU 2 α0 φ(s) (12.50)
−∞ ik
where s ≡ Ut/b (the aerodynamic time variable).
“chapter12(new)” — 2005/11/23 — page 470 — #24

The Wagner function and the exponential approximation attributed to Jones are
depicted in Fig. 12.13. Use of the Duhamel or superposition integral enables the air-
loads resulting from any arbitrary angle-of-attack time history to be mathematically
constructed from the Wagner function:
L = πρb2 [ḧ + U θ̇ − baθ̈ ] + 2πρbU Wφ (12.51)

Mθ = πρb [baḧ − Ub( 21 − a)θ̇
2
− b2 ( 18 + a )θ̈]
2
+ 2πρUb( 21 + a)Wφ (12.52)
where
! s dw3c/4 (σ )
Wφ = w3c/4 (0)φ(s) + φ(s − σ ) dσ (12.53)
0 dσ
Although these results represent a considerable improvement over the pure Fourier
integral approach, they are as yet sufficiently complex to preclude their use in a
flutter (instability) analysis. This type of formulation typically finds application
mainly in the calculation of transient loads wherein the airfoil motion is known a
priori.
Generalized Theodorsen function. The concept of using the Theodorsen

function with a complex argument (as a Laplace transform variable p) in a more
general eigenvalue analysis is certainly attractive from an analysis point of view.
However, for many years, as attractive as the concept was, it was found that
the direct use of the Theodorsen function with a complex argument led to sin-
gularities with regard to using Hankel functions with complex arguments. This
Fig. 12.13 Wagner function.
“chapter12(new)” — 2005/11/23 — page 471 — #25

difficulty was found to be caused by the lack of compatibility of initial con-

ditions between the original frequency-domain formulation of the Theodorsen
function and what is required in a true Laplace transform formulation. Subse-
quently, Edwards et al. resolved this difficulty by a reformulation that includes
considerations of the initial conditions. The resulting formulation retains the basic
form that one would obtain with a straightforward (but erroneous) direct substi-
tution of p̂[=(b/U)(σ + iω) = σ̂ + ik] for ik in Theodorsen’s expressions for lift
and moment, except that the Laplace-transformed lift-deficiency function must
now use modified Bessel functions rather than Hankel functions:
L( p) = πρb2 [ p2 h( p) + (Up − ba)p2 θ( p)]

+ 2πρUbC( p̂){ph( p) + [U + b( 21 − a)p]θ( p)} (12.54)
Mθ ( p) = πρb {ap h( p) − [U( 21 − a)p − b( 18 + a )p ]θ( p)}
3 2 2 2
+ 2πρUb2 ( 21 + a)C( p̂) · { ph(p) + [U + b( 21 − a)p]θ( p)} (12.55)
where C( p̂), the generalized Theodorsen function, is given by

K1 ( p̂)
C( p̂) = F( p̂) + iG( p̂) = (12.56)
K1 ( p̂) + K0 ( p̂)
where
b b
p̂ = p = (σ + iω) (12.57)
U U
The principal disadvantage of this method is its lack of sufficient generality in that
the required modified Bessel functions are not readily defined in terms of ratio-
nal functions. The general inability of transcendental (irrational) functions to be
inverse Laplace-transformed limits their practical usage in routine dynamic ana-
lyses. However, the principal usefulness of this form of the Theodorsen problem
is that it gives a firm mathematical foundation to techniques already used to incor-
porate Laplace transform techniques in aeroelastic problems. Such techniques are
based on devising and using approximations to the Laplace transforms involving
transcendental functions, which are then of a rational function form and are there-
fore easily inverse Laplace transformable. An example of such a method is given
in the next section. Note that the approximation techniques used are applicable to
a wide range of lift-deficiency functions for which a Laplace transform invertable
form is desired.
Laplace transform with approximate Wagner function. In Fig. 12.13 an

exponential approximation to the Wagner function was given as
φ(s) ≈ 1 − 0.165e−0.0455s − 0.355e−0.3s (12.58)
Besides giving a fairly accurate fit to the “exact” Wagner function, this approxi-
mation has two desirable characteristics. First, it has the character of a decaying
functionality, which is entirely consistent with the physics of the unsteady flow
“chapter12(new)” — 2005/11/23 — page 472 — #26

Fig. 12.14 Time histories of aerodynamic angles with variations in quasi-steady

angles of attack.
“chapter12(new)” — 2005/11/23 — page 473 — #27

phenomenon; disturbances at the airfoil section are convected away from the sec-
tion by the airstream so that their influences are monotonically lessened. Second,
from a mathematical point of view the approximation has a straightforward easily
calculable Laplace transform. The Laplace transform is defined as
! ∞
L[ f (t)] = F(p) ≡ f (t)e−pt dt (12.59)
0
Therefore,
! ∞
b
L[φ(Ut/b)] = [1 − 0.165e−0.0455s − 0.355e−0.3s ]e−psb/U ds
U 0
0.5p2 + 0.281(U/b)p + 0.0136(U/b)2
= (p) = (12.60)
p[ p + 0.3(U/b)][ p + 0.0455(U/b)]
where p behaves like d( )/dt.
Then, the Laplace transform for lift and moment can be written as
L( p) = πρb2 [ p2 h( p) + (Up − bap2 )θ ( p)] + 2πρUbp( p){ ph( p)

+ [U + b( 21 − a)p]θ( p)} (12.61)
Mθ ( p) = πρb {ap h( p) − [U( 21 − a)p − b( 18 + a2 )p2 ]θ( p)}
3 2
+ 2πρUb2 ( 21 + a)p( p) · { ph( p) + [U + b( 21 − a)p]θ( p)} (12.62)

Thus, it can be seen that the product of the Laplace transform variable with the
Laplace transform involving the Wagner function [ p( p)] is exactly equal to the
generalized Theodorsen function [C( p̂)]. The preceding expressions for lift and
moment distribution, although utilizing a practical approximation for the Wagner
function (and for the generalized Theodorsen function), provide a sufficiently
general description of the airloads to enable one to make linear flutter formulations
of multiple-degree-of-freedom systems involving pitch and plunge of the airfoil
sections. These expressions can then be written in matrix form involving the vector
of Laplace-transformed plunge and pitch degrees of freedom of the airfoil section
[h(p), θ ( p)] :

L( p) 1 −ba 0 U
= πρb2 2 ( 1 + a2 ) p + 0 −Ub( 1 − a) p
2
Mθ ( p) ba −b 8 2
0.5p2 + 0.281(U/b)p + 0.0136(U/b)2
+ 2πρUb
[ p + 0.3(U/b)][ p + 0.0455(U/b)]
# #
1 b( 21 − a) 0 U h( p)
× p+
2 + a)
1 θ( p)
b( 2 + a) b ( 4 − a )
1 2 1 2 0 Ub(
(12.63)
The preceding expressions for the loads (lift and pitching moment) are stated
in the form of transfer functions, wherein the loads are expressed as Laplace
“chapter12(new)” — 2005/11/23 — page 474 — #28

transform operations of the responses. This has the advantage that the (input)
responses do not need to be explicitly stated, thus making the expressions quite
useful for eigenvalue analyses.
12.3.2 Unsteady Decay Parameter

The Laplace-transformed approximate Wagner function, and the Duhamel inte-
gral use of the function, form the basis for more general tools and additional
analysis capabilities that are particularly useful to rotary-wing applications. It is
highly desirable to retain and extend the use of quasi-steady airload descriptions to
include true unsteady effects. The approach followed here and by Leishman is to
define an effective angle of attack, which, when used in a quasi-steady formulation,
produces the lift-deficiency effects (attenuation and phase lags) that characterize
unsteady aerodynamics. One method for accomplishing this is to define a parame-
ter that incorporates the unsteady effects and acts to modify the usual quasi-steady
formulation. We will define this parameter as the unsteady decay parameter. As
we will see, because this parameter is a “dynamic” quantity it can further serve
to model other unsteady aerodynamic phenomena that traditional methods cannot
address. The approach begins with the expression for circulatory lift caused by an
arbitrary quasi-steady angle of attack αQS :
! s
dαQS (σ )
Lcirc = 2πρbU 2 αQS (0)φ(s) + φ(s − σ ) dσ (12.64)
0 ds
which can be written in a more compact notation, replacing the theoretical value
of the lift-curve slope 2π by the actual two-dimensional value clα :
Lcirc = 1/2ρcclα U 2 αE (12.65)
where, after performing an integration by parts on Eq. (12.64), the effective angle
of attack αE can be put into the following form:
! s
1
αE = αQS + αQS (σ )φ (s − σ ) dσ
2 0
! s
1
= αQS − αQS − αQS (σ )φ (s − σ ) dσ
2 0
= αQS − αW (12.66)
where the quantity αW is herein defined as the unsteady decay parameter.

! s
1
αW ≡ αQS − αQS (σ )φ (s − σ ) dσ (12.67)
2 0
With the inclusion of the unsteady decay parameter, the effective angle of attack
αE can then be used in a quasi-steady manner. It can be easily verified that the
Wagner problem is duplicated using the unsteady decay parameter when the quasi-
steady angle of attack is a step function. Moreover, the role of this parameter is
“chapter12(new)” — 2005/11/23 — page 475 — #29

seen as an “adjuster” to the quasi-steady angle of attack: as a quasi-static condition

is approached (unsteady effects are dying out), the unsteady decay parameter also
approaches zero and the effective angle of attack then tends to the quasi-static
one. Figures 12.14a–12.14c present examples of the use of the unsteady decay
parameter in defining time-histories of effective angles of attack: a) two cycles of
quasi-steady angle of attack, b) one cycle with an abrupt cutoff, and c) one-quarter
cycle with a maintaining of the maximum value.
Figure 12.14a presents the case of a startup of a sinusoidal variation in the quasi-
steady angle of attack. The figure clearly shows the convergence of the effective
angle of attack αE to the sinusoidal variation that would accrue when using the
Theodorsen function. Note that the Theodorsen effective angle of attack is shown
as an ongoing purely sinusoidal function. Figure 12.14b shows the variations in
the unsteady decay parameter and the resulting effective angle of attack for the
case wherein the quasi-steady angle of attack is cut off after one cycle. The decay
feature of the unsteady decay parameter and that of the resulting effective angle
of attack are clearly seen. Finally, Fig. 12.14c presents the aerodynamic angle
behaviors for the case wherein the quasi-steady angle of attack is held at the unit
value after the cycle has reached the end of its first quarter of oscillation.
Approximate compressibility corrections. Beddoes suggests a compress-

ibility correction to the Wagner function in terms of the Prandtl–Glauert correction
factor β:
1
(1 − 0.165e−0.0455β s − 0.355e−0.3β s )
2 2
φc (s, M) = (12.68)
β
where β = (1 − M 2 ).
As stated earlier, and as the following development will make more apparent,
we wish to work with the actual lift-curve slope clα instead of the theoretical
incompressible value (≈2π). Hence, the division by β in the preceding expression
will be omitted because it will be inherently included with the actual (Mach-
number-dependent) lift-curve slope. The compressibility corrected unsteady decay
parameter then becomes
! s
αW ≡ 1/2 αQS − αQS (σ )βφc (s − σ ) dσ (12.69)
0
Time-history formulations. The preceding expression for the unsteady decay

parameter serves mainly as a definition and, more importantly, as a starting point
for formulating practical tools for rotor aeroelastic calculations. The principal
advantage of the effective angle-of-attack approach (utilizing the unsteady decay
parameter) is that it provides a convenient method for including unsteady effects
in aeroelastic analyses utilizing time-history solutions.
A general characteristic of this class of aeroelastic analyses is the use of a
quasi-steady airloads formulation wherein the airfoil coefficients (cl , cd , and
cmac ) are obtained using table look-ups. That is, at any instant of time, for any
arbitrary blade section the instantaneous section angle of attack α and Mach
number M are calculated. Typically, the angle of attack is calculated using the
“chapter12(new)” — 2005/11/23 — page 476 — #30

quasi-steady assumption wherein the inflow angle is calculated using the air
velocity components at the three-quarter-chord point. These two variables are then
used (with appropriate interpolation schemes) to look up the aerodynamic section
coefficients using available static airfoil tables. The wind-tunnel testing of airfoils
over ranges of α and M appropriate to the operational envelope of the rotorblade
sections has become a routine part of new helicopter development. Consequently,
a large body of airfoil data, as functions of α and M, currently exists.
Rotor aeroelastic analyses incorporating such table look-up schemes generally
have the advantage of providing a more accurate modeling of the low- (reduced-)
frequency airloads needed for trim calculations and for low-frequency aero-
mechanical instability analysis. Furthermore, these time-history analyses provide
a straightforward, practical vehicle for solving rotor aeroelastic problems, both
vibration and stability related, which are governed by strong nonlinearities. The
analyses of such nonlinear phenomena, using time-history solution techniques, are
treated in subsequent sections. Note that, as developed in the preceding section,
eigenvalue analyses that use a linear form of airloads (lift-curve slope times angle
of attack) together with a Laplace-transformed Wagner function can be formulated.
For such analyses the effective angle-of-attack approach described herein offers
no advantage since the two methods are actually identical.
The basis for the use of the effective angle-of-attack method is the facility in
solving for the time history of this variable quantity using an appropriate differ-
ential equation. To this end, the convolution integral representation given earlier
is an ideal definition because it allows a transfer function to be formed and can
also be calculated using a recursive numerical time-marching scheme. When we
take the Laplace transform of the effective angle of attack (relative to aerodynamic
time s), the following relationships are formed:
1 0.007508 0.1005
ÂE = ÂQS + β 2 + ÂQS (12.70)
2 p̂ + 0.0455β 2 p̂ + 0.3β 2
where
ÂE = L(αE ); L( ) ≡ Laplace transform with respect to s
ÂQS = L(αQS )
p̂ ≡ (aerodynamic time) Laplace transform variable
[equivalent to d( )/ds]
To complete the formulation, the aerodynamic time s must be related to real
time t. For fixed-wing problems wherein the streamwise air velocity is a constant,
the aerodynamic time is merely proportional to real time by the factor (U/b). For
rotary-wing problems, however, the streamwise velocity is generally variable, and
the relationship between the two timescales must be taken differentially.
ds = [2U(t)/c] dt (12.71)
Then the real-time differential operator p[=d( )/dt] can be formed by means of
the following substitution:
d( ) d( ) dt
p̂ = = · = [b/U(t)] p (12.72)
ds dt ds
“chapter12(new)” — 2005/11/23 — page 477 — #31

which then gives
1 0.007508 0.1005
AE = AQS + β 2 (U/b) + AQS
2 p + 0.0455β 2 (U/b) p + 0.3β 2 (U/b)
(12.73)
Numerical solution for the effective angle of attack in the time domain.
The calculation of the effective angle of attack in time-history analysis typically
requires some form of (time-marching) numerical integration algorithm for the
unsteady decay parameter. The appropriate time-marching variable for our pur-
poses is the aerodynamic time variable s. Thus, the kth time-marching integration
interval sk must be related to real time t as follows:
2U(t)
(s)k = t (12.74)
c k
Taking this (nondimensional) time interval sufficiently small to ensure both numer-
ical stability and accuracy, and the freestream velocity U(ψ) constant over the
time interval, leads to a recurrence relationship for calculating the unsteady decay
parameter. The numerical recursive representation of the unsteady decay parameter
begins by breaking the parameter into a sum of two parts:
(αW )k = Xk + Yk (12.75)
where the formulations for Xk and Yk begin with the following Duhamel integral
expressions:
! s
X(s) = A1 αQS + a1 β 2 αQS e−a1 (s−σ ) dσ (12.76a)
0
! s
Y (s) = A2 αQS + a2 β 2 αQS e−a2 (s−σ ) dσ (12.76b)
0
where the constants A1 and A2 are respectively given by 0.165 and 0.335 and a1
and a2 are respectively given as 0.0445 and 0.3. Following development which
parallels that given in Leishman, and averaging the quasi-steady angle of attack
over each integration step, leads to the following expressions for marching the
solution from the (k − 1)th step to the kth:
Xk = Xk−1 exp [−0.0455β 2 (s)k ] + 0.165(αQSk − αQSk−1 )

1 + exp [−0.0455β 2 (s)k ]
× (12.77a)
2
1 + exp [−0.3β 2 (s)k ]
Yk = Yk−1 exp [−0.3β 2 (s)k ] + 0.335(αQSk − αQSk−1 )
2
(12.77b)
“chapter12(new)” — 2005/11/23 — page 478 — #32

Usage in time-history aeroelastic analyses. The effective angle of attack,

once calculated using the recursive algorithm given earlier, can then be used to
approximate the unsteady aerodynamic section coefficients from the static ones.
Thus, the unsteady airloading can be approximated by combining the (incom-
pressible) noncirculatory virtual mass contributions with what would otherwise
be the circulatory contributions, now represented by the static coefficients, as
determined by the effective angle of attack:
 π 

 
  
* + 
 

cl  2  cθ̇  c (αE ) 
cd ≈ 0 + cd (αE ) (12.78)
cmac unsteady  

 π 1 x a




U 
c (α )

− −  mac E static
airfoil data
2 2 c

virtual mass effects
Results of early investigators. The two-dimensional theory with a variable-

stream velocity, as first formulated by Isaacs et al., approaches more nearly the
helicopter rotor blade in forward flight by allowing for a sinusoidal variation in
the streamwise air velocity. Early investigators have treated this problem in the
context of a multiharmonic response problem wherein the disturbances, as well as
the streamwise variation, are sinusoidal at the same frequency (or integer multi-
ples of it). As depicted in Fig. 12.15, these formulations typically coincide with
the Theodorsen problem with the exception of the inclusion of the streamwise
velocity variation and at most simple harmonic motion of the pitch angle. In these
formulations the sinusoidal variation of the streamwise velocity, as well as the
motion of the airfoil section, is at the rotor circular frequency .
Prior to 1994, all of the existing analytic approaches were either formu-
lated solely in the frequency domain or did not assume a harmonic variation in
streamwise velocity. The most rigorous approach is that of Isaacs, which entails
no simplifications based on the elimination of higher orders of “small numbers”.
The formulation offers a complete Fourier series representation for the unsteady
lift and moment, as presented in the form of ratios with the steady values. It
is, however, limited to the case of rotations about the midchord and does not
Fig. 12.15 Schematic of the Isaacs problem.
“chapter12(new)” — 2005/11/23 — page 479 — #33

include any plunge motion. But within the limitations of these assumptions, it
is a rigorous theory and forms a credible basis for evaluating all of the various
theoretical methods. This theory is sufficiently involved to preclude its detailed
description in this text.
Van der Wall and Leishman performed a comprehensive comparative study of
all of the various methods in the literature for addressing this problem in 1994 and
introduced a new indicial response methodology based on the use of Duhamel’s
integral [Eq. (12.64)]. As regards the evaluations of the preexisting theories, van
der Wall and Leishman concluded that within the range of reduced frequencies
typically encountered by helicopter rotor blades the various theories examined
gave equivalent results. Figures 12.16a–12.16c compare the results obtained with
the newer indicial response methodology with those obtained with the Isaacs theory
for constant, cosine, and sine variations in pitch angle. The results clearly show
the indicial response method to be quite accurate compared with the “touchstone”
Isaacs theory for a) constant pitch angle, b) sine variation, and c) cosine variation.
Indicial response methodology. The methodology for obtaining the indicial

response solution requires an efficient method for solving Eq. (12.53) subject to
the use of the more general definition of aerodynamic time s. For this case the
aerodynamic time variable s is generalized and related to the variability of the
freestream velocity:
!t
2
s= U(t) dt (12.79)
c
0
Both the apparent mass terms and the circulatory terms must be generalized.
Thus, Eqs. (12.51) and (12.52) can be rewritten as
d(Uθ)
L = πρb2 ḧ + − baθ̈ + 2πρbWφ (12.80)
dt
d(Uθ)
Mθ = πρb2 baḧ − b(1/2 − a) − b2 (1/8 + a2 )θ̈ (12.81)
dt
+ 2πρUb(1/2 + a)Wφ
where Wφ is the Duhamel integral, as given by Eq. (12.53), but the derivative
of the normal velocity at the three-quarter point [Eq. (12.45)] with respect to the
aerodynamic time variable σ is now given by a more general form:
dw3c/4 (σ ) dḣ d [U(σ )α] 1 − 2a dα̇(σ )
= + +b (12.82)
dσ dσ dσ 2 dσ
As presented by Leishman, the approximate Wagner function, defined by Eq.
(12.58), and Duhamel’s integral are used to calculate the time variation of the
equivalent angle of attack αe , that is,
! s
dα
αe (s) = α(s0 )φ(s) + (σ )φ(s − σ ) dσ (12.83)
0 dσ
“chapter12(new)” — 2005/11/23 — page 480 — #34

Fig. 12.16 Comparison of results from Isaacs theory with results using the indicial
response method, varying amplitudes of freestream oscillation, and form of pitch angle
oscillation (after van der Wall and Leishman).
“chapter12(new)” — 2005/11/23 — page 481 — #35

where αe can then be used in a quasi-steady sense. Leishman’s method for obtaining
a solution assumes first that the α(s0 ) term can be neglected because it represents
short-lived startup transient conditions. (Although somewhat different in detail,
this formulation leads to an expression for effective angle of attack that is com-
pletely compatible with that defined using the unsteady decay parameter.) The
solution is taken to be the sum of three terms, where here α is taken to be the
quasi-steady value:
αe (s) = α(s) − X(s) − Y (s) (12.84)
where
! s dα
X(s) = A1 (σ )e−b1 (s−σ ) dσ (12.85a)
s0 ds
! s
dα
Y (s) = A2 (σ )e−b2 (s−σ ) dσ (12.85b)
s0 ds
and where, as in a preceding subsection, A1 and A2 and b1 and b2 are constants

defining the approximate Wagner function [Eq. (12.58)]. Leishman gives a series
of comprehensive derivations for a series of algorithms for solving for X(s) and
Y (s) based on different numerical schemes for evaluating the derivative of α
with respect to σ and for effecting the actual numerical integration. Two prin-
cipal time-marching schemes are defined that have contrasting cost vs accuracy
characteristics. Although one scheme is simpler and commensurately lower in
computational cost, its cost characteristics are offset by lower accuracy. The pre-
ferred solution offered by Leishman is claimed to have relative errors of less than
1% from an exact solution if both b1 s and b2 s are less than 0.25:
X(sk ) = Xk = e−b1 s Xk−1 + A1 (αk − αk−1 )e−b1 s/2 (12.86a)

Y (sk ) = Yk = e−b2 s Yk−1 + A2 (αk − αk−1 )e−b2 s/2 (12.86b)
where s is given by
!
2 t+t Uk + Uk−1
s = U(t) dt = t (12.87)
c t c
A major significance of the indicial response methodology is that it is not lim-

ited to any of the earlier assumptions of harmonic variation in either the freestream
variation or angle of attack. Since the earlier work, Jose, Leishman, and Baeder
have extended the methodology to the compressible case and compared the results
with Euler computational-fluid-dynamics (CFD) calculations. Figure 12.17 com-
pares applications of the indicial response method and an Euler CFD method
respectively to an incompressible flow case with a sinusoidal variation in velocity.
Generally, the comparison is quite good.
“chapter12(new)” — 2005/11/23 — page 482 — #36

Fig. 12.17 Results for a constant angle of attack in an oscillating flow, compari-
son between the indicial response method and using an Euler CFD code (after Jose,
Leishman, and Baeder).
12.3.3 Padé Formulations

The developments of the preceding section can be generalized by use of the
so-called Padé approximants, which are essentially relatively simple mathematical
forms. These forms have been found to facilitate spanning the bridge from the
frequency domain to the Laplace operator domain (i.e., time-history solutions). In
contrast to the generally transcendental nature typical of unsteady aerodynamic
formulations, the Padé approximants are defined in terms of rational functions,
which are known to have simple Laplace inversions or inverse Fourier transforms.
Because of this mathematical advantage, it becomes desirable to approximate
either theoretical or experimental unsteady aerodynamic data (originally defined
in the frequency domain) using the Padé approximant. The Padé approximant also
provides a quick method for interpolating and/or extrapolating the frequency-
domain data, which are usually limited to only discrete frequencies.
The Padé approximant of any unsteady frequency-domain aerodynamic coeffi-
cient C is defined as the ratio of two polynomials in the variable (iω):
$
N+1
aj (iω) j
j=1
C(ω) = (12.88)
$
N
bj (iω) j
j=1
where N is the order of the approximant. The degree of the denominator is lower
than that of the numerator by 1 because of the known asymptotic behavior at large
frequencies. Note that the coefficient C(ω) is generally complex.
As extracted from the work of Bielawa et al., Fig. 12.18 presents an example of
the lift coefficient variations with reduced frequency for a two-dimensional airfoil
“chapter12(new)” — 2005/11/23 — page 483 — #37

Fig. 12.18 Example of unsteady aerodynamic airfoil characteristics and associated

Padé approximation.
at a high subsonic Mach number. These data were obtained first from a theoretical
prediction, using the work of Jordan, and then from a suitably calculated Padé
approximant (approximating the Jordan results).
Physical constraints. The Padé coefficients aj and bj in the preceding

equation are determined by imposing the following requirements:
1) The zero-frequency data must be satisfied exactly.
2) The Padé approximants should approach the piston theory results asymptot-
ically for high frequencies.
3) The available data (except for that at zero frequency) are approximated in
the least-squared sense.
4) The resultant poles (roots of the denominator polynomial) must be aperiod-
ically stable (i.e., all lie on the negative real axis).
Basic methodology. The reason that the Padé approximant is so useful is that
the Padé form, when approximating the aerodynamic coefficient’s variation with
respect to iω, can actually provide a reasonably accurate analytical continuation
of the coefficient off the imaginary axis into the positive and negative halves of
the complex plane. Thus, after the Padé form has been successfully found, subject
to the preceding itemized constraints, then iω can be replaced by λ(= σ + iω),
or, alternatively, by the Laplace transform variable p. Most unsteady aerodynamic
data are presented in the form of variations with respect to reduced frequency k,
which then become analytically continued with respect to the Laplace transform
variable based on aerodynamic time p̂. The steps in utilizing the approximant are
basically as follows:
1) Obtain the variation of the (complex-valued) aerodynamic coefficient(s) with
respect to frequency ω (or k). This variation represents the coefficient taken as
“chapter12(new)” — 2005/11/23 — page 484 — #38

a complex-valued function of a complex variable evaluated along the positive

imaginary axis.
2) Fit the Padé approximant form to the data, subject to the constraints given
previously. (Note: This procedure is presently not a well-established one, and
various authors have used a variety of “tricks” to make the approximants well
behaved.)
3) Make the substitution ik, (iω) → p̂, ( p).
4) Using partial fractions, rewrite the Padé form of the aerodynamic coefficients
in the following generic form:
A2 p A3 p
Cx = A1 + + + ··· QS
p + p̃1 p + p̃2
B3 p B4 p
+ B1 + B2 p + + + ··· QS (12.89)
p + p3 p + p4
where for this most general case QS and QS are, respectively, the Laplace
transforms of the quasi-steady inflow and geometric pitch angles comprising
the quasi-steady angle of attack and Cx refers to lift or moment coefficient, as
appropriate.
As with the formulations for the unsteady decay parameter, the Padé formula-
tions can be used either in Laplace variable eigenvalue solutions or in time-history
solutions. The approximants also provide a convenient interpolation formula
for obtaining values of the complex coefficients at frequencies other than those
available from the analytical or experimental source.
12.4 Bending-Torsion Flutter

The blade flutter problem typically occurs at reduced frequencies (based on
conditions at the 75% radius point) that make the neglect of true unsteady aerody-
namic effects unacceptable from an accuracy standpoint. However, the mechanism
of flutter is not necessarily dependent on the use of unsteady airloads. The diver-
gence instability is an aperiodic one and therefore occurs at essentially zero
frequency; hence, static airloads can be applied quite accurately. For these reasons
the material presented in this section is based upon an incompressible (hovering)
quasi-steady formulation. To a great extent the derivations of the appropriate equa-
tions of motion for blade flutter and divergence are extensions of the classical beam
formulation equations originally used for fixed-wing flutter and divergence. The
objective of the material presented herein is to formulate the basics of the prob-
lem so that the important interactional forces can be identified. Where practical,
modifications for including more rigorous unsteady airloads are also presented.
12.4.1 Basic Considerations

In contrast to the physics of the pitch-flap-lag instability problem, that of blade
flutter and divergence is dictated by the interactions of the blade out-of-plane
motion z and the torsional motion θ . Important couplings that exist between these
“chapter12(new)” — 2005/11/23 — page 485 — #39

Fig. 12.19 Typical blade section showing parameters defining flutter and divergence
characteristics.
degrees of freedom require that chordwise offsets of the aerodynamic and mass
centers relative to the elastic (shear) center yAC and yCG , respectively, be taken into
consideration. These features are depicted in Fig. 12.19, which shows a typical
blade section.
12.4.2 Basic Flutter Analysis: Two-Dimensional Wing

As an introduction to the flutter of a flexible rotor blade, a short exposition of
basic fixed-wing flutter principles is in order. Assuming that the wing (of effective
span length ) is describable using the properties at the typical section, as shown
earlier, the resulting equations of motion are as follows:
Vertical motion:
mz̈ + myCG θ̈ + kz z = L (12.90)
Pitching motion:
myCG z̈ + (Iθ + myCG

2
)θ̈ + kθ θ = LyAC + Mθnc (12.91)
where the perturbational lift L and (noncirculatory) moment Mθnc are respectively
given by
* #
c2 c
L= πρ −z̈ + U θ̇ + − yAC θ̈
4 -- - 4- - - - - - - - - -
c c (
+ clα ρU C(k) −ż + Uθ + − yAC θ̇ (12.92)
4 2
  
 c2 c c c 3c 
Mθnc = πρ  z̈ − U θ̇ − − yAC θ̈  (12.93)
 4 4- - 2 4 8 
---------------
“chapter12(new)” — 2005/11/23 — page 486 — #40

The underlined terms are neglected because they are small compared with the
already present inertia terms. Furthermore, the circulatory term in the expression
for lift [∼C(k)] can be treated as a frequency-domain lift-deficiency function (for
frequency-domain solution techniques), or as a Laplace transformed function using
the approximate Wagner function (for general matrix eigenvalue solutions). For
example, the preceding flutter equations can be expressed in the following form
using the Laplace transform variable approach:
m myCG c2 0 −1 kz 0
p2 + πρU p+
myCG (Iθ + myCG
2 ) 4 0 c/2 0 kθ
c 0.5p2 + 0.281(U/b)p + 0.01365(U/b)2
+ clα ρU
2 [ p + 0.3(U/b)][ p + 0.0455(U/b)]

1 −(c/2 − yAC ) 0 −1 z̄
× p+ =0 (12.94)
yAC −yAC (c/2 − yAC ) 0 −yAC θ̄
This equation can then be expanded into a matrix eigenvalue problem of canonical
form.
Stability solution. The flutter solution can be obtained using a variety of

the methods outlined in Chapter 9, for both frequency-domain and Laplace
variable-domain formulations. In addition, a variety of specialized flutter solution
methods has been devised, as described by Rodden and Johnson. Additionally,
see Bisplinghoff et al. for a good discussion of the physical mechanism of flut-
ter. In particular, however, the preceding Laplace variable-domain formulation
will yield an eigenvalue solution having eight eigenvalues, of which four will be
real, reflecting the aerodynamic modes associated with the Padé representation
of the unsteady aerodynamics. In addition to these four real-valued eigenvalues,
Fig. 12.20 Typical variation of flutter eigenvalues with a nondimensional airspeed

parameter.
“chapter12(new)” — 2005/11/23 — page 487 — #41

two complex conjugate pairs will generally result. Figure 12.20 shows the typical
behavior of these two root pairs with increases in the nondimensional parameter
2U/cωθ , where
ωθ2 = kθ /(Iθ + myCG

2
) (12.95)
12.4.3 Flutter of a Flexible Rotor Blade

A modal approach is used wherein the z and θ physical motions are respectively
described in terms of flatwise bending and elastic torsion natural modes γw (r) and
γθ (r), as follows:
"
z= γwi (r) · qwi (t) (12.96a)
i
"
θ= γθj (r) · qθj (t) (12.96b)
j
Presented without derivation are the following two general equations for the
flatwise bending and elastic torsion modal responses. Note that the equation
derivation assumes that the effects of control pitch angle θ0 and steady coning
β0 are negligible. The equations initially show only the elastomechanical terms in
detail.
Flatwise bending:
m γwm (r)q̈wm + m yCG γθk (r)q̈θk + [EIγwm (r)] qwm

! R

− γwm (r)
2
m r1 dr1 qwm
r

1
dL(r)
− m ryCG γθk (r) qθk =
2
(12.97)
dr
2
Elastic torsion:
m yCG γwm (r)q̈wm + m (kθ2 + yCG

2
)γθk (r)q̈θk + 2 m ryCG γwm (r)qwm

3
− [GJ + Tka2 ] γθk (r)qθk + 2

m kz2 γθk (r)qθk

4 5
dL(r) dMθ
= yAC + (r)nc (12.98)
dr dr
Explanation of terms above braces: These terms represent effects that are new to
the rotary-wing case relative to the fixed-wing case.
“chapter12(new)” — 2005/11/23 — page 488 — #42

1) This term represents the flatwise stiffening caused by centrifugal force and
is the source of the already considered Southwell coefficient or rise factor.
2) This term represents a moment caused by the centrifugal force operating in
the presence of a moment arm provided by the torsional deflection. A loading is
produced by the spanwise derivative of this moment:
3) This term represents a torsional moment about the (inclined) elastic axis,
where the inclination is caused by flatwise bending. The moment is thereby a
result of a component of the radially directed centrifugal force in the direction
normal to the elastic axis taken together with the c.g. offset.
4) This additional torsional stiffening, denoted bifilar stiffening, is caused by

nonparallel alignment of the tensile stress filaments as a result of blade torsion:
“chapter12(new)” — 2005/11/23 — page 489 — #43

5) This term represents the propeller moment torsional stiffening caused by

radially aligned centrifugal forces having, as a consequence, components in the
plane of the rotor, which are in opposing directions. Torsional deformations provide
moment arms for these opposing forces to produce a moment:
Unsteady aerodynamic load distribution. The key features of the quasi-

steady airloads description used in this formulation are that the angle of attack α
is defined by the sum of the geometric pitch angle and the inflow angle resulting
from air velocities normal to the blade chord at the 75% chord location. Upon
substituting the expressions for the blade section vertical velocity and the torsional
motion in terms of modal response, the descriptions for the lift and (noncirculatory)
aerodynamic pitching-moment distributions are as follows:
Lift distribution:
dL(r) c2 ρ
= πρ r γθk (r)q̇θk + clα cr C̄(k) − γwm (r)q̇wm
dr 4 2
c
+ r γθk (r)qθk + − yAC γθk (r)q̇θk (12.99)
2
Noncirculatory pitching-moment distribution:
dMθ (r)nc c3
= −πρ r γθk (r)q̇θk (12.100)
dr 8
where C̄(k) = C(k) (Theodorsen function or some appropriate) lift-deficiency

function (evaluated at r = 0.75R).
“chapter12(new)” — 2005/11/23 — page 490 — #44

Modal equations of motion. A standard application of Galerkin’s method is

then used to reduce the flatwise bending and torsion equations to ordinary differen-
tial equations of motion: the flatwise bending equation is multiplied by [γwm (r) dr],
and the torsion equation is multiplied by [γθk (r) dr], and both equations are
integrated over the span of the blade. Because of the orthogonality property of
the natural mode shapes, the resulting equations of motion (in tensor notation) are
then given as follows:
ith Flatwise bending mode:
S10i (q̈wi + ωi2 qwi ) + A1im q̇wm + S31ij q̈θj

− A2ij q̇θj + (−A3ij + S38ij )qθj = 0 (12.101)
jth Torsion mode:
S31ij q̈wi + A4ji q̇wi + S38ij qwi + S78j (q̈θj + ωθ2j qθj ) + A5jm q̇θm − A6jm qθm = 0
(12.102)
where
! R
S10i = m γw2i (r) dr (12.103a)
0
! R
S31ij = m yCG γwi (r)γθj (r) dr (12.103b)
0
! R
S38ij = m ryCG γw i (r)γθj (r) dr (12.103c)
0
! R
S78j = m (kθ2 + yCG
2
)γθ2j (r) dr (12.103d)
0
where ωwi and ωθj are, respectively, the natural frequencies in the ith flatwise
bending and jth torsion modes (at rotor speed). The aerodynamic related integrals
are defined as follows:
! BR
ρ
A1im = clα C̄(k) crγwi (r)γwm (r) dr (12.104a)
2 0
! BR 2 c
ρ c
A2ij = π + clα cC̄(k) − yAC γwi (r)γθj (r) dr (12.104b)
2 0 2 2
! BR
ρ
A3ij = clα C̄(k) cr 2 γwi (r)γθj (r) dr (12.104c)
2 0
! BR
ρ
A4ij = clα C̄(k) cyAC rγwi (r)γθj (r) dr (12.104d)
2 0
“chapter12(new)” — 2005/11/23 — page 491 — #45

! BR c c2
ρ
A5jm = − yAC π + clα cC̄(k) rγθj (r) γθm (r) dr (12.104e)
2 0 2 2
! BR
ρ
A6jm = clα C̄(k) cyAC r 2 γθj (r)γθm (r) dr (12.104f)
2 0
Effective c.g. offset. For the purpose of examining the results, it is convenient
to define an effective c.g. offset (relative to the ith flatwise mode and the jth torsion
mode):
S31ij
( yCG )eff = (12.105)
S10i S78j
12.4.4 Methods of Solutions

The methods of solution available can be summarized as follows:
Selection of a lift-deficiency function. A form of the lift-deficiency function

C̄(k) must be selected. The appropriate form is dependent on the range of reduced
frequencies prevailing in the analysis, which, in turn, is dependent on which modes
are selected as the coupled set. The form also depends on the type of solution
scheme used. Specifically, three forms of lift-deficiency function are available:
1) For low reduced frequencies, the lift-deficiency function approaches unity;
hence, C̄(k) = 1.0 is an appropriate value to use. Furthermore, for divergence
calculations this is the exact value to use.
2) A Padé form of the lift-deficiency function can be used. Such a form is
appropriate for solutions using Laplace transform methods. In this method the
reduced frequency k is replaced by the aerodynamic time derivative pb/U [where
p = d( )/dt] as the dynamic parameter. Thus, a general eigensolution can be formu-
lated wherein the operator p can be replaced by the eigenvalue λ. A major difficulty
with this method is the lack of Padé forms for all of the lift-deficiency functions of
interest to rotary-wing aeroelasticity. However, one major lift-deficiency function
that is available in Padé form is the Theodorsen function [C̄(k) = C(k)]:
, 0.5p2 + 0.281(U/b)p + 0.01365(U/b)2
,
C(k), = (12.106)
k=bp/(iU) [ p + 0.3(U/b)][ p + 0.0455(U/b)]
3) The actual frequency-domain form of the lift-deficiency function C̄(k) can

be used in a frequency-domain solution scheme such as the Nyquist criterion. Use
of this form of the lift-deficiency function has the advantage that a wide variety of
such functions are readily available without recourse to any approximations such
as are inherent in Padé approximants.
Selection of method of solution. Typically, the flutter boundaries are of

most concern. Frequency-domain characterizations of the airloads are most readily
appropriate for this purpose, but for some applications damping levels might be
required and other methods must be used.
“chapter12(new)” — 2005/11/23 — page 492 — #46

Characteristic-equation-based solutions. These require that the equations of

motion, including the airloads, are easily Laplace transformable and that the system
is simple enough that the expanded characteristic equation can be formed and that
it be a reasonably low-order polynomial. When examining for the divergence
boundaries, the lift-deficiency function C(k) is unity, and, when fully expanded,
the characteristic equation is of the form
a0 λ4 + a1 λ3 + a2 λ2 + a3 λ + a4 = 0 (12.107)
then the conditions for divergence are given by a4 = 0.

When examining for flutter, we seek conditions where λ = ±iω, which
define the stability boundary. Use of the Routh or Hurwitz criteria is a stan-
dard method when an expanded characteristic equation be available. For cases
wherein the lift-deficiency function is set to unity, [C(k) = 1], a low-order poly-
nomial typically results. If a complex form of the lift-deficiency function is
used, [C(k) = F(2cω/3R) + iG(2cω/3R)], then the expanded characteristic
equation becomes complex:
(iω) = R (iω) + iI (iω) = 0 + i0 (12.108)
and both real and imaginary parts of the characteristic equation must equal zero
(as is discussed in an earlier section). The two parts of the characteristic equation,
R (iω) and I (iω), each define two real-valued equations in the unknown variable
Z(=ω2 ). If some single parameter (such as the torsion mode frequency ωθ ) is
simultaneously varied in both of these parts of the characteristic equation and
the two equations are solved for Z, then neutral stability points (i.e., the stability
boundary values of ω2 ) will be the intersection points of the two curves.
Conventional-matrix-eigenvalue-based solutions. The use of Laplace trans-

form space eigenvalues (with the attendant use of Padé airload descriptions) will
give estimates of damping levels in addition to indirectly giving flutter boundaries.
Even for relatively simple systems, as typified by Eqs. (12.101) and (12.102), the
matrix statement of the problem becomes too large to use a characteristic equation
approach. For these cases the equations of motion must be rewritten to give a
canonical form of the matrix eigenvalue problem. The use of Padé airload descrip-
tions will necessitate defining additional elements of the state vector, which will
then lead to aerodynamic modes in the final solution.
Multiple-degree-of-freedom Nyquist criterion. This stability analysis tech-

nique, as described in Chapter 9, is ideally suited for analyzing rotors (and wings)
for flutter by virtue of the fact that most of the unsteady aerodynamic tools avail-
able are formulated in the frequency domain. Also, as regards application to wing
flutter, it is not uncommon to have wing structure frequency-response-function
information available from test. The use of this technique is conceptualized in Fig.
12.21.
As the figure shows, the two principal dynamic systems interacting are the blade
elastomechanics and the unsteady airloads. All flutter equations, including those
“chapter12(new)” — 2005/11/23 — page 493 — #47

Fig. 12.21 Conceptualization of the interaction of the dynamic components defining

blade and/or wing flutter.
developed herein for rotor blade flutter, can be written as

[ZB ] {X} = {FA } = 21 ρU 2 [QA (k, m, . . .)] {X} (12.109)
where the [ZB ] matrix would contain all of the flutter equation terms excluding the
unsteady airloads {FA }. The unsteady airloads description is typically proportional
to some definition of dynamic pressure and a function of (reduced) frequency, and
in the case of the Loewy function to harmonic number m. The [QA ] matrix can
be made as rigorous as desired. In the case of rotor blades, although there will
generally be a spanwise variation of freestream velocity, thereby resulting in a
commensurate spanwise variation in reduced frequency, this need not pose any
inherent limitation to a rigorous evaluation of the matrix. With the variation in
spanwise freestream velocity, the dynamic pressure descriptor in Eq. (12.109)
would be based on the tip speed R.
Following the development of Chapter 9, Eq. (12.109) can be written as
{X1 } = [ZB (ω)]−1 {FA } (12.110)

{FA } = 21 ρ(R)2 [QA (k(r, ω), m, . . .)] {X2 } (12.111)
Then, upon relating {X1 } to {X2 }, using the techniques described in Chapter 9 for
this method, the final useful form is obtained:
- .
(ω) {X} = [ZB (ω)]−1 21 ρ(R)2 [QA (k(r, ω), m, . . .)] {X} (12.112)
The analysis for flutter then becomes the determination of the locus of characteristic
multipliers as the frequency ω is varied. If any of the loci cross the real axis at
a value greater than unity, the system is unstable.
Summary of results. The typical results for a simple two-degree-of-freedom

system (one flatwise mode and one torsion mode) are shown in Fig. 12.22. This
figure shows the variations of the two stability boundaries associated, respectively,
“chapter12(new)” — 2005/11/23 — page 494 — #48

Fig. 12.22 Typical blade flutter and divergence characteristics.
with divergence and flutter, with the two major dynamic parameters: (nondimen-
sional) blade torsion mode frequency ωθ / and c.g. offset yCG . Note that for a
given range of aft (–) c.g. offsets, two values of torsion frequency typically exist
for which the blade is neutrally stable in flutter.
12.4.5 Comments and Interpretation

1) The presence of rotational effects makes the issue of divergence more sig-
nificant for the rotor blade than it is for the nonrotating fixed wing. This is because
the mass center can develop moments on a rotor blade even in a nonoscillatory
state and with coincident aerodynamic and shear centers:
“chapter12(new)” — 2005/11/23 — page 495 — #49

2) Use of the simplified lift-deficiency function C(k) = 1.0 will generally pro-
duce conservative results. However, this generality should prove to be unreliable
when the flutter frequency is integral harmonic.
3) Based on the use of the force-phasing matrix technique, it can be shown that
for both the flutter and divergence instabilities similar drivers of the motion are at
work. The torsion degree of freedom is driven by inertia moments associated with
blade vertical acceleration ∼z̈ (flutter) or by centrifugal force and bending ∼2 z
(divergence), each taken together with the c.g. offset. The flatwise bending degree
of freedom is driven in each case by the aerodynamic lift generated by the torsion
deflection θ .
12.5 Three-Dimensional Aerodynamic Theories

A variety of three-dimensional theories has been formulated for rotorcraft
unsteady airloads both in hover and forward flight. The range of applicability of
any one of these theories is generally limited by virtue of the assumptions made in
their respective formulations with regard to four basic issues: 1) frequency content,
2) inclusion of compressibility effects, 3) inclusion of the forward-flight condition,
and 4) the type of wake geometry. Analyses addressing the hovering condition are
typically concerned with the rotor flutter aeroelastic problem and relax the assump-
tion regarding frequency content to the extent that it can be integral or nonintegral
with respect to the rotor rotational frequency. Linearity is generally assumed to
the extent that the airloads for different frequencies can be superimposed. These
theories invariably take the Loewy theory as a starting point.
The emergence of quality analyses that address the inclusion of compressibility
effects is relatively recent and is in part a result of the continuing proliferation
of computer resources with profoundly increased speed and memory capabilities.
With the complementary development of computational-fluid-dynamics (CFD)
technology in general, application of CFD to the rotor problem is becoming an
increasingly viable approach. CFD has the promise of simultaneously relaxing all
of the restrictive assumptions currently being made in rotor unsteady aerodynamic
problems, including those dealing with the nonlinearities of transonic and subsonic
stalled flow.
On the other hand, analyses addressing the forward-flight condition are con-
cerned principally with the harmonic loads aeroelastic problem. Generally, these
theories are formulated for situations wherein the unsteadiness of the airloads is
confined to frequencies that are integer multiples of the rotor rotational frequency.
A major concern of the analyses formulated for the forward-flight condition is the
geometry of the rotor wake, that is, whether it is rigid, distorted, and/or prescribed.
Again, the effects of wake distortion are most important and pronounced as they
then relate to the harmonic loads response problem. Several authors have estab-
lished the premise that accurate aerodynamic rotor loads require some form of wake
distortion for the accurate prediction of rotor blade loads. An unsteady flutter-type
aerodynamic loads formulation valid for forward flight is not yet well-established
and is in need of further methodology development. It can be argued that for flutter
instability calculations the effects of wake distortion are probably not as impor-
tant. The inclusion of a rigorous calculation of rotor wake distortion utilizing a
multiple application of the Biot–Savart theorem is presently cost prohibitive, and
“chapter12(new)” — 2005/11/23 — page 496 — #50

the development of methods for “prescribing” the distortion from a priori cal-
culations is currently an active area of methodology development. Despite the
differences in assumptions made in each of the various three-dimensional theories
regarding these four basic issues and in other procedural assumptions, underlying
similarities and unifying principles exist. These unifying principles are issues that
are addressed herein. Each of the various unsteady rotor aeroelastic theories devel-
oped assuming three dimensionality requires a basic understanding of one or more
of the following considerations.
12.5.1 Basic Mathematical Structure of the Formulation

Regardless of whether the theory in question is two or three dimensional, the
basic mathematical problem is inherently that of a boundary condition problem.
Typically, something is known about the geometry of the problem. This is usually
stated in some form of the nonpermeability of the blade sections with regard to
the instantaneous air particle velocities at the surface of the blade sections as
they move within the air mass. Sometimes the statement regarding the attenuation
of all disturbances at infinity is made. The quantities that are unknown are the
airloads themselves in some form of the pressure loading at the blade surface.
Additionally, some form of an elemental solution that relates the known boundary
conditions (the blade surface velocities) to the unknown quantities (the pressure
loadings on the blades) is identified. This appropriate elemental solution often has
some form of singular behavior so that some care must be exercised to ensure
satisfactory convergence. A large part of the mathematics of the problem is in the
devising of practical ways for defining the points in space for distribution of these
singularities. This problem structuring is exemplified by the conceptualization
given in Fig. 12.23.
Fig. 12.23 Conceptualization of the unsteady rotor aerodynamic problem.
“chapter12(new)” — 2005/11/23 — page 497 — #51

The functions f and F are generally complicated integro-differential operations.

The inversion process is essentially that of finding the solution function(s) F .
12.5.2 Mathematical Tools

Potential flow. A basic assumption typically invoked is that the flow is irro-
tational, inviscid, and (usually, but not necessarily) incompressible. Such a set
of assumptions leads to a condition of potential flow, that is, the mathematical
starting point is of the following form:
∇2 = 0 (Laplace equation) (12.113)
where the velocity field is then defined by
q = ∇ (12.114)
The kinematics of the flow (velocity field) is then related to the pressures using
Kelvin’s equation:
( p∞ − p) D ∂
= = + (q · ∇) (12.115)
ρ Dt ∂t
Elemental solutions. Either of two basic elemental or building-block solu-

tions to Laplace’s equation are assumed: vortex flow and pressure-doublet flow.
Vortex flow is conveniently described by the theorem of Biot and Savart. As
usually interpreted, this theorem relates the velocity “induced” at some test point
in space q by some vortical element of strength . In truth, the equation relates
a basic solution to the Laplace equation in terms of a singularity (vortical flow at
one point) to a consistent velocity field everywhere else. The theorem of Biot and
Savart can be expressed mathematically by either of the following forms:
1) Vector formulation: For a vortex of strength , having an elemental length
of ds, the elemental velocity field is given by

dq = ds × R (12.116)
4π R3
“chapter12(new)” — 2005/11/23 — page 498 — #52

2) Scalar formulation: Any one component of the velocity field can be expres-
sed in terms of the vortical strength of an elemental length of line vortex normal
to the component of velocity:
r sin κ
dq = ds = ds (12.117)
4π R3 4π R2
The theorem of Biot and Savart is commonly interpreted as a relationship

between some entity called a vortex (of strength ) and the velocity of the
fluid induced by it. However, it is more correctly interpreted as a statement
of a consistent flow pattern that satisfies the Laplace equation. Of the several
elemental solutions of Laplace’s equation possible, that of vortex flow is used
because a vortex moving in stationary fluid is mathematically equivalent to an
element of lift.
Various ground rules have been formulated for the use of vortices as elemental
mathematical solutions for building up solutions to real fluid flow problems. Those
relating to lifting surfaces follow from elementary wing theory and are given in
the following without proof:
1) The (airfoil-section-like) lifting body can be modeled alternatively by
elemental vortices distributed over the surface of the body, on the mean cam-
ber line, or concentrated as a line vortex at some chordwise point such as the
quarter-chord. This system of vorticity moves with the translational velocity of the
lifting body U.
2) Vortices (or vortex distributions γ ) must be placed over the wake extending
from the trailing edge of the lifting body to ensure continuity of the vortical system.
This system of vorticity remains imbedded in the flow and does not move with the
translational velocity U.
3) The total blade circulation b is the chordwise integration of the chordwise
vorticity γb . The chordwise (or trailing) and spanwise (or shed) components of
wake vorticity (as defined at the trailing edge) γtw and γsw , respectively, as shown
in Fig. 12.24, are related to the total blade section circulation b by the following
expressions:
“chapter12(new)” — 2005/11/23 — page 499 — #53

Fig. 12.24 Schematic of vortex structure for a lifting airfoil section in an unsteady
environment.
Trailing-wake vorticity:
∂b
[γtw ]T.E. = − (12.118)
∂r
Shed-wake vorticity:
1 ∂b
[γsw ]T.E. = − (12.119)
U ∂t
Blade-section circulation:
! T.E.
b = γb (r, ξ , t) dξ (12.120)
L.E.
4) As in the case of two-dimensional aerodynamic theories, the pressure differ-

ential between the upper and lower blade surfaces is given by the unsteady form
of Bernoulli’s law, wherein the required integration is taken in the streamwise
direction:
! x
p(r, x, t) ∂
= Uγb (r, x, t) + γb (r, ξ , t) dξ (12.121)
ρ ∂t L.E.
Pressure-doublet flow is defined by means of the acceleration potential ,
which is related to the velocity potential , discussed earlier by the following
relationships:
D
= ; hence ∇ 2 = 0 (12.122)
Dt
“chapter12(new)” — 2005/11/23 — page 500 — #54

Thus, the acceleration and velocity potential both satisfy the same differential
equation. The acceleration doublet is the appropriate elemental solution because
this solution element is directly proportional to an incremental pressure or lift. The
mathematical expression for such a doublet solution to the Laplace equation is as
follows:
p ∂ 1 p 1
ψ= = ∇ ·n (12.123)
4πρ ∂n R 4πρ R
where, as shown in the accompanying sketch, the n vector is normal to the surface
with the incremental pressure p:
Even for pressure-doublet-flow approaches to the problem, the velocity field is

required because it is in the definition of the velocities at the airfoil section that the
boundary conditions on the solution must be applied. The difficulty in applying
the pressure-doublet method of solution is that a velocity potential must be
formulated from the acceleration potential (using a suitable integration of the
substantial derivative equation given earlier).
Thus, because q = ∇ [as per Eq. (12.119)], must therefore be an appropriate
integral function of , which itself is built up from elemental solutions:
= I(ψ) (12.124)
where ψ is given by Eq. (12.123). Generally, the pressure (acceleration) potential is

used to formulate a frequency-domain description of the unsteady airloads caused
by the attendant mathematical complexity in effecting the required integration to
obtain the requisite normal velocities. The principal advantage of the pressure
potential method is that it leads to a direct kernel function representation of the
standard inversion process defined earlier.
Normal velocity:
va (r, x, t) = v̄a (r, x)eiωt (12.125)
Pressure distribution:
p(r, x, t) = p(r, x)eiωt (12.126)
“chapter12(new)” — 2005/11/23 — page 501 — #55

Then
!!
v̄a (r, x) = p(η, ξ ) K(η, ξ , r, x, ω) dη dξ (12.127)

blade kernel function
surface
One advantage of a pressure potential formulation is that pressure differentials

can exist only on the blade surface; hence, the preceding integral only involves
integrations over the blade surface.
12.5.3 Rotor Wakes

The work of Loewy teaches that an inescapable facet of rotary-wing unsteady
aerodynamics is the pervasive influence of the rotor wake structure. Hence,
some accounting of the wake must be made in any successful formulation of
rotor aerodynamics. A variety of assumptions of varying complexity can be made
concerning the structure of this wake. The simplest such assumption is that of uni-
form (momentum-induced) inflow, wherein the wake position is determined by the
inflow velocity that is constant with distance traversed downward from the rotor
disk. This assumption is generally referred to as a rigid wake and gives rise to a
regular cylindrical wake wherein the trailing tip vortices describe sheared helices
(see Fig. 12.25).
Simple momentum considerations can be used to provide the required wake
skew angle χ. This angle is determined by using the average components of air
Fig. 12.25 Structure of a “rigid” helical rotor wake.
“chapter12(new)” — 2005/11/23 — page 502 — #56

velocities flowing through the rotor disk. Consider the rotor advance and inflow
ratios:
V
µ= cos αR (12.128)
R
where αR is the rotor angle of attack. Rotor inflow λ is then obtainable from simple
momentum considerations:
CT
λ = −µ tan αR + (12.129)
2B2 λ2 + µ 2
The skew angle of the wake χ is then given by
χ = tan−1 (λ/µ) (12.130)

The completely nonrigid (or distorted) wake assumes that the vortices compris-
ing the wake are free to move in accordance with the induced velocity compatibility
requirements. One result of this assumption is, as in the fixed-wing case, the math-
ematical reproduction of two strong rolled-up vortices trailing from the rotor plane
tips (see Fig. 12.26).
From the results of basic fluid mechanics, this result is actually to be expected
because, at a far distance from the rotor, it is indistinguishable from any other
lifting surface (such as a fixed wing). In this light the rotor is seen to approximate
a lifting line with trailing vortices (see Fig. 12.27).
12.5.4 Miller’s Theory for Hovering Rotors

This theory is remarkable in that, although it is a three-dimensional solution
of closed form, it exhibits some of the same features as those of the Loewy
theory, which is two-dimensional. The Miller theory assumes that the rotor can
be represented by a disk of vorticity distributed both radially and azimuthally.
Fig. 12.26 Structure of a “nonrigid” distorted rotor wake.
“chapter12(new)” — 2005/11/23 — page 503 — #57

Fig. 12.27 Schematic representation of lifting element in the far field.
This is basically an assumption of an infinite number of blades. More specifi-

cally, the Miller theory is formulated using the following explicit assumptions:
1) The vorticity on any blade is azimuthally (integrally) harmonic but constant
along the span and can be represented as
∞
"
= [nc cos nψ + ns sin nψ] (12.131)
n=1
2) The vortex system consists of the bound vortices on the (infinite in number)
blades, the trailing tip vortices, and the spanwise constant shed vortices. This
vortex system forms a disk and an attached semi-infinite cylinder “filled” with
shed vorticity (see Fig. 12.28).
The pertinent details of the derivation are based on applications of the Biot and
Savart equation and the geometry shown in Fig. 12.29.
The development of the theory follows in the classical approach outlined ear-
lier wherein the air velocities induced at the blade sections (the known boundary
conditions) are formulated as functions of the (unknown) strengths of the assumed
system of vorticity.
Vertical air velocity induced by the trailing vortices is
t [ξ − r cos(ψ − φ)] ds
dvt = (12.132)
4π [ξ − 2rξ cos (ψ − φ) + r 2 + h2 ]3/2
2
“chapter12(new)” — 2005/11/23 — page 504 — #58

Fig. 12.28 Vortical structure of the Miller theory.
where the differential arc length ds and vertical spacing of the trailing vorticity h
are respectively given by
2π v0
ds = ξ dφ; h= (12.133)
QR
Vertical air velocity induced by the shed vortices is
s r sin(ψ − φ) ds
dvs = − (12.134)
4π [ξ − 2rξ cos(ψ − φ) + r 2 + h2 ]3/2
2
where in this instance the arc length is a radial element:

ds = dξ (12.135)
Using Fig. 12.24 as a guide, we can then relate the shed vorticity to the bound (and
hence, trailing) vorticity in the following manner:
,
db ,, dt
s = − , =− (12.136)
dψ φ=ψ dφ
The integrated effect of the trailing vorticity is as follows: because it is assumed

that the bound circulation is constant along the blade, the only trailing vorticity
that is to be considered are these which are attached to the blade tips. The basic
assumption of an infinite number of blades thereby results in the trailing vortical
system forming a wake structure consisting of a cylindrical surface. To facilitate
the solution, we first make the following nondimensionalizations:
ξ = R; r = ηR; h = zR (12.137)
“chapter12(new)” — 2005/11/23 — page 505 — #59

506
ROTARY WING STRUCTURAL DYNAMICS
“chapter12(new)” — 2005/11/23 — page 506 — #60

Fig. 12.29 Detailed geometry of rotor and wake vorticity structure of Miller theory.
Then the integrated trailing vortex-induced velocity is given by

! ! ∞
2π 1 [1 − η cos(ψ − φ)](d/dz) dz dφ
vt =
0 0 4π R [1 − 2η cos(ψ − φ) + η2 + z2 ]3/2
!
1 d [1 − η cos(ψ − φ)] dφ
2π
=
4π R 0 dz [1 − 2η cos(ψ − φ) + η2 ]
! 2π " ∞
1 QR
= v0 [nc cos nψ + ns sin nψ]
4π R 2πv0
n=1
× [1 + η cos(ψ − φ) + η2 cos 2(ψ − φ) + · · · ] dφ (12.138)
or, finally,
∞
1 QR " n
vt = [η nc cos nψ + ηn ns sin nψ] (12.139)
4π R 2πv0
n=1
Thus, the effect of the trailing vorticity on the induced velocity is seen to depend on
the span and becomes more concentrated at the blade tips with increasing harmonic
content.
Integrated effect of the shed vorticity. The effects of the shed vorticity must
be taken not only vertically down the wake and azimuthally, but radially as well:
! 1 ! 2π ! ∞
1 d d [η sin(ψ − φ)] dz dφ dη
vs =
4π R 0 0 0 dφ dz [η2 − 2ηη cos(ψ − φ) + η2 + z2 ]3/2
! 1 ! 2π
1 QR d η sin(ψ − φ) dφ dη
= (12.140)
4π R 2πv0 0 0 dφ [η − 2ηη cos(ψ − φ) + η2 ]
2
In this case the integration must be broken up into two parts to avoid the singularity
when η = η :
∞
1 QR "
vs = [nc cos nψ + ns sin nψ]
4π R 2π v0
n=1
! η (n−1) ! 1 #
η ηn
× n dη + n (n+1) dη (12.141)
0 η η η

= (2−ηn )
Therefore,
∞
1 QR "
vs = (2 − ηn )[nc cos nψ + ns sin nψ] (12.142)
4π R 2πv0
n=1
“chapter12(new)” — 2005/11/23 — page 507 — #61

Combination of effects. The combination of the effects of the trailing

vorticity and shed vorticity then yields the following remarkably simple form:
Q
vn = vt + vs = n (12.143)
4π v0
The shorthand notation in this equation implies the correspondence between the
nth harmonic of inflow and the nth harmonic of vorticity “producing” it. Consider
now the nth harmonic of lift on one of the rotor blades. The total unsteady lift
on the blade can be considered to be the sum of the quasi-steady lift LQS and the
wake-induced lift Lw ; hence, any one harmonic of blade lift can be written as
Ln = LnQS − Lnw (12.144)
where
vn
Lnw = 21 ρU 2 × 2b × 2π × (12.145)
U
The results of the preceding derivation, together with the simple expression for the
total nth harmonic of lift Ln (=ρUn ), then yield, after considerable cancellation
of terms,
σπ
Lnw = Ln (12.146)
4|λ|
where σ is the rotor solidity (=2bQ/π R2 ) and λ is the (uniform) inflow ratio. Then
the nth harmonic of (total) lift can be written as
σπ
Ln = LnQS − Ln (12.147)
4|λ|
or, in the final most useful form,
Ln 1
= (12.148)
LnQS 1 + (σ π /4|λ|)
This result is thus seen to have the form of a lift-deficiency function, applicable
to harmonic airloads in hover. Note that this result duplicates a result obtained by
Loewy using an entirely different approach (two-dimensional vs three-dimensional
for the Miller theory). From a practical standpoint this result is highly useful for
determining the loss of cyclic (1P) damping in a rotor in a hover or vertical flight
regime.
12.6 Dynamic Stall and Stall Flutter

12.6.1 Introduction
A singular characteristic of rotorcraft operation is the wide cyclic variations
in airfoil section angles of attack along the span of the rotor blade, especially
“chapter12(new)” — 2005/11/23 — page 508 — #62

as the vehicle operates at increasingly higher advance ratios. The requirement

to maintain roll trim at the high advance ratios creates the situation wherein the
blades operate with relatively high dynamic pressures and correspondingly low
angles of attack on the advancing side of the disk (ψ ≈ π/2) and low dynamic
pressures and correspondingly high angles of attack on the retreating side of the
disk (ψ ≈ 3π/2). Two speed-limiting phenomena occur as a result of this disparity
in aerodynamic operation: 1) the increasingly higher airspeeds on the advancing
side create transonic-flow conditions with attendant critical Mach-number drag
rises; and 2) the increasingly higher angles of attack on the retreating side create
airfoil-stall conditions. Furthermore, each of these conditions is highly dynamic
in nature.
The inclusion of the effects of these two phenomena in practical aeroelastic
analysis of rotor blades has continued to be a particularly demanding challenge.
One reason for the intensity of this challenge is the lack of a well-developed
technology base within the fixed-wing field that could be tapped, as has been
the case with other aeroelastic problems. Typically, the fixed-wing vehicle does
not routinely operate in dynamic conditions of transonic flow and/or stall. The
effects of accelerating Mach numbers on transonic flow appear to be of the form
of modifications to the basic, already present nonlinear effects of transonic flow
itself. However, the dynamics of an airfoil oscillating in stall is found to present
unique characteristics not present with static stall and forms the subject matter of
this section.
Historically, the inclusion of transonic flow and stall effects in airloads and
aeroelastic response prediction analyses was accomplished using static airfoil char-
acteristics in the form of extensive tabulations of lift, drag, and pitching-moment
coefficient data as functions of angle of attack and Mach number. As such, they
could only be used quasi-statically. For hovering and low-speed flight this practice
has justification because the section angles of attack are more or less constant
with time. However, at increasingly higher airspeeds it was found that the airfoil
section loadings, as measured in flight, do not correspond to those measured for
the same blade sections statically in wind-tunnel tests. To gain an understanding
of the dynamic-stall effects, let us first briefly review the static-stall characteristics
of two-dimensional airfoil sections.
12.6.2 Static-Stall Characteristics

With the extensive utilization of airfoil characteristics for rotorcraft analysis,
two-dimensional tests of airfoils now typically include the testing of the airfoils at
high angles of attack, including up to ±180 deg for at least one low subsonic Mach
number. The following material describes the typical behavior of airfoils from the
linear low-angle-of-attack range into deep (static)-stall conditions.
Lift coefficient. As shown in Fig. 12.30, airfoils generally exhibit a high

degree of linearity in lift coefficient with angle of attack up to some value αs ,
wherein a maximum lift coefficient cl (= clmax ) is encountered. Although somewhat
subjectively defined, this angle of attack is typically referred to as the static stall
angle. Above this angle of attack the lift coefficient initially decreases, but then
“chapter12(new)” — 2005/11/23 — page 509 — #63

Fig. 12.30 Typical static two-dimensional airfoil cl (lift) characteristics.
again increases to a second maximum lift coefficient at an angle of attack of

approximately 45 deg.
Drag coefficient. Within the linear angle of attack, as described earlier, the
drag coefficient cd typically assumes a more or less constant value cd0 up until αs
is encountered. For angles of attack beyond αs , the drag increases markedly with
angle of attack up to a value of approximately 90 deg, the point at which the airfoil
is acting essentially as a plate flat to the wind. For further increases in angle of
attack, the drag coefficient then decreases (see Fig. 12.31).
Moment coefficient. The static-moment coefficient, as taken about the aero-

dynamic center (at or near the quarter-chord point), will remain essentially a
constant small value up to the stall angle. Above the static-stall angle the airfoil
will experience a sharp nose-down moment indicating a rearward shift in center
of pressure (see Fig. 12.32).
Fig. 12.31 Typical static two-dimensional airfoil cd (drag) characteristics.
“chapter12(new)” — 2005/11/23 — page 510 — #64

Fig. 12.32 Typical static two-dimensional airfoil cmac (moment) characteristics.
12.6.3 General Dynamic-Stall Characteristics

The most conspicuous characteristic of dynamically stalled airfoils is the
appearance of hysteretic behavior, which depends on whether the angle of attack is
increasing or decreasing through the stall onset range of angle of attack. Essentially,
the airfoil stalls later for increasing angles of attack, and unstalls later for decreas-
ing angles of attack. This hysteretic behavior is most pronounced and consequently
most important with respect to only the lift and moment coefficients.
Lift coefficient. The most striking characteristic of airfoil lift in the presence
of a positive angle-of-attack rate α̇ is that the airfoil stalls both at an angle of attack
and at a corresponding lift coefficient that are respectively higher than those for
the static case (see Fig. 12.33).
Fig. 12.33 Typical dynamic two-dimensional airfoil cl (lift) characteristics.
“chapter12(new)” — 2005/11/23 — page 511 — #65

Fig. 12.34 Typical time-averaged two-dimensional airfoil cl (lift) characteristics.
The “overlift” characteristic is important to rotorcraft operations because it

means that the rotor can potentially pull somewhat greater load factors momentarily
than those normally calculated using static airfoil characteristics. Furthermore,
Halfman et al. have shown experimentally that, for oscillatory angles of attack,
the time-averaged lift will be far greater than the static lift at the deep penetration
angles of attack (see Fig. 12.34).
Moment coefficient. Airfoils exhibit several important dynamic stall char-

acteristics in pitching moment:
1) The airfoil will again exhibit a pronounced hysteretic stalled behavior in
stalling at a higher-than-static (stall) angle of attack for positive angle-of-attack
rates and unstalling at a lower than static (stall) angle of attack for positive angle-
of-attack rates.
2) When the airfoil does stall (dynamically) in moment, the moment coefficient
achieved will be greater in magnitude than the static value. This “overmoment” is
directly related to the overlift also achieved.
3) When this “greater-in-magnitude” (negative) stall moment is occurring,
the lift coefficient can still be producing lift as if it were unstalled (the overlift
condition).
4) The hysteretic behavior of the moment characteristic is such as to trace a
“figure eight” in the α−cmac plane with both “clockwise” and “counterclockwise”
directed segments. The clockwise portion of the hysteretic loop is capable of
producing negative effective damping for cyclic motion.
5) At high angles of attack, in the deep-stall penetration angle-of-attack
range, the hysteretic loops are counterclockwise, and the effective damping again
becomes positive.
These characteristics are illustrated in Fig. 12.35.
Physical explanation of dynamic stall. Through both analytical and

experimental flow-visualization studies, a fairly accurate understanding of the
mechanism of (leading-edge) dynamic stall has emerged. From potential flow
wing theory it is known that the flow around the leading edge of a lift-producing
airfoil is significantly accelerated around the sharp curvature characteristic of lead-
ing edges. This accelerated flow, in fact, accounts for the leading-edge “suction”
“chapter12(new)” — 2005/11/23 — page 512 — #66

Fig. 12.35 Typical dynamic two-dimensional airfoil cmac (moment) characteristics.
loading needed to align the lifting load normal to the freestream direction. When
an airfoil is pitching upward into the high-angle-of-attack stall region, this high
curvature flow region on the leading edge begins to separate and in doing so forms
a vortex near the leading edge. As the angular rate goes to zero and then to negative
values, the vortex detaches from the leading edge and is convected downstream
but near the upper surface of the airfoil. These events are graphically shown in
Fig. 12.36. The suction formerly associated with the (unstalled) leading edge now
becomes initially a mechanism for a concentrated source of increased lift that tra-
verses along the upper surface of the airfoil. The magnitude of the increase depends
on a variety of factors such as the strength of the vortex and its distance from the
upper surface. The streamwise movement of the vortex depends on the airfoil
shape and pitch rate. The relative distance between the vortex and the airfoil varies
according to the kinematics of the airfoil.
As the vortex leaves the leading edge and traverses the chordwise length of the
airfoil, the overlift mechanism caused by the vortex suction changes its moment
arm relative to the mean aerodynamic center point xAC so that the overlift vortex
suction produces an increasingly more negative (nose-down) pitching moment, that
is, dynamic moment stall. The moment coefficient during dynamic moment stall
is significantly greater in magnitude than that normally obtainable during static-
stall conditions. As the vortex leaves the trailing edge, its effect on the airfoil is
subsequently diminished so that the lift abruptly drops off, and dynamic lift stall
occurs. Coincidentally, the moment coefficient typically increases to values that
are more typical of a static-stall condition, and the airfoil then remains stalled until
the angle of attack drops sufficiently so that reattachment of the flow can occur.
The dynamic-stall phenomenon is strongly governed by the rate of change of the
angle of attack, and the higher this rate, the higher the strength of the vortex, and
the higher the degree of overlift and overmoment.
12.6.4 Effective Pitch Damping

The pitching-moment characteristics of the airfoil can be idealized mathemat-
ically by retaining the concept of the (positive damping) noncirculatory moment
“chapter12(new)” — 2005/11/23 — page 513 — #67

Fig. 12.36 Flow visualization of a NACA 0012 airfoil undergoing dynamic stall:
Re = 3 × 10−1 ; α0 = 10 deg, α = 10 deg, and k = 0.049 (courtesy of U.S. Army
Aeroflightdynamics Laboratory).
“chapter12(new)” — 2005/11/23 — page 514 — #68

and introducing a correction caused by dynamic stall:

Mc/4 = Mnon- + Mdue to (12.149)
circulatory dynamic stall

= −(π/8)ρc3 U α̇ = F (α, α̇, α̈, Ü,...)
The combination of these two contributions is sufficiently general to simulate

the hysteretic pattern shown in Fig. 12.35. The two enclosed areas represent mea-
sures of work; a clockwise rotation represents positive work, and counterclockwise
rotation represents negative work. To see how the moment hysteresis creates neg-
ative damping, consider the work done by the air mass upon the airfoil per cycle
of airfoil oscillation:
! ! π
dα
W= Mc/4 dα = Mc/4 dωt (12.150)
cycle −π dωt
Let ωt = ψ. Also, substitute the preceding expression for noncirculatory pitching
moment. Then the expression for work per cycle becomes
! π
π
W = − ρc3 Uωα 2 + F (α, α̇, α̈, . . .)α cos ψ dψ (12.151a)
16 −π
due to noncirculatory due to moment hysteresis
effects (positive damping)
π
= − ρc2 Ukα 2 × [1 − K(k, α0 , α)] (12.151b)
8
= Q(k, α0 , α)
Typically, it is found that K(k, α0 , α), the work caused by moment hysteresis,
is positive. This means that the airfoil is capable of extracting energy (or work)
from the airstream. The extent to which the airfoil can experience positive work
(negative damping) is indicated by the reduction of the quantity Q(k, α0 , α) to
zero and then to negative values. This characteristic is shown qualitatively by
Fig. 12.37.
The actual range of negative damping thus depends principally on reduced
frequency and mean incidence angle. Carta and Ham present, for a NACA 0012
airfoil, the tabulation of an equivalently defined function. Figure 12.38, which
shows a two-dimensional damping surface plot of this function, is reproduced
from their findings.
An alternate and perhaps more useful method for describing the airfoil pitch
damping is to formulate an equivalent damping coefficient. Let the two-
dimensional aerodynamic pitch-damping coefficient be denoted as Cθ , where
1 1 kπ
Mθ = − ρU 2 c2 × Cθ × θ̇ = ρU 2 c2 × ϒ × × θ̇ (12.152)
2 2 2
and where, from potential flow theory, Cθ → kπ/2, as α0 → 0. As gleaned
from the results of Carta and Ham, the pitch-damping effectivity factor ϒ can
be plotted vs k, for varying values of the mean angle of attack, as shown in
Fig. 12.39.
“chapter12(new)” — 2005/11/23 — page 515 — #69

Fig. 12.37 Qualitative (moment) work per cycle characteristics of a dynamically

stalled airfoil.
12.6.5 Synthesized Unsteady Airfoil Data

The general features of dynamic stall discussed in the preceding section were
essentially gleaned from wind-tunnel tests of oscillating two-dimensional air-
foils. The effective damping approach based on energy dissipation per cycle is
a basic approach to solving what are, in essence, truly nonlinear processes (see
Chapter 13). For problems that are describable as simple harmonic motion and
are therefore solvable in the frequency domain, this approach has the advantages
of being simple to implement and reasonably accurate. However, for the more
realistic (forward-flight) operational environment of the rotor-blade airfoil section
Fig. 12.38 Surface representation of the two-dimensional aerodynamic damping of

a NACA 0012 airfoil (after Carta and Ham).
“chapter12(new)” — 2005/11/23 — page 516 — #70

Fig. 12.39 Equivalent damping characteristics for a NACA 0012 airfoil: M = 0.2,
α = 6 deg, and αstall = 13.5 deg (after Carta and Ham).
the assumption of simple sinusoidal motion is suspect. For analyses of the rotor
in the time domain, with the attendant nonlinearities, an analytical representation
of the dynamic-stall process is necessary. To this end, such a modeling should
adhere to the following guidelines:
1) The formulation should be expressible in the form of differential equations
that are solvable in the time domain.
2) The formulation should account for the various aspects of the physics as
outlined earlier. In particular, the modeling of the physics should account for the
delay in the stall with overlift characteristics, moment stall preceding lift stall, and
suitable reattachment to potential flow conditions.
3) Being a “dynamic” phenomenon, the formulation should contain function-
ality with respect to angle of attack α, angle of attack rate α̇, and a time-variable
parameter that gives a decay-like behavior with time. The aforementioned unsteady
decay parameter αW (see Section 12.3.2) satisfies this requirement and can be used
for this purpose.
4) The methodology should be “tuned” to fit existing measured (real-world)
dynamic-stalled airfoil data. This suggests a semi-analytical approach wherein
suitable physical phenomena are mathematically modeled and then calibrated
using experimental data. Such an approach is not uncommon in aerodynamic
theory; the use of a measured static lift-curve slope clα , in place of the theoretical
value, 2π , is just such a case in point. The use of a lift-curve slope implies a linear
functionality, and the use of a measured clα instead of 2π is just such a calibration
of a functionality deduced from both theory and experimentation.
5) The use of an assumed (nonlinear) functionality to model generically the
physics together with measured unsteady stall airfoil data implies a curve-fitting
procedure. Use of some form of nonlinear least-squares fitting procedure can be
used.
“chapter12(new)” — 2005/11/23 — page 517 — #71

6) Although dynamic stall does exhibit unique characteristics over and apart
from those of static stall, it still exhibits many of the qualitative characteristics of
static stall. The static-stall functionality in a sense provides a nonlinear “signature”
for the airfoil, which is already available and which could be used to help define
the dynamic-stall functionality. To this end, the desired dynamic-stall functionality
could incorporate a distortion of the static characteristics.
7) In addition to dynamic-stall functionality aimed at distorting the static-stall
characteristics, other supplementary functionality is required to model the vortex
shedding phenomena as discussed earlier.
Specific modeling procedures. Several investigators have addressed the

problem of modeling dynamic stall with varying degrees of success. A complete
description of these methodologies is beyond the scope of this text, but the reader is
referred to Leishman for a comprehensive review of the subject. A comprehensive
semi-empirical method is presented by Leishman and Beddoes, which utilizes
indicial response functions for both the circulatory and noncirculatory forces.
One example of a modeling procedure that can be described relatively easily is
that of Gangwani. This method takes the tack of mathematically defining suitable
distortions of the static coefficients to achieve the observed dynamic-stall behavior.
The following is a synopsis of the method. It begins by first identifying the basic
dynamic parameters upon which all of the subsequent modeling must be based:
(α/αss ), A ≡ bα̇/U, αW (12.153)
where αss , is the static-stall angle of attack. These parameters are then used to
define supplementary time-domain functionality of the unsteady coefficients:
Lift:
clU (t) = clS (α − α1 − α2 ) + a0l α1 + cl1 + cl2 (12.154)
Pitching moment:
cmU (t) = cmS (α − α2 ) + a0m α2 + cm (12.155)
Drag:
cdU (t) = cdS (α − α2 ) + cd (12.156)
where the additional dynamic-stall-related parameters are described as follows:

1) α1 , cl1 : Define the dynamic distortions of the static curves and
(conventional) potential flow unsteady effects.
2) α2 , cl2 , cm : Define the reattachment and overlift and overmoment
effects.
3) a0l , a0m : Define the static lift and moment curve slopes, respectively, at very
small angles of attack (linear range values).
The specific events modeled are as follows: 1) angle of attack for moment stall
inception αDm , where the airfoil is taken to be stalled in a moment sense when
α exceeds αDm ; 2) (nondimensional) delay time for vortex reaching trailing edge
“chapter12(new)” — 2005/11/23 — page 518 — #72

smt ; and 3) angle of attack for flow reattachment αRE , where the flow is taken to
be reattached when α̇ is negative and α is less than αRE .
The specific dynamic-stall modeling parameters in Eqs. (12.154–12.156) are
all taken to be rational functions of the three basic parameters [i.e., (α/αss ),
A ≡ bα̇/U, and αW ) using a variety of unknown constants that are evaluated (cali-
brated) using the experimental data. The selection of the rational functions is based
on engineering judgement and providing sufficient algebraic terms to enable the
mathematics to find the more important ones. For example, the following linear
and nonlinear terms have been selected with constants that must be evaluated:
α1 = (P1 A + P2 αW + P3 )αss (12.157)
2
α α
cl1 = Q1 A + Q2 αW + Q3 + Q4 (12.158)
αss αss
The constants P1 , P2 , . . . , Q4 are all determined using nonlinear least-squares
curve-fitting techniques with the experimental data. Examples of the ability of this
approach to fit the experimental hysteretic loops with a high degree of accuracy
are shown in Fig. 12.40, as extracted from the work of Bielawa and Gangwani
et al. Because the synthesized method is formulated in the time domain, it can
easily address nonsinusoidal angles of attack. Figure 12.41 shows the results of
applying the methodology to a ramp function in angle of attack. Although exam-
ples of applying the method to nonsinusoidal motion are sparse in the literature,
the results of Fig. 12.41 show the potential of the method to give accurate simula-
tions for conditions that were not actually used to quantify the parameters in the
method. Gangwani presents results of applying this method to four airfoils often
used in rotary wing applications. Although variation was found in the various
empirical parameters for the airfoils synthesized, the parameter orders of magni-
tude and signs were found to be consistent. Leishman presents a comprehensive
survey of the various methods that have been devised for modeling the dynamic
stall phenomenon.
Within any given Mach-number Reynolds-number test condition, the accuracy
of any one of the loops curve-fitted by this process is comparable to the accuracy of
any other loop because all of the loops are curve fitted at the same time. Generally,
the various constants falling out as a result of the nonlinear curve-fitting calculation
are therefore tabulated functions of Mach-number. This method has the advantage
that it is relatively easy to implement in a time-history form of solution for a
nonlinear aeroelastic formulation. The number of Mach-number variable constants
is relatively small compared with the high degree of accuracy obtainable for a
fairly complex aerodynamic phenomenon. The principal disadvantage is that, for
any new (dynamically) untested airfoil, the unsteady coefficients needed for the
analytical functionality are not available. The use of unsteady parameters from a
similar airfoil taken together with the static coefficient data could be used as an
approximation, but one of uncertain accuracy.
12.6.6 Stall Flutter of Rotor Blades

Knowledge of, and the eventual reduction of, high oscillatory load phenomena
continues to be a major goal for successful rotorcraft design. One such oscillatory
“chapter12(new)” — 2005/11/23 — page 519 — #73

Fig. 12.40 Comparisons of synthesized lift and pitching-moment coefficient loops

with test data for a Vertol modified NACA 0012 airfoil: M = 0.6, and Re = 6.2 × 106 .
“chapter12(new)” — 2005/11/23 — page 520 — #74

Fig. 12.41 Correlation of synthesized unsteady aerodynamic data with test data for
a ramp function in angle of attack (after Gangwani).
load phenomenon that afflicts both helicopters and propellers is that of stall flutter,
which is the major result of the unsteady stall phenomenon described earlier. The
phenomenon typically manifests itself in these two elements of rotorcraft by the
substantial resulting increases in blade torsional and pitch link loads. The important
characteristics of stall flutter are as follows:
1) The phenomenon essentially involves a single-degree-of-freedom oscillation
in one of the blade torsion/feathering modes.
2) In propellers the phenomenon principally occurs at the start of the air-
craft takeoff roll at the time when the propeller blades are at maximum angle of
attack because of the lack of inflow angle, which normally results from propeller
translation through the air.
3) In rotorcraft the phenomenon principally occurs in main rotor blades only
for periodically recurring periods of time when the blade is in the retreating blade
(third and fourth) azimuthal quadrants. The requirement of the rotor to maintain
roll trim in high-speed flight necessitates that the blades operate with high angles
of attack in the retreating blade quadrants (see Fig. 12.42).
4) Because the conditions for the phenomenon are typically of short duration,
the oscillations are normally divergent only for short periods of time, and, hence,
the oscillations are not usually destructive in a catastrophic sense. The oscillation
generally assumes the character of a complex multiharmonic limit-cycle oscillation
as depicted in Fig. 12.43.
5) For rotorcraft the occurrence of stall flutter is strongly dependent on the
advance ratio µ and the blade loading, as measured by CT /σ (see Fig. 12.44).
From the results of numerous experimental studies, the source of the self-excited
oscillations has been identified to be the substantial loss of pitching-moment
damping because of the hysteretic nature of dynamic stall. An understanding of why
a stalled rotor blade should lose pitch damping is given in the preceding section.
As can be appreciated by the mathematical strategems required to synthesize the
“chapter12(new)” — 2005/11/23 — page 521 — #75

Fig. 12.42 Typical angle-of-attack distribution over the roter disk for a helicopter in
high-speed flight—140 kn with nonuniform inflow; µ = 0.33.
dynamic-stall phenomenon, stall flutter is governed by highly nonlinear processes.

Consequently, to date stall flutter has been approached in two ways: 1) by defining
(amplitude-dependent) equivalent dampings and stiffnesses and working in the
frequency domain and 2) by numerically solving the nonlinear equation of motion
for blade torsional response in the time domain.
Fig. 12.43 Typical time history of a blade pitch link showing the occurrence of stall
flutter in a helicopter.
“chapter12(new)” — 2005/11/23 — page 522 — #76

Fig. 12.44 Stall flutter limited rotor lift capability in forward flight based on full-scale
wind-tunnel and flight tests.
Based on the observation that stall flutter is a single-degree-of-freedom torsional

response phenomenon, the governing differential equation of motion for the jth
torsion mode can be written as
! 1 π
S78j q̈θj + R γθj (x) ρc3 R(x + µ sin ψ)γθj (x)q̇θj
0 8

+ F (θ , θ̇ , . . .) dx + ωθ2j S78j qθj = 0 (12.159)
where
! 1
S78j = R γθ2j (x)m (ky210 + kz210 ) dx (12.160)
0
Although expressible in this relatively compact equation, the stability solution for
stall flutter is generally complicated by the complexity of the function F (θ , θ̇ , . . .).
Because solution of this equation represents an exercise in nonlinear analysis,
further consideration of stall flutter is deferred until the next chapter.
References
Section
Fung, Y. C., An Introduction to the Theory of Aeroelasticity, Wiley,
New York, 1955.
“chapter12(new)” — 2005/11/23 — page 523 — #77

Kemp, N. H., “On the Lift and Circulation of Airfoils in Some
Unsteady-Flow Problems,” Journal of the Aeronautical Sciences,
Vol. 19, No. 10, 1952, pp. 713, 714.
Loewy, R. G., “A Two-Dimensional Approximation to the Unsteady
Aerodynamics of Rotary Wings,” Journal of the Aeronautical
Sciences, Vol. 24, No. 2, 1957, pp. 81–92.
Luke, Y. L., and Dengler, M. A., “Tables of the Theodorsen Circula-
tion Function for Generalized Motion,” Journal of the Aeronautical
Sears, W. R., “Some Aspects of Non-Stationary Airfoil Theory and Its
Practical Application,” Journal of the Aeronautical Sciences, Vol. 8,
No. 3, 1941, pp. 104–108.
12.3 Beddoes, T. S., “A Synthesis of Unsteady Aerodynamic
Effects Including Stall Hysteresis,” Vertica, Vol. 1, 1976,
pp. 113–123.
Bielawa, R. L., Johnson, S. A., Chi, R. M., and Gangwani, S. T.,
“Aeroelastic Analysis for Propellers,” NASA CR-3729, 1983.
Bisplinghoff, R. L., Ashley, H., and Halfman, R. L., Aeroelasticity,
Edwards, J. V., Ashley, H., and Breakwell, J. B. “Unsteady Aero-
dynamic Modeling for Arbitrary Motions,” Proceedings of the AIAA
Dynamic Specialist Conference, Paper 77-451, 1977.
Greenberg, J. M., “Airfoil in Sinusoidal Motion in a Pulsating Stream,”
NACA TN-1326, 1947.
Isaacs, R., “Airfoil Theory for Flows of Variable Velocity,” Journal of
the Aeronautical Sciences, Vol. 12, No. 1, 1945, pp. 113–117.
Isaacs, R., “Airfoil Theory for Rotary Wing Aircraft,” Journal of the
Aeronautical Sciences, Vol. 13, No. 4, 1946, pp. 218–220.
Jones, R., Flannelly, W. G., Nagy, E. J., and Fabunmi, J. A., “Exper-
imental Verification of Force Determination and Ground Flying on
a Full-Scale Helicopter,” U. S. Army Applied Technology Lab.,
USAAVRADCOM-TR-81-D-11, Fort Eustis, VA, May 1981.
Jones, R. T., “Operational Treatment of the Non-Uniform Lift Theory
in Airplane Dynamics,” NACA Technical Note 667, 1938.
Jordan, P. F., “Aerodynamic Flutter Coefficients for Subsonic, Sonic
and Supersonic Flow (Linear Two-Dimensional Theory),” Royal
Aircraft Establishment Reports and Memoranda No. 2932, London,
April 1953.
Jose, A. I., Leishman, J. G., and Baeder, J. D.,“On Calculating
Unsteady Airloads with Time-Varying Free-Stream Mach Num-
bers,” Proceedings of the 61st Annual National Forum of the
American Helicopter Society, 2005.
“chapter12(new)” — 2005/11/23 — page 524 — #78

Leishman, J. G., Principles of Helicopter Aerodynamics, rev. ed.,
Cambridge Univ. Press, New York, 2005.
McGuire, D. P., “The Application of Elastomeric Lead-Lag Dampers
to Helicopter Rotors,” Lord Library No. LL2133, 1976.
McGuire, E. P., “Fluidlastic Dampers and Isolators for Vibration
Control in Helicopters,” Lord Library No. LL-6502, 2001.
van der Wall, B., and Leishman, J. G., ”The Influence of Variable Flow
Velocity on Unsteady Airfoil Behavior,” Journal of the American
12.4 Ham, H. D., “Helicopter Blade Flutter,” AGARD Rept. 607 (revi-
sion of Pt. III, Chap. 10 of the AGARD Manual of Aeroelasticity),
1973.
Miller, R. H., and Ellis, C. W., “Blade Vibration and Flutter,” Jour-
nal of the American Helicopter Society, Vol. 1, No. 3, 1956,
pp. 19–38.
Rodden, W. P., and Johnson, E. H., “MSC/NASTRAN—Aeroelastic
Analysis,” Users Guide, Ver. 68, MacNeal-Schwendler Corp.,
Los Angeles, CA, 1994.
12.6 Beddoes, T. S., “Onset of Leading Edge Separation Effects Under
Dynamic Conditions and Low Mach Number,” Proceedings of the
34th Annual National Forum of the American Helicopter Society,
May 1978.
Carlson, R. G., Blackwell, R. H., Commerford, G. L., and Mirick, P. H.,
“Dynamic Stall Modeling and Correlation with Experimental Data
on Airfoils and Rotors,” Proceedings of Specialists’ Meeting on
Rotorcraft Dynamics, NASA Ames Research Center, 1974.
Carta, F. O., and Ham, N. D., “An Analysis of the Stall Flutter Insta-
bility of Helicopter Rotor Blades,” Proceedings of the 23rd Annual
National Forum of the American Helicopter Society, May 1967.
Dowell, E. H., Clark, R. C., Cox, D., Curtiss, H. C. Jr., Scanlan, R. H.,
Edwards, J. W., Hall, K. C., Peters, D. A., Scanlan, R., Simiu, E.,
Sisto, F., and Strganac, T. W., A Modern Course in Aeroelasticity,
Fourth Edition, Kluwer Academic Publishers, Norwell, MA, 2004.
Gangwani, S. T., “Prediction of Dynamic Stall and Unsteady Air-
loads for Rotor Blades,” Journal of the American Helicopter Society,
Vol. 27, No. 4, 1982, pp. 57–64.
Gangwani, S. T., “Synthesized Airfoil Data Method for Predicition of
Dynamic Stall and Unsteady Airloads,” NASA CR-3672, 1983.
Halfman, R. L., Johnson, H. C., and Haley, S. M., “Evaluation of High
Angle-of-Attack Aerodynamic—Derivative Data and Stall-Flutter
Prediction Techniques,” NACA TN-2533, 1951.
“chapter12(new)” — 2005/11/23 — page 525 — #79

Problems
12.1 The pitching moment that a rotor can generate is directly limited by atten-
uations caused by unsteady aerodynamic considerations. Expressions for
the rotor pitching moment, including unsteady aerodynamic effects, can
be obtained using both the Loewy/Miller lift-deficiency function (for har-
monic airloads) and the air mass dynamics theory in the preceding chapter
(Section 11.5).
(a) Derive expressions for the ratio of rotor pitch damping (with unsteady
aerodynamic effects) to that using only the quasi-steady assumption
using i) the Loewy/Miller lift-deficiency function and ii) air mass
dynamics, respectively.
(b) Explain any differences obtained using each of the two unsteady
aerodynamic formulations considered in part a.
12.2 Consider the tail rotor of a helicopter in yawing (hovering) flight at sea level
(see Problem 4.1). In this maneuver the (two-bladed) tail rotor develops
once per rev (1P) flapping responses as a result of the gyroscopic excita-
tions. A properly designed tail rotor must account for these responses with
regard to determining an appropriate setting of the flapping limit stops. The
1P excitations created by the gyroscopics are equilibrated by the aerody-
namic loads, which are necessarily unsteady. The rotor and flight condition
specifications are as follows:
radius R = 3.25 ft; thrust T = 275 lb; tip speed R = 712 ft/s
chord c = 0.65 ft; tip loss B = 0.95; ρ = 0.002378 lb-s2 /ft4
tail rotor position aft of aircraft c.g. TR = 20.5 ft
(a) Using the Loewy/Miller lift-deficiency function for harmonic airloads,

estimate the percent loss in first harmonic aerodynamic effectivity for
respective helicopter yaw rates z of ±100 deg/s.
(b) Using the results of part a and the blade properties given in Problem 4.1,
calculate the maximum teetering angles expected in this yaw maneuver.
Note that the rotor inflow for a rotor ascending at ascent speed Vz is
given by the momentum equation:
λ + Vz
λ = CT / 2B2 (12.161)
R
12.3 Consider a rotor in high-speed flight with an edge-on orientation (helicopter

mode) with an advance ratio µ of 0.45. The rotor blades, which are assumed
to be effectively rigid, have a radius of 20 ft and a chord of 2.4 ft. In this flight
condition a collective angle θ0 produces a rolling moment. Using simple
approximations, estimate the percent loss in the quasi-steady calculated
value of this (rolling moment/collective angle) effectivity which arises from
unsteady (variable stream velocity) effects. [Hint: The rolling moment arises
“chapter12(new)” — 2005/11/23 — page 526 — #80

principally from the airloads on the advancing and retreating sides, where
at the 0.75R spanwise location the tangential velocity UT is respectively
proportional to R(0.75 ± µ).]
12.4 The design of a high-speed (articulated) helicopter rotor blade with a radius
of 20 ft is to incorporate a swept tip of 35 deg over the outer 10% of the
radius. It is to have an aspect ratio (R/c) of 12 and use 11% thick airfoil
sections. The blade is to have a uniform spanwise mass distribution, be
chordwise balanced on the line of section quarter-chords (including the aft-
swept quarter-chord in the tip region of the blade), and weigh 100 lb. A flutter
analysis of the blade shows that, because of the swept tip, the blade will have
bending-torsion flutter susceptibility involving the first bending mode with
the first torsion mode. However, the torsion mode does not couple unstably
with the second bending mode. Devise a design modification incorporating
an incremental mass distribution m and appropriate chordwise location
y10CR , which will decouple the torsion mode from the first flatwise mode,
but leave the favorable (stable) coupling with the second flatwise mode
intact. [Note that the devised incremental mass distribution must be physi-
cally realistic and compatible with the established blade geometry (i.e., the
additional mass must “fit” within the selected airfoil section).] The first and
second bending and torsion mode shapes are, respectively, given by
γw1 (x) = −3.2x + 14.0x 3 − 14.0x 4 + 4.2x 5

γw2 (x) = 6.8225x − 114.8611x 3 + 258.0401x 4
− 206.2732x 5 + 57.2716x 6
γθ1 (x) = sin[(π/2)x]
“chapter12(new)” — 2005/11/23 — page 527 — #81

13
Analysis of Nonlinear Systems
13.1 Introduction
The aeroelasticity of rotating wings is an especially abundant source of non-
linear dynamics problems. In this branch of aeroelasticity, significant nonlinearities
can be found in each of the three basic types of forces traditionally defining the
subject matter (i.e., inertia, elastic, and aerodynamic). Nonlinearities arise in the
inertia loadings of rotor blades because of their freedom to flex with signifi-
cant amplitudes in a rotating coordinate frame in at least four modes of elastic
deformation: transverse (beam) bending in two directions, axial torsion, and axial
extension. The internal elastic forces in rotor blades are also rich in nonlinearities.
The relatively high aspect ratios of rotor blades taken together with the inherent
twist (both built-in and elastic) and the aforementioned flexing amplitudes define
a variety of elastic nonlinearities.
However, perhaps of all of the nonlinearities present in rotor aeroelastic pheno-
mena, those arising from aerodynamic sources are the least tractable. Principal
sources of aerodynamic nonlinearities are the dynamic stall effects treated in
the preceding chapter and those of transonic flow occurring on the advancing
blades in high-forward-flight conditions. Indeed, even the unsteady aerodynamic
phenomena not associated with dynamic stall and advancing blade compressibility
effects are subject to nonlinearities because blade motion in the air mass is not
infinitesimal.
Beyond these basic sources of nonlinearities, other sources exist as well.
An important one is that afforded by the (articulated) blade lead-lag dampers.
Although these dampers do provide a measure of “linear” damping, they are
still essentially hydraulic devices and are characterized by velocity squared and
saturation-type nonlinearities. The control systems of rotors are additional sources
of nonlinearities because of the kinematics of the swash plate and pitch control
push rods, as well as inherent nonlinearities (both intentional and necessary) in
the power boost and stability augmentation systems.
The study of nonlinear dynamics, generally, and nonlinear vibrations, specif-
ically, has evolved into an extensive field of technology. Much is now known
of nonlinear dynamics, and a variety of analysis techniques have been formu-
lated. Nonlinear dynamics, generally exhibit characteristics that are quite unique
to the point that they can be considered to be “hallmarks,” that is, they have
become virtually synonymous with nonlinear phenomena. The nonlinear dynamic
characteristics of rotors are quite consistent with these more general hallmark
529
“chapter13(new)” — 2005/10/20 — page 529 — #1

characteristics. Indeed, the rotary-wing nonlinear phenomena that presently have

been either identified to be potentially real or actually observed to be real typi-
cally exhibit the most well-defined of these general characteristics: limit cycles,
subharmonics, higher harmonic distortion, and response “jump” phenomena.
For the most part the nonlinear differential equations of motion appropriate to
rotor blades have the following general characteristics:
1) They are multivariate and are conveniently expressed in matrix form.
2) They are nonhomogeneous, with the nonlinear portion of the excitation being
multiharmonic in the rotor rotation frequency .
3) They contain a principally linear form that is replete with periodic coeffi-
cients (i.e., a strongly imbedded Floquet problem).
Mathematically, the rotor aeroelastic problem can be expressed by the following
general matrix form:

[M] Ẍ + [C(ψ)] Ẋ + [K(ψ)] X = F(ψ) + FNL (ψ, X, Ẋ) (13.1)
where the [C] and [K] matrices and the {F} vector are periodic with period 2π :
[C(ψ + 2π )] = [C(ψ)] (13.2a)

[K(ψ + 2π )] = [K(ψ)] (13.2b)
{F(ψ + 2π )} = {F(ψ)} (13.2c)
The vector {FNL (ψ, X, Ẋ)} represents the system nonlinearities that might or might
not be expressed in simple analytic form.
Because the nonlinearities existing in rotors are not all explicit or analytically
defined, a variety of methods must be used for their analysis. Thus, in the fol-
lowing material the distinction between explicit and implicit nonlinearities will
be stressed. Three major methods for analyzing the nonlinear dynamics typical
of rotary-wing systems are presented in this chapter: 1) simple linearizations,
2) direct numerical solutions, and 3) quasi-linearization. Furthermore, in support
of the direct numerical solution methodology, a section is devoted to the numerical
extraction of stability descriptors from the resulting response time histories.
13.2 Simple Linearization

13.2.1 Equivalent Damping and Stiffness
Perhaps the simplest form of linearization possible is the approximation of the
nonlinearity fNL in the form of an equivalent damping and/or stiffness. The basis
for the equivalent damping approximation CE is the equivalencing of the energy
dissipated per cycle of oscillation by the simple equivalent (linear) viscous damper
WD to that dissipated by the nonlinear element WNL :
2π/ω
WD = CE X̄ ωπ =
2
(−)fNL (X, Ẋ) · Ẋ dt = WNL (13.3)
0
“chapter13(new)” — 2005/10/20 — page 530 — #2

ANALYSIS OF NONLINEAR SYSTEMS 531
where the motion X(t) is taken to be simple harmonic with a frequency of ω and
an amplitude of X̄. The required expression for the equivalent damping can then
be easily written as
WNL
CE = (13.4)
X̄ 2 ωπ
The simplified analysis of stall flutter offered in the preceding chapter constitutes
a solution of this type.
In contrast to the equivalent damping representation given earlier, the equivalent
stiffness approximation to the nonlinearity cannot be predicated on the basis of
equivalence of energy dissipation. Thus, nonlinearities of this type should not have
dependencies on velocity. The techniques used to obtain an equivalent stiffness
approximation can best be formulated by moving on to the next topic.
13.2.2 Describing Functions

The approximations of an equivalent damping and/or stiffness can be general-
ized by the concept of a describing function, when used in the frequency domain.
If the motion of the system is principally sinusoidal, then the element comprising
the nonlinearity will generally produce a sinusoidal output (force) in response to a
sinusoidal input (motion). The higher harmonics of the output of the element can
be neglected for either of the following reasons:
1) The transmission of the higher harmonics by the linear portion of the system
is significantly less than the transmission of the fundamental. This type of system
behavior resembles that of a low-pass filter.
2) The force generated by the nonlinearity is small, and the higher harmonics
present are negligible.
If either or both of these characteristics are present, then a describing func-
tion approach can be taken. This technique consists of casting the nonlinearity in
the alternate form of an equivalent linearized (but amplitude-dependent) transfer
function GN :
complex amplitude of output ( force)
GN ≡ (13.5)
complex amplitude of input sinusoidal (motion)
which we denote the describing function. This quantity, which can be real or
complex, is generally evaluated by expanding the output force caused by a simple
harmonic input motion in the form of a Fourier series based on the period of the
assumed frequency of motion ω:
∞

fNL (X, Ẋ) = Re An e inωt
(13.6)
n=1
for
X = X0 + X̄eiωt (13.7)
“chapter13(new)” — 2005/10/20 — page 531 — #3

where
2π/ω
1
An = fNL (X0 + X̄ eiωt , iωX̄ eiωt )einωt dt (13.8)
π 0
The describing function is then formed by retaining only the first-harmonic

coefficient:
GN = A1 /X̄ (13.9)
The preceding formulations then allow one to extract from the analytic (or even
tabulated) representation of the nonlinearity an amplitude-dependent value for
an equivalent stiffness. For an equivalent stiffness KE the appropriate source of
stiffening would come from the real part of the describing function:
KE = −(A1 )R /X̄ (13.10)
where
A1 = (A1 )R + i(A1 )I (13.11)
It can be verified that the imaginary part of the describing function is equivalent
to the calculation of energy dissipated per cycle; hence, this component of the
describing function also yields the value of the equivalent damping:
CE = −(A1 )I /X̄ω (13.12)
The great advantage of the describing function approach is that describing func-
tions can be found for nonlinearities that cannot be stated analytically (as is the
case with various nonlinear aerodynamic phenomena such as dynamic stall). The
disadvantage of the technique is that it is only valid in the frequency domain and
it discounts the effects of the higher harmonics on the fundamental.
13.2.3 Example: Stall Flutter

The nonlinear phenomenon of stall flutter can be addressed in two ways: 1) a
direct numerical assault on the nonlinear equations of motion using a time-domain
formulation of the unsteady stalled airloads (e.g., the synthesized unsteady stalled
airfoil data presented in the preceding chapter) and 2) the use of suitable lineariza-
tions. Using measured data presented by Carta and Ham, for a NACA 0012 airfoil
in oscillatory pitching motion, for a range of angles of attack from zero to deep
stall, values of equivalent damping and stiffness can be calculated. The results of
these calculations are presented in part in Fig. 12.39 of the preceding chapter.
Note that Fig. 12.39 was constructed by taking the tabulated values of faired
two-dimensional aerodynamic damping data presented in the work of Carta and
Ham as a basis. For each reduced frequency the tabulated values are first divided
by the respective values for zero angle of attack. The equivalent damping (which is
actually used in the linearized problem statement) is then formed from the values
of damping at zero angle of attack (as defined by virtual mass considerations)
multiplied by the values given in this figure. Because the experimental setup did
“chapter13(new)” — 2005/10/20 — page 532 — #4

not allow for extensive variations in oscillatory amplitude (only one amplitude,
6 deg, was used), these results are prima facie independent of amplitude. Thus, the
results inherently assume that the effective damping moment varies linearly with
pitch amplitude, as does the linear (virtual mass) damping moment. The validity
of this approximation is moot, but, in the absence of more substantial unsteady
airfoil data, it does not appear unreasonable, especially because helicopter stall
flutter limit-cycle instabilities have been observed to be of this order of magnitude.
The variation of the measured zero-angle-of-attack damping values with
reduced frequency ( for this airfoil and for a pitch amplitude of 6 deg), as normal-
ized by the theoretical Theodorsen value of π k/2, is also presented in Fig. 12.35.
This variation is presented as a reference and is indicated by the dashed line in
the figure. Note that the measured variation is significantly less than the theo-
retical values, but improves with reduced frequency. Because the test article that
generated these data was an airfoil section having an aspect ratio of only 1.35, it
is reasonable to expect that inherent three-dimensional end effects could reduce
the two-dimensional pitch damping that might otherwise be available. Also, the
recovery trend of the experimental two-dimensional damping values to the theoret-
ical ones with reduced frequency is consistent with findings of various researchers
to the effect that increasing unsteadiness (k → ∞) tends to reduce the relative
importance of spanwise cross talk and thereby makes the airfoil sections behave
more two-dimensionally.
Because the full Fourier decomposition of the unsteady moment coefficient is
not available, the static-moment coefficient characteristics (Fig. 13.1) must be used.
Figure 13.2 is constructed by using the preceding expression for KE , but as applied
to Fig. 13.1. This task represents an approximation of unknown accuracy, as taking
the imaginary component of the first harmonic of the Fourier decomposition would
truly represent that component of the unsteady moment in phase with displacement.
In the absence of such data and the fact that the equivalent stiffness calculation is
independent of hysteresis, the use of the static characteristics for the calculation
of the equivalent stiffness offers a reasonable approximation.
Fig. 13.1 Static pitching coefficient characteristics for a NACA 0012 airfoil: M = 0.2.
“chapter13(new)” — 2005/10/20 — page 533 — #5

Fig. 13.2 Equivalent stiffness characteristics for a NACA 0012 airfoil: M = 0.2, and
αstall = 13.5 deg.
The determination of the stall flutter characteristics using Figs. 13.2 and 12.39,
then, is as follows:
1) Determine the linear elastomechanical dynamic characteristics in terms of
generalized inertia Iθ and blade torsion natural frequency ωθ .
2) Determine the linear pitch damping (as given by the virtual mass effects) for
a selected rotor azimuth angle ψ̂ (typically taken to be 3π/2):
1
πρR4 c 3 2
Cθ = R γθ (x)[x + µ sin ψ̂] dx (13.13)
8 µ R
3) Select a range of mean angles of attack α0 , as determined for conditions at

a representative (nondimensional) spanwise location x̃ and corresponding values
of azimuth (bracketing ψ̂). Each mean angle of attack is approximated by the sum
of components deriving from the blade built-in and elastic twist angles θB and θE ,
respectively, the rotor collective and lateral cyclic control angles θ0.75R and B1s ,
and the (quasi-steady) inflow angle φ0 , all as measured at the representative blade
spanwise location x̃(= r̃/R):
α0 = θB (x̃) + θE (x̃) + θ0.75R − B1s sin(ψ̂)

+ tan−1 [UP (x̃, ψ̂)/UT (x̃, ψ̂)] (13.14)
= φ0
and where x̃ represents a mean aerodynamic radius, which can be approximated as

1 2
µ x γθ (x) dx
x̃ = 1 2
(13.15)
µ γθ (x) dx
“chapter13(new)” — 2005/10/20 — page 534 — #6

4) For each mean angle of attack within the selected range, iterate on amplitude
of motion ᾱ and reduced frequency k to determine a combination that will give
a value of equivalent damping (= ϒ · Cθ , where ϒ is obtained from Fig. 12.35),
which is either zero or negative for each of the selected mean angles of attack
α0 . Specifically, for each iteration value of amplitude ᾱ calculate the effective
frequency of motion:

ωθ2 + Kθ (α0 , ᾱ)aero
νθ = (13.16)
Iθ
where
1 c 2
ρR3
(Kθ )aero = (R)2 Kcm (α0 , ᾱ) γθ (x) (x + µ sin ψ̂) dx (13.17)
2 µ R
and where Kcm (α0 , ᾱ) is obtained from Fig. 13.2. The effective reduced frequency
keff is then given by the following expression:
c
keff = νθ /[R(x̃ + µ sin ψ̂)] (13.18)
R
which, together with each of the selected values of α0 , define the values of the
parameter ϒ, as per Fig. 12.35. Stall flutter instability is indicated by a zero or
negative value for ϒ.
Note that the limits of integration for the various integrals given earlier are
from µ to 1. Because the stall flutter instability is confined to the reversed-flow
quadrants of the rotor disk, these limits define the blade spanwise extent that is still
operating in forward-flow conditions. The development already given constitutes,
at best, only a rough cut at a rotor blade stall flutter calculation. For a more rigorous
calculation of stall flutter, as well as providing solutions to a variety of complicated
nonlinear instability phenomena, recourse must be made to numerical transient
time-history solutions of the full nonlinear equations of motion. Techniques for
the numerical integration of these equations, which have been found to be well
suited to rotary-wing aeroelastic applications, are presented in the next section.
13.3 Transient Solutions Using Numerical Integration

The most general dynamic description of rotor blade aeroelasticity, as indicated
earlier, involves a variety of sources of nonlinearities. Some of these, although
yielding to reasonably accurate modeling empirically, can nevertheless still not
be easily expressed in simple analytic form. For this reason recourse must be
made to accurate solution techniques using numerical integration of the complete
nonlinear differential equations of motion. Additionally, it sometimes happens
that some new rotor aeroelastic phenomenon (i.e., instability) might be experi-
enced in the field. To gain insight to the physics of the phenomenon quickly, a
full simulation of the complete nonlinear dynamics is warranted, especially if
such a simulation capability is operational and readily available. Once the physics
is understood using the complete simulation, suitable simplifications and/or lin-
earizations can then be formulated to give an analytical capability for making
“chapter13(new)” — 2005/10/20 — page 535 — #7

rapid stability boundary estimations. In a sense this scenario defines two phases:
physical modeling and mathematical modeling. Both of these phases are important
for a thorough understanding of the phenomenon at hand. The thrust of the present
section is in describing some of the basic techniques found useful in obtaining
transient solutions of the complete nonlinear equations of motion, given earlier as
[M]{Ẍ} + [C(ψ)]{Ẋ} + [K(ψ)]{X} = {F(ψ)} + {FNL (ψ, X, Ẋ)} (13.19)
Most practical solution methods currently operational use some form of numer-
ical integration in time (or alternatively, nondimensional time, ψ = t) with a
uniform step size dt (or dψ). In the following material time will be assumed to be
nondimensionalized as per this definition.
13.3.1 Basic Schemes for Numerical Integration of Equations

The basic issues or attributes on which the subject matter has developed are
those of accuracy, practicality (or, alternatively, cost), and stability. It is on the
basis of one or more of these attributes that the methods described herein have been
advocated and made operational. These attributes therefore conveniently serve as
“yardsticks” for comparing the various methods. All numerical integration schemes
consist of four basic operations, which are either explicitly or implicitly stated:
1) Evaluation of highest differentiated terms: All numerical integration meth-
ods are of two basic types:
a) The equations of motion are used to evaluate the highest differentiated
terms, and numerical quadrature techniques are then used to integrate responses
to calculate the lower-order differentiated terms; or
b) The equations are used as part of the selected quadrature technique itself
to evaluate the lowest-order differentiated terms, and then appropriate numerical
(differencing) techniques are used to differentiate the responses to obtain all higher-
order differentiated terms.
Either of these two method types can be made to yield accurate solutions, and,
correspondingly, each has advantages and disadvantages. Both basic types must
start with the same differential equation, which is used for numerical evaluation
at each discrete point in time (non dimensional) ψk .
For those methods of the first type, the actual nonlinear differential equations
are used to evaluate the highest differentiated quantities {Ẍ}k , in terms of the
quantities of lesser order. For the class of differential equations considered herein
(second order), this can be expressed in terms of the acceleration vector, which
is generally coupled by an inertia matrix. Thus, our original differential equation
can be rewritten to the following form:
[M]{Ẍk } = −[C(ψk )]{Ẋk } − [K(ψk )]{Xk }

+ {F(ψk )} + {FNL (ψk , X k , X˙k )} = { (ψ, X, Ẋ)k } (13.20)
For methods of the second type, the equations of motion are used as part of the
actual numerical integration scheme, and the resulting matrix equation set is highly
method dependent. An example of this equation transformation is presented in the
following material.
“chapter13(new)” — 2005/10/20 — page 536 — #8

2) Numerical integration scheme: The state conditions (the vectors of state

variables and/or their time derivatives) at the next point in time { }k+1 are related
to higher differentiated conditions at the present time { }k and at one (or optionally
more) time step(s) in the past { }k−m . This is typically achieved using any of a
wide variety of differencing schemes, with or without specific assumptions on
specialized characteristics of the responses. Four diverse practical schemes that
have been successfully implemented are given in the following as examples.
The first example is the Adams method class of methods, which use variations
of the Newton backward-difference formulas:

xk+1 ≈ xk + ψ 1 + 21 ∇ + 12 ∇ + 38 ∇ 3 + · · · ẋk
5 2
(13.21a)

xk+1 ≈ xk + ψ 1 − 21 ∇ − 12 ∇ − 24
1 2 1 3
∇ − · · · ẋk+1 (13.21b)
These general formulas give rise to a “family” of methods all identifiable as Adams
methods:
a) First-order backward-differencing:

ẋk+1 = ẋk + ψ 23 ẍk − 21 ẍk−1 (13.22a)
xk+1 = xk + 1
2 ψ(ẋk+1 + ẋk ) (13.22b)
b) Second-order backward-differencing:

ẋk+1 = ẋk + ψ 12 23
ẍk − 43 ẍk−1 + 5
12 ẍk−2 (13.23a)

xk+1 = xk + ψ 12 5
ẋk+1 + 23 ẋk − 1
12 ẋk−1 (13.23b)
c) Third-order backward-differencing:

ẋk+1 = ẋk + ψ 2455
ẍk − 24
59
ẍk−1 + 37
24 ẍk−2 − 38 ẍk−3 (13.24a)

xk+1 = xk + ψ 24 9
ẋk+1 + 24
19
ẋk − 5
24 ẋk−1 + 1
24 ẋk−2 (13.24b)
The Adams method is a technique of the first basic type wherein the highest-
order differentiated terms are calculated from the equations of motion, and
lower-order terms are subsequently integrated using one of the integration schemes
already given. Note that the use of increasing orders of backward differencing does
not necessarily ensure more accuracy or stability. For many applications first-order
differencing is preferable by virtue of its relative simplicity.
A second example is the numerical integration scheme used in the Newmark
method. This scheme can be seen to be based on the assumption of linear
acceleration over each time interval:
ẋk+1 = ẋk + ψ[(1 − δ)ẍk + δ ẍk+1 ] (13.25a)

xk+1 = xk + ψ ẋk + ( ψ)2 21 − α ẍk + α ẍk+1 (13.25b)
“chapter13(new)” — 2005/10/20 — page 537 — #9

where α and δ are parameters that can be determined in order to guarantee integra-
tion accuracy and stability. One set of values producing an unconditionally stable
integration, which has found wide application, is δ = 21 and α = 41 . The Newmark
method is a technique of the second basic type, wherein the equations of motion
are used in the integration algorithm itself, and the accelerations must be calculated
using differencing techniques. Typical usage of the Newmark method requires the
use of appropriate differencing equations to form the {˙}k+1 and {¨}k+1 quantities.
A third example is the harmonic acceleration method, wherein the accelera-
tion is assumed to be simple harmonic over each time interval, with a specified
characteristic frequency ω̄. This frequency can be thought of as an “integration”
frequency. For modal formulations the integration frequencies will take on dif-
ferent values for each (modal) degree of freedom and are typically taken to be
the natural frequencies of each of the respective modes. The resulting integration
algorithms are as follows:

cos ω̄ ψ − cos 2ω̄ ψ 1 − cos ω̄ ψ
ẋk+1 = ẋk + ẍk − ẍk−1 (13.26a)
ω̄ sin ω̄ ψ ω̄ sin ω̄ ψ

1 − cos ω̄ ψ
xk+1 = xk + (ẋk+1 + ẋk ) (13.26b)
ω̄ sin ω̄ ψ
This scheme can be seen to be a modified form of the first of the optional forms
of the Adams method (involving only the first backward difference) given earlier.
For the harmonic acceleration method the acceleration is assumed to follow a sine
wave, whereas for the first backward-difference Adams method the acceleration
inherently follows a parabola. Indeed, for ω̄ ψ → 0 the algorithms are equivalent.
The harmonic acceleration method has the ability to duplicate exactly the free
responses of the degrees of freedom oscillating at their respective intrinsic natural
frequencies. Thus, although calculated responses at other frequencies would be
less accurate, such responses would inherently be attenuated by the second-order
response characteristics of the modal degrees of freedom.
The fourth example is a variant of one of the versions of the Runge–Kutta
method. The version described herein is based on a fourth-order Runge–Kutta
scheme with Gill coefficients and relates to a first-order form of the differential
equations of motion. This method was described in Chapter 9 (Section 9.4.5), and
the basic equations are duplicated here for completeness:
{Y } = [Ẋ, X]T (13.27)
which results in a first-order form of the differential equations. The actual solution
algorithm involves the use of the usual integration interval h = ψ, as well as
one-half of this interval h/2:

h 1
{Yk+1 } = {Yk } + {k1 } + 2 1 − √ {k2 }
6 2

1
+ 2 1 + √ {k3 } + {k4 } (13.28)
2
“chapter13(new)” — 2005/10/20 — page 538 — #10

where the vectors {k1 }, {k2 }, {k3 }, and {k4 } are given by
∗
{k1 } = {Yk } = { f (ψk ,Y k )} (13.29a)

h h
{k2 } = f ψk + , Y k + k1 (13.29b)
2 2

h h 1 √
{k3 } = f ψk + , Y k + √ 1 − √ k1 − (1 − 2)k2 (13.29c)
2 2 2

h √
{k4 } = f ψk + h, Y k + √ −k2 + (1 + 2)k3 (13.29d)
2
One distinction of Runge–Kutta methods is that they are self-starting, that is,
solutions can be advanced in time ({ }k → { }k+1 ) without the need for knowledge
of the response in the past { }k−m .
3) Uncoupling of the responses: The basic (matrix) differential equation of
motion set typically contains a mass matrix [M] and a stiffness matrix [K], both
of which are generally well-populated and often unsymmetrical. In all of the
methods discussed earlier, at some point in the solution flow, comprising any
single time-integration step, a decoupling must be performed. The methods that
work explicitly with accelerations, such as the Adams, harmonic acceleration, and
Runge–Kutta methods, must first have the accelerations decoupled so that the
“present-time” accelerations {Ẍk } can be made available for use in the numerical-
integration schemes. Thus, simultaneous equation solutions must be performed
at each time step to uncouple the accelerations; this is indicated by the follow-
ing expression (note that, although a premultiplication by [M]−1 is indicated, the
proper operation should nonetheless be a simultaneous equation solution):
{Ẍk } = [M]−1 { (ψ, Ẋ, X)k } (13.30)
Other methods, of which the Newmark method is a prime example, combine

the statement of the differential equations directly with the numerical-integration
scheme to solve for the “time-advanced” displacements {Xk+1 }. Such methods
typically require a decoupling calculation on some linear combination of the three
basic linear descriptor matrices [M], [C], and [K]. As in the case with the New-
mark method, this typically leads to displacement decoupling. Let δ = 21 , so that,
after some manipulations, the Newmark method equations reduce to the following
coupled equation set (following a simultaneous equations solution):
{Xk+1 } = [D]−1 {[B]{Xk } − [F]{Xk−1 }

+ ah2 {{ fk+1 } + (1/α − 2){ fk } + { fk−1 }}} (13.31)
“chapter13(new)” — 2005/10/20 — page 539 — #11

where
h
[D] = [M] + [C0 ] + ah2 [K0 ] (13.32)
2
[B] = 2[M] − (1 − 2α)h2 [K0 ] (13.33)
h
[F] = [M] − [C0 ] + ah2 [K0 ] (13.34)
2
Because the Newmark method is defined for constant linear dynamic matrices, the
M, C 0 , and K 0 matrices in the preceding expressions must be the components of
the respective total system matrices that are constant. Consequently, the excitation
vector { fk } must be defined as follows:

{ fk } = − [C(ψk )] − [C0 ] {Ẍk } − [K(ψk )] − [K0 ] {Ẍk }
+ {F(ψk )} + {FNL (ψk , X k , Ẋ k )} (13.35)
where C 0 and K 0 represent the constant components of the total damping and
stiffness matrices, respectively. From this development two difficulties arise when
the algorithm is applied to the general nonlinear problem. First, the constant parts
of the damping and stiffness matrices must be formed and extracted from the total
matrices (at each discrete time point). Second, the solution for {Xk+1 } requires
knowledge of { fk+1 }, which is not generally available within the time-step solution
flow. One way to solve this problem is to extrapolate from { fk } to { fk+1 } using
backward-differencing techniques.
4) Initial conditions and startups: The operations already described relate to
repetitive calculations made within the solution flow defined by an arbitrary kth
point in time. Inherent in these operations is the assumption that somehow the
solution has been going on for a sufficient period of time so that all required past
quantities { }k−m are available. However, for start-up calculations of the solution
such quantities would not generally be available. Therefore, the fourth basic opera-
tion needing exposition is the definition of adequate start-up techniques. It should
be stressed that the need for defining start-up techniques is limited to only the
schemes that require quantities in the past { }k−m . Thus, because the Runge–Kutta
method does not require such quantities, start-up calculations are not a consider-
ation for this method. However, the other methods clearly require some sort of
startup procedure.
Three methods can be defined for starting the solution. First, an integration
method that does not require previous time quantities, such as the Runge–Kutta
or some lower-order method, can be used to establish sufficient points in time as
might be required by the nominal technique. An example of such a lower-order
method is Euler’s method, which essentially uses a zeroth-order differencing on
the integration of the acceleration:
ẋ1 = ẋ0 + ψ ẍ0 ; x1 = x0 + 1
2 ψ(ẋ1 + ẋ0 ) (13.36)
The second method is to use appropriate backward-differencing techniques to
approximate the missing required previous time quantities:
ẋ−1 = ẋ0 − ψ ẍ0 ; x−1 = x0 − 1
2 ψ(ẋ−1 + ẋ0 ) (13.37)
“chapter13(new)” — 2005/10/20 — page 540 — #12

A third method is to rewrite the algorithm in such a way as to utilize the ini-
tial conditions. This method is especially useful for application to the Newmark
method. Because the difference equation contains displacement terms for three
consecutive time intervals and none for velocity, it is possible to formulate an
alternate algorithm that utilizes the velocity information. For this purpose the
following formulas are offered:
{X}1 = [D]−1 {[P]{X}0 + [Q]{Ẋ}0 + αh2 { f }1 + [Q]{F}0 } (13.38)
where

[P] = [M] + (h/2)[C] − 21 − α h2 [K]

− 41 − α h3 [C][M]−1 [K] (13.39)

[Q] = h [M] − 41 − α h2 [C][M]−1 [C] (13.40)

[R] = h2 21 − α h2 [I] + 41 − α h[C][M]−1 (13.41)
where [I] denotes a unit matrix.
13.3.2 Comparison of Methods

The basic operations, as well as the details of specific numerical-integration
schemes presented earlier, serve to give a rough overview of the mathematical
details involved in obtaining time histories of the solutions to the nonlinear equa-
tions of motion. The transient responses produced by each of the various schemes
have different characteristics with respect to the basic attributes identified earlier:
accuracy, practicality, and stability. An exhaustive comparison of the four methods
described in the preceding section is beyond the intent of the present text. Further-
more, it should be stressed that other methods exist that are not described herein,
which have been and currently are implemented to good advantage, but again are
not discussed for the same reason.
Accuracy vs cost vs stability. In a sense the three attributes are not indepen-
dent of each other in that all of the attributes are directly related to the integration
step size h. Generally, the solution can be made as accurate as desired and/or the
algorithm might (or might not) be made stable (if unstable) by the selection of a
suitably small step size. Conversely, the cost of obtaining a numerical solution is
inversely related to the step size; smaller step sizes require more points in time
for the calculations and commensurately more computer CPU time (=$). Another
factor impacting on computer costs is the degree of complexity of the algorithm.
The need to make many time-history solution calculations would make simplicity
an obviously desirable characteristic.
Thus, the proper selection of an integration method (and, correspondingly, step
size) would entail making some sort of a trade-off study. The fact that a multiplicity
of methods is presently in use would indicate that several factors are at work in
making this trade-off study: those relating to technical issues (form of equations
“chapter13(new)” — 2005/10/20 — page 541 — #13

used, type of excitation, etc.), as well as those of a more subjective nature (e.g.,
familiarity with any given method and desirability for compatibility with other
existing methods).
13.3.3 Examples of Time-History Solutions of Nonlinear

Aeroelastic Phenomena
Example 1: helicopter stall flutter. Results obtained using various nonlinear
representations of the dynamic stall phenomenon exist in the literature. Several
theoretical/analytical models have been proposed and formulated for the phe-
nomenon, all of which need some form of numerical integration of the equations
of motion for a solution. Figure 13.3, as obtained from the work of Carlson et al.,
shows the results obtained by considering two conceptually different models. The
first such model represents a “brute-force” approach wherein the dynamic-stall
airloads are considered to be describable by a tabulated function of α (section
angle of attack) A and B, where A and B are, respectively, the nondimensionalized
(but implicitly sinusoidal) angle-of-attack rate and acceleration:
α̇b α̈b2
A≡ ; B≡ (13.42)
U U2
The second method considered (for which results are shown in the figure) is based
on the so-called time-delay method for modeling the unsteady effects of dynamic
stall. This method is based on the basic hypothesis that a maximum quasi-steady
angle of attack exists at which the pressure distribution and the boundary layer are
in equilibrium. During further increases in angle of attack beyond this static stall
angle, there are finite time delays before a redistribution of pressure causes, first,
a moment break, and then, a loss of lift corresponding to flow separation. This
method is characterized by an heuristically derived nonanalytical computational
scheme that can only be modeled using a nonlinear time-history solution approach.
The stall flutter signature shows that over one rotor revolution the response
grows exponentially, but is then quenched in the first quadrant to grow once again
in the retreating blade quadrants, etc. This response pattern is highly periodic and,
as such, would have to be strictly classified as a harmonic response. However,
because locally the response is in an energy absorption mode of operation, it
should be considered to be an instability. Although Fig. 13.3 shows the calculated
responses using two different analytical time-domain simulations of the unsteady
stalled airfoil data, other more rigorous methods of simulating these data have been
developed and have been applied to both helicopter and propeller stall flutter. The
calculation for the response could use any relatively straightforward numerical-
integration scheme for the elastomechanics. One challenge in predicting stall flutter
is in calculating the highly azimuthally variable and inflow-dependent impressed
angle of attack. A significant portion of this angle of attack is as a result of the
inflow angle, which in turn, is a function of the variable inflow environment of the
rotor. This impressed angle of attack would be expected to function as a highly
sensitive triggering mechanism for the instability.
“chapter13(new)” — 2005/10/20 — page 542 — #14

Fig. 13.3 Correlation of CH-53A helicopter blade stresses and push rod loads using
alternate nonlinear calculation schemes for dynamic stall (after Carlson et al.).
Example 2: propeller stall flutter. Figure 13.4 presents results obtained,

again using a numerical solution of the nonlinear equations of motion, with
the incorporation of the synthesized unsteady stalled airfoil data (described in
Section 12.6.5), for the case of a statically thrusting propeller. The experimental
results are presented in the form of isostress curves for the one-half peak-to-peak
( 21 PTP) torsion stresses measured at a spanwise location on the blade. The analyt-
ical results again obtained using a numerical time-history solution are represented
“chapter13(new)” — 2005/10/20 — page 543 — #15

Fig. 13.4 Stall flutter correlational results for a statically thrusting

propeller—experiment data vs time-history solutions using synthesized unsteady
stalled airfoil data (after Bielawa et al.).
by the five open or closed square symbols on the figure. These five cases were
selected with blade pitch angle-rotor speed combinations that would appear to
result in both stable and unstable calculated responses. Two extreme cases were
selected to overlap partially (either in blade pitch angle or rotor speed) with one case
that was deemed to be a representative condition strongly associated with signifi-
cantly higher pitch angles. These two extreme cases were deemed to be probably
stable conditions. Additional cases were introduced to enhance the definition of
the apparent stall flutter boundary. The calculated results for the five conditions
are shown either with an open or closed symbol denoting stability or instability,
respectively. Where appropriate, the calculated 21 PTP torsion limit-cycle torsion
stresses are shown parenthetically.
These results, while showing the practical usefulness of such an approach to
analyzing nonlinear phenomena, can identify special considerations that must be
made in order to obtain reasonable solutions. In the course of performing the
calculations, it was noted that the degree of stall flutter response obtainable was
a strong function of the value selected for structural damping. With a sufficiently
high value of damping (0.02), the blade was unconditionally stable, and the stall
flutter condition could not be induced. The final value of damping used (0.008) was
selected on the basis that it produced stable limit-cycle oscillations that neither grew
nor diminished once the instability “locked in.” The more detailed a description of
the phenomenon that is used, the more detailed all of the aeroelastic descriptions
need to be. In the figure a pitch-angle disparity of approximately 6 deg is noted
“chapter13(new)” — 2005/10/20 — page 544 — #16

in the apparent stall flutter boundary pitch angle between the experimental and
analytical results. Such a disparity can come from several sources and represents a
general lack of refinement in predicting either the stall characteristics of the airfoil
or the detailed inflow description and the resulting mean angles of attack of the
various blade sections.
Example 3: subharmonic (helicopter) rotor instability. Another rotor

instability somewhat related to stall flutter but dependent on advancing blade
compressibility phenomena is the subharmonic advancing blade instability
reported by Paul. This instability was observed on a high-speed research helicopter
under conditions of high advance ratio, µ = 0.5, and high advancing blade-tip
Mach number, Mtip ≈ 0.94. This Mach-number condition is well above the criti-
cal Mach number for the airfoil used. The instability was observed on a high-speed
test aircraft as a 1/2P (half per rev) blade response that appeared to the test pilot
as a “split” rotor disk (see Fig. 13.5).
The rotor blade was a 1960s vintage design with a NACA 0012 blade section,
and no compressibility relief (in any form of sweep or thin sections) was built
into the tip sections. The instability was successfully analyzed using a compre-
hensive aeroelastic analysis of the blade with as complete a quasi-steady airloads
description as was possible at that time. This airloads description used look-up
tables for the static airfoil characteristics and included the high-Mach-number
pitching-moment coefficient properties shown in Fig. 13.6.
The findings from this study were interesting from a number of standpoints:
Based on the conditions of the airfoil sections on the advancing side, it was not sur-
prising that the statically unstable pitching-moment characteristics at low-angle,
high-subsonic-Mach-number conditions coupled with blade torsional flexibil-
ity were found to be the principal sources of the 1/2P blade-tip oscillations.
Other findings were somewhat surprising, however. First, the statically unstable
pitching-moment characteristic itself was found to be the important contributor to
Fig. 13.5 Pictorial description of “split” tip-path plane oscillation (after Paul).
“chapter13(new)” — 2005/10/20 — page 545 — #17

Fig. 13.6 Influence of Mach number on NACA 0012 airfoil pitching-moment

characteristics.
the instability and not the unstable slope of the pitching-moment characteristic.
Second, blade mode coincidence with even or odd multiples of 1/2P was not
a necessary condition for instability. Blade mode detuning for odd multiples
of 1/2P would in some instances raise the threshold of the blade oscillation to
a higher Mach number.
13.4 Quasi-Linearization for Explicit Nonlinearities

13.4.1 Basic Ideas
For systems wherein the nonlinearity can be stated as an analytic amplitude-
dependent describing function, the equations of motion can then be cast in a
quasi-linearized form. Let us assume a sinusoidal characteristic motion for the
system degrees of freedom:
{X(t)} = {X̄}eiωt (13.43)
which then allows us to formulate a quasi-linearized form of an eigenvalue problem

to be posed, as indicated in the following compact form:
[L(X̄, ω)]{X̄} = 0 (13.44)
The stability solution is achieved by then expanding the characteristic determinant

|L(X̄, ω)| and setting it to zero, thereby obtaining a characteristic equation. This
characteristic equation is generally complex valued and a function of both ω and X̄.
Because the real and imaginary parts of the characteristic equation fR and fI must
both equal zero, a (nonlinear) system of two algebraic simultaneous equations in
as many variables is defined. Such a system of equations can be solved using a
“chapter13(new)” — 2005/10/20 — page 546 — #18

Newton–Raphson method. This procedure requires one to form a Jacobian matrix

[J( fR , fI )], which is herein defined as

∂fR /∂ X̄ ∂fR /∂ω
[J] = (13.45)
∂fI /∂ X̄ ∂fI /∂ω
The equation

X̄ f
[J] − R (13.46)
ω fI
and the expressions
{X̄}new = {X̄}0 + { X̄}; ωnew = ω0 + ω (13.47)
are then used to iterate from initial guesses of {X̄} and ω, (i.e., {X̄}0 and ω0 , respec-
tively) to the correct values of {X̄} and ω. The resulting solution for these variables
generally defines a limit-cycle condition that might or might not be stable. This
is illustrated in the following example, wherein for a certain type of nonlinearity
the locus of limit-cycle amplitudes has been plotted vs a system parameter . An
instability is known to exist for values of this parameter between the values of
1 and 2 for a zero value of the amplitude {X̄}. For the parameters selected in
this example, it is found that in the region near 2 the solution for {X̄} is triple
valued. Here the range of limit-cycle oscillations increases beyond 2 to the point
A beyond which the system abruptly stabilizes to point B. For subsequent reduc-
tions in the parameter , to values below 2 , the system again becomes unstable,
and {X̄} grows until point C is reached, wherein the response abruptly jumps to
the limit-cycle locus at point D. Such behavior is a classic example of a hysteretic
nonlinear response characteristic (see Fig. 13.7).
Fig. 13.7 Locus of limit-cycle amplitudes for a nonlinear system with a hysteretic
response behavior.
“chapter13(new)” — 2005/10/20 — page 547 — #19

13.4.2 Generalized Quasi-Linearization for Systems

with Periodicity
The basic idea behind quasi-linearization is 1) to obtain some solution to the
full-blown nonlinear equation system (as given earlier); 2) to linearize the sys-
tem of equations about that equilibrium solution; and 3) to solve the resulting
linear equation set for stability. The stability results then indicate how stable the
equilibrium solution obtained is. For rotary-wing problems the complete system
of equations typically contains a nonhomogeneous part and linear and nonlinear
terms that are periodic.
This conceptualization can be mathematically stated using the following
development. Let the total solution vector {X} be expressed as follows:
{X(ψ)} = {X̂(ψ)} + {δX(ψ)} (13.48)

where {δX(ψ)} represents the perturbation of the solution from the equilibrium
solution {X̂(ψ)}. When the equilibrium solution is substituted into the original
equations of motion and second- and higher-order terms in {δX(ψ)} are neglected
[valid because of the infinitesimal size of {δX(ψ)}], the resulting equation set will
be a linear system of equations generally with periodic coefficients. This resulting
equation set is easily recognized to constitute a standard problem of Floquet theory.
The periodicity of the coefficients will arise principally from the periodic nature
of the equilibrium solution (even with an originally formulated equation set that
might have only constant coefficients!). Once such a linearized equation set is
formed, the various tools available for Floquet theory can then be brought to bear.
As straightforward a description of the problem as this might be, it is still
fraught with difficulties. In the first place, the ability to calculate the equilibrium
solution is seriously compromised by the fact that standard methods for solving
nonlinear differential equations are unable to distinguish between equilibrium and
perturbational solutions. Consequently, if the nonlinear system has an unstable
perturbational solution, a straightforward numerical solution of the nonlinear dif-
ferential equations might not be possible because the calculated responses will
necessarily be divergent anyway. Although this result does accurately indicate an
unstable system, it is generally not a practical, cost-effective way to assess sta-
bility because of the relatively high computer times required to obtain numerical
solutions to the nonlinear equation set. Indeed, eigensolutions are generally pre-
ferred because they are not subject to the convergence characteristics typical of
direct assaults on the nonlinear equations using any of a wide variety of numerical
solution schemes.
13.5 Numerical Methods for Stability Estimation

The solution of differential equations of motion by numerical integration, as
developed earlier, offers a powerful generic method for obtaining accurate “simu-
lations” of the dynamic behavior of structures and of rotor systems, in particular.
Clearly, the result of using such simulations is a series of numerical values at
discrete points in time that represent time histories of the responses. One problem
with such simulations is similar to the problem identified earlier: Time histories
“chapter13(new)” — 2005/10/20 — page 548 — #20

(actual or accurate simulations) often contain responses that are ambiguous as to

whether they are resonant responses or true exponentially growing instabilities.
Furthermore, they often contain more response information than we desire (all
modes are typically in evidence), and it is commensurately difficult to extract
by inspection specific stability information about a particular mode from them.
The need then exists to process the time histories in a quantitative way in order
to extract specific stability information from them selectively. Four basic meth-
ods are presently in use for achieving this objective: 1) the log decrement, 2) the
moving-block technique, 3) Prony’s method, and 4) the Ibrahim time-domain
(ITD) method. Within this section, we will examine the first three and briefly
touch on the last of the methods.
13.5.1 Log Decrement

This method is a well-established, basic, simple method for cases wherein
the time-history responses are dominated by one response mode, which is lightly
damped or outright unstable. Such responses typically show a readily discernible
exponential growth (or decay) so that simple methods can be used. Although a
discussion of the log decrement is provided in an earlier chapter, a review is
presented here for ready comparison with other methods. Assuming that a visual
trace of the computed time history can be obtained from the numerical solution
(e.g., using a computer-driven plotter), an estimate of the damping requires the
following steps:
1) Obtain the damped frequency ωD by counting cycles and dividing by the
elapsed time.
2) Pick a point in the time history t1 , where the sinusoidal response is a
maximum (x = maximum = A0 ). Thus, the response is to be approximated by
x(t) = exp(−ζ ωD t)[A sin ωD t] (13.49)
Therefore,
A0 = exp(−ζ ωD t)A (13.50)
3) Record the maximum amplitude after the nth swing past the median value
An ; this value is given by
An = exp[−ζ ωD (t + nπ/ωD )]A (13.51)
4) Form the ratio A0 /An (=Xn ), and then take its natural logarithm. It can then
be seen that the critical damping ratio ζ is related to this logarithm as follows:

ζ / 1 − ζ 2 = n(Xn )/nπ (13.52)
It can further be seen that for cases wherein one mode is contributing practically
all of the response (i.e., all other modes have been damped out by then) and the
damping is small (positive or negative), then
ζ ≈ n(Xn )/nπ (13.53)
“chapter13(new)” — 2005/10/20 — page 549 — #21

13.5.2 Moving-Block Technique

The moving-block technique is, in a sense, an extension of the log-decrement
method but with selective filtering. As with the log-decrement method, it too
works best when one mode is lightly damped (positive or negative). This tech-
nique, furthermore, represents an “engineering approach” to the problem and
lacks some of the mathematical elegance of other methods. As the name sug-
gests, a key component of this method is the definition and manipulation of a data
block of a certain size that is used to “move” through the data to identify certain
changes in parameters with time. It can best be described by stating the steps
taken:
1) The assemblage of points constituting the time history is input to a standard
fast Fourier transform (FFT). [Note that, to optimize the FFT calculations, the
number of points to be used should be 2 to some power (e.g., 512 = 29 ).] The FFT
calculation is performed for the purpose of identifying critical frequencies in the
response. These frequencies are deemed initial critical frequencies and are used
as starting points for “fine tuning” for the more correct ones.
2) For each of the initial (FFT determined) frequencies, the results are
optimized by repetitively performing a discrete Fourier analysis (DFA) of the
time history wherein the number of time-history points constituting a multiple of
the period of the frequency is varied. Thus, if from the FFT calculation, one of
the initial frequencies had a period corresponding to 23 times the sampling time
interval t, then the number of points might be varied from 20 to 26 × N, where
N is a user-selected number of periods to be used. The variation is performed until
a maximum amplitude is obtained with the DFA. The number of points so selected
defines the optimized value of that frequency. This frequency is then used as the
basis for sizing the moving block.
3) Again, for each of the identified frequencies the number of points deter-
mined from the DFA to give an optimized amplitude at that frequency (i.e., in our
example, say the number came out to be 25) is held fixed as a “block size,” and
the DFA is then repeatedly made but with different sequential points as the block
moves over the time trace (i.e., shifts in time) starting at one end of the trace (usually
the start). As the DFA calculations for (complex-valued) amplitude are made at
each shifted time, they are recorded as an additional assemblage of numbers An
(n = 1, 2, . . .).
4) The variation of the natural logarithms of the absolute values of the DFA
obtained numbers n|An | is then plotted vs the shifted time values. It can be seen
that, if a straight line is plotted through these data, the slope of the line is equal to
an “equivalent” characteristic exponent σ , where the equivalent root λ is given by
σ ± iω.
All of the preceding four steps comprising the moving-block technique are
schematically illustrated in Fig. 13.8.
There are three basic advantages to this method. First, stability information is
inherently available from the transient time history. There is no need to obtain
equilibrium solutions or trim states; the method does not “know” whether the
time-history assemblage of points is actual experimental data or calculations.
Second, the time-histories can contain forced excitations that the method will
treat as responses at a critical frequency equal to the excitation frequency. A priori
“chapter13(new)” — 2005/10/20 — page 550 — #22

Fig. 13.8 Schematic illustration of the moving-block technique.
knowledge of these frequencies can enable the method to ignore such frequencies
selectively beyond the FFT phase. Third, the method is relatively simple to
implement; FFT and DFA algorithms are readily available.
One basic disadvantage of the method is that it is relatively expensive. Typically,
the method would find use in low-frequency stability investigations, and several
rotor revs would be needed to obtain enough points to enable the FFT and DFA
calculations to be reasonably accurate. A second disadvantage is that it is not a
mathematically rigorous method, and some “art” is sometimes needed to obtain
credible stability predictions. Often the plot of n|An | vs time is not a straight
“chapter13(new)” — 2005/10/20 — page 551 — #23

line, and the calculation is thereby somewhat ambiguous. This deficiency can
sometimes be overcome by “weighting” one end of the trace or another (or even
the middle portion). Third, as with the log-decrement method, the moving-block
method will only identify the lightly damped modes and is blind to the highly
stable modes.
One technique that facilitates obtaining a good smoothing of the An values
is to use a filtering window such as a Hamming window or a Hanning window.
The method has proven capable of identifying aliased frequencies for time histories
that were generated by dynamic simulations with a high Floquet-like behavior; this
fact should be used for filtering out aliased frequencies in the critical frequency
selection part of the implementation.
13.5.3 Prony’s Method

This method is a form of curve-fitting procedure wherein the assemblage of
numbers constituting the time history is fitted to exponential functions. The method
consists of two main procedures: 1) determining the exponential constants for the
exponential functions and 2) implementing a standard least-squares curve-fitting
procedure. For stability purposes the first step actually suffices. The starting point
for this method is a statement of the curve-fitting approximation:
f (t) ≈ C1 ea1 t + C2 ea2 t + · · · + Cn ean t (13.54)
or, in a more useful equivalent form,
f (t) ≈ C1 µt1 + C2 µt2 + · · · + Cn µtn (13.55)
where
µk = eak (13.56)
From the second form of the preceding approximation, it can be shown that each
of the constants µ1 , . . . , µn is a solution of the following algebraic equation:
µn − α1 µn−1 − α2 µn−2 − · · · − αn−1 µ − αn = 0 (13.57)
where the coefficients α1 , α2 , . . . , αn are determined from the linear simultaneous

equations constructed from the values of the function for which the approximation
is being sought:
fn−1 α1 + fn−2 α2 + · · · + f0 αn = fn (13.58a)

fn α1 + fn−1 α2 + · · · + f1 αn = fn+1 (13.58b)
············
fN−2 α1 + fN−3 α2 + · · · + fN−n−1 αn = fN−1 (13.58c)
“chapter13(new)” — 2005/10/20 — page 552 — #24

Each of the fk ordinates is known (they are the numerical values that are being
curve fitted), and the equation set can be solved for the nα values providing that
N ≥ 2n. If N = 2n, a direct solution arises; if N > 2n, then the solution would be
approximate in a least-squares sense. Up to this point the solution is concerned with
obtaining the n exponential functions, µk (= eakt ). To complete the calculation,
the Cn coefficients must be calculated; this can be accomplished using a stan-
dard least-squares calculation. One idiosyncrasy of the method is that it does not
handle nonviscous damping well (most likely because nonviscous damping does
not produce responses of a simple exponential form). However, one substantial
advantage of this method is that it will detect more highly damped modes than will
the moving-block method.
13.5.4 Ibrahim Time Domain (ITD) Method

This method, like Prony’s method, is an exponential curve-fitting method, with
the difference that this method processes all of the channels of calculated response
on the structure at the same time. Note that the three preceding methods are directed
to analyzing only one channel of response time history at a time. The advantages
of the ITD method are, first, that rigorous methods of testing the accuracy have
been devised to increase the confidence of the predictions and, second, that it
will process all modes with little regard for damping level. It will, furthermore,
obtain not only the complex frequencies, but also the spatial (again complex-
valued) mode shapes. The details of this method are beyond the scope of this text,
and the reader is referred to the bibliography for implementation information. One
consideration to be made using the ITD method is that it assumes that the responses
are free vibrational with damping present. Thus, time histories containing forced
responses could be expected to cause problems.
13.6 Future Directions

The subject of nonlinear dynamics, especially as applied to rotorcraft aero-
elasticity, is yet an expanding technology. New computational concepts are and
will be developed to capitalize on the ever-expanding computer resources being
made available to the dynamicist. Although some techniques have not yet even
been conceived, two new technology areas that are evolving because of essential
nonlinearities bear some description: finite element in time and chaos theory.
13.6.1 Finite Elements in Time

This methodology for addressing nonlinear dynamics to a large measure takes
the form of a state transition matrix formulation (see Section 9.4.5). The basic idea
is to treat time as one of the independent variables in what would be an otherwise
partial differential equation. The difference is that the powerful energy method-
ologies are brought together with a like treatment of time to give a boundary-value
problem on the variable displacements and velocities. Basically, the time variability
can be treated like a Galerkin problem wherein a series of assumed functions of
time are used for the basic solution.
“chapter13(new)” — 2005/10/20 — page 553 — #25

13.6.2 Chaos Theory

The study and, more importantly, the engineering application of chaos theory
are even more in their infancy than is the method of finite elements in time. Chaotic
motion has the prima facie appearance of random motion, but, in fact, has order (of
a higher degree, to be sure). Rigorously defined, chaotic motion is motion wherein
the details of the responses are highly dependent on the initial conditions. For most
applications the existence of chaotic motion in a system would be, like vibrations
and aeroelastic instability, a response condition to be avoided and/or designed
out of the system. However, the quantification of the appropriate parameters for
and the subsequent prediction of chaos will require a new dynamics vocabulary
and a new technology base. For now it is a new direction for study and a new
methodology awaiting significant application.
References
Section
13.1 Urabe, M., “Galerkin’s Procedure for Nonlinear Periodic Systems,”
Archives of Rational Mechanics and Analysis, Vol. 20, 1965,
pp. 120–152.
Bathe, K.-J., Finite Element Procedures in Engineering Analysis,
Prentice–Hall, Englewood Cliffs, NJ, 1982.
Ralston, A., and Wilf, H. S. (eds.), Mathematical Methods for Digital
Computers, Wiley, New York, 1960.
Hildebrand, F. B., Introduction to Numerical Analysis, McGraw–Hill,
New York, 1956.
13.3 Paul, W. F., “A Self-Excited Rotor Blade Oscillation at High Sub-
sonic Mach Numbers,” Journal of the American Helicopter Society,
Vol. 14, No. 1, 1969, pp. 38–48.
13.4 Hildebrand, F. B., Methods of Applied Mathematics, Prentice–Hall,
New York, 1952.
13.5 Hammond, C. E., and Doggett, R., “Demonstration of Subcriti-
cal Damping by Moving-Block/Randomdec Applications,” NASA
SP-415, 1976, pp. 59–76.
Hildebrand, F. B., Introduction to Numerical Analysis, McGraw–Hill,
New York, 1956.
Ibrahim, S. R., and Mikulcik, E. C., “A Method for the Direct Identifi-
cation of Vibration Parameters from the Free Response,” Shock and
Vibration Bulletin, No. 47, Pt. 4, 1977, pp. 183–198.
“chapter13(new)” — 2005/10/20 — page 554 — #26

Problems
13.1 Although the action of a viscous damper can be simply approximated as a
resisting force proportional to velocity (by a constant of proportionality A),
a more realistic modeling of this action should include “saturation” effects at
high velocities. Consider a rotor lead-lag damper whose (nonlinear) damper
force is modeled by
A
FNL = − tan−1 (k )
˙ (13.59)
k
where the saturation value is given by
Aπ
FNLsat = − (13.60)
2k
Derive an expression for the (amplitude-dependent) effective linear damp-

ing rate for sinusoidal motion of the damper (i.e., = ¯ eiωt ).
13.2 An often specified operational “flight condition” for helicopters is the abil-
ity to land and take on passengers while perched on a sloping terrain with
the rotor rotating (at design rotor speed, though not actually lifting). In such
a condition the rotor’s tip-path plane must be maintained horizontally. Con-
sequently, in this condition the rotor blades are flapping with amplitude β̄
(= slope angle of the terrain) relative to the shaft. Because there are no Cori-
olis accelerations in the tip-path plane, the nominal lead-lag motion in this
plane is zero. However, relative to the shaft axis (where the lead-lag dampers
are located and must function), there is nominal blade lead-lag motion
(Coriolis accelerations) as a result of the flapping. This nominal lead-lag
motion can be thought of as a Hook’s joint effect, which manifests itself as
a periodic variation in the rotor speed of any one of the rotor blades. The
definition of the effective (linear) blade lead-lag damper characteristics is
required for the ground-resonance calculations for this operating condition.
The effective variability of the rotor speed, as seen by the damper, and the
specific assumed lead-lag damper saturation characteristics defined earlier
in Problem 13.1 are to be included in this calculation.
(a) Show that the lead-lag motion as seen by the shaft axis lead-lag dampers
is given by
[1 − (1 − cos β) sin2 ψ]
˙4 = (˙ + ) − (13.61)
[1 − (1 − cos β) cos2 ψ]
where
β = β̄ cos ψ (13.62)
“chapter13(new)” — 2005/10/20 — page 555 — #27

and ˙1 and ˙4 are the lead-lag motions measured in the tip-path plane
and shaft axis plane, respectively. [Hint: Consider the blade motion
with three consecutive coordinate transformations going from the 1
(rotating tip-path plane), 2 (nonrotating tip-path plane), 3 (nonrotating
shaft axis plane), and 4 (rotating shaft axis plane) coordinate systems.]
(b) Derive an expression (valid to fourth order in β̄) for the effective (rotor
period, time-averaged) linear lead-lag damping coefficient (small
values of ¯1 ) subject to the inclusions of the two aforementioned effects.
13.3 As developed in detail in the work of Bathe, the “stability” of any numeri-
cal integration algorithm can be determined from the eigenvalues of the
integration approximation operator matrix Â defined for that algorithm.
With this matrix any vector of a suitable combination of dynamic quantities
(displacements, velocities, accelerations, etc.) at one time step {X}t can be
defined in terms of the same vector at the next time step {X}t+δt . For an
unforced system this matrix is defined by the following relationship:
{X}t+δt = [Â]{X}t (13.63)
Determine an Â matrix for harmonic acceleration method for numerical

integration of equations of motion, for a standard spring-mass-damper
system.
13.4 A conventional rigid rotor blade (articulated only in flapping about a flapping
hinge) is operating at high thrust in hover. Its lifting airloads are assumed
to be quasi-steady and governed by a two-dimensional lift coefficient
approximated by
c (α) ≈ 5.73α + 22.48α 3 − 611.57α 5 (13.64)
where the local angle of attack α (in radians) is assumed to be spanwise

invariant and expressible as
1 iω
α = α0 − β̇ = α0 − β̄eiωt (13.65)

Solve for (a) the conditions for limit-cycle oscillations in terms of steady
angle of attack α0 and flapping amplitude β̄ and (b) the resulting frequency
of motion.
“chapter13(new)” — 2005/10/20 — page 556 — #28

14
Model Rotor Testing for Aeroelastic Stability
14.1 Introduction
Testing for rotorcraft aeroelastic and aeromechanical stability continues to be a
necessary tool for the successful development of new designs. Although aeroelastic
analyses have continued to develop with increased breadth and sophistication, they
are still not accurate enough to be used routinely as design tools for the analyses of
man-rated aircraft. Not only must these analyses be validated experimentally, but
instability issues inherent in new rotorcraft concepts must be identified as a guide
to new or continued aeroelastic methodology development. Furthermore, although
testing of a prototype at full scale for aeroelastic stability is still performed, it is
with the intention of providing proof-of-design validations rather than exploring
new stability issues or sizing the design parameters to preclude the occurrence of
any possible instability. Such testing at full scale is typically very cost prohibitive
and is best done only where absolutely necessary. Thus, multiple reasons exist for
testing rotorcraft for aeroelastic instability at model scale.
For such testing to have relevance to full-scale behavior, it must be performed
with the correct attention to the scaling of the appropriate parameters, both with
the rotor and the airframe. Such scaling is generally an exercise in knowing which
issues can be compromised and which cannot, for it is impossible to scale every-
thing at scale factors of anything less than unity. Construction of aeroelastic models
requires attention to issues of practical construction technique. Often a properly
scaled model can end up being literally impossible to fabricate for certain important
ranges of parameters. Although certain test practices continue to be standard, the
state of the art in test procedures is constantly growing, and new resources must be
integrated into the test engineer’s tool kit. As with any other type of testing, there
are certain test preparations, test procedures, and data-reduction techniques appro-
priate to rotorcraft aeroelastic stability testing that must be understood. All of these
issues are covered in this chapter, along with material on aeroelastic considerations
to be given to nominally nonaeroelastic testing.
14.2 Scaling Laws

14.2.1 Scaling of the Rotor
For a complete aeroelastic modeling of the rotor, five basic scaling considera-
tions exist relating to the proper interactions of the aerodynamic, elastic, inertial,
557
“chapter_14(new)” — 2005/10/20 — page 557 — #1

and gravity forces (Hunt). Assuming complete geometric modeling, these inter-
actions can be stated mathematically in terms of the following nondimensional
parameters, which should be maintained invariant:
Frequency scaling:
E E
λf = = (14.1a)
m2 ρ(R)2
Lock number:
ρacR4
γ = (14.1b)
Ib
Advance ratio:
V
µ= (14.1c)
R
Froude number:
2 R
F= (14.1d)
g
Mach number:
R
M= (14.1e)
a∞
The frequency scaling parameter ensures that the blade has the correct natural
frequencies in bending in relationship to rotor frequency. The Lock number ensures
that the rotor has the correct aerodynamic damping and aerodynamic coupling char-
acteristics, and the advance ratio ensures that the scaling of forward-flight speed
is correct in relationship to rotor rotational speed. The Froude number ensures that
the gravity effects, in terms of gravity springs and the rotor thrust, are properly
scaled in relation to the other three basic forces, and the Mach number ensures
that the correct compressibility effects are encountered on the advancing blade
portions of the rotor disk. The Froude number is typically in the order of 500–700
and becomes increasingly important with rotor size. Because the Froude number is
relatively large compared with the other nondimensional parameters, strict scaling
of the gravitational terms can sometimes be relaxed if their effects (as they relate
to the phenomenon at hand) can be approximated.
14.2.2 Scaling of the Pylon

The model pylon itself must be properly scaled relative to the attached rotor in
that it should present to the rotor a properly scaled impedance. As discussed in
Chapter 11, proper scaling of the impedance of the pylon, for instabilities involving
rigid-body motion of the pylon, can be achieved by addressing three basic char-
acteristics: 1) the various mass ratios between the rotor and pylon masses must
be maintained; 2) any geometrical couplings existing between the hub degrees of
freedom must also be maintained; and 3) the pylon natural frequencies must be
scaled in order to give the same frequencies (as nondimensionalized by the rotor
frequency).
“chapter_14(new)” — 2005/10/20 — page 558 — #2

TESTING FOR ROTOR AEROELASTIC STABILITY 559
Mass ratios. Inspection of the air-resonance equations (as given in

Appendix D) shows that two mass ratios of importance to the proper scaling of
rigid-body motion phenomena are 1) the (translational motion) mass coupling
parameter (involving rotor in-plane blade mode generalized mass and in-plane
pylon effective mass) 3 , and 2) the (rotational motion) inertia coupling parameter
4 . This latter parameter involves the rotor flapping blade mode generalized mass
and pylon inertia (rotational about some focal point zfoc , located a distance heff
below the rotor hub):
2
bS48 (generalized ground-resonance
3 = ; coupling parameter) (14.2)
2Meff S49
2
bS12
4 = (14.3)
2Ieff S10
2 , and where the integrals S , S , S , and S are integrals

where Ieff = Meff heff 10 12 48 49
of appropriately mode shape weighted blade mass distribution and are defined in
Appendix D.
Couplings. Couplings of the pylon degrees of freedom can occur because of

longitudinal center-of-gravity locations off the rotor rotation axis and because of
the focusing of the roll and pitch rotations about some position below the rotor
plane. A reasonable approximation to the free-flight condition is to take the focal
point to be the point on the rotor rotation axis that intersects the horizontal plane
containing the total aircraft gravity center. Indeed, it can be seen that only for a
scaling of the focal point at the aircraft center of gravity do the already described
mass ratios 3 and 4 scale commensurately. The choice of the aircraft c.g.
as a focal point is somewhat arbitrary but is nonetheless convenient because it
eliminates the gravity forces as contributors to the roll and pitch spring rates and
thereby minimizes the effects of any non-Froude scaling.
Pylon natural frequencies. As developed in Section 11.4, the requirement to

match the pylon impedance also requires that the natural frequencies (with respect
to rotor rotational frequency) of the pylon (with the constraint of it being focused
at a point heff below the hub) be maintained. In the present context this scaling
principle becomes important when we might wish to alter the properties of an
unscaled rotor also having an unscaled focal point. Thus, for two configurations
( )1 and ( )2 , which have the same rotor properties and the same effective mass
Meff , but are operating at different rotor speeds 1 and 2 , the frequency criterion
then becomes that of maintaining the same effective rotational stiffness.
The pylon stiffnesses in pitch and roll can each be approximated as a sum of
an explicit spring rate (around the focal point) Kp and an implicit one caused by
rotor flapping flexibility Kr . Furthermore, this rotor flapping spring rate Kr can be
conveniently expressed as a factor K̄r multiplying the rotor speed squared. This
factor is frequency dependent and is proportional to the number of blades and the
blade integration constant S12 defined earlier. The K̄r factor depends principally on
the Lock number γ , the nondimensional frequency of vibration ω̄, and the blade
“chapter_14(new)” — 2005/10/20 — page 559 — #3

flapping natural frequency ω̄w , again nondimensionalized by rotor speed :

b
Kr = 2 K̄r S12 , γ , ω̄, ω̄w (14.4)
2
Then the invariancy of pylon frequency criterion can be written as
Kp + K̄r 2
= invariant (14.5)
Ieff 2
2 , and M
But Ieff equals Meff heff eff itself must also be invariant. Therefore, for two
different configurations that must both present the same impedance to the rotor,

Kp / 2 + K̄r Kp / 2 + K̄r
2
= 2
(14.6)
heff heff
1 2
Noting that K̄r must be the same for both configurations, we can then rewrite the
expression to separate out the explicit stiffness rate for the second configuration:
2 2
heff Kp1 h eff
Kp2 = 22 2
+ K̄r 2
−1 (14.7)
heff 1 21 heff 1
Thus, with this equivalency relationship it is possible to achieve a proper scaling of

pylon natural frequencies even for conditions where mismatches in rotor speed
and/or focal point heff are present.
For the model pylon to be properly scaled relative to the rotor, it must present
to the rotor a properly scaled impedance. This can be achieved by matching 1) the
mass ratios between the rotor mass and that of the pylon, 2) any couplings existing
between the hub degrees of freedom, and 3) the pylon natural frequencies (as
nondimensionalized by the rotor frequency).
14.2.3 Scale Factors

Of the five pertinent rotor scaling considerations, only (at most) four can be
simultaneously satisfied for rotors at model scale. Generally, for testing in air,
rotors cannot satisfy both Froude and Mach scaling. Typically, the rotors are scaled
for one or the other, or they are scaled somewhere in between using a combination
of geometric scaling, as quantified by λ , and velocity scaling, as quantified by λV :
λ ≡ MS /FS ; λV ≡ VMS /VFS (14.8)
From these two scale factors all other aeroelastic-, aeromechanical-, and gravity-
related properties can be scaled using Table 14.1, which was extracted from the
work of Hunt. Note that, √ for Froude-scaled rotors, the velocity scale factor is
uniquely determined to be (λ ). Also, the scale factors that relate to atmospheric
sonic speed, gravity, kinematic viscosity, and air density (λa , λg , λµ , and λρ ,
respectively) are normally equal to unity. Only in special pressurized wind tunnels
(which also typically use a heavy gas such as Freon instead of air) can these scaling
factors be varied substantially from unit values.
“chapter_14(new)” — 2005/10/20 — page 560 — #4

Table 14.1 Scaling formulas for aeroelastic quantities
Value of scale factor for unit values of λa , λg , λµ , and λp
Exact Model designed Model designed

expression for for full-scale for full-scale
Quantity to be scaled scale factor General Froude numbers Mach numbers
Dimensionless parameters:
3/2
Reynolds number Re λV λ λp /λµ λV λ λ λ
√
Mach number M λV /λa λV λ 1
Froude number F λ2V /λg λ λ2V /λ 1 1/λ
Velocities √
Linear V λV λV √λ 1
Angular λV /λ (= λ ) λV /λ 1/ λ 1/λ
Structural characteristics
Density σ λσ 1 1 1
Elastic modulus E λσ λ2V λ2V λ 1
Forces   
Aerodynamic A λp λ2V λ2 
 
 



Elastic B λE λ2 λ2V λ2  3 λ2

 λ 

Inertial (radial & oscillatory) C λσ λ2 λ4 



Gravitational W λσ λ3 λ3  λ3
Model responses
Displ. amplitude a λ /λ λ λ λ λ
TESTING FOR ROTOR AEROELASTIC STABILITY
“chapter_14(new)” — 2005/10/20 — page 561 — #5

√ A E √
Frequency ω (λE /λσ )/λσ λ 1/ λ 1/λ
Acceleration ä λ2ω λ λ2 λ 1 1/λ
Dynamic stress p λE λ2V λ 1
Dynamic strain δ 1 1 1 1
561
14.3 Model Construction Considerations

14.3.1 Pylon Constraint
The requirement for both inertia and Froude scaling, the available techniques
for lightweight, low-damping model construction, and the need to approximate
free flight with a constrained nonflying pylon in a wind-tunnel environment all
invariably drive the model design to the same simplified pylon configuration: the
pylon system is typically designed for articulation only in pitch and roll about
some effective total aircraft center-of-gravity point using a gimbal arrangement,
as shown in Fig. 14.1. The design and construction of such a configuration, with
a scaling of its dynamics as close to that of the full-scale vehicle as possible,
represents a substantial accomplishment both in engineering and craftsmanship.
However, the intrinsic deficiencies and/or difficulties in this approach consist of
the following:
1) The modeling of the pylon mass as a gimbaled rigid-body mass articulated
in pitch and roll about some effective center-of-gravity point is an approximation
subject to inaccuracies.
2) Some ranges in pylon parameters can be simply impractical to construct
at model scale. This difficulty is compounded by the fact that relative internal
damping tends to vary in an inverse manner with model scale.
Fig. 14.1 Schematic of a 1/5.86-scale BMR/BO-105 air-resonance model.
“chapter_14(new)” — 2005/10/20 — page 562 — #6

3) The need to approximate the gravity springs in pitch and roll, together
with the concurrent need to approximate free-flight trim conditions, places con-
straints on both the elastic restraints and preloads about the gimbal. For some
combinations of required spring rates and preloads, special gadgetry can be either
impractical or too expensive.
An additional constraint on the design of the gimbal arrangement is the need for
providing some form of “snubbing” of the model pitch and roll amplitudes so that
unstable motion, when it occurs, is delimited and thereby kept to noncatastrophic
values.
14.3.2 Blade Construction

Reference to Table 14.1 shows that for Froude-scaled vs Mach-scaled blades
the blade density scaling is the same, but that the elastic force (stiffness) scaling
(EIMS /EIFS ) is different by a factor of the length scale factor. Furthermore, the
Mach-scaled blades are seen to exhibit full-scale stresses, whereas the Froude-
scaled blades exhibit stresses that are scaled by the length scale factor. Practically
speaking, this means that Mach-scaled blades are considerably more stiff than those
that are Froude scaled. Furthermore, because the modulus of elasticity scales as
the geometric scale factor for Froude-scaled blades and blades must be fashioned
from similar materials as full-scaled blades, care must be exercised in order to
achieve a proper scaling of EI.
14.3.3 Aerodynamic Performance

For scaling purposes the total rotor lift varies as
Lift ∼ ρccl R(R)2 (14.9)
where the (representative blade section) lift coefficient cl is nondimensional and

dependent on geometric angles that should be independent of type of scaling.
Thus, using the results of Table 14.1, one can write the ratio of model-scale lift to
full-scale lift as follows:
3
LMS λ , for Froude scaling
= (14.10)
LFS λ2 , for Mach scaling

But according to Table 14.1, the ratio of rotor lift coefficients must always be unity
for either type of scaling:
CLMS
=1 (14.11)
CLFS
This scaling of total rotor lift and rotor lift coefficients becomes difficult because
of the nonlinear relationships defining the inflow and consequently the local inflow
angle distributions. In this regard three points should be noted:
1) The inflow ratio λ typically varies as the rotor thrust coefficient and would
thus be invariant with Froude vs non-Froude (e.g., Mach) scaling.
“chapter_14(new)” — 2005/10/20 — page 563 — #7

2) By virtue of the aforementioned difference in scaling of the rotor lift for the
two types of scaling, for the same geometry (including blade pitch angles) the two
types of scaling will produce different inflow velocities because of differing values
of lift.
3) Only for Froude scaling does weight scale in the same ratio as does aero-
dynamic lift.
Hence, in general the unit scaling of (required) lift coefficients for non-Froude-
scaled rotors would be difficult to achieve for the same geometric pitch angles. This
feature would be especially important for model tests wherein both rotor/fuselage
interactions and blade pitch angle similarity are important.
An additional aspect of aerodynamic performance is the matching of (scaled)
hub moments to achieve trimmed flight. In the usual implementation of model
fuselage pitch and roll articulation, gimbals with rotational springs about the gim-
bal axes are used (see Fig. 14.1). These springs generally have fixed scaled rates
according to the requirements for invariance of the pitch and roll frequencies
relative to the rotor rotational frequency and usually have low spring rates. The
maintenance of scaled hub moments in forward-flight (wind-tunnel) conditions
requires that some trim moments must be equilibrated by moments about the gim-
bal axes. The dual constraints of low spring rates and transmissibility of variable
trim moments pose special problems in the design of the gimbal axes.
14.3.4 Damping
The sources of damping occurring in the full-scale helicopter in free flight are
either structural [complex modulus of elasticity E = (1 + ig)Enom ], aerodynamic,
and/or frictional. Table 14.1 shows that for either type of scaling the aerodynamic,
elastic, and inertial forces all scale by the same scale factors. Consequently, sources
of damping that are either of the (classical) structural and/or aerodynamic types
should, to first-order effects, scale properly. However, the damping caused by fric-
tional sources tends to scale inversely with the scale factor. This can be appreciated
by noting how this form of damping would scale based on simple ideas using the
results of Table 14.1. Assuming that in a rotational joint (or bearing) the shear
stress caused by friction is constant with scale, then the nondimensional damping
is governed by the following relationships:
−1
[(M/θ̇ )/Iω]MS −2 λ , Froude scaling
= λV (14.12)
[(M/θ̇ )/Iω]FS 1, Mach scaling
Thus, the use of bearings and/or constructions having frictional joints with Froude-
scaled models should be avoided or minimized as much as possible for an accurate
scaling of internal damping.
14.4 Instrumentation and Test Procedures

14.4.1 Typical Stability-Measuring Instrumentation
A wide variety of transducers is currently available for the usual measurements
of quantities relating to rotor aeroelastic stability. Although significant variation
“chapter_14(new)” — 2005/10/20 — page 564 — #8

in objective and/or scope can exist between various rotor aeroelastic tests, the
required instrumentation generally consists of elements in each of the following
types.
Motion transducers. Measurement of the motion is essential to all aero-

elastic testing and can be achieved using a variety of transducers using different
measurement principles.
Potentiometers. These devices are relatively low-technology devices that

present an electrical resistance proportional to angular motion. Potentiometers
are most often used for measuring rotational motion where damping is not a
consideration.
LVDTs. These devices (linear variable differential transformers) present an

electromagnetic induction proportional to linear motion. LVDTs are useful in low
load factor environments in situations where zero damping must be maintained by
a noncontacting transducer.
Strain gauges. These devices are relatively low-cost, low-technology devices

that produce an electrical resistance proportional to linear strain. Strain gauges
have a film-like construction and are attached to the surface of an elastic element.
Strain gauges are generally available in a variety of forms so that the basic types of
strain occurring on the surface of the structure are measurable. Motion is typically
measured using strain gauges by measuring the strain in some one or more of the
elastic elements of the test apparatus whose motion-strain relationship is known.
Accelerometers. Accelerometers come in a variety of sizes, types (number of

axes), and construction types, depending on the physical principle used to create the
electrical output. All accelerometers typically utilize some form of internal elastic
structure to produce an electrical impedance or an electrical charge proportional
to linear acceleration.
Force transducers. The measurement of forces is somewhat optional in that

either static (trim) loads can be measured and/or the dynamic loads participating
in the instability can be measured. Forces are typically measured using load cells,
which come in three basic types:
Strain-gauge types. Load cells can be fabricated in a variety of complexities

ranging from simple beams whose strain is proportional to moment produced by a
load at an antinodal point to complicated devices that are designed for measuring
all six components of load, each of which is proportional to the strain in various
of the structural elements, all with minimal elastic coupling. Strain-gauge-type
load cells are subject to the complexities required for temperature compensation
in the essentially Wheatstone bridge circuitry and to the relatively large mechanical
compliance of the cell (i.e., sufficient strain; hence, motion must be developed in
order to change the resistance of the strain gauges).
“chapter_14(new)” — 2005/10/20 — page 565 — #9

Piezoelectric types. Load cells based on a piezoelectric principle generate an

electrical charge proportional to a stress state in a crystal having piezoelectric
characteristics. Although they are generally more expensive, load cells of this type
have the advantages that multiple components of loads can be readily measured by
the same compact device and that they have low compliance characteristics. Gen-
erally, piezoelectric devices have poor low-frequency and dc measuring properties,
but are quite accurate for sinusoidal loads of sufficient frequency.
Piezoresistive types. Load cells based on a piezoresistive principle generate

an electrical impedance proportional to the stress state in a crystal having piezo-
electric characteristics. Piezoresistive load cells (and accelerometers) generally
operate using an effective Wheatstone bridge circuit and are similar to strain gauges
in operation. Piezoresistive load cells also have low mechanical compliance but
are not subject to the limitations on low-frequency and dc loads characteristic of
piezoelectric devices.
Pressure sensors. Most aeroelastic phenomena entail some form of

unsteady aerodynamic loadings that are essential to the motion. Such loadings can
now be measured quite accurately using various miniature pressure sensors.
These devices typically utilize some form of structural element that converts pres-
sure to a stress in a crystal having piezoelectric characteristics that can then be
exploited in the same manner as described earlier for force transducers.
Rotor azimuth indexing. Knowledge of the position of the rotor-blade’s

azimuthal position is needed for ascertaining the rotor rotational speed and for
correlating the blade motion with the aerodynamic environment. Rotor hub (and
shaft) azimuthal measurements are obtained three basic ways.
Magnetic pickups. Devices of this type generally utilize a magnetized gear-

like wheel attached to the shaft together with a nonrotating coil with a ferro-
magnetic core in close proximity to a point in space traversed by the teeth of
the gear-like wheel. The device thereby generates an electromagnetically induced
impulsive voltage in the coil proportional to the velocity of passage of the gear
teeth past the ferromagnetic core. The instances when the impulsive voltages occur
thereby define the instances when the gear teeth are at the point in space near the
coil core.
Inductive pickups. Inductive pickups are mechanically similar to magnetic

pickups but instead incorporate a nonmagnetic gear-like wheel that generates an
increase in inductive capacity when the gear teeth are at the point in space near
the coil core. Measurement of inductance minimizes the dependency of rotational
speed on the measurement.
Phototachometers. These devices achieve time markers by producing a

change in optical reflectivity, as measured by a photosensitive device, when the
optical reflectivity differentiated sections of the circumference of the rotor shaft
are at the point in space to activate the photosensitive device.
“chapter_14(new)” — 2005/10/20 — page 566 — #10

Rotating to nonrotating data transmission. Testing for the physics of

aeroelastic phenomena in rotor blades might include the measurements of stresses
(as measured by strain gauges) and aerodynamic pressures (as measured by pres-
sure sensors). Because these quantities are in the rotating coordinate system, means
must be provided for transmitting the outputs of the transducers to the nonrotating
coordinate system. Two devices exist for addressing this issue.
Slip ring assemblies. Slip ring assemblies essentially consist of low-noise,

low-electrical-resistance rings that rotate with the rotor shaft and nonrotating low-
resistance brushes that are in contact with the rings and thereby form an electrical
circuit. The electrical signals required for transducer operation are transmitted from
the rotating system to the fixed and vice versa through this ring-brush electrical
circuit.
Resolvers. Some quantities, although measured in the rotating coordinate

system (and transmitted to the fixed using slip rings), need to be reviewed and
subsequently used in the fixed coordinate system. An “online” conversion from
the rotating to nonrotating coordinate system is thus often required. For example,
the blade flapping angle β, as measured in the blade coordinate system, is often
more meaningfully interpreted by conversion to rotor modes. This can be accom-
plished by appropriately combining the blade flapping angle measurements for the
separate blades and then feeding these combinations into a simple trigonometric
resolution based on rotor azimuth angle. Devices that accomplish this trigonomet-
ric resolution are referred to as resolvers wherein the resolution is implemented
either electromechanically or digitally using a microprocessor chip.
Exciters. Although aeroelastic instability phenomena are nominally self-

excited, the need exists for exciting appropriate subsystems that might not in
of themselves be unstable and/or for providing a “standard” input to calibrate all
the various responses to be measured. Devices that can be used to provide such
excitations are as follows.
Electrical and hydraulic shakers. Shakers are most often used for vibration
related testing of structures. However, they have a valid application to aeroelas-
tic stability testing by providing the means for measuring component impedance
from which total system stability can be calculated (using Nyquist techniques).
Furthermore, for some response applications shakers can be used as sources of
standard input, either sinusoidal or pulse. Such shakers consist of two basic parts,
the base and a plunger, with a sinusoidal linear relative motion of two parts of the
shaker whose amplitude and frequency are electronically (servo) controlled.
Piezoelectric crystals. The usual operation of piezoelectrically active solids

(crystals) is to use the charge produced by an applied stress. The phenomenon is
actually reversible in that a strain can be induced in a piezoelectric solid by the
application of a voltage. In practice this is accomplished by attaching a piezoelec-
tric film to an elastic structure in the form of a strain gauge, one that is now a strain
producer, however. Again, such a device tends to operate best for sinusoidal input
strains rather than steady ones.
“chapter_14(new)” — 2005/10/20 — page 567 — #11

14.4.2 Test Preparation

Preparations that should be made for aeroelastic stability testing are typically
those to be made for most other forms of testing. When practical and as appro-
priate, these preparations generally consist of three basic items. First, ranges of
the appropriate parameters must be defined. This selection would be made based
principally on the purpose of the tests. For cases where the testing is in support
of design validation for a new aircraft design, the range of parameters would be
largely defined by the characteristics of the full-scale design. On the other hand,
for tests that are more exploratory in nature the range of parameters would be as
wide as possible commensurate with available resources and time. The appropriate
nondimensional parameters resulting from this process would then be used to scale
the dimensional parameters of the model.
Other considerations that should be taken into account are any limitations on
the construction of the model. For some configurations the selection of a geometric
scale factor λ might be made on the basis of available test equipment and facilities.
However, with the selection of the geometric scaling factor it might be found
that a strict Mach-number or Froude-number scaling is not possible because of
practical construction considerations. In such cases a compromise must be made
in the selection of the velocity scale factor λV to a value other than that uniquely
determined by either Mach-number or Froude-number scaling.
A second principal item of test preparation is the calibration of the test instru-
mentation. Such calibrations are, of course, largely dependent on the type of
instrumentation used. Generally, such calibrations should be kept as simple as
is practical. Methods for calibrating the instrumentation should be devised that
are as direct as possible for obtaining the necessary tares and calibration factors.
One point to keep in mind is that aeroelastic stability testing entails dynamic
responses as opposed to static readings; hence, the calibrations should be made
using dynamic proof loads and/or deflections rather than static ones (as would
generally be simpler).
A third necessary item in the test preparations are the pretest calculations (using
some appropriate analysis) of the dynamic phenomena being tested using the
selected ranges of the test parameters. To be sure, if a truly accurate, reliable analy-
sis were available for analyzing the dynamic phenomenon in question, then testing
would not be necessary. Because this supposition is not true, a less than 100% per-
fect analysis must be used for the pretest calculations. Consequently, the pretest
calculations must be used as a guide in identifying trends in the responses to be
measured. Furthermore, a complementary analysis to be brought to the test process
is some form of online stability evaluator to assess how much stability margin is
available at any one test condition to guide the testing near a stability boundary. The
log-decrement and moving-block techniques described in the preceding chapter
are often used for this purpose.
14.4.3 Test Procedures and Data Reduction

Excitations. Although unstable systems can typically be relied on to exhibit
instability caused by the small disturbances always present in the test environment,
a standard form of excitation is often employed in order to systematize the process.
“chapter_14(new)” — 2005/10/20 — page 568 — #12

For tests wherein actual stable or unstable time-history traces are sought, a standard
excitation is truly required in order to produce measurable responses all with the
same initial energy level. This uniformity of initial conditions thereby ensures that
the use of the stability evaluator already discussed will produce consistent results.
For cases of this type of (transient) testing, one appropriate form of excitation
would be an impulse, suitably calibrated to some standard magnitude. Actually,
any form of standard excitation could be used, even sinusoidal wherein after steady
sinusoidal motion is established over some fixed period of time, the sinusoidal
excitation is then sharply terminated.
Another form of stability testing is to test each of the various subsystems sep-
arately in the frequency domain and then to use the multiple-degree-of-freedom
Nyquist criterion (see Section 9.3.3). In this form of testing, frequency-response
functions (either in the form of mobilities or impedances) are obtained. Thus,
with tests in the frequency domain the excitation must generally be some form
of sinusoidal force and/or moment. However, in such cases although the pri-
mary responses will be sinusoidal and thus not generally unstable with regard to
the phenomenon being tested, other instability issues can arise. In this case an
appropriate form of monitoring of the responses would be to perform online fast-
Fourier-transform (FFT) calculations to determine if there are harmonic responses
at frequencies other than the fundamental one being used for the basic har-
monic excitation. Such monitoring is well within the state of the art of current
digital-signaling-processing (DSP) techniques and can be accomplished using
high-speed (digital) implementations of FFT algorithms on microprocessors whose
architectures are optimized for DSP operations.
Instability quenching. An important aspect of aeroelastic stability testing is

that the experimental article (i.e., the test model) can potentially destroy itself
in the process of providing useful results regarding its stability (or lack thereof ).
Therefore, suitable means must be identified and implemented in the design stage
of the test rig to ensure that the model will yield the desired stability information
and still be intact and therefore able to be tested again. Generally, there are two
basic methods for achieving a control of the unstable motion, that is, instability
quenching: passive and active.
A simple method of passive instability quenching is the use of nonlinear stiffness
and/or nonlinear damping elements so that, for small motion, the model “sees” the
correct values of key stiffnesses and/or dampings, and for motion above certain
(predetermined) threshold values greater stiffness and/or damping values would
become effective. Where feasible, one simple way of achieving this would be to
devise amplitude limiters, which would be rigidly grounded and would thereby
provide essentially an infinite stiffness. An alternate approach would be to use a
frictional damper as the contacting surface of the amplitude limiter. In this case
the frictional damping would be engaged only after the response of the struc-
ture exceeded the threshold displacement value(s). Both of these approaches are
schematically described in Fig. 14.2.
Active instability quenching requires that provision be made in the model con-
struction for the operation of three instrumented activities: 1) sensors to ascertain
whether the motion of the model is exceeding some predetermined threshold
value(s); 2) a caging or restraining mechanism to lock out key degrees of freedom
“chapter_14(new)” — 2005/10/20 — page 569 — #13

Fig. 14.2 Schematic representations of methods for passive instability quenching

using amplitude delimiting.
(motion) of the model using some form of active power; and 3) a source of intel-
ligence to activate the caging motion when appropriate. These three activities can
now be implemented with state-of-the-art electronics and real-time microproces-
sors. However, a detailed description of generic methods for such implementations
is highly situation dependent and beyond the scope of this text. Generally, as shown
in Fig. 14.3, these three activities can be implemented using any of the various
motion sensors identified earlier along with some kind of powered mechanical
“chapter_14(new)” — 2005/10/20 — page 570 — #14

Fig. 14.3 Schematic representation of an example of an active instability quencher.
actuator (electrical, hydraulic, or pneumatic), all of which can be controlled by

a real-time microprocessor that would identify the presence of an instability and
thereupon actuate the powered restraining mechanism.
A critical item in the methodology just given is the ability of the intelligence
source to detect the presence of unstable motion at the incipient motion level. For
this task a variety of methods can be implemented assuming the programmability
of the selected microprocessor. See Section 13.5 for a discussion of instability-
estimating algorithms.
14.4.4 Image Derotator Instrumentation

Modern derotator instrumentation was originally developed by Stetson and
Elkins to provide unique optical measurements in the vibration and flutter testing of
bladed-disk compressor stages of turbojet engines. Derotators allow the continuous
stationary viewing of a rotating structure. The development of the derotator system
begins with the unique optical properties of the folded Abbé inverting prism. The
folded Abbé prism is one of a class of image-rotating prisms that can yield a
stationary image of a rotating object when it is synchronized to rotate at half the
speed of the object. As shown in Fig. 14.4, the unit has two prisms with 30-,
60-, and 90-deg angles and one prism with 30-, 120-, and 30-deg angles, and they
are separated by two air gaps. Light makes five reflections in transit, thus giving
an inverted image. In the work of Stetson and Elkins, the folded Abbé prism was
selected over other image-rotating prisms because it posseses a maximum anglular
field of view (14.3 deg).
Making the image-rotating prism operate in an experimental test environment
was found to require a variety of special design considerations. The prism had
to be mounted in a suitably hollow shaft because the derotator has to make its
optical measurements coaxially with the rotating object. The angular velocities
of the object and the derotator had to be be strictly maintained, which required
“chapter_14(new)” — 2005/10/20 — page 571 — #15

Fig. 14.4 Folded Abbé inverting prism (after Stetson and Elkins).
special indexing on the two shafts with computer control to maintain the 2 to 1
rotation relationship. The coaxial alignment had to be maintained free of significant
vibration, which necessitated the use of an air bearing support. Figure 14.5 presents
a pictorial of the arrangement selected for imbedding the prism within the hollow
shaft motor. Apart from the special mechanical design considerations, a significant
effort was expended to “orchestrate” the components electronically. Figure 14.6
presents a simplified schematic view of the interconnective block diagram for the
system.
Although the system was originally used to investigate the aeroelastic char-
acteristics of turbojet engine bladed-disk compressor stages, the system has
Fig. 14.5 Installation of the prism within the hollow-shafted motor (after Stetson and
Elkins).
“chapter_14(new)” — 2005/10/20 — page 572 — #16

Fig. 14.6 Schematic representation of the system block diagram (after Stetson and
Elkins).
Fig. 14.7 Derotated image test: object speed = 5000 rpm (after Stetson and Elkins).
“chapter_14(new)” — 2005/10/20 — page 573 — #17

considerable potential application to any rotating dynamic structure. Stetson and

Elkins applied the special optical capability of the derotator to take measurements
using hologram interferometry and speckle photography. The details of this type
of instrumentation involve considerable attention to the optical physics and are
beyond the scope of this text. These and other forms of optical measurement can
be made with the derotator, as appropriate to the experimental task. Figure 14.7
presents a measure of the derotation possible with this system. This figure shows
the stationary view of a rotating object whose rotational speed was 5000 rpm. The
object made 667 revolutions during the exposure.
14.5 Aeroelastic Considerations for Nonaeroelastic Testing

14.5.1 Aerodynamic Performance Testing
Model rotor whirl tests are performed for a variety of reasons, one of the prin-
ciples of which is to establish aerodynamic performance. Such tests are often
performed with the aerodynamic shell of the fuselage mounted beneath the geo-
metrically scaled rotor. Not only the rotor but all other airframe components (i.e.,
fuselage, empennage, tail rotor, etc.) exhibit distinct aerodynamic force char-
acteristics (lift, drag/propulsive force, moments). Because all of these airframe
components are typically of interest, each of these airframe components will often
be separately attached to a mechanical ground point by means of some form of
load cell. Consequently, although the model might be admirably designed and
fabricated to measure aerodynamic loads, it would generally be highly improp-
erly scaled in a dynamic sense and would not necessarily exhibit the same aero-
elastic characteristics as the full scale flight vehicle. Thus, although the model
is not intended for aeromechanical or aeroelastic testing, the aeroelastic issue
must still be addressed to ensure that the model will still be stable at the result-
ing dynamic scaling. In this context, two general issues can be easily defined.
First, the rotor by itself at least should be aeroelastically stable. Second, the
rotor-fuselage combination should be stable from both ground resonance and air
resonance standpoints.
Blade aeroelastic stability. Because for aerodynamic (performance) tests

aeroelastic scaling is not generally invoked, any means that stabilize the rotor
blades are acceptable (as long as aeroelastic effects are not an integral part of
the aerodynamic tests). From a flutter standpoint a simple means of stabiliz-
ing the blades would be to control the chordwise mass center to eliminate the
inertial coupling needed to produce flutter. Also, the blades could be fabricated
to be relatively very stiff in torsion in order to provide a maximum separation
between the pertinent frequencies. From a pitch-flap-lag instability standpoint the
blades could be provided with explicit excessive amounts of lead-lag damping
and/or with any form of kinematic coupling that analysis would show to be
stabilizing.
Rotor-fuselage coupled stability. As was indicated earlier, a comprehensive

aerodynamic model test facility for rotorcraft performance could include a separate
load balance for each of the modeled rotorcraft substructures. As such, the rotor
“chapter_14(new)” — 2005/10/20 — page 574 — #18

Fig. 14.8 Inclusion of pylon damping in a nonaeroelastic test facility to preclude the
occurrence of ground resonance.
would be attached through its load cell to a structural ground point and would
form the basis of a configuration that might experience classical ground resonance
instability even though the full-scale rotorcraft would be instability free. In such a
case sufficient care should be put into the design of the support structure (including
the stiffness characteristics of the load cell) to ensure a stable configuration. As
shown in Fig. 14.8, one method is to fabricate the support structure so that an ample
source of pylon damping is available between the load cell and the ground point.
In this manner the stabilizing loads of the damper would not introduce errors in
the measurement of rotor hub loads (except for the effects of hub motion itself on
the loads).
An underlying issue in all of these considerations is that some form of aeroelastic
analysis should precede the fabrication of the test articles and that, on the basis
of this analysis, appropriate features must be designed into these test articles to
ensure stability. Inasmuch as the stability must be guaranteed, the inclusion of the
stabilizing features should be conservative.
14.5.2 Vibration-Related Model Testing

The testing of rotors for vibration characteristics at model scale is generally
a difficult task because of the many features upon which vibration depends. To
the extent that the vibration source is purely aerodynamic (i.e., very stiff blades
and zero rotor-fuselage coupling effects present), then geometric scaling can offer
a basis for the validity for vibration testing. However, generally, the full-scale
rotorcraft does have significant flexibilities both in the rotor and fuselage, and
appropriate scaling laws must be invoked (see Section 14.2). In addition to the basic
blade scaling issues identified in that section, two important effects that especially
impact on the proper scaling are the flexibilities of the control system and the pylon
(i.e., fuselage) mobility characteristics. Both of these structural issues present
formidable challenges to the model designer. The control system, as typified by the
swash plate, is a structure with nonlinear time-variable stiffness characteristics.
“chapter_14(new)” — 2005/10/20 — page 575 — #19

The pylon mobility characteristics of a full-scale rotorcraft are predominated by

the elastomechanical dynamic characteristics of the elastic airframe in free-free
boundary conditions (as appropriate to free-flight conditions). Typically, at rotor
blade passage frequencies the elastic airframe is responding at vibratory conditions
in a range of frequencies above its first few normal modes. Incorporating correct
fuselage mobilities in this frequency range is extremely difficult to incorporate in
a model, and it is, moreover, equally difficult even to know the natural frequencies
a priori on the basis of finite element analysis at the design stage of any rotorcraft’s
development.
References
Section
14.1, 14.2 Bielawa, R. L., “Validation of a Method for Air Resonance Testing
of Helicopters at Model Scale Using Active Control of Pylon
Dynamic Characteristics,” Journal of the American Helicopter
Society, Vol. 34, No. 2, 1989, pp. 33–42.
Hunt, G. K., “Similarity Requirements for Aeroelastic Models of
Helicopter Rotors,” Aeronautical Research Council, London,
CP-1245, 1972.
Stewart, H. L., Pneumatics and Hydraulics, Audel, (Div. of Bobbs-
Merrill), Indianapolis, IN, 1966.
14.4 Stetson, K. A., and Elkins, J. N., “Optical System for Dynamic
Analysis of Rotating Structures,” Air Force Aero Propulsion
Laboratory Technical Report, AFAPL-TR-51, 1977.
Problems
14.1 A given full-scale rotor blade has the following spanwise (constant) section
property distributions:
Edgewise bending stiffness:
EIe = 362 × 106 lb-in.2
Flatwise bending stiffness:
EIf = 5.29 × 106 lb-in.2
Mass distribution:
m = 0.1118 lb-s2 /ft2
“chapter_14(new)” — 2005/10/20 — page 576 — #20

Also,
c = 1.5 ft, R = 20 ft, R = 650 ft/s
Determine appropriately scaled, similar properties each for a Mach-scaled

and a Froude-scaled model rotor 9 ft in diameter.
14.2 Consider the full-scale rotor defined in Problem 14.1, wherein a scaled
model aeroelastic stability test is to be performed. This test is to determine
the effects of advance ratio µ as they impact on flap-lag stability, espe-
cially with regard to the periodicity of the aerodynamic coefficients (Floquet
theory effects). A test range of advance ratios of 0.4 to 0.6 is required. The
design and construction of the model rotor is to be, as in Problem 14.1,
scaled to a 9-ft diameter. However, the tests are to be performed in a wind
tunnel limited to a useful speed range of 150–240 ft/s.
Determine appropriately scaled bending stiffness (both flatwise and
edgewise) and mass distributions needed to accomplish these tests.
14.3 A 1/5-scale free-flying Mach-scaled helicopter model is to be designed,

fabricated, and tested for the purposes of measuring rotor noise. Using
simple momentum strip theory aerodynamic analysis, estimate the scale
factor required for the collective control angle range in order for the model
to fly with an appropriately scaled gross weight.
14.4 As discussed in Section 14.2.2, the proper scaling of the natural frequen-
cies of the pylon structure of a model requires knowledge of the implicit
rotational spring caused by the rotor as quantified by K̄r . Using material
from Appendix D, derive an expression for this spring rate for the hovering
flight condition.
14.5 Figure 14.4 defines a typical installation of a multiaxis load cell for mea-
suring rotor loads. Generally such load cells are strain-gauge instrumented
and therefore require a finite degree of compliance in order to achieve
accurately measurable strains. Successful designs of these load cells main-
tain linearity despite exhibiting considerable coupling between the degrees
of freedom. Thus, the hub load vector {F}, [= Fx , Fy , Fz , MxF , MyF T ]
can be related to the incremental hub displacement degree of freedom vec-
tor, {δXh }, [= δ x , δ y , δ z , , δθxF , δθyF T ] by means of a stiffness matrix
defined by the load cell compliance:
{F} = [KLC ]{δXh } (14.13)
Using the development in Appendix D as a basis, formulate the addi-

tional terms required in the equations of motion to account for load-cell
compliance using this [KLC ] matrix.
“chapter_14(new)” — 2005/10/20 — page 577 — #21

15
Elastomeric Devices for Rotorcraft
15.1 Introduction
The use of elastomeric materials in the control of vibration and in rotor-
stabilization devices is well established in current rotorcraft design practice. In
this chapter we examine the basic properties of elastomeric materials and identify
the critical issues relating their application to rotorcraft. The key component in
all elastomeric devices is the elastomer itself, some form of a rubbery material
that behaves very much like a structurally damped metal, only with significantly
reduced elastic modulus but with a correspondingly enhanced loss factor. The
material is typically bonded in thin layers alternately with equally thin layers of
metal (or composite) to form a device that can serve both as a limited rotation
bearing and/or as a source of damping. The majority of elastomeric devices used
in rotorcraft applications typically utilize the elastomeric material in the shearing
mode of deformation.
15.2 Examples of Elastomeric Devices

Figure 15.1 shows a typical spherical elastomeric bearing used to retain a rotor
blade to the rotor hub.
Fig. 15.1 Typical spherical elastomeric bearing for radial retention of the rotor blade
and provision for articulated motion (Reproduced courtesy of the Lord Corporation).
579
“chapter_15(new)” — 2005/11/16 — page 579 — #1

This use of an elastomeric bearing serves several functions most elegantly.

First, it serves as an efficient radial retention of the blade to the hub. An important
property of elastomeric materials, when used in a laminated configuration, is that
the compressive modulus normal to the laminates is relatively high. Hence, the
spherical elastomeric bearing is ideal for this application. An alternate method for
blade retention would be to use two elastomeric bearings, one to provide only radial
thrust capacity and another configured as an equivalent angular contact bearing.
The second function of the spherical bearing is to provide coincident flapping
and lead-lag hinges. The use of elastomeric bearings significantly reduces the
requirements of vigilant periodic lubrication and maintenance. Lastly, the third
function of the bearing is to provide for the limited rotational motion of the blade
in feathering. Although the bearing does demonstrate inherent stiffnesses in each
of the three axes, these stiffnesses constitute acceptable increments relative to those
that are already present as a result of inertial and aerodynamic sources.
Elastomerics are also widely used in the design of lead-lag dampers. Such
devices have the advantage of requiring significantly lower maintenance relative
to the hydraulic dampers they replace. The substitution of elastomerics in the design
and construction of a lead-lag damper does not generally alter the outer geometry
or attachment configuration of the damper. Often, retrofits of hydraulic dampers
with elastomeric ones are feasible. Figure 15.2 presents a pictorial representation
of a typical design and installation of elastomeric components on a bearingless
rotor system.
The figure shows a spherical elastomeric bearing to provide centering and angu-
lar articulation and a system of flat elastomeric layers to provide centering of the
Fig. 15.2 Section view of a typical installation of an elastomeric damper/pivot device

in a bearingless rotor system (after McGuire).
“chapter_15(new)” — 2005/11/16 — page 580 — #2

ELASTOMERIC DEVICES FOR ROTORCRAFT 581
flexure in the flapwise direction and a source of lead-lag damping. Thus, this
use of elastomerics allows a considerably more compact and lower maintenance
rotor system. An additional advantage claimed for elastomeric devices is that their
failure modes are “graceful” and visually evident in the form of degradation of
the elastomeric material. As attractive as the use of elastomeric devices is, how-
ever, they do have downsides that have to be addressed, as discussed later in this
chapter.
15.3 Basic Characteristics of Elastomeric Materials

Elastomeric materials are essentially nonlinear in nature and dependent on three
main properties: amplitude of motion, temperature, and frequency. The damping
provided by elastomerics is quite similar to structural damping in metals and com-
posites. On a material characterization basis, the damping force (stress) is not
generally affected by the magnitude of the time rate of strain, but rather only by
the sign of the time rate. As with structural damping, the elastomeric material
properties are most usefully described in the frequency domain with a complex
modulus wherein the damping stresses are proportional to response strain ampli-
tudes, but in phase with the time derivative of strains. Thus, for a sinusoidal tensile
strain, wherein ε = ε0 eiωt , the total material stress is described with a loss factor η,
resulting in a complex modulus:
σ = E(1 + iη)ε (15.1a)
or, in the case of shear strain and stress,
τ = G(1 + iη)γ (15.1b)
A basic characteristic of elastomeric materials is that they have three behavioral

states: rubbery, glassy, and a transition state in between. As shown in Fig. 15.3,
the variations of the in-phase moduli E or G and the loss factor η vary with
frequency and temperature in an inverse manner, and the loss factor takes on
different characteristics depending on in which state it is operating.
The inverse relationship between temperature and frequency gives rise to the
frequency-temperature correspondence principle, which is quantified by a shift
factor αT . Figure 15.4 shows how the shift factor is hypothetically defined as a
function of temperature.
The use of the shift factor is best illustrated by considering the replotting of
the moduli and loss factors, as obtained (experimentally) at different tempera-
tures, on the same graph. This is achieved by appropriately shifting the data along
the frequency axis so that all of the data curves fall on the same two curves. The
result is the generation of the so-called reduced frequency, which is defined as the
multiplication of the actual frequency by the shift factor, as shown in Fig. 15.5.
In a sense, the plotting of the data to achieve the results indicated in Fig. 15.5
inherently defines the shift factor. In practice, the results of tests of elastomeric
materials often incorporate the generation of the shift factor and are presented in
the form of nomograms, as typified in Fig. 15.6.
“chapter_15(new)” — 2005/11/16 — page 581 — #3

Fig. 15.3 Effects of frequency and temperature on viscoelastic behavior.
The use of the nomogram is as follows: A value of actual frequency is selected

from the right-hand frequency scale, and a horizontal line is drawn to intersect
the line representing the temperature of operation. Vertical lines are then drawn
from that intersection to the curves for modulus and loss factor. By this means, the
effects of the frequency-temperature correspondence principle are invoked without
actually having to calculate the shift factor.
Fig. 15.4 Hypothetical definition of shift factor as a function of temperature (after

Jones).
“chapter_15(new)” — 2005/11/16 — page 582 — #4

Fig. 15.5 Hypothetical plots of modulus amplitude and loss-factor data vs reduced
frequency (after Jones).
Generally, the use of any particular elastomer is guided by utilizing it in its

transition region wherein the loss factor is a maximum. With this caveat the effects
of frequency are usually minimized, and the resulting device is considered to have
moduli and loss factors that are generally independent of frequency within the
bandwidth of operation.
Fig. 15.6 Hypothetical example of a reduced temperature nomogram of elastomeric

modulus and loss factor incorporating temperature and frequency variations (after
Jones).
“chapter_15(new)” — 2005/11/16 — page 583 — #5

15.4 Elastomeric Lead-Lag Dampers

Using the material properties for the selected elastomer, the design of a lead-lag
damper is then a matter of incorporating the appropriate geometry (i.e., number of
elastomer plies and adjoining metal plies and thicknesses of each) to achieve the
desired stiffness and damping rates. The resulting damper force-displacement char-
acteristics are then expressible in the form of a complex spring rate K = K + iK ,
as defined in the frequency domain. Assuming sinusoidal motion for the damper
(x = δ cos ωt), the resulting sinusoidal force at frequency ω is expressed as
K
F(t) = K x + ẋ (15.2)
ω
or, as expressed completely in the frequency domain,
F(ωt) = δ(K cos ωt − K sin ωt) (15.3)
15.4.1 Single-Frequency Operation

The operation of a specific, but typical, elastomeric lead-lag damper (for the Bell
Model 412) was extensively investigated by Felker et al. Their investigation was
motivated by concerns relating to operational characteristics at two frequencies
at the same time. The basic damper operation is predicated on stabilizing the
blade’s lead-lag motion at a frequency of 3.3 Hz. Figure 15.7 presents for this
damper the variations of the K and K characteristics with amplitude for nominal
single-frequency operation.
15.4.2 Dual-Frequency Operation

Background. One of the downsides of using elastomeric dampers was found
to be the reduction of damping properties at the nominal stabilization frequency by
Fig. 15.7 Bell 412 Damper complex spring rate characteristics for single-frequency
motion (after Felker et al.).
“chapter_15(new)” — 2005/11/16 — page 584 — #6

the dual simultaneous motion at this nominal frequency and at another frequency.
The normal operation of rotors will generate once-per-rev in-plane oscillations as
a result of drag airloads on the advancing side in forward flight. Thus, the lead-
lag dampers will of necessity undergo once-per-rev motion at the same time that
stabilization might be needed at the lower lead-lag natural frequency. In the case
of the Bell 412, the once-per-rev frequency is 5.4 Hz. Experimental data were
obtained and published by Felker et al. for variations in the dual-frequency motion.
For this case experimental data were obtained for the damper to be driven by a
displacement composed of two sinusoidal motions:
x(t) = δ1 cos ω1 t + δ2 cos ω2 t (15.4)
The resulting data and the analytical predictions presented indeed confirmed the
degradation of the 3.3-Hz damping in the presence of varying degrees of oscil-
latory motion at 5.4 Hz. The analytical approach used by Felker et al. was to
curve fit time-domain expressions for effective spring and damper coefficients
k and c, respectively, using the measured single-frequency data. The resulting
analytical expressions for both k and c were simple polynomials in the instan-
taneous displacement x. The final expressions for damper force then combines
these polynomial expressions together with appropriate signum functions on
x and ẋ.
Numerous researchers have also endeavored to devise an accurate modeling
of the damper characteristics in the time domain (e.g., Gandhi and Chopra,
Brackbill et al.) in order to address this problem. Two difficulties with this
approach, however are, first, the need for extensive detailed testing of prospective
dampers to provide sufficient data to be curve fitted to the nonlinear time-domain
modeling, and, second, the lack of universality in applying the modeling coef-
ficients so obtained to alternate damper configurations. An alternate approach is
presented herein that uses the usual single-frequency data, as is normally obtained
directly from the damper manufacturer. However, this method does not require
intermediary transformations of K and K into the time domain and is essentially
formulated in the frequency domain.
Alternate methodology for dual-frequency analysis. The method is predi-

cated on a singular assumption: The elastomer has no memory regarding past
cycles, so that each cycle of motion exists as an isolated single-frequency cycle.
The method then begins with Eq. (15.4). The objective is to transform this equa-
tion into an amplitude and phase-angle-modulated sinusoidal function. Thus, for
each of the two sinusoidal functions, with respective amplitudes and frequen-
cies of δ1 and ω1 and δ2 and ω2 , the two frequencies can each be rewritten as
follows:
δ1 + δ2 δ2 ω2 − δ2 ω2 δ1 ω1 + δ2 ω2 ω1 − ω1
ω1 = ω1 + = + δ2 (15.5a)
δ1 + δ2 δ1 + δ2 δ1 + δ2 δ1 + δ2
δ1 + δ2 δ1 ω1 − δ1 ω1 δ1 ω1 + δ2 ω2 ω1 − ω2
ω2 = ω2 + = − δ1 (15.5b)
δ1 + δ2 δ1 + δ2 δ1 + δ2 δ1 + δ2
“chapter_15(new)” — 2005/11/16 — page 585 — #7

The combined dual-frequency motion of the damper can then be written as

follows:

δ2 δ2
δ(t) = δ1 cos ω̄t cos ω̃t − sin ω̄t sin ω̃t
δ1 + δ2 δ1 + δ2

δ1 δ1
+ δ2 cos ω̄t cos ω̃t + sin ω̄t sin ω̃t (15.6)
δ1 + δ2 δ1 + δ2
where
δ1 ω1 + δ2 ω2
ω̄ = (15.7a)
δ1 + δ2
ω̃ = ω1 − ω2 (15.7b)
It can be easily confirmed that when either of the amplitudes δ1 or δ2 goes to

zero, the correct expression for single-frequency motion is regained. The barred
frequency ω̄ thus represents an “averaged” frequency. Equation (15.6) can then be
rewritten in its final useful form as
1/2
δ(t) = δ12 + δ22 + 2δ1 δ2 cos ω̃t cos(ω̄t − φ) (15.8)
where the phase angle φ is given by

 
δ2 δ1
−δ1 sin ω̃t + δ2 sin ω̃t
 δ + δ2 δ + δ2 
φ = tan−1  1 1  (15.9)
δ2 δ1
δ1 cos ω̃t + δ2 cos ω̃t
δ1 + δ2 δ1 + δ2
Equations (15.8) and (15.9) thus represent a sinusoidal function that is both ampli-
tude and phase modulated. These equations are then used in a single-frequency
formulation, recognizing that, with the basic assumption just presented, the single-
frequency amplitude, as expressed by Eq. (15.8), is the instantaneous amplitude
and the appropriate frequency is ω̄. Thus, the expression for the instantaneous
damper force is given by
F(t) = δinst. [K (δinst. ) cos(ω̄t − φ) − K (δinst. ) sin(ω̄t − φ)] (15.10)
where the instantaneous amplitude δinst. is given by

1/2
δinst. (t) = δ12 + δ22 + 2δ1 δ2 cos ω̃t (15.11)
Note that K and K are evaluated, as sinusoidally varying functions of time, using
the single-frequency data presented in Fig. 15.7, along with the instantaneous
amplitude δinst .
Following the procedure formulated by Felker et al., one then retrieves the
resulting values of K and K (appropriate to one of the frequencies) by using a
“chapter_15(new)” — 2005/11/16 — page 586 — #8

harmonic analysis of F(t). Thus, taking the stabilization pertinent frequency as the
“1” frequency, the resulting K and K values can be evaluated as follows:
2 1 T
K1 = F(t) cos ω1 t dt (15.12)
δ1 T 0
2 1 T
K1 = F(t) sin ω1 t dt (15.13)
δ1 T 0
where T is taken to be a multiple of the period of the “1” frequency, 2π/ω1 .

The results of applying Eqs. (15.8–15.13) together with the data of Fig. 15.7 are
presented in Figs. 15.8a and 15.8b, again for the Bell 412 damper, wherein the “1”
frequency is taken to be the blade stabilization frequency of 3.3 Hz.
Fig. 15.8 (a) Damper stiffness, (b) Damper damping. Comparisons of predicted
and measured Bell 412 damper characteristics at 3.3 Hz for dual-frequency motion
(experimental data from Felker et al.).
“chapter_15(new)” — 2005/11/16 — page 587 — #9

Figures 15.8a and 15.8b clearly show the degradations in both stiffness and
damping that motion at 3.3 Hz sustains when increasing amplitudes of 5.4 Hz are
present. The correlation of predicted values with experiment appears to be compa-
rable with what was achieved with the curve-fitted damper and spring coefficient
method of Felker et al.
Aside from the damper attenuation characteristics caused by dual-frequency
motion, the use of elastomeric dampers does introduce an additional stiffness
into the lead-lag dynamics. Even with only single-frequency motion the effective
stiffness and damping values decrease with damper amplitude (see Fig. 15.7).
Accordingly, the lead-lag dynamics must be examined for suitable frequency
placement and damping levels for worst-case scenarios involving both high initial
displacements and the degree of dual-frequency motion.
15.4.3 Low-Amplitude Operation

A further consideration in the use of elastomeric dampers is that for very
small motion the damping has been found to attenuate. This characteristic has
led to conditions of small-amplitude limit-cycle oscillations. Along with the com-
plications of dual-frequency motion, this characteristic has also been a driver of
the several efforts by investigators to quantify the nonlinear damper characteris-
tics in the time domain. One solution to the low-amplitude problem has been the
design of hybrid dampers that incorporate both elastomeric and hydraulic compo-
nents. Outwardly, these dampers look a lot like conventional hydraulic lead-lag
dampers; only the hydraulic fluid is sealed in, and they are devoid of the pistons and
seals typical of strictly hydraulic dampers. In addition to showing a blade-to-blade
Fig. 15.9 Comparison of loss factors of typical Fluidlastic® and purely elastomeric
lead-lag dampers as functions of dynamic amplitude (after McGuire).
“chapter_15(new)” — 2005/11/16 — page 588 — #10

installation. Besides showing an application of blade-to-blade dampers, Fig. 11.17

also illustrates a typical installation of Fluidlastic hybrid dampers.
Figure 15.9 compares the characteristics of a conventional elastomeric
damper with those of the Fluidlastic damper. The figure clearly demonstrates
the attenuation of damping at low values of amplitude for the conventional
elastomeric damper, as well as the improvements obtained with the hybrid
damper.
15.5 Other Applications of Elastomerics

In addition to bearings and dampers, elastomeric devices find other useful appli-
cations in rotorcraft. Elastomeric couplings in drive shafting are useful in providing
a measure of damping in the drive-train torsional dynamics. Another important
application of elastomerics is in the fabrication of tuned absorbers for the control
of fuselage vibration. As discussed in Chapter 7, a tuned vibration absorber has
optimum attenuation characteristics with a specific combination of stiffness and
damping values.
Because vibration absorbers do not constitute part of the “useful” loads in an
airframe, lightweight design is imperative. The use of elastomeric tuned absorbers
is, therefore, an elegant design solution. Not only is the elastomeric absorber com-
pact and of minimal weight, but the application of elastomerics offers a wide range
of suitable configurations, and the damping is not subject to high periodic mainte-
nance procedures. Figure 15.10 illustrates an example of such a tuned elastomeric
absorber.
Fig. 15.10 Example of an elastomeric tune vibration absorber (Reproduced courtesy

of the Lord Corporation).
“chapter_15(new)” — 2005/11/16 — page 589 — #11

References
Section
15.3 Jones, R. M., Mechanics of Composite Materials, McGraw–Hill, New
York, 1975.
15.4 Brackbill, C. R., Lesieutre, G. A., Smith, E. C., and Ruhl, L. E.,
“Characterization and Modeling of the Low Strain Amplitude and
Frequency Dependent Behavior of Elastomeric Damper Materials,”
Journal of the American Helicopter Society, Vol. 45, No. 1, 2000,
pp. 34–43.
Felker, F. F., Lau, B. H., McLaughlin, S., and Johnson, W., “Non-
linear Behavior of an Elastomeric Lag Damper Undergoing Dual-
Frequency Motion and its Effect on Rotor Dynamics,” Journal of
Gandhi, F., and Chopra, I., “Analysis of Bearingless Main Rotor Aeroe-
lasticity Using an Improved Time Domain Nonlinear Elastomeric
Damper Model,” Journal of the American Helicopter Society,
Vol. 41, No. 3, 1996, pp. 267–277.
McGuire, D. P., “The Application of Elastomeric Lead-Lag Dampers
to Helicopter Rotors,” Lord Library No. LL2133, 1976.
“chapter_15(new)” — 2005/11/16 — page 590 — #12

16
Blade Section Properties
16.1 Introduction
As lifting devices go, rotary wings differ considerably from conventional fixed
wings. Basically, rotary wings are devoid of the various cutouts associated with
fixed wings and are principally designed to take the attendant tension loads and
to have favorable frequency placements. Beyond these issues, the analysis of
rotorcraft stability issues is seen to be highly dependent on the interactions of the
blade kinematics and the resulting coupled blade responses. At the basic level, the
calculation of blade stiffness and mass distributions is mandatory. Also, even for
blades constructed of conventional (isotropic) materials we need to know where
the various elastomechanical points of interest (e.g., mass center, shear center)
are located on the blade sections. Finally, with the introduction of composite
construction, additional couplings accruing from anisotropic properties become
available, which are either detrimental or to be exploited for aeroelastic response
enhancement. This chapter addresses all of these issues with a special emphasis
on the use of composite material construction. In most of what is presented, the
use of conventional (isotropic) materials can be considered as a special case of
composites.
With the current availability of quality finite element tools, the temptation to
restrict one’s efforts solely to their use is quite persuasive. However, blind use of
only one tool is often a prescription for a failed or, on a more positive note, a more
expensive design. The material we will examine serves two functions: first, it will
provide useful first cuts at any new blade design to identify the critical issues that
must be addressed. And second, it will provide the basis for reality checks on the
more comprehensive methods the designer will eventually apply.
Figure 16.1 identifies the basic issues addressed in this chapter. The underlined
quantities are those needed to define the elastic characteristics of the blade. In their
more basic form, these are the stiffness distributions. More general relationships
are possible, however, and will be investigated.
16.2 Generalized Blade Elastic Properties

16.2.1 Typical Rotor-Blade Construction
All rotor blades generally have some form of main structural spar element,
either of metallic (isotropic) composition or composite material. This spar is also
generally hollow, to minimize weight and typically forms a shell-like structure.
591
“chapter_16(new)” — 2005/11/17 — page 591 — #1

Fig. 16.1 Basic elastomechanical properties of the blade section.
As a rule of thumb, the blade section needs to be mass balanced to some point
near the quarter-chord and therefore has some form of balance weight near the
section leading edge that can contribute to the section’s elastic properties. Lastly, a
common design practice is to use a very thin skin for the aft portion of the section,
which is then strengthened and stabilized with the use of filler material (often
honeycomb). The aft skin is capable of contributing to the bending stiffness, and
the honeycomb filler, while providing negligible bending stiffness, can provide
shear stiffness.
16.2.2 Composite Construction Characteristics

The structural advantages afforded by using composite construction in rotor
blades are widely recognized. Not only does this technology present the capacity
for improved aerodynamic performance and potentially lower fabrication costs,
but it also offers the design flexibility for including elastic couplings not previ-
ously available with isotropic (metal) construction. Here we examine the impact
of composites on blade design from the standpoint of defining the elastic stiffness
properties. The very important topics relating to strength and failure mechanisms
are beyond the scope of this text and hence not covered, however.
The use of composite materials in the design and fabrication of rotor blades
starts with an attention to the basics of uniaxial composites. The material presented
herein draws on the basics presented in Appendix E. The main issues we address
here are directed at formulating a completely general description of the elastic
stiffness properties of beam configurations that relate to rotor blades. In this section
we examine the section properties for three basic composite beams, single- and
“chapter_16(new)” — 2005/11/17 — page 592 — #2

BLADE-SECTION PROPERTIES 593
multiple-cell beams of arbitrary geometry and built-up prismatic bars. Figure 16.1
begins the examination by defining the forces (T , Vy , and Vz ) and moments (Mx ,
My , and Mz ), as well as the average section strains (u , γxy
o , γ o ) and curvatures, (φ ,
xz x

v , and w ) that define the section stiffness properties. Initially, we will neglect
warping effects.
16.2.3 Fully Populated Stiffness Matrix

General stiffness matrix. The elastic relationship between the six compo-
nents of (force/moment) and six components of elastic motion (strains/curvatures)
defines the section properties. We seek formulations of the components of the [C]
matrix in the following equation:
   

 T 
 
 u 


 
 
 


 Mx 
 
 φx 


 
 
 


 
 
 

 My   −w 
(6)
= Cij (16.1)

 Mz  
 v  

 − − −
 
− − −

 
  o 
 

 V 
 
 γ 
xy 


y 
 
 
   o  
Vz γxz
With the exclusion of warping effects, Eq. (16.1) represents the most general
expression for the section stiffness properties. In the sections to follow, the var-
ious elements of the [Cij(6) ] matrix will be developed. Note that, as written, this
definition for the beam stiffness is the composite equivalent of an isotropic (metal)
Timoshenko beam. For an isotropic beam, most of the elements in this matrix
would be zero. Such is not generally the case for composite beams, however.
Simplification to equivalent Euler–Bernoulli form. For most cases a com-

posite equivalent of an Euler–Bernoulli beam is the more appropriate form. To
o and γ o
this end, Eq. (16.1) must be rewritten so as to retain the shear strains γxy xz
while removing the shears, Vy and Vz . To accomplish this, we first invert and then
partition Eq. (16.1):
     

 u 
 
 T 
 
 T  

 
 
 
  

 φx 
 
 Mx 

 
 M




 
 
 
 | 

x 


 
 
 
  
  My 

(4×4) (4×2)
 −w  −1  My 
 11
| 12 
(6)  
v = Cij = − − − − − − − − −− 

 
 
 Mz   (2×4) |  Mz 

 − − −
 
 − − −
 (2×2) −−
 


    21 |  
 γxy     
22
 o 
 
 V 
 
 V 


 
 

y 
 

y 

 o     
γxz Vz Vz
(16.2)
“chapter_16(new)” — 2005/11/17 — page 593 — #3

Then, upon setting Vy and Vz to zero the following reduced equation is obtained:
       

 u   
 T  
 T  
 u  

       
 φ   Mx 
  
Mx  −1

 φx 

x (4 × 4) (4 × 4)

= 11 −→ = 11
(16.3)

 −w  
 
M 
 
 M 
 
−w  

   y  y   
    
  
   
  
v Mz Mz v
where the resulting (reduced) 4 × 4 stiffness matrix with only the significant
nonzero terms identified is
 (4) (4) 
C11 C12 0 0
 .. (4) 
 .
(4)
C22
(4)
C23 C24 
(4 × 4) −1  
11 = Cij (4)
=  (16.4)
 ···
(4)
C33 0 
 
(4)
(symm.) ··· C44
(4) (4) (4) (4)

The diagonal elements C11 , C22 , C33 , and C44 are equivalent to the famil-
iar section properties for an isotropic beam: AE, GJ, EIf , and EIe , respectively.
(4) (4) (4)
The remaining coefficients C12 , C23 , and C24 represent elastic coupling stiff-
nesses relating to extension-torsion coupling, flatwise bending-torsion coupling,
and edgewise bending–torsion coupling, respectively. All of these couplings are
available only with the use of anisotropic composite lay ups.
16.3 Beams with Thin-Shell Closed-Cell Construction

16.3.1 Single-Cell Beams
True realization of the superior strength-to-weight properties of composites is
generally achieved by using thin-shelled configurations. Figure 16.2 presents a
pictorial of the principal assumptions used in a single-shell beam configuration.
The development presented herein follows the pioneering work of Rehfield.
16.3.2 Assumptions
All of the basic features of composites, as presented in Appendix E, can be
brought to bear in formulating a theory for single-cell beams. In particular, the
following assumptions are made:
1) The beam is composed of “thin-shell” elements, where everywhere along the
contour of the section s, the thickness of the beam wall approaches a very small
value compared with the beam’s outer dimensions, thereby defining structural
membranes.
“chapter_16(new)” — 2005/11/17 — page 594 — #4

Fig. 16.2 General idealized features of a closed single-cell thin wall beam.
2) The constitutive relations for the structural membrane stress resultants and
strains are given by
    
Nxx 
  A11 A12 A16  εxx 
 
Nss =  A22 A26  εss (16.5)
 
  
  
Nxs (sym.) A66 γxs
where the Nxx and εxx are the axial stress resultant and strain, respectively, Nss
and εss are the “hoop” stress resultant and strain, respectively, and Nxs and γxs
are the shear flow and shear strain along the contour s. For a laminate of N plies,
the various Aij stiffness coefficients in Eq. (16.5) are calculated by summing the
products of the plane off-axis stress stiffnesses Qij for each of the k plies together
with its thickness tk (as described in Appendix E):

N
(k)
Aij = Qij tk (i, j = 1, 2, 6) (16.6)
k=1
3) The beam forms a closed section.

4) All of the laminates comprising the beam wall are “collapsed” down to the
mid-thickness points.
5) Initially, no warping occurs in any of the thin-shell elements.
6) At the section face, all of the shell elements present only a tensile stress and
a shear flow in the direction of the shell element contours.
7) The membrane axial strain across the cross section εxx o is independent of the
o o
transverse shear strains γxy and γxz .
“chapter_16(new)” — 2005/11/17 — page 595 — #5

8) The hoop stress resultant is zero, enabling the elimination of the hoop strain:
εss = −(A12 εxx + A26 γxs )/A22 (16.7)
The elemental constitutive equation can then be written as

Nxx K11 K12 εxx
= (16.8)
Nxs K21 K22 γxs
where
K11 = A11 − A212 /A22 (16.9a)

K12 = K21 = A16 − A12 A26 /A22 (16.9b)
K22 = A66 − A226 /A22 (16.9c)
16.3.3 Axis Locations

Tension axis. The location of the x axis in the plane of the beam section can
be conveniently defined as the tension center or the axis for which no bending
occurs. This can be expressed mathematically as

K11 y ds = 0 (16.10a)

K11 z ds = 0 (16.10b)
Principal axes. The y and z axes that uncouple the respective bending
deformations are defined as the principal elastic axes and are determined by the
condition

K11 yz ds = 0 (16.11)
16.3.4 Elements of the C (6) Matrix

The various elements comprising the C (6) matrix, as defined by Eq. (16.1), are
formed by appropriately integrating the K11 , K12 , and K22 quantities [as defined
by Eqs. (16.5) and (16.9)] along s, the circumferential contour of the thin-shelled
spar. The resulting elements of the (symmetric) C (6) matrix can then be written as
“chapter_16(new)” — 2005/11/17 — page 596 — #6

follows:

C11 = K11 ds (extensional stiffness) (16.12a)

2Ae
C12 = K12 ds (extension—twist coupling stiffness) (16.12b)
c
C13 = C14 = 0 (choice of axis) (16.12c)

dy
C15 = K12 ds (extension—Y -shear coupling stiffness) (16.12d)
ds

dz
C16 = K12 ds (extension—Z-shear coupling stiffness) (16.12e)
ds

4Ae
C22 = K22 ds (torsional stiffness) (16.12f)
c

2Ae
C23 = K12 z ds (torsion—Y -bending coupling stiffness) (16.12g)
c

2Ae
C24 = − K12 y ds (torsion—Z-bending coupling stiffness) (16.12h)
c

2Ae dy
C25 = K22 ds (torsion—Y -shear coupling stiffness) (16.12i)
c ds

2Ae dz
C26 = K22 ds (torsion—Z-shear coupling stiffness) (16.12j)
c ds

C33 = K11 z2 ds (flatwise bending stiffness) (16.12k)
C34 = 0 (choice of axis) (16.12l)

dy
C35 = K12 z ds (bending—Y -shear coupling stiffness) (16.12m)
ds

dz
C36 = K12 z ds (bending—Z-shear coupling stiffness) (16.12n)
ds

C44 = K11 y2 ds (edgewise bending stiffness) (16.12o)

dy
C45 = − K12 y ds (bending—Y -shear coupling stiffness) (16.12p)
ds

dz
C46 = − K12
y ds (bending—Z-shear coupling stiffness) (16.12q)
ds
2
dy
C55 = K22 ds (transverse Y -shear stiffness) (16.12r)
ds
“chapter_16(new)” — 2005/11/17 — page 597 — #7


dy dz
C56 = K22 ds (transverse shear coupling stiffness) (16.12s)
ds ds
2
dz
C66 = K22 ds (transverse Z-shear stiffness) (16.12t)
ds
where

2Ae
= rn ds ds (16.13)
c
Ae is the enclosed area of the single closed-cell beam, and c is the circumference.
16.3.5 Practical Discrete Implementation of Integrals

In actual usage, a practical implementation of the line integrals defined by
Eqs. (16.12a–16.12t) must be used. One scheme for accomplishing this is illus-
trated in Fig. 16.3, wherein the contour is described by a series of planar segments,
each with its own laminate configuration, tied together circumferentially.
The figure indicates that each segment has an axial stress resultant N11 and a
shear flow N12 that are taken to be constant over each segment. [Note that because
of the sign convention defined in Appendix E for determining off-axis stresses and
strains N12 and ε6 are now opposite in sign to Nxs and γxs , respectively, as given
in Eq. (16.8).] The stress is then combined with the geometry of the segments to
Fig. 16.3 Contour points and geometry of segments comprising a composite

closed-cell, thin wall beam.
“chapter_16(new)” — 2005/11/17 — page 598 — #8

form summation approximations to the Eqs. (16.12) integrals:
(i)
C11 = K11 si (16.14a)
i
(i)
C12 = − K12 ρi si (16.14b)
i
(i) (i)
C13 = K11 zmid si (16.14c)
i
(i) (i)
C14 = − K11 ymid si (16.14d)
i

(i) dy
C15 = − K12 si (16.14e)
ds i
i

(i) dz
C16 = − K12 si (16.14f)
ds i
i
(i)
C22 = K22 ρi2 si (16.14g)
i
(i) (i)
C23 = − K12 ρi zmid si (16.14h)
i
(i) (i)
C24 = K12 ρi ymid si (16.14i)
i

(i) dy
C25 = K22 ρi si (16.14j)
ds i
i

(i) dz
C26 = K22 ρi si (16.14k)
ds i
i
(i) (i)
C33 = K11 [zmid ]2 si (16.14l)
i
(i) (i) (i)
C34 = − K11 zmid ymid si (16.14m)
i

(i) (i) dy
C35 = − K12 zmid si (16.14n)
ds i
i

(i) (i) dz
C36 = − K12 zmid si (16.14o)
ds i
i
(i) (i)
C44 = K11 [ ymid ]2 si (16.14p)
i
“chapter_16(new)” — 2005/11/17 — page 599 — #9

(i) (i) dy
C45 = K12 ymid si (16.14q)
ds i
i
(i) (i) dz
C46 = K12 ymid si (16.14r)
ds i
i
(i) dy 2
C55 = K22 si (16.14s)
ds i
i
(i) dy dz
C56 = K22 si (16.14t)
ds i ds i
i
(i) dz 2
C66 = K22 si (16.14u)
ds i
i
Note the following points:

1) Coefficients C13 , C14 , and C34 are nonzero because the constraints defined
by Eqs. (16.10) and (16.11) are not implemented a priori. All of the segment
positions are taken relative to an arbitrary origin. To achieve the decoupling, the
origin must be translated and rotated after the coefficients have been evaluated.
2) The distance from the origin normal to each of the segments ρi takes the
place of the quantity 2Ae c.
3) The y axis has been taken in the aft direction to conform to the origi-
nal Rehfield formulations. In all of the developments in the preceding chapters,
the y axis is taken in the forward (lead-edge direction), and the x axis is
taken to be in the radial direction. Compensation must, therefore, be made for
compatibility.
16.3.6 Example of Usage of the Single-Cell Beam Theory

As was mentioned earlier, the availability of quality finite element analyses
poses a temptation to work exclusively with such codes and neglect the sin-
gle cell theory. Hodges et al. have explored this issue by making a comparison
between the results of using the already developed single-cell analysis and an
MSC/NASTRAN finite element analysis. To this end, a model rotor-blade con-
figuration was selected. The single-cell spar for this blade configuration has
a 35.23-in. radius, is rigid up to radial station 5.23, and has a constant sec-
tion to the blade tip. The cross section for this blade configuration is shown in
Fig. 16.4.
As the figure shows, the laminate ply lay up is unbalanced (anisotropic), which
gives rise to couplings not normally present with metallic (isotropic) fabrication.
The objective here was to achieve variations in twist with rotor speed. The compar-
ison between the two analytic approaches was made on the basis of three loading
cases: 1) lifting combined with weight, 2) a concentrated torque at the blade tip,
and 3) a distributed centrifugal load, as a result of rotor speed. The results of the
comparison for these three cases are presented in Figs. 16.5–16.7.
“chapter_16(new)” — 2005/11/17 — page 600 — #10

Fig. 16.4 Model rotor cross section (after Hodges et al.).
Although the agreement between the two analytical methods is generally

quite good, discrepancies between them were attributed to wall thickness
considerations.
16.3.7 Relaxation of the Warping Assumption

Although the results presented in the preceding subsection were based solely
on the methodology outlined in the preceding subsection, the work of Rehfield
Fig. 16.5 Beam deflection calculations as a result of combined lift and blade weight
(after Hodges et al.).
“chapter_16(new)” — 2005/11/17 — page 601 — #11

Fig. 16.6 Beam twist calculations as a result of applied torque at the blade tip (after
Hodges et al.).
also included the effects of warping. This subsection presents the additional for-
mulations need for including warpage. This additional formulation requires the
inclusion of an additional term to the expression for the axial deformation:
u(x, y, z) = U(x) + yβz + zβy + ψφx (16.15)
Fig. 16.7 Beam twist calculations caused by centrifugal force (after Hodges et al.).
“chapter_16(new)” — 2005/11/17 — page 602 — #12

where U(x) is the (average) extension of the section, βy and βz are the section
rotations, and ψ is the torsion-related warping function, defined as
s
2Ae
ψ= s− rn ds (16.16)
c 0
and where

dy dz
rn = − z −y (16.17)
ds ds
(Note: It can be readily verified that ψ is identically zero for circular sections.)
The warping function is included in the formulation by expanding the constitu-
tive equations to a 7 × 7 matrix representation, where the generalized strains are
now given by
{E} = [u , φx , −w , v , γxy , γxz , φx ]T (16.18)
Thus, the constitutive equation matrix then becomes

   


T 
 
 u  

   

 Mx  



 φ 



 
 
 x 


 
   
 


 M y 
 | 
 −w 


 
 (6) 
 

Mz Cij | C i7  v
= − − − − − − −− (16.19)

 Vy   | 
 


 

T
Ci7 | C77 

o
γxy 


 
 
 


 V z   
 o 


 
 
 γ xz 


 − − − 
 
− − −− 

 Qw   
 

φ x
The quantity Qw is defined as the generalized warping function and for all practical
cases is zero. One exemption would be the imposition of a constraint to keep
a section from warping. This could alternatively be achieved by setting φx to
zero. The additional Cij terms are then given by the following:

C17 = K11 ψ ds (16.20a)

2Ae
C27 = K12 ψ ds (16.20b)
c

C37 = K11 zψ ds (16.20c)

C47 = − K11 yψ ds (16.20d)
“chapter_16(new)” — 2005/11/17 — page 603 — #13


dy
C57 = K12 ψ ds (16.20e)
ds

dz
C67 = K12 ψ ds (16.20f)
ds

C77 = K11 ψ 2 ds (16.20g)
Elimination of the warping deformation terms from the reduced constitutive equa-
tions can still be achieved using the techniques defined by Eqs. (16.2) and (16.3),
wherein the matrix partitioning now includes the generalized warping function Qx
along with the horizontal and vertical shears Vy and Vz .
16.3.8 Honeycomb Filler

Filler material is often used in the airfoil section aft of the main load-carrying
spar. The principal function of the filler is to stabilize the geometry of the aft part of
the airfoil, so that relatively thin skin can be used. Filler material is principally of
two types: foam and honeycomb. Foam has the advantage that it can be very easy to
install and inherently adheres to the inside of the outer shell material. It is typically
inserted into the blade structure in semiliquid form and, with suitable additives,
develops gaseous voids that force the semiliquid into compact low-density cellular
foam. However, foam has the disadvantages that it requires the completed (closed)
outer structure, requires special application tools for deep insertion into one end of
the blade, and requires special techniques to maintain proper mass quality control.
The use of honeycomb effectively addresses these shortcomings. Honeycomb filler
is widely used in blade construction for a number of reasons:
1) It is lightweight, but with consistent mass properties.
2) It is relatively easy to machine to achieve the proper inner contour of the
airfoil. This is accomplished by machining the honeycomb in the compacted
form.
3) Fabrication does not require the outer structure to be completed and closed.
Honeycomb is typically bonded together with the other blade parts all at one time.
4) Honeycomb can carry loads and has stiffness in the out-of-plane direction.
(see Fig. 16.8).
Neither foam nor honeycomb have any significant stiffness in the in-plane
(X1 -X2 ) direction and thus cannot provide any contributions to the bending stiff-
ness of the section. But because of the out-of-plane stiffness of honeycomb, it can
contribute to the shear stiffness properties of the section. As shown in Fig. 16.8,
honeycomb does have an effective Young’s modulus E3∗ for normal loading in the
X3 direction. However, this stiffness is not a property that can be exploited for
our purposes. Indeed, this stiffness is effective only in compression, as any tensile
load would need to be transmitted through the bonding material, which would be
expected to be considerably softer and weaker than the honeycomb. The shear
stiffness, however, is a property that does play a role and should be addressed in
analyses of the beam section properties. Gibson and Ashby present formulations
for the effective (averaged) shear stiffness for hexagonal honeycomb of arbitrary
cell wall widths based on the geometry depicted in Fig. 16.8. For regular hexagon
“chapter_16(new)” — 2005/11/17 — page 604 — #14

Fig. 16.8 (a) Loads carried by hexagonal honeycomb on the faces normal to the X3
axis, (b) single shell with unequal wall widths (after Gibson and Ashby).
cells the honeycomb is essentially isotropic resulting in a relatively simple expres-

sion for the ratio of effective shear stiffness G∗23 to shear stiffness of the cell wall
material G:

G∗23 t
= 0.577 (16.21)
Gs l
where t and l are defined in Fig. 16.8. Gibson and Ashby present experimental
results showing good correlation with Eq. (16.21).
16.4 Multicell Beams

Multicell beam construction arises for those cases wherein either the main
structural spar has supplementary webs, or the aft portions of the section (aft non-
spar skin, trailing-edge stiffener, and/or honeycomb filler) can contribute to the
shear stiffness of the section. The extension of the single-cell theory to the more
general multicell case was addressed by various authors (e.g., Rehfield et al.).
“chapter_16(new)” — 2005/11/17 — page 605 — #15

Their development draws heavily on the single-cell development of Rehfield

described earlier in the preceding section. For the multicell beam, special attention
must be paid to the interaction of the various cells in defining the torsional stiffness
and corresponding torsion-related couplings. In particular, the work of Rehfield et
al. presents original formulations for calculating the combined torsional stiffness
and the location of the shear center. This work makes the usual assumption that
all of the cells must rotate by equal amounts and must equal that of the section as
a whole. However, this assumption represents only half of the problem because,
additionally, all of the cells must translate commensurately with the rotation about
the shear center to maintain the section’s shape. When this element of the prob-
lem is included, the calculation of the shear center actually falls out as part of the
solution and does not need to be calculated separately, as is done in Rehfield et al.
The formulations to follow for determining the torsional stiffness and shear center
are equally applicable to beams of composite construction as well as of the more
conventional isotropic (metallic) construction.
16.4.1 Torsional Stiffness and Shear Center

In this section we again build upon the development of Rehfield et al. to formu-
late methodology for calculating the torsion stiffness of a multicell beam section
and in the process, for finding its shear center. The concept of shear center becomes
obscured for composites because of the possible elastic couplings available. In par-
ticular, with composites it is possible to find the point where shear is decoupled
from torsion (i.e., the classic definition of shear center), only to find that the bend-
ing moment is still coupled with torsion. In such a case one would have to abandon
using the shear center in favor of some other reference axis and use the full elastic
couplings that result from the anisotropy. In the material to follow, although the
calculation of a shear center is an integral part of the formulation, the rigor of
the elastic torsion moment calculation is of prime importance for all construction
materials.
The blade section then is assumed to be stressed by a torsion moment Mx and,
using St. Venant torsion theory, is resisted by a two systems of cellular shear flows
that are constant in each wall segment qk∗ , k = 1, . . . , n, and q̃k∗ , k = 1, . . . , n − 1.
The first of these shear flow systems provides the solution elements for constraining
each of the cells to have the same rotation as the section as a whole (as is typically
done by previous analysts). The second shear flow system provides the elements
for translating the cells individually to maintain their side, contiguous to each other
and to preserve the section shape. These shear flow systems are depicted in Fig.
16.9 for the total section and for an individual cell.
Note that there are only n − 1 of the q̃k∗ shear flows because that which would
have been for q̃n∗ would duplicate qn∗ . Upon reflection of the multicell problem
generally, one can see that there are actually 2n − 1 independent shear flows to be
determined: the usual n for the cell “skin” walls and the n − 1 for the septa dividing
the cells. By this reasoning it is appropriate to include only n − 1 additional q̃k∗
shear flows. Because all of the selected shear flows are closed and circular with
neither “sources” nor “sinks,” the section supports no aggregate shear force, but
readily supports the applied torsion moment Mx .
“chapter_16(new)” — 2005/11/17 — page 606 — #16

Fig. 16.9 Torsion model for the beam section depicting the two systems of shear flows
and details for the kth cell.
The starting point for the development is the relationship between the shear
strain in each wall and the shear-stress resultant (i.e., shear flow). For composite
walls with nonzero shear-extension coupling [i.e., K12 = 0, in Eq. (16.8)], the
shear flow can be rewritten by eliminating the axial strain from the constitutive
“chapter_16(new)” — 2005/11/17 — page 607 — #17

relations, resulting in the following expressions:
Nxs = K22 (1 − β)γxs = q∗ (or, q̃∗ ) (16.22)
where β is an elastic tailoring parameter defined as
K122
β= (16.23)
K11 K22
for conventional (isotropic, metallic) walls the constitutive relationship becomes
Nxs = tGγxs = q∗ (or, q̃∗ ) (16.24)
To make the following development more universal, we rewrite Eqs. (16.22) and
(16.24) as
Nxs = Hγxs = q∗ (16.25)
where it is understood that H can take on either of the expressions, as appropriate.
Rotation constraints. Figure 16.9b shows the combined shear flows on each
wall of the kth cell. The two different shear flows combine over each wall to produce
for each cell a rotational twist rate and a vertical shearing strain. As regard the cell
twist rates, they must all be compatible with that of each of the other cells and to
the total section:
φ1 = φ2 = φ = · · · = φk = · · · φn = φ (16.26)
The twist rate of each cell is based on the integration of its shear strain γxsk around
its contour sk . Well-established theory gives for the arbitrary kth cell the following
expression for this twist rate:

1
φk = γxsk ds (16.27)
2Ak ck
where, as Fig. 16.9b shows, the contour circumference is given by
ck = ckU + ck−1,k + ckL + ck,k+1 (16.28)
“chapter_16(new)” — 2005/11/17 — page 608 — #18

The collection of twist rates for all of the cells can then be put into a matrix form as
   ∗ 
 φ1   q1   ∗ 

 
 
 
  q̃1 

 φ2   
 q2  ∗  
 


 
 
 
 
 ∗ 


 
 
 
 
 q̃ 2 


 . 
 
 . 
 
 


 ..   
 ..   
 .. 

 
 
 
 
 . 

 
 
 ∗ 
 
 


 φ 
 
 q 
 
 

 k−1   k−1   q̃ ∗ 
∗ ∗ ∗ ∗ ∗ ∗ ∗ k−1
φ = φk = S q + S̃ q̃ = S qk + S̃

     q̃k∗ 
φ 
 



 ∗ 
qk+1 








 k+1 
 
 
 
 ∗  

 
 
 
 
 q̃ 


 ..  
 ..  

k+1 

 . 
  
 . 
 
 . 


 
 
 
 
 . 

 
  
 ∗ 
 
 . 


φ 
 
q 
 
 



n−1 
 

n−1 
 q̃ ∗
φn qn ∗ n−1
(16.29)
where
 
 1
 
 


 1

 

 . 
φ = .. φT = {U} φT (16.30)

 


 

 1


  
1
and the matrices given in Eq. (16.29) are defined as follows:
S ∗ ∗
S12 0 0 ··· 0 ··· 0

11
S ∗ ∗ ∗ ··· 0 
 21 S22 S23 0 0 
 
 .. .. .. .. .. .. .. .. .. 
 . . . . . . . . . 
∗  
[S ] =  ∗ ∗ ∗ 
 0 0 0 Sk−1,k Sk,k Sk,k+1 0 0 
 
 .. 
 .. .. .. .. .. .. .. .. .. 
 . . . . . . . . . . 
0 ··· ··· ···0 ∗
Sn−1,n ∗
Sn,n
(16.31)
The elements of Eq. (16.31) are defined as

∗ 1 ds
Sk,k+1 =− (16.32)
2Ak ck,k+1 H
“chapter_16(new)” — 2005/11/17 — page 609 — #19


∗ 1 ds
Sk,k = (16.33)
2Ak ck H

∗ 1 ds
Sk−1,k =− (16.34)
2Ak ck−1,k H
where it is understood that the evaluation of H is made using the properties for
the wall indicated by the line integral. The matrix for the secondary (translation
constraint) shear flows is given by
 ∗ ∗ 
S̃11 S̃12 0 0 0 ··· 0 0 0
 ∗ ∗ ∗ 
 S̃21 S̃22 S̃23 0 0 ··· 0 0 0 
 
 . .. 
 . .. .. .. .. .. .. 
∗ 
. . . . . . . . 

S̃ =  .. .. 
 ∗ ∗ ∗ ∗ ∗ ∗ 
S̃k,1 S̃k,2 S̃k,3 · · · S̃k−1,k S̃k,k S̃k,k+1 . . 
 
 . .. .. .. .. .. .. .. .. 
 .. . . 
 . . . . . . 
∗
S̃n,1 ∗
S̃n,2 ∗
S̃n,3 ··· ··· ∗
S̃n,n−2 ∗
S̃n,n−1
(16.35)
where

∗ 1 ds
S̃1,1 =− (16.36)
2A1 cLE H

∗ 1 ds ds ds
S̃k,k =− + + (16.37)
2Ak ckU H ck−1,k H ckL H

∗ 1 ds ds
S̃k,i =− + 1≤i ≤k−1 (16.38)
2Ak ckU H ckL H

∗ 1 ds
S̃k,k+1 =− (16.39)
2Ak ck,k+1 H

∗ 1 ds
S̃n,m =− 1≤m ≤n−1 (16.40)
2An cTE H
where k = 2, … , n − 1.
Translation constraints. Figure 16.10 depicts the translations that the cell
area centroids must undergo in order 1) to maintain the centroids on a straight
line that also passes through the shear center and 2) to keep the cell septa walls
contiguous.
From the geometry shown in Fig. 16.10, one can see that the (translational)
shear strain that each cell must undergo can be related to the twist rate and the
“chapter_16(new)” — 2005/11/17 — page 610 — #20

Fig. 16.10 Pictorial representation of the translational shearing strains of the cells
needed to maintain section shape and the cells properly contiguous to each other.
shear center location, as follows:
(γxz )k = (yk − ySC )φT (16.41)
The z-direction shear strains are likewise related to the shear flows using the H
factor defined by Eq. (16.25). The following matrix equation defines the required
translational shear relationships:
 
y1 − ySC 0 0  
..
  .
 y2 − ySC 0   
   
  φ = yk − ySC  {U} φT
 ..
.   
 0  ..
.
0 0 yn − ySC
 ∗ 
 q1   ∗ 

 
  q̃1 

 q2  ∗  
 


 
 
 ∗ 


 
 
 q̃ 


 ..   

2 


 . 
 
 . 


 
 
 . 


 ∗ 
 
 . 


 k−1 
q 
 
 

  q̃ ∗ 
∗ ∗ ∗ ∗ ∗ ∗ k−1
= [T ] q + [T̃ ] q̃ = [T ] qk + [T̃ ∗] (16.42)

 ∗ 
 
 q̃k∗ 


 q 
 
 


 k=1   
 ∗  

 
 
 q̃ 


 ..  

k+1 


 . 
 
 . 


 
 
 . 


 ∗ 
 
 . 


 n−1 
q 
 
 ∗  

 ∗  q̃n−1
qn
“chapter_16(new)” — 2005/11/17 — page 611 — #21

where
T ∗ ∗
T12 0 0 ··· 0 ··· 0 
11
T ∗ ∗
T22 ∗
T23 0 0 ··· 0 
 21 
 
 .. .. .. .. .. .. .. .. .. 
∗ 
 . . . . . . . . . 

T = ∗ ∗ ∗ 
 0 0 0 Tk−1,k Tk,k Tk,k+1 0 0 
 
 .. 
 .. .. .. .. .. .. .. .. .. 
 . . . . . . . . . . 
0 ··· ··· ···0 ∗
Tn−1,n ∗
Tn,n
(16.43)
 ∗ ∗ 
T̃11 T̃12 0 0 0 ··· 0 0 0
 ∗ ∗ ∗ 
 T̃21 T̃22 T̃23 0 0 ··· 0 0 0 
 
 . 
 . .. .. .. .. .. .. .. 
∗ 
. . . . . . . . 

T̃ =  .. .. 
 ∗ ∗ ∗ ∗ ∗ ∗ 
T̃k,1 T̃k,2 T̃k,3 · · · T̃k−1,k T̃k,k T̃k,k+1 . . 
 
 . .. .. .. .. .. .. .. .. 
 .. . 
 . . . . . . . 
∗
T̃n,1 ∗
T̃n,2 ∗
T̃n,3 ··· ··· ∗
T̃n,n−2 ∗
T̃n,n−1
(16.44)
The elements of these matrices are defined as follows:

Rotation constraint shear flows:

∗ 1
Tk,k−1 = ds (16.45)
ck−1,k H

∗ dz 1
Tk,k = ds (16.46)
ck ds H

∗ 1
Tk,k+1 =− ds (16.47)
ck,k=1 H
Translation constraint shear flows:

∗ dz 1
T̃1,1 = ds (16.48)
cLE ds H

∗ dz 1 dz 1
T̃k,i = ds + ds 1≤i ≤k−1 (16.49)
ckU ds H ckL ds H
“chapter_16(new)” — 2005/11/17 — page 612 — #22


∗ dz 1 dz 1 1
T̃k,k = ds + ds − ds (16.50)
ck ds H
U ck ds H
L ck−1,k H

∗ 1
T̃k,k+1 =− ds (16.51)
ck,k=1 H

∗ dz 1
T̃n,n−1 = ds (16.52)
cTE ds H
The various elements can then be brought together to form a matrix eigenvalue-like
statement of the problem:
   
   q∗     q∗ 
| | 
 
 | | 
 
∗ S̃ ∗ | − {U}  − −∗ − − − − 
S | −0− − | −0 − |− 0− ∗
− − − |
− − − − − −−
| q̃ + y SC  | |  q̃ =0
− − −  
T ∗ | T̃ ∗ | − {yk }  
 

 0 | 0 | {U} 
 − −
  − 

φT φT
(16.53)
or, more compactly as
 ∗ 
 q   −1

 
 | |
− − − ∗ ∗
S | −S̃ − −| −−−{U} 
q̃∗ = −ySC − − − | |
−−

 

− − −
∗ ∗
  T | T̃ | − {yk }
φT
 
   q∗ 
| | 
 
− − − 
−0− − | −0 − |− 0− ∗
+ | |   q̃ 
0 | 0 | {U}  − − −
 

φT
 
 w1 

 


 w2  

 

0 
 .  
= −ySC [V ]−1 − − − φT = −ySC {W } φT = −ySC .. φT
{U} 
 


 w2n−1 

 


− − −  
 
w2n
(16.54)
The shear center location can then be immediately calculated from the last row of
Eq. (16.54):
1
φT = −ySC w2n φT −→ ySC = − (16.55)
w2n
Once the shear center is found using Eq. (16.55), the solution for the torsional
stiffness proceeds from Eq. (16.54). We rewrite Eq. (16.54) so that the shear flows
“chapter_16(new)” — 2005/11/17 — page 613 — #23

are direct functions of the section twist rate:

   

 w1  
 w1  
∗ 
 
 
 

q 
 w2   
 w2  
1
− − − = −ySC . φT = . φT = Q∗ φT (16.56)
∗ 
 . 
.  w 
 .. 
q̃ 
 

2n

 


 
 
 

w2n−1 w2n−1
Solution for the torsion stiffness. With the solution for the shear flows in
hand, the calculation of the torsion moment resisting the applied moment Mx
can be readily accomplished. The torsion moment per twist rate is obtained by
summing the individual torques within each cell and the torques accruing from
the secondary translational constraint shear flows. Because all of the assumed shear
flows are circular, the resisting moment can then be easily written as

q∗
Mx = 2 A Ã − − − = 2 A Ã Q∗ φT
--
--
(16.57)
q̃∗
where the n elements of the {A} vector contain the individual cell areas:
 


A1 


 


 A2  



 


 .. 

 . 

 

{A} = Ak (16.58)

 


 .. 


 


 . 


 


A 


 n−1 

 
An

The n − 1 elements of the Ã vector consist of the areas defined in Fig. 16.9a for
the secondary shear flows:
 !n 

 1 Ai 


 !n 

 Ai 




2 


 


 .
. 

.
Ã = !n (16.59)

 
k Ai 

 


 


 .
. 


 . 


!n 

A
n−1 i
“chapter_16(new)” — 2005/11/17 — page 614 — #24

Equation (16.57) then defines the torsion stiffness C22 ∗ analogous to the term
similarly defined for the single-cell beam [Eq. (16.12f)]:
∗ Mx
C22 = (16.60)
φT
16.4.2 Coupling Strains

All of the development presented in the preceding section is completely valid for
either isotropic (metallic) or anisotropic composite section construction. However,
for composite materials the γxs strains that go into defining the torsion moment can
also be accompanied by axial strains and stress resultants εxx and Nxx , respectively
[as per Eq. (16.8)]. These quantities in turn give rise to the coupling terms in
the stiffness matrix. In this section we develop the expressions needed for these
quantities. The two cellular shear flow systems must be expanded to define the
shear strains in each element. Reference to Fig. 16.9b shows that the shear flows
in the n − 1 septum elements (beginning with k = 2) are given by
qk,k−1 = qk∗ − qk−1

∗
+ q̃k∗ (16.61)
And the shear flows in the outer skin elements (both top and bottom) are given by

k
qk = qk∗ + q̃i∗ (16.62)
i=1
where both shear flows are proportional to the rate of twist, as per Eq. (16.56). We
next reorder the element shear flows as follows:
 
1 0 ···
------------- -------------
 1 1 0 ··· 
 
 I 1 1 1 0 
 n 
 . 
 .. 
" #   
qk  −−−−−−−−−−−− 1 −1− − · −1− −1
· ·−  ∗
{Q} = = −1  Q
qk,k−1  1 0 
 0 −1 
 1 0 
 
 .. .. 
 0 . . 
 In−1 
 −1 1 0 
−1 1
(16.63)
Note that here In refers to an n × n identity matrix. Within each element the
shear strain is then given by
{γxs } = {λ} φT (16.64)
“chapter_16(new)” — 2005/11/17 — page 615 — #25

where
 
------------- -------------
..
 . $ % 
 
 1 
 0 
 K22 (1 − β) j 
 
 .. 
 . 
 
{λ} = − − − − − − − − − − − − − − − − − − − − − − −− {Q}
 .. 
 . $ 
 % 
 1 
 
 0
K22 (1 − β) j,j−1 
 
 .. 
.
(16.65)
and where j = 1, . . . , n and ( j, j − 1) = (2, 1), . . . , (n, n − 1). The axial strain for
each element is then calculated with the realization that with only an applied
moment Mx , Nxx must be identically zero for all elements:
" #
K
{εxx }= − 12 λ φT
K11
 
-------------- --------------
..
 . 
 $ % 
 K12 
 0 
 K11 K22 (1 − β) j 
 
 .. 
 . 
 
= −  − − − − − − − − − − − − − − − − − − − − − − − − − − − {Q}φT
 
 .. 
 .
$ % 
 K12 
 
 0
K11 K22 (1 − β) j,j−1 
 
 .  ..
(16.66)
This strain distribution can then be used to calculate the various section strains that
couple with the torsion moment.
16.5 Total Section Properties

16.5.1 Secondary Section Components
Secondary components that go into making up an airfoil section include the
leading-edge balance weight, the aft skin, the trailing-edge stiffener, any additional
vertical septum webs, and any filler (typically only) aft of the main structural spar.
“chapter_16(new)” — 2005/11/17 — page 616 — #26

Clearly, for determining the chordwise mass center location, all of the components
must be considered. For certain loading conditions some, but not necessarily all,
of these components can contribute to the stiffness characteristics. For bending
stiffness considerations, the leading-edge balance weight, aft skin, trailing-edge
stiffener, and the additional webs can all contribute and should be included. For
conditions relating to shear, the filler (if it is honeycomb) can also play, and its
contribution should be examined.
16.5.2 Calculation of Principal Elastomechanical

Distributions and Section Centers
Most of the blade aeroelastic analyses presented in the preceding chapters
require the locations of the various section centers relative to a reference point on
the airfoil, usually taken to be the shear center. In the material to follow, however,
the locations of all of the various centers are presented relative to the leading edge
(see Fig. 16.1), and the reader can transfer the centers relative to other reference
points, as appropriate.
Mass properties. The calculations of the mass distribution and mass cen-
ter of the airfoil section are elementary exercises in statics. The mass densities
and cross-sectional areas of the various components ρj and Aj , respectively, and
(LE)
(from the geometry of the section) the component mass centers yCGj are typically
readily available quantities. From these quantities all of the pertinent mass-related
properties can be found:
Mass distribution:

m= ρj Aj (16.67)
j
Mass center:
(LE) 1 (LE)
yCG = yCGj ρj Aj (16.68)
j m
j
Chordwise mass moment of inertia:

(LE)
(LE)
IM Z
= (yCGj )2 + kz2j Aj (16.69)
j
Flatwise mass moment of inertia:

IMY = ky2j Aj (16.70)
j
where (ky )j and (kz )j are, respectively, the radii of gyration of the component about
axes parallel to the y and z axes through its mass center:
Torsional inertia (about the section mass center):
(LE) (LE)
JM0 = IMY + IMZ − m(yCG )2 (16.71)
“chapter_16(new)” — 2005/11/17 — page 617 — #27

Elastic properties. As described in preceding sections, a variety of couplings

is possible with the anisotropic properties of a composite spar. In this section we
address only those properties and couplings that would result from both isotropic
and anisotropic components:
Neutral axis point (tension center):
! (LE)
(LE) j yT (AT E)j
yNA = !j (16.72)
j (AT E)j
(LE)
where ATj , and [ yT ]j are, respectively, the area and centroid location of the jth
component, if it can take tension.
(Nonprincipal axis) flatwise bending stiffness:

EIyy = (EIy )j (16.73)
j
(Nonprincipal axis) edgewise bending stiffness:
(LE)
EIzz = Ej [IAz − (yNA )2 AT ]j (16.74)
j
Product of inertia:

(LE)
EIyz = Ej z[y − yNA ] dA (16.75)
j j
(Principal axis) flatwise bending stiffness:
&
2
EIyy + EIzz EIyy − EIzz
EIf = − 2
EIyz + (16.76)
2 2
(Principal axis) edgewise bending stiffness:
&
2
EIyy + EIzz EIyy − EIzz
EIe = + 2 +
EIyz (16.77)
2 2
“chapter_16(new)” — 2005/11/17 — page 618 — #28

Rotation angle defining the principal axes:

1 2EIyz
φ= tan−1 (16.78)
2 EIzz − EIyy
Relative to the axes defined in Fig. 16.1, φ would be a positive (counterclockwise)

rotation about the x axis.
Shear center. Because other components other than just the spar can contribute
to the shear bearing capacity, the results of Section 16.4 can be applied in an
approximate manner. Figure 16.11 shows a possible approximation of the complete
blade section to a series of closed section cells.
Note that the shear bearing capacity of the honeycomb material has been approx-
imated as a series of thin-web septa, the number of which must be made on the
basis of the convergence of calculated values with variation of the assumed spac-
ing. The H value that would be needed for the approximate honeycomb septa
would be the product of the effective shear modulus G∗23 , as given by Eq. (16.21),
and the assumed spacing sHC .
Torsion stiffness. In addition to the use of a multicellular approximation to

the airfoil section, any discrete sources of torsional stiffness should be included.
For example, if the leading-edge balance weight has inherent torsional stiffness,
it should be included; thus,
∗
GJ = C22 + GJj (16.79)
Fig. 16.11 Multicell approximation to the shear bearing components of an airfoil

section.
“chapter_16(new)” — 2005/11/17 — page 619 — #29

16.6 Hygrothermal Effects in Composites

A significant characteristic of composites is that they can absorb moisture
and are structurally affected by heat. As Fig. 16.12 shows, however, the effects
of hygrothermal variations on the basic moduli are not great. However, for
anisotropic (unbalanced) composites, the differences in variations between the
moduli with either heat or moisture content will generally produce warpage of the
structure. In particular, the intentional incorporation of tension-torsion coupling
[i.e., nonzero C12 (Eq. 16.12b)] would be seriously compromised by hygro-
thermal effects. Winckler has devised a method for negating these effects for
Fig. 16.12 Typical effects of temperature and moisture concentration on tensile and
shear moduli (after Tsai and Hahn, data from Browning et al.).
“chapter_16(new)” — 2005/11/17 — page 620 — #30

the intentional incorporation of tension-torsion coupling in thin-wall tubes. The

method that achieves this elimination of effects is to build up all laminates from
symmetric ply groups of the following form:

[ · · · ]T = θ, (θ + 90◦ ) · · · s (16.80)
The interpretation of Eq. (16.80) is that to achieve hygrothermal stability each of

the ply groups is to be composed of four symmetric layers of the same material and
thickness, whose ply pairs are oriented 90 deg to each other. The local laminates
anywhere around the spar can be made of multiple ply groups, with different
orientations and materials, as long as each ply group follows the symmetric four-
layer rule. The total local laminates can then be expressed as follows:

NPG

[ · · · ]T = θi , (θi + 90◦ )s · · · T (16.81)
i=1
Figure 16.13 compares the tension-torsion coupling of two laminates of equal

thickness: one with a [θ4 , −θ4 ]T , ply structure and another incorporating the
hygrothermal stability constraint.
Fig. 16.13 Comparison of tension-torsion couplings achieved with stabilized and

nonstabilized hygrothermal laminates of equal thickness (after Winckler).
“chapter_16(new)” — 2005/11/17 — page 621 — #31

It is evident that little coupling capacity is sacrificed with the incorporation of

the hygrothermal stabilization constraint. The model rotor blade examined in Sub-
section 16.3.5 (see Fig. 16.4) approximately follows the hygrothermal stabilization
constraint.
16.7 Prismatic Bars

The objective here is to define the usual engineering section constants for a bar
with width c and thickness h, as shown in Fig. 16.14.
The appropriate starting point is the stress-strain relationship for a general
composite laminate, as presented in Appendix E as Eq. (E.28)
 | 
" #
Ni A B
 ij | − −− εjo
im
− − − = − − −|  k (16.82)
Mk m
Bkj | Dkm
where all of the indices are for 1, 2, and 6. Ni and Mk are, respectively, the compo-
nents of stress resultant and moment per unit width, εjo are the average strains, and
km are the curvatures. The appropriate integrations over the thickness have already
been achieved with the definitions for the Aij , Bim , and Dkm coefficients. For sym-
metric laminates, the Bim terms are zero. However, Eq. (16.82) does not address
all of the terms required for Eq. (16.1) because, as developed in Appendix E, only
w displacements were considered. Also, the k2 and M2 curvature and moment,
respectively, relate to curvatures and moments on the sides of the beam. Addi-
tional development is therefore required for ascertaining the in-plane moment and
Fig. 16.14 Idealization of a simple laminated composite prismatic bar.
“chapter_16(new)” — 2005/11/17 — page 622 — #32

curvature characteristics. Also, the out-of-plane shear strain and stress resultant γxzo
and Vz , respectively, involve a careful consideration of interlaminate shear charac-
teristics, which are beyond the intended scope of the present text, but addressed by
Whitney. Furthermore, in practice the simplistic lay up shown in Fig. 16.14 would
be enhanced either by laminates normal to those shown in the figure and/or with
layers of spiral wound composite on the outside to give environmental protection
as well as a deterrent to interlaminate separation.
References
Section
16.3 Gibson, L. J., and Ashby, M. F., Cellular Solids—Structure and
Properties, 2nd ed. Cambridge Univ. Press, New York, 1997.
Hodges, R. V., Nixon, M. W., and Rehfield, L. W., “Comparison of
Composite Rotor Blade Models: A Coupled-Beam Analysis and an
MSC/NASTRAN Finite-Element Model,” NASA TM-89024, 1987.
Rehfield, L. W., “Design Analysis Methodology for Composite Rotor
Blades,” Proceedings of the Seventh DoD/NASA Conference on
Fibrous Composites in Structural Design, AFWAL-TR-85-3094,
U.S. Air Force, 1985, pp. V(a)-1–V(a)-144.
16.4 Rehfield, L. W., Atilgan, A. R., and Hodges, D. H., “Structural
Modeling for Multicell Composite Rotor Blades,” AIAA Paper
No. 88-2250, 1988.
16.6 Browning, C. E., Husman, G. E., and Whitney, J. M., “Moisture Effects
in Epoxy Matrix Composites,” Composite Masterials: Testing and
Design (Fourth conference), ASTM STP 617, American Society for
Testing and Materials, 1977, pp. 481–496.
Tsai, S. W., and Hahn, H. T., Introduction to Composite Materials,
Technomic Publishing, Lancaster, PA, 1980.
Whitney, J. M., Structural Analysis of Laminated Anisotropic Plates,
Winckler, S. J., “Hygrothermally Curvature Stable Laminates with
Tension-Torsion Coupling,” Journal of the American Helicopter
Society, Vol. 30, No. 3, 1985, pp. 56–58.
“chapter_16(new)” — 2005/11/17 — page 623 — #33

17
Cross-over Topics
As helicopter dynamics technology has advanced, so too has the size and scope
of the dynamist’s tool kit. Clear-cut structural-dynamics-related topics have been
addressed in the preceding chapters. However, in this chapter we examine topics
that are not strictly structural dynamics and/or aeroelasticity, but crossover into
and have their impact on these subjects:
Interactions of the rotor/drive system with the engine/fuel control system: The
requirements for higher-bandwidth engine control systems have introduced prob-
lems requiring a blending of those tools typically in the purview of the control
systems dynamist with those of the aeroelastician/structural dynamist.
Aeroelastic optimization: Optimization methodology has matured and is slowly
making its way into standard design methodology. The goal of optimization
methodology is to mechanize the design process, if even only partially, in order to
attack vibration and instability-related problems earlier in the design cycle.
Both of these topics are addressed in this chapter to at least a basic degree so that
the problem areas are well defined, and current solution methods are identified.
Where possible, additional reference material has been identified for more in-depth
study of these topics.
17.1 Interactions of the Rotor Drive and Engine/Fuel

Control Systems
The engine fuel control system of any modern helicopter is typically com-
posed of a variety of servo-controlled subsystems. To a great extent, the details
of these subsystems are dictated by the operating characteristics of the free power
turbine engine powerplant presently used in all medium- to large-sized helicopters.
The detailed dynamic description of any one engine/fuel control system is conse-
quently most meaningfully described by means of a block diagram that can include
any number of gain scheduled parameters, limiters, and other nonlinear features.
In addition to and partly motivated by these characteristics, present strategies
for fuel control system design include a digital implementation of the selected
control system dynamics. All of these factors make it difficult to define a “typical”
engine/fuel control system transfer function. For present purposes, suffice it to say
that the coupled dynamics of the rotor/engine/fuel control system is dominated
by responses in two frequency ranges: a low-frequency mode of operation that
characterizes the responsiveness of the rotor speed to control inputs and the higher-
frequency modes associated with the torsional dynamics of the drive train. These
625
“chapter17(new)” — 2005/10/20 — page 625 — #1

Fig. 17.1 Frequency spectrum of rotorcraft dynamic related phenomena of interest

(after Kuczynski et al.).
features are put into the context of the complete frequency spectrum of interest to
the helicopter dynamist in Fig. 17.1, as extracted from Kuczynski et al.
The design of modern helicopter engine/fuel control systems is driven by
two main considerations: maximizing control system responsiveness in the low-
frequency mode and stabilization of the higher-frequency drive system torsion
modes, which experience has shown to be potentially unstable. Although the
details of the low-frequency operations are important, we will focus on the
higher-frequency modes that impact on the dynamics of the drive train.
17.1.1 Analysis Techniques

Prerequisities. The key to analyzing the combined drive/engine/fuel control
system is to have suitable transfer function representations for each of the major
subsystems. Using basic control system techniques, the component subsystems
are integrated into a total system and analyzed. For the systems that are basi-
cally linear, root-locus techniques would be applicable along with appropriate z
transform methods for systems that use a sampled, computer-controlled implemen-
tation. Other system configurations often lend themselves to Bode plot analysis
techniques, wherein parameter variations are made to ensure maximum control
system responsiveness along with suitable gain and phase margins at the high-
frequency end. Finally, because of the details involved (i.e., nonlinearities) and the
present computational resources available, the total system is often analyzed using
“chapter17(new)” — 2005/10/20 — page 626 — #2

CROSS-OVER TOPICS 627
time-history simulations. These simulations are constructed by combining the dif-

ferential equations inherent in the block diagrams of the transfer functions for the
engine and fuel control systems with a set of time-dependent integro-differential
equations defining the torsional motions of the drive system. The transfer func-
tion block diagram (together with the applicable time constants and gains) for the
engine and fuel control system are normally provided by the engine manufacturer.
Dynamic subsystems. The detailed analysis of the composite control sys-

tem with digital implementation is beyond the scope of this text, and the reader
is referred to any of several excellent texts on computer-controlled systems (e.g.,
Astrom and Wittenmark) for appropriate analysis techniques. The intent here is to
present the basic dynamic characteristics needed to define the system in anticipa-
tion of the application of the more exact theory and design practices appropriate
to computer-controlled systems. Additionally, because filters play a significant
role in these systems, some introductory material is presented on the characteris-
tics and use of filters. For present purposes, Fig. 17.2 defines the major dynamic
subsystems comprising the combined rotor drive and engine/fuel control system.
Control system. The up-front control system, which produces a fuel flow com-
mand in response to a throttle input, can be thought of as a classic regulator and can
typically be implemented in a variety of ways. At the heart of this subsystem would
be some form of servo-controlled flow valve, some form of transducer to convert
throttle position to a command signal, and a servo amplifier. The details of such
an arrangement is subject to the controller specifications and design assumptions
made by the control system designer. For present purposes it is sufficient to assume
that the flow valve could be represented by a first-order lag and that the remainder
of the controller could be assumed to be of the form of a standard position, integral,
derivative (PID) controller. Thus, let us denote the throttle position as X0 , the fuel
Fig. 17.2 Flow block diagram for the composite engine, drive, and control system.
“chapter17(new)” — 2005/10/20 — page 627 — #3

flow valve command as X1 , and the commanded fuel flow as X2 . Then the Laplace
forms of these two portions of the control system could be written in the following
manner:

1 TD p
X1 = KP 1 + + X0 (17.1)
TI p 1 + TD p/N
KV
X2 = X1 (17.2)
1 + TV p
where KP is the proportional gain; TI is the integral time constant; and TD is the
derivative time constant. In the derivative portion of the controller, there is typically
a filter, with time constant TD /N, with N residing in a typical range of 3–10. For
the simple first lag KV and TV are the appropriate gain and time constants. Also,
as shown in Fig. 17.1 a notch filter is often included for system stabilization. The
characteristics and role of this filter are addressed in a subsequent section.
Engine/fuel metering system. The dynamics of the engine and its attendant
fuel control system are describable by means of the transfer function provided by
the engine manufacturer. In general, the engine/fuel control dynamic system is
excited by the pilot throttle control (both directly and indirectly through correlator
coupling with the collective pitch angle) and by the turbine speed feedback loop
(governor).
Drive system and rotor lead-lag dynamics. The drive system, as we have seen
earlier, consists of the engine power turbine spool, rotors and/or propellers, and
all interconnecting shafts, inertias, and gearing. Sources of damping that must
be considered include the turbine itself, all aerodynamic elements (rotors and
propellers), and, in the case of articulated rotors, the lead-lag dampers. Excitation
of the subsystem consists of the controls acting on the aerodynamic elements (blade
pitch angle, etc.) and the gas torque acting on the power turbine disk of the engine.
System coupling. Intersystem coupling exists by virtue of the blade (collec-

tive) pitch angle, which is often correlated with the throttle so that both the control
system and aeromechanical dynamics receive excitations. Excitation of the total
system caused by blade pitch angles thus consists of gas torque acting on the power
turbine disk of the engine together with the torques generated by the aerodynamic
lead-lag forces on the blades.
17.1.2 Stabilization of the Combined Torsion Drive System

The aforementioned two basic objectives of a successful design of the combined
torsion drive system are both speedy responsiveness of the control system to deliver
torques and stabilization of any higher-frequency drive system torsion modes.
Unfortunately, the means of achieving the former (i.e., to operate at as high a gain
in the control system as possible) are detrimental to the latter. Two basic methods
that have evolved for providing the stabilization, without resorting to unacceptably
low gain values, are 1) to decouple the higher-frequency from the low-frequency
“chapter17(new)” — 2005/10/20 — page 628 — #4

Fig. 17.3 Implementation of artificial damping in the torsion system through the
engine/fuel metering system.
response and 2) to provide some form of damping. The decoupling is typically

accomplished in the control system by means of a suitable notch filter tuned to
the frequency of the higher-frequency torsion mode to be stabilized. Addition
of damping to the system can be accomplished mechanically, using the lead-lag
dampers and damped couplings in the drive system, as well as artificially in the
control system. The use of the lead-lag dampers is problematic, however, in that
the loss of one of the dampers must not result in an unstable system and the use
of blade-to-blade dampers removes the lead-lag dampers from the torsion system.
These two scenarios for stabilization are indicated, respectively, in Figs. 17.2
and 17.3.
In the latter case of the introduction of artificial damping electronically, the
key is the use of a high-pass filter to suppress the low frequencies of shaft speed
associated with normal engine usage. The choice of location as to where the speed
sensor should be placed must be driven by identification of the antinodes of the
torsional mode to be suppressed. In both stabilization scenarios, the key is the use
of appropriate filtering. The following section presents basic material for analyzing
and designing filters, both analog and digital.
17.1.3 Basic Filter Analysis

The technology of filters in dynamic systems is quite extensive, and only a
basic exposition of the subject is presented herein. The reader is referred to any of
a wide selection of texts on the subject. The material presented in this subsection
is gleaned principally from Johnson and Antoniou. Filters can be implemented
“chapter17(new)” — 2005/10/20 — page 629 — #5

either in analog or digital form. Although originally defined in terms of networks

of passive electronic elements (i.e., resistors, capacitors, and inductors), modern
analog filter design makes use of operational amplifier implementation together
with precision resistors and capacitors. Digital filters must be implemented using
sampling techniques either in software in a control computer or in dedicated
hardware.
Low-pass filter. The starting point for all filter design is the low-pass filter,
whose loss or attenuation specifications are shown in Fig. 17.4. As indicated in
Fig. 17.4, there is a range of frequencies from zero to ωp , where the loss A(ω) is
no greater than Ap , which defines the “pass” band. Similarly, there is a range of
frequencies from ωa to ∞ that defines an “attenuation” or “stop” band, wherein
the loss must be no less than Aa . The frequency band, between ωp and ωa defines
a transition band that would ideally be equal to zero, but for all practical filters
is finite and to be minimized. The transfer function that relates input to output,
the gain H(ω), is a complex-valued function that is the reciprocal of the loss {i.e.,
H(ω) = [A(ω)]−1 }.
Several basic low-pass filters have evolved for the low-pass filter: the
Butterworth, the Chebyshev, the Inverse Chebyshev, the Bessel, the Cauer, and the
elliptic. Of these, the Butterworth is the most popular design used in circuits. All
of these filters produce a low-pass filtering and can be defined mathematically with
transfer functions that are rational functions with numerators and denominators
that are each specific polynomials of the Laplace transform variable. The polyno-
mials for each of these transfer functions are variable in degree n, which thereby
defines the “order” of the filter. (See Baker for numerical values of the coeffi-
cients defining various orders of Buttersorth, Chebyshev, and Bessel filters.) Each
Fig. 17.4 Practical loss specifications for the low-pass filter.
“chapter17(new)” — 2005/10/20 — page 630 — #6

of these filters has unique characteristics relating to either flatness of magnitude

in the pass band, rate of attenuation in the transition band, and/or constancy of
time (i.e., group) delay, defined as the rate of phase angle change with frequency.
Figure 17.5 shows typical loss and amplitude response characteristics for three of
Fig. 17.5 (a) Typical loss and (b) amplitude response characteristics for the
Butterworth, Chebyshev, and elliptic low-pass filters.
“chapter17(new)” — 2005/10/20 — page 631 — #7

these filter types. Each filter has been configured to have the same loss specifica-
tions and was selected because it can easily be configured to match the end of the
pass band condition (see Antoniou).
Satisfying the specification for the start of the attenuation band (i.e., ωa and
Aa ) is generally achieved by selecting an appropriate order of the filter. Antoniou
presents inequalities for determining the appropriate orders for these filters to meet
these conditions.
Frequency transformations. Filters are also classified with regard to the

function they are to serve. Here there are four basic types: low-pass, high-pass,
band-pass, and band-stop. (The band-stop is essentially the notch filter alluded to
already.) The reason that the low-pass filter receives the most attention in the litera-
ture is that it can be used as a building block for constructing the other three types.
This is achieved by virtue of appropriate frequency transformations. For these
transformations, we first start with a normalized low-pass filter (any type will do),
whose transfer function is defined as Hn (P), and amplitude response |Hn (j)|, has
a cutoff frequency (frequency in the middle of the transition frequency range) equal
to unity. The idea is to “map” the low-pass filter’s pass and attenuate frequency
ranges into the similar ranges for the required derived filter. Thus, if the frequency
range for the low-pass filter is defined by the frequency , the frequency range
for the derived filter will be defined by the frequency ω, which will have a specific
functional relationship to :
P = F( p) and j = F( jω) (17.3)
so that
H( p) = Hn [F( p)] (17.4)
and, therefore
|H( jω)| = |Hn ( j)| = |Hn [F( jω)]| (17.5)
Figure 17.6 summarizes the transformations required to transform the low-pass

filter to each of the three other basic functional types. An example of the use of the
low-pass filter to synthesize band-stop filters is presented in Fig. 17.7. This figure
compares the notch filter response characteristics for a standard biquadratic notch
filter with those for two Butterworth “derived” filters.
The biquadratic filter has the usual transfer function expressed as:
eout p2 + ω02
= 2 (17.6)
ein p + 2ζD ω0 p + ω02
“chapter17(new)” — 2005/10/20 — page 632 — #8

Fig. 17.6 Frequency transformations required for converting a low-pass filter to

other derived filters.
17.1.4 Digital Implementations of Filters

Modern practice is to design control systems within a digital framework. All
of the basic filters considered in the preceding sections can be implemented as
analog filters using operational amplifiers whose various gains can be digitally
scheduled. However, an all-digital implementation of the filtering is a worthy
objective because of the superior stability, repeatability, and accuracy of digital
systems. Also, a digital implementation provides maximum commonality with
the rest of the control system and can be more easily programmed to achieve
gain scheduling. There are important considerations, however, in converting an
analog design to a digital one. The most important of these relates to the fact that
digital systems operate in a sampled-time environment, rather than a continuous
one. Special tools are therefore required. A detailed examination of the design
of digital filters is clearly beyond the intended scope of the present text, and the
interested reader is referred to Antoniou. Only a brief exposition of the key ideas
is presented in this section.
The design of digital filters can proceed in alternate valid pathways, and the
methodology presented herein is not exhaustive. One such valid way is to take a
practical analog design and convert it to a digital one. Care must be taken to ensure
“chapter17(new)” — 2005/10/20 — page 633 — #9

Fig. 17.7 Notch filter frequency responses a) amplitude, b) phase angle, normalized
with respect to the notch frequency ω0 .
that the sampling rate in the system is high enough to preclude aliasing problems
(see Chapter 8). We can assume here that a practical choice of sampling rate has
already been made for the digital control system and that adequate anti-aliasing
filters have been incorporated. The first problem in the conversion of an analog
“chapter17(new)” — 2005/10/20 — page 634 — #10

filter to a digital one is the fact that the analog filter is mathematically defined
using the Laplace transform, which is not appropriate to a sampled time system.
For this case the z transform is the appropriate tool.
The z transform. Stated simply, the z transform of a sampled function f is

given by
∞

F(z) = f (nT )z−n (17.7)
n=−∞
where T is the sampling interval and z is a complex-valued variable. In simplified

notation this equation is often written as
F(z) = Zf (nT ) (17.8)
Some of the properties that make the z transform useful are given (without
proof) as follows:
1) It is a linear operation:
aF(z) + bG(z) = Z[af (nT ) + bg(nT )] (17.9)
2) Translation:
Zf (nT + mT ) = zm F(z) (17.10)
3) Complex differentiation:
dF(z)
Z[nTf (nT )] = −Tz (17.11)
dz
4) Convolution:
∞
∞

Z f (kT )g(nT − kT ) = Z f (nT − kT )g(kT ) = F(z)G(z) (17.12)
k=−∞ k=−∞
Just as with Laplace transforms a set of standard z transforms exists for a selection
of typical functions (see Antoniou).
Digital filter elements. The most general description of a digital filter is that
it is recursive, which means that the response of the filter is a function of the
excitation as well as the response itself:

N
N
y(nT ) = ai x(nT − iT ) − bi y(nT − iT ) (17.13)
i=0 i=1
The nonrecursive filter is the special case of the recursive type, wherein the
response is not a function of itself. The value of N defines the order of the filter.
“chapter17(new)” — 2005/10/20 — page 635 — #11

In the case of analog filters, the elements used to construct them include resisters,
capacitors, and inductors for passive systems and, for active filters, operational
amplifiers and selectable resistance and capacitance networks. In the case of digital
filters, there are only three (mathematically defined) elements: the unit delay, the
adder, and the multiplier. All of these are defined in terms of an input variable(s)
x(nT ) and an output variable y(nT ). The implementation of these can be performed
via a shift register and/or appropriate NAND or NOR gates. Mathematically, these
elements are defined as follows:
1) Unit delay:
y(nT ) = x(nT − T ) (17.14)
which can be generalized by means of the shift operator, E:
E −r x(nT ) = x(nT − rT ) (17.15)
The z transform of the unit delay is given as
Y (z) = z−1 X(z) (17.16)
2) Adder:

k
y(nT ) = xi (nT ) (17.17)
i=1
whose z transform is given by

k
Y (z) = Xi (z) (17.18)
i=1
3) Multiplier:
y(nT ) = mx(nT ) (17.19)
whose z transform is given by
Y (z) = mX(z) (17.20)
These basic elements are illustrated in Table 17.1. The transfer function relat-
ing x(nT ) to y(nT ) for the recursive filter [Eq. (17.13)] can be obtained by z
transforming this equation:

N
N
Zy(nT ) = ai z−i Zx(nT ) − bi z−i Zy(nT ) (17.21)
i=0 i=1
or

N
ai zN−i
Y (z) i=0
= H(z) = (17.22)
X(z)
N
zN + bi zN−i
i=1
“chapter17(new)” — 2005/10/20 — page 636 — #12

Table 17.1 Basic digital filter elements
Bilinear transformation. In this section we tie together the characterizations

defined for the analog filter with those needed to produce a digital filter that approxi-
mates the performance of the analog filter. Based on the impulsive response of an
integrator, a difference equation can be written describing the response caused by
excitation over a time step:
T
y(nT ) − y(nT − T ) = [x(nT − T ) + x(nT )] (17.23)
2
which, upon applying the z transformation, yields the following transfer function:
Y (z) T z+1
HI (z) = = (17.24)
X(z) 2 z−1
Because the (Laplace transform) transfer function for an integrator is given by

(1/s), the following relationship between the two transform variables is inferred:
2 z−1
s= (17.25)
T z+1
“chapter17(new)” — 2005/10/20 — page 637 — #13

Now, if a given analog filter is characterized by a (Laplace transform) transform

function,

N
ai sN−i
i=0
HA (s) = (17.26)

N
sN + bi sN−i
i=1
then an approximation to a derived digital (z transform) transfer function is given by
HD (z) = HA (s)| 2 z−1 (17.27)

s= T z+1
Thus, the derived digital filter having this transfer function will have approxi-
mately the same time-domain response as the given analog filter. Discrepancies will
arise because of the choice of sampling rate and the effects of digital quantization
(effects of fixed digital word length).
Warping effect. Because we are concerned only with sinusoidal excitations

and responses, the Laplace transform variable will lie on the imaginary axis (i.e.,
s = jω). The bilinear-transformation defined by Eq. (17.25) will then map the z
transform variable onto the unit circle. This can be expressed as
HD (e jT ) = HA ( jω) (17.28)
where is the excitation frequency for the digital filter and ω is the corresponding
frequency for the analog filter. The mapping of jω on the imaginary axis to j on
the unit circle can be expressed with the help of Eq. (17.23) as
2 T
ω= tan (17.29)
T 2
For T sufficiently small (i.e., <0.3), the two frequencies are approximately
equal. However, when this term is not small, a warping of the frequencies will
result. That is, the filter performance of the analog and that of the derived digital
filter will be similar, but the frequency spectrum for the digital filter will be warped
relative to the analog filter at the higher frequencies. This can be seen by inverting
Eq. (17.29):
2 ωT
= tan−1 (17.30)
T 2
One method for overcoming this problem is by designing the analog filter to
have prewarps of the pass-band and stop-band edge frequencies ωi so that the
correct edge frequencies will occur, as specified, in the digital filter i . This is
accomplished with the use of Eq. (17.29):
2 i T
ωi = tan (17.31)
T 2
“chapter17(new)” — 2005/10/20 — page 638 — #14

Filter realization. In the case of analog filter realization, the Laplace trans-
form transfer function can be directly used to define the appropriate differential
equations, which can then be implemented either passively or actively. In the case
of digital filters, the starting point must be the z-transformed transfer function,
which will result from the use of the bilinear transformation and prewarping, as
per Eqs. (17.27) and (17.31). The one difficulty is that the z-transformed transfer
function will generally be a rational function, with polynomials in z in both the
numerator and denominator. The final filter design must have a realization consist-
ing of the basic digital filter elements, that is, unit delays, adders, and multipliers.
As it turns out, there is no unique methodology for doing this (see Antoniou). What
is presented herein is a variant of the direct method. This method first decomposes
the transfer function into two simpler transfer functions representing, respectively,
the numerator and denominator. Thus,
Y (z) N(z) N(z)
H(z) = = = (17.32)
X(z) D(z) 1 + D (z)
where

N
N(z) = ai z−i (17.33)
i=0
and

N
D (z) = bi z−i (17.34)
i=1
The z transform of the response Y (z) can be then written as
Y (z) = U1 (z) + U2 (z) (17.35)
where
U1 (z) = N(z)X(z) (17.36)
and
U2 (z) = −D (z)Y (z) (17.37)
This decomposition is illustrated in Fig. 17.8. Each of the simpler transfer functions
shown in Fig. 17.8 then represents a polynomial in (1/z). The second part of the
direct method is then to implement the realizations of each of the polynomials.
To this end, let us consider the numerator polynomial and rewrite Eq. (17.32) as
follows:
U1 (z) = [a0 + z−1 N1 (z)]X(z) (17.38)
“chapter17(new)” — 2005/10/20 — page 639 — #15

Fig. 17.8 Decomposition of H(z) into two simpler transfer functions representing the
numerator and denominator.
Fig. 17.9 Basic unit for realizing a polynomial transfer function (after Antoniou).
Fig. 17.10 One of several possible realizations of a polynomial transfer function N(z)
(after Antoniou).
“chapter17(new)” — 2005/10/20 — page 640 — #16

Fig. 17.11 Realization of the complete transfer function H(z) (after Antoniou).
where

N
N1 (z) = ai z−i+1 = a1 + z−1 N2 (z)X(z) (17.39a)
i=1

N
N2 (z) = ai z−i+2 = · · · (17.39b)
i=2
The basic unit for realization, as defined by Eq. (17.38), is illustrated in Fig. 17.9.
The complete realization of the polynomial N(z) is inferred by Eqs. (17.39a)
and (17.39b), wherein the basic unit is cascaded. This is illustrated in Fig. 17.10.
The realization of the complete transfer function then first realizes the denominator
function −D (z) to realize the U2 (z) transfer function (using the technique outlined
previously) and then combines them, as shown in Fig. 17.11.
Figures 17.9–17.11 illustrate only one of several possible implementations
using unit delays, adders, and multipliers that can be programmed within a digital
environment to achieve a specified z transform for any given filter.
17.1.5 FADEC Systems

The design of computer-implemented control systems for the engine/fuel sys-
tem is becoming more comprehensive and increasingly tailored to the vagaries
of critical flight conditions. Such computer control systems are denoted by the
acronym of FADEC, which stands for full authority digital electronic (or engine)
control. Besides its principal functions, the FADEC system would incorporate all
of the drive system stabilization features with appropriate filtering described in
the preceding sections. It would do considerably more, however. Coyle presents
an in-depth exposition of the helicopter turbine engine and its governing systems
“chapter17(new)” — 2005/10/20 — page 641 — #17

Fig. 17.12 Typical interfaces and functions of the FADEC computer (diagram
courtesy of Shawn Coyle).
including the principal features and practical benefits of the FADEC system. As
shown in Fig. 17.12, the FADEC is integrated into the engine/fuel control system
using a variety of sensors and performs several calculations regarding the meter-
ing of the fuel, as well as detection of imminent and actual engine failure and the
initiation of remedial action.
The principal benefit of the FADEC is that it makes the rotorcraft easier to fly,
as it relieves the pilot(s) from having to spend time monitoring the several engine
gauges. Equally important is that it keeps the wear and tear on the engine(s) to
a minimum and thereby reduces maintenance costs. FADEC systems are now
standard equipment on nearly all rotorcraft turbine engines and on at least one
helicopter reciprocating engine.
17.2 Aeroelastic Optimization

Aeroelastic optimization is the application of general optimization techniques
to the design process of any part of the rotorcraft whereby some desirable aero-
elastic aspect of the design is either maximized or minimized. This process falls
under the well-established category of constrained optimization. The techniques
for constrained optimization are now quite mature (see Vanderplaats), and efficient
computer implementations (e.g., the CONMIN computer code) are available for
a variety of computer platforms. The material presented herein will touch only
“chapter17(new)” — 2005/10/20 — page 642 — #18

briefly on the optimization process itself and then show how it can be applied to
rotorcraft structural dynamics and aeroelasticity.
17.2.1 Constrained Optimization Problem

In any optimization problem there are three main considerations to be addressed:
1) the definition of the design (modification) parameters, 2) the definition of a suit-
able cost function J, which is to be optimized (either maximized or minimized),
and 3) the identification and mathematical definition of the constraints. Addition-
ally, because the optimization process is of necessity an iterative one, it must start
with an initial “guess” for a set of design parameters that satisfies the constraints.
The condition that a design satisfies all of the constraints is defined by stating that
the vector of design parameters lies within the feasible design space. These ideas
are illustrated in Fig. 17.13.
Figure 17.13 shows, for a simple two-parameter (i.e., D1 and D2 ) design space,
that the feasible design space is bounded first by side constraints, which are typi-
cally upper and lower limits that constrain each of the design parameters to realistic
values. Further bounding of the feasible design space arises from both equality and
inequality constraints. The figure shows two inequality constraints G1 and G2 . Also
indicated in the figure are the contour curves defining the functional variation of the
cost function J with the two design parameters. The absolute optimization point
for the function is shown as outside the feasible design space and thus unreachable
by any combination of the design parameters. The optimized design point is the
one that simultaneously lies within or on a boundary of the feasible design space
Fig. 17.13 Illustration of a hypothetical simple feasible design space and the
optimization process.
“chapter17(new)” — 2005/10/20 — page 643 — #19

and provides the best value of the cost function. This defines the constrained opti-
mization (minimum) point. Generally, the constrained optimization process will
result in a solution wherein the design parameters reside somewhere on a constraint
boundary.
Some of the challenges in formulating a practical optimization program relate to
inherent numerical and discretization issues and to the need for trend information.
As relates to the hypothetical example presented in Fig. 17.12, depending on where
one starts in the design space, the problem is one of where to go to next and by how
much. To a large degree, this depends on knowing gradient information of both
the cost function and the constraints. Although the trend of information might be
available analytically, generally it is not, and numerical evaluations must be made.
Numerical evaluations of gradients (i.e., partial derivatives) will depend on the size
of the increments used for approximating the partial derivatives. Also, depending
on the starting point and the nonlinearity of the cost function and/or constraints,
the increments used by the optimization code for iteratively wending its way to the
optimization point can become erratic. Various techniques have been devised for
avoiding these erratic iterations and making the iteration converge more directly
to the optimization point. All of these issues have been addressed more or less by
several analysts. Several optimization resources are now available, and a few are
included in the list of references.
17.2.2 Application to Rotorcraft Structural Dynamics

and Aeroelasticity
Application of the constrained minimization process to rotorcraft has mostly
been with regard to vibration minimization and/or stabilization of any number of
aeroelastic and aeromechanical instabilities. The material to follow shows how the
various elements of constrained minimization can be defined and applied.
Design parameters. The set of design parameters pm for the rotor-blade

problem could include a variety of elastomechanical and aerodynamic descriptors.
One problem of defining a practical set of design parameters is that of keeping
the number down to a sufficiently practical size. Also, design parameters must
be selected that are truly independent and not potentially contradictory. Thus,
gross blade distributions, such as mass and stiffness must be physically consistent.
A safe approach would be to deal with physical dimensions of blade components
comprising the various distributions.
Cost function. For the vibration minimization problem several appropriate

cost functions could be defined, depending on the ability to define the func-
tion quantitatively. Because we wish to minimize the components of vibration
in an essentially narrowband sinusoidal process, we could define the cost function
in terms of the cosine and sine (Fourier) components of some vibratory quantity
Q at a point in the structure. Thus, alternative formulations for the cost function J
“chapter17(new)” — 2005/10/20 — page 644 — #20

could be

3
J= (Qj2nc + Qj2ns )Wj where n = kb, k = 1, 2, . . . (17.40)
j=1
or

3
J= (|Qjnc | + |Qjns |)Wj again, where n = kb, k = 1, 2, . . . (17.41)
j=1
and where Wj is a set of arbitrary weighting factors that must be (user) selected
to reflect the priorities assigned to the various components. Ideally, one would
want to use some “bottom-line” description of the actual vibratory accelerations
at some critical point (or points) in the aircraft structure for the Fourier compo-
nents of Q. However, the ability to analyze the entire aircraft, including the effects
of rotor-fuselage coupling, would require a substantial formulation and computa-
tional chore. Furthermore, as we have seen, such a formulation is not presently
a reliably accurate basis. An alternate selection of quantity for the basis of the
cost function might be the various components of the hub shears and/or moments
H(x, y, z) and M(x, y, z) , respectively. For the vibration minimization problem several
appropriate cost functions could be defined, depending on the ability to accurately
formulate them quantitatively. An even simpler formulation might be to identify
and formulate appropriate dynamic magnification factors for a selection of blade
bending modes.
For the stabilization of aeroelastic phenomena, the cost function could take the
form of a combination of stability descriptors for any number of phenomena, again
with appropriate weighting functions. Thus, if we were to require that both ground
resonance and air resonance instabilities be stabilized or kept to stable levels, the
cost function could take the following form:
  2 

 

 (GR)
(SC)
σm 
J= Wm ζreq + 

 (SC) 2 (SC) 2 

m [σ m ] + [ω ] m
GroundResonance
  2 

 

 (AR)
(SC)
σn 
+ Wm ζreq +  (17.42)

 (SC) 2 
(SC) 2 
n [σ n ] + [ω ] n
AirResonance
where ζreq would be a required damping ratio to be applied to all stability critical
modes and the various stability critical eigenvalue components are denoted as
(. . .) (SC) .
Constraints. Constraints must be applied to two different groups of quan-

tities: the design parameters themselves and those quantities that define blade
response.
“chapter17(new)” — 2005/10/20 — page 645 — #21

Constraints on the design parameters. The constraints on the design para-

meters would consist of more or less generic limitations on any set of detailed
design parameters:
m ≤ pm ≤ pm
p(L) m = 1, 2, . . . , M
(U)
(17.43)
which specifies lower and upper bounds on each parameter’s range of variability
and
Gn ( pm ) ≥ 0 n = 1, 2, . . . , N (17.44)
which specifies that there shall be no interference between the design parameters.
Constraints placed on supplementary responses. For those cases wherein the

cost function is to be a vibration-related quantity, the imposition of system stabi-
lization can be achieved by means of constraints placed on the dynamic descriptors.
Besides achieving a minimum vibration solution, the designer of the rotor blade
must also achieve one that is aeroelastically stable. The vibration minimization
process can therefore be augmented at this point by constraining the blade modal
damping to be positive and the natural frequencies to have values separated from
integral harmonic values:
(L)
ζk ζk − 1 ≥ 0 k = 1, 2, . . . , K (17.45)
which specifies the lower bounds on aeroelastic stability for all of the coupled
aeroelastic modes and
(L)
ωk2 ωk2 − 1 ≥ 0
(U) k = 1, 2, . . . , K (17.46)
1 − ωk2 ωk2 ≥ 0
which specifies the frequency placements within bounds and away from nearest
integral harmonics.
Iteration process. Beginning with some initial design (one whose design
parameters lie within the feasible design space, i.e., those that actually do satisfy the
constraints), the optimization process can then iterate to those design parameters
that minimize the cost function J. Typically, this minimization process ultimately
must be capable of generating and using “trend information” regarding both the
constraints and the cost function. This trend information takes the form of various
gradient or partial derivative matrices. These expressions can be calculated by any
of a number of methods available, as appropriate to the specific problem. The
choice of methodology for calculating these matrices, as well as using them to
effect iterative changes in the right direction to the optimum value, is subject to a
variety of schemes and implementations. A straightforward approach would be to
evaluate them using numerical differentiation (as is done in the CONMIN code).
There is a tradeoff to make, however, in that small increments needed for the
differentiation process are conducive to accuracy, but can lead to increased CPU
time to arrive at an optimization solution.
“chapter17(new)” — 2005/10/20 — page 646 — #22

17.2.3 Practical Implementation Scheme

Figure 17.14 illustrates the various elements that would comprise an actual
practical use of aeroelastic optimization within a design environment.
This figure stresses that the primary element in the optimization process would
be the design engineer(s). Design is not so much a deterministic science, as a
creative process that must allow for innovative thought. All of the results of the
optimization process must be evaluated by the human element to determine feasi-
bility and/or update existing constraints or impose new ones. The figure also shows
that within the optimization program it is desirable (from an economical CPU time
standpoint) to incorporate a simplified engineering model of the rotor aeroelas-
ticity for a close “ballpark” optimization result. The results of the optimization
program would then be verified by using a more comprehensive aeroelastic code.
The end result of this calculation would be the blade loads, stability descriptors,
and an assessment of some aspect of the fuselage vibrations. Before the design
could proceed to fabrication and test, the design engineer would pass judgement
of the optimized design for practicality.
Aeroelastic optimization has received increasingly more positive interest
because of the emergence of increasingly more powerful (and yet less expensive)
computer resources and the increasingly more user-friendly implementations of the
Fig. 17.14 Practical implementation scheme for rotorcraft design with aeroelastic
optimization.
“chapter17(new)” — 2005/10/20 — page 647 — #23

optimization schemes. Thus, it is to be expected that computer-aided (optimiza-

tion) design systems will be increasingly accepted in the design room. Several
researchers have applied variants of this process to the rotorcraft design process,
and some of their published results are listed in the references.
References
Section
17.1 Alwang, J. R., and Skarvan, C. A., “Engine Control Stabilizing
Compensation - Testing & Optimization,” Journal of the American
Antoniou, A., Digital Filters: Analysis and Design, McGraw–Hill
Book Company, New York, 1979.
Astrom, K. J., and Wittenmark, B., Computer Controlled Systems,
Theory and Design, (Information and System Sciences Series),
Prentice–Hall, Englewood Cliffs, NJ, 1984.
Baker, B. C., “Anti-Aliasing, Analog Filters for Data Acquisition
Systems,” Microchip Application Note AN699, 1999.
Corliss, L. D., “The Effects of Engine and Height-Control Character-
istics on Helicopter Handling Qualities,” Journal of the American
Coyle, S., Cyclic & Collective: Second Edition of the Art and Sci-
ence of Flying Helicopters, Helobooks, a Division of Mojave Books
Limited, Mojave, CA, 2004.
Greensite, A. L., Elements of Modern Control Theory, Spartan Books,
New York, 1970.
Howlett, J. J., Morrison, T., and Zagranski, R. D., “Adaptive Fuel
Control for Helicopter Applications,” Journal of the American
Johnson, D. E., Introduction to Filter Theory, Prentice–Hall Inc.,
Englewood Cliffs, NJ, 1976.
Kuczynski, W. A., Cooper, D. E., Twomey, W. J., and Howlett,
J. J., “The Influence of Engine/Fuel Control Design on Heli-
copter Dynamics and Handling Qualities,” Journal of the American
17.2 Davis, M. W., and Weller, W. H., “Application of Design Optimization
Techniques to Rotor Dynamic Problems,” Journal of the American
Kresselmeier, A., and Steinhauser, R., “Systematic Control Design
by Optimizing a Vector Performance Index,” Proceedings of IFAC
Symposioum on Computer Aided Design of Control Systems, 1979,
pp. 113–117.
“chapter17(new)” — 2005/10/20 — page 648 — #24

Haftka, R. T., Gurdal, Z., and Kamat, M. P., Elements of Struc-
tural Optimization, Kluwer Academic Publishers, Dordrecht, the
Netherlands, 1990.
McCarthy, T. R., and Chattopadhyay, A., “A Coupled Rotor/Wing Opti-
mization Procedure for High Speed Tilt-Rotor Aircraft,” Journal of
Vanderplaats, G. N., Numerical Optimization Techniques for Engi-
neering Design, with Applications, McGraw-Hill, New York,
1984.
Vanerplaats, G. N., “CONMIN - a Fortran Program for Constrained
Function Minimization, User’s Manual,” NASA TMX-62282, 1987.
Weller, W. H., and Davis, M. W., “Wind Tunnel Tests of Helicopter
Blade Designs Optimized for Minimum Vibration,” Journal of the
Young, D. K., and Tarzanin, R. J. (Jr.), “Structural Optimization and
Mach Scale Test Validation for a Low Vibration Rotor,” Proceed-
ings of the 47th Annual National Forum of the American Helicopter
Society, May, 1991.
“chapter17(new)” — 2005/10/20 — page 649 — #25

18
Concluding Thoughts
18.1 Key Historical Milestones

This text was written in part by gleaning many of the contributions of other
researchers and dynamists. A complete perspective of all significant contribu-
tions in the field, however, would surely create information overload and obscure
those contributions that can truly be called seminal, and without which this text
would have been impossible to write. These contributions are briefly described
and presented in approximate chronological order.
18.1.1 High-Level Languages

The first high-level language to be created was FORTRAN, invented by John
Backus and released commercially in 1957. This language allowed the direct usage
of actual syntactical formulas and human sounding words. This freed the dynamist
not only to formulate his equations, but also to combine the definition of the
dynamic problem along with the mathematical solution to produce engineering
relevant results. The use of high-level languages spawned the many (at times quite
comprehensive) analyses that constitute the present literature of this technology.
Following the success of FORTRAN, other high-level languages that emerged
were Ada, Algol, BASIC, COBOL, C, C++, LISP, Pascal, PL-1, and Prolog.
18.1.2 Early Key Analyses

The complete literature of rotary wing structural dynamics and aeroelastic anal-
yses is now well beyond what is presentable in any one book’s bibliography. Indeed,
many well-formulated contributions currently exist and more continue to find their
way into the literature. However, two analyses especially, that addressed actual
catastrophic instabilities observed in full-scale aircraft need to be put into histori-
cal perspective: ground resonance and propeller-nacelle flutter. The basic dynamic
equations describing ground resonance, as well as both an insightful statement of
the physics involved and the introduction of pertinent solution techniques, were
presented by Coleman and Feingold in 1958. Although several aircraft were totally
damaged by ground resonance, no loss of life was involved.
The case of propeller-nacelle flutter was not so fortunate; aircraft were lost in
flight with substantial loss of life. It is unfortunate that although the physical aspects
of the instability were identified earlier, knowledge of the instability passed into
651
“chapter_18(new)” — 2005/10/20 — page 651 — #1

obscurity. At that time, the instability was deemed to be improbable for the early
class of piston engine, propeller-driven aircraft then in existence. The instability
resurfaced in fact with the advent of turbine powered aircraft with high inertia
propellers. Identification of the instability was, of course, difficult with the break-
up of the stricken aircraft. It was the work of Houbolt and Reed in 1962 that
correctly inferred that this instability was the culprit and it was subsequently
reintroduced into the dynamics literature.
18.1.3 Credible Blade Equations of Motion

By virtue of their relatively high aspect ratio, rotor blades can exhibit a high
degree of coupling between both flatwise and edgewise bending and each with
torsion. Without credible equations for the elasto-mechanical behavior of rotor
blades, no accurate calculations of rotor dynamics would be possible. The first
contribution to the definition of blade dynamics was that of Houbolt and Brooks
in 1958. Subsequently, the work of Hodges and Dowell in 1974, extended the
work of Houbolt and Brooks by rigorously addressing the nonlinearities. This set
of equations has become the starting point for many subsequent investigators.
18.1.4 Fast Fourier Transform

As described in Chapter 8, the modern dynamic signal analyzer is the principal
tool universally used for frequency domain testing. All of these analyzers are based
on the extensive use of the Fourier transform. The need for speed in calculating the
transform is essential for these devices to be practical. The fast Fourier transform
(FFT) was the enabling calculation algorithm. As it turns out, the basic idea for the
FFT was independently re-invented by several analysts, going as far back as Gauss
(circa 1805). However, the first successful development of a high-level language
algorithm for modern computer-oriented applications is credited to Cooley and
Tukey in 1965.
18.1.5 Efficient Matrix Eigenvalue Algorithms

With the emergence of high-level computer languages came the ability to obtain
accurate eigensolutions for large dimensional symmetric matrices. This enabled
the efficient calculation of several modal frequencies and mode shapes for dynamic
systems with relatively many degrees of freedom. Several methods emerged includ-
ing the Givens, Householder, Jacobi, and Lanczos methods. Most of these routines
were documented by Wilkinson and Reinsch in 1971 and later incorporated in the
EISPACK set of programs in 1976.
As is apparent from the previous chapters, the development of rotary wing
structural dynamics and aeroelasticity theory entails the extensive need for obtain-
ing eigenvalue solutions of non-symmetric matrices. Although robust symmetric
matrix eigenvalue solutions have been available for many years, it was not until
1971 that a robust solution for the non-symmetric problem, that could be readily
programmed in a high-level language, was published by Wilkinson and Reisch.
Besides providing accurate solutions to the usual equations of motion in Laplace
transformed matrix form, this solution capability enabled the development of
“chapter_18(new)” — 2005/10/20 — page 652 — #2

CONCLUDING THOUGHTS 653
efficient solutions to the Floquet stability analysis problem as well as the devel-
opment of the multiple-degree-of-freedom frequency-domain stability analysis
methodology (as described in Chapter 9).
18.1.6 Personal Computer

Although the first freely programmable computer (the Z1 computer) was devel-
oped by Zuse in 1936, all of the previously described milestones did not come
together until the mid 1960s when mainframe computers became commercially
available and analysts had increasingly more access to them. Even more significant
was the introduction of personal computers in 1981, together with the imple-
mentation of the DOS operating system on these computers. From that point on
the increase in capability and productivity along with the continual decrease in
cost/calculation “is all history.”
18.1.7 Finite Element Analyses

A major point stressed in this text is that simpler analyses, i.e., those that
demonstrate to the fullest the physics of the vibration and/or stability problem
should always be used as “first cuts” at solving the problem. Indeed, this text is
comprised of such simpler analyses. Eventually, the solution of the problem will
progress to the next step wherein more comprehensive analyses, which can improve
the focus and increase the accuracy, can be used. The finite element analysis is
often this next step in the process by virtue of its ability to incorporate a wide range
of details that can approximate “real-world” rotorcraft structures. However, the use
of such analyses often comes with the price of losing perspective and/or spending a
great deal of computer time and yet still missing an important result. Once the point
has been reached to invoke such analyses, a variety of well-developed codes is now
available. Besides NASTRAN, this selection menu now includes a wide variety
of comprehensive and capable codes, such as ALGOR, ANSYS, COSMOS/M,
I_DEAS, and MARC. A comprehensive listing of available finite element codes
is presented by Stewart.
The development of finite element codes began principally with NASTRAN,
which has found almost universal application within the rotorcraft industry.
Because NASTRAN demonstrated leadership in this development, a brief his-
torical review of it is in order. NASTRAN evolved from specifications put forth
by NASA for a general purpose digital program for structural analysis in 1965.
The first release of NASTRAN was in 1970. Since then, the NASTRAN code has
continued to grow to include a wide variety of analysis capabilities. Only those
capabilities directly impacting on rotorcraft are identified herein. In 1972 com-
plex eigenvalue analysis and frequency response calculations were introduced. In
1979 the aeroelastic supplement was introduced, thereby enabling accurate sub-
sonic wing flutter calculations to be made. Layered composites modeling was
introduced in 1981. Later at the end of that decade (1989), the Lanczos method
for real eigenvalues was introduced. In 1991 methods for structural optimization
were included. A feature of the NASTRAN code is the ability to tailor the code to
specific user requirements using the DMAP feature. As early as 1984 the DMAP
alteration was applied by Lawrence et al. to allow the calculation of blades in
“chapter_18(new)” — 2005/10/20 — page 653 — #3

a centrifugal force field. This capability is to be included as a regular standard

feature in 2006.
18.2 What’s on the Horizon?

18.2.1 Vibrations
The quest to reduce airframe vibrations to the level producing “jet-smooth” rides
will continue. The use of active blade-appended devices will either mature to the
point of acceptable reliability or be abandoned as a vibration-reduction methodol-
ogy. Active vibration control, as applied to the airframe will continue and become
more viable with the design of more compact, lightweight implementations. With
increased computer resources, finite element analyses should experience increased
accuracy in predicting airframe mobilities and rotor aerodynamic analyses should
also meet with increased accuracy in calculating rotor airloads. Because noise
is a continuing problem area with rotorcraft, the development and integration of
acoustoelasticity methodology with that of standard vibration analyses, to reduce
interior noise, will become practical.
18.2.2 Aeromechanical and Aeroelastic Instabilities

The subject matter of rotary wing structural dynamics and aeroelasticity is quite
mature and it is unlikely that any new generic instabilities will be discovered.
Instead, the existing analyses will be applied to every increasing “real-world”
operational conditions. This ultimately means the inclusion of a wide variety of
nonlinear dynamic components and boundary conditions. To a certain extent the
comprehensiveness of any stability analysis is limited to the number of degrees
of freedom that can be accurately handled with computer eigenvalue routines.
This limitation must be addressed with more efficient numerical routines, or with
analyses that can utilize smaller problem sizes, such as the multiple-degree-of-
freedom frequency-domain methodology described in Chapter 9. Another fall-
out from increased computer resources will be the evolution of more accurate
and representative unsteady airloads for rotorcraft. This will become especially
important as new blade tip designs are devised and the speed envelope of rotorcraft
is expanded.
References
Section
18.1 Bellis, M., Inventors, The History of Computers, available online
at http://inventors.about.com/library/blcoindex.htm (cited June
2005).
Bellis, M., Inventors of the Modern Computer, FORTRAN—The First
Successful High Level Programming Language, available online
at http://inventors.about.com/library/weekly/aa072198.htm (cited
June 2005).
“chapter_18(new)” — 2005/10/20 — page 654 — #4

CONCLUDING THOUGHTS 655
Coleman, R. P., and Feingold, A. M., “Theory of Self-Excited
Mechanical Oscillations of Helicopter Rotors with Hinged Blades,”
NACA Rept. 1351, 1958.
Cooley, J. W., and Tukey, J. W., “An Algorithm for the Machine
Calculation of Complex Fourier Series,” Mathematics of Compu-
tation, Vol. 19, 1965, pp. 297–301.
Hodges, D. H., and Dowell, E. H., “Nonlinear Equations of Motion
for the Elastic Bending and Torsion of Twisted Nonuniform Rotor
Blades,” NASA TN D-7818, 1974.
Houbolt, J. C., and Brooks, G. W., “Differential Equations of Motion
for Combined Flapwise Bending, Chordwise Bending and Torsion
of Twisted Non-uniform Rotor Blades,” NACA Rept. 1346, 1958.
Houbolt, J. C., and Reed, W. H., “Propeller-Nacelle Whirl Flutter,”
Journal of the Aeronautical Sciences, Vol. 29, No. 3, 1962, pp.
333–346.
Lawrence, C., and Kielb, R. E., “Nonlinear Displacement Analysis
of Advanced Propeller Structures Using NASTRAN,” NASA TM-
83737, 1984.
Lawrence, C., Aiello, R. A., Ernst, M. A., and McGee, O. G., “A
NASTRAN Primer for the Analysis of Rotating Flexible Blades,”
NASA TM-89861, 1987.
Smith, B. T. et al., Matrix Eigensystem Routines - EISPACK Guide,
2nd ed., Vol. 6, Lecture Notes in Computer Science, Springer–
Verlag, New York, 1976.
Stewart, C. D., Finite Element Analysis/Method Resources, available
online at http://www.steelynx.net/fea/html (cited July 2005).
Wikipedia, the free encyclopedia, Fast Fourier Transform, available
online at http://en.wikipedia.org/wiki/Fast_Fourier_transform
(cited June 2005).
Wilkinson, J. J., and Reinsch, C., Handbook for Automatic Com-
putation, Vol. II—Linear Algebra, Springer–Verlag, New York,
1971.
“chapter_18(new)” — 2005/10/20 — page 655 — #5

Appendix A:
Glossary of Rotorcraft-Related Terms
Words and Phrases

airfoil = shape of a cross section of a rotor blade
alpha-two hinge = actual hinge arrangement or kinematic coupling of
the blade pitch link geometry so that lead-lag motion
of the blade produces pitch change, as well, at blade
sections
American rotor = main rotor turning counterclockwise when viewed
rotation from above
angle of attack = angle between the relative wind and the chord of the
blade section
antinode point = point in a dynamic system, oscillating at one of
its natural frequencies, where the response is a
maximum
antiresonance = response condition in a dynamic system wherein a
condition small sinusoidal excitation force applied at some
point in the system will cause a zero-amplitude
response at some other point in the system
articulated rotor = rotor system in which individual blades have hinges
for flapping and lead-lag motion
aspect ratio = ratio of the blade radius to the average blade chord
length
bearingless rotor = form of a hingeless rotor wherein the feathering
bearing is replaced by a torsionally soft elastic
member
bifilar absorber = type of pendulum absorber characterized by the
presence of two points of support forming a
parallelogram-like ganged translational motion, but
no rotational motion
blade chord = blade width, that is, local dimension perpendicular
to blade radius
blade damper = device suppressing the lead-lag motion about an
articulated rotor blade hinge utilizing friction,
hydraulic, or elastomeric energy dissipation
camber = upward convexity of the mean line of the airfoil
657
“appendix-a(new)” — 2005/10/20 — page 657 — #1

centrifugal force = force created by the tendency of a body to follow a

straight-line path even while being forced to move in
a curved path, resulting in a force that tends to pull
it away from the center of rotation
collective pitch = the part of a helicopter’s control system wherein all
blades change pitch simultaneously by the same
amount
compressibility = aerodynamic effects acting on portions of the blade
as a result of the closeness of air velocities to the
speed of sound
coning = constant value of upward flapping of a blade caused
by lift and centrifugal force
control axis plane = plane of the rotor wherein there is no cyclic feathering
(i.e., the plane of the swash plate)
Coriolis force = force created in a rotating body tangential to the rota-
tion by the tendency of the body to maintain angular
momentum when at the same time some of its mass
elements are being moved radially
critical damping = ratio of damping in a dynamic system to that required
ratio for periodic motion to become aperiodic
cyclic pitch = part of a helicopter’s control system wherein the
blades change pitch in a sinusoidal fashion as they
traverse around the rotor azimuth, each blade having
the same amount of such pitch change when it reaches
any given rotor azimuth
damping = suppression of translational or rotary velocities in a
blade, structural component, or completely coupled
aircraft
delta-three hinge = actual hinge arrangement or kinematic coupling of
the blade pitch link geometry so that flapping motion
of the blade produces pitch change, as well, at blade
sections
disk loading = ratio of the gross weight of the rotorcraft to the area
of the lifting rotor disk(s)
dynamic stall = aerodynamic stalling of an airfoil when the airfoil
enters stall with significant motion, usually accom-
panied by momentary increases in lift and moment
caused by shedding of vorticity from the leading edge
feathering = rotation of the blade about its spanwise axis to change
its pitch
flapping = vertical motion of blade sections caused by either
rotation about the horizontal hinge (i.e., normal to the
plane of rotation) at the blade root or elastic bending
of the blade
flutter = any dynamic instability involving mechanical
dynamic forces with unsteady aerodynamic forces,
usually occurring in wings and rotors

APPENDIX A 659
forward whirl = whirl motion in the same direction as the spin rotation
of the body
frequency response = complex-valued ratio of output response (amplitude
function and phase) to sinusoidal input force, as a function of
frequency (can be a scalar or matrix quantity)
Froude scaling = special type of model scaling wherein the velocities
are scaled to the square root of full-scale values
gyroscopic effect = tendency for particles in a spinning body to react
to a moment applied to the body in a nonspinning
coordinate frame so as to achieve a rotational velocity
a quarter of a revolution removed from the applied
moment
hingeless rotor = rotor with no actual mechanical hinges that
achieves flapping and lead-lag motion by elastically
flexing
hinges = mechanisms that hold the blades proper to the hub
and allow free angular motion with zero moment
transfer
hub = structure that holds the blades to the rotor shaft
hub moment = couple produced on the hub by the blades at their hub
attachment
impedance = ratio of force quantity(s) to response quantity(s), for
sinusoidally excited systems. The term is used gen-
erally to describe such a ratio wherein the response is
taken to be either displacement, velocity, or acceler-
ation, and used specifically wherein the response is
velocity (see mechanical impedance). Whether used
generally or specifically, impedance is the reciprocal
(inverse) of mobility as long as the response quanti-
ties are consistently defined. Impedance and mobility
can be scalar or matrix quantities.
indicial response = dynamic response caused by a unit input
induced velocity = downward air velocity generated by the rotor in
the process of developing (upward) rotor thrust, or
locally by a vortex
inflow = downward component of air velocity through and
perpendicular to the rotor disk
inflow ratio = ratio of total average velocity normal to the plane
of rotation to the tip speed, normally taken positive
when flowing downward through disk
isotropic = having the same properties in any direction
laminate = in a composite structure, the stacking of any number
of ply groups with dissimilar properties (fiber angle,
thickness, fiber type, etc.) to form a two-dimensional
sheet of finite thickness
lateral cyclic control = component of cyclic control used to roll the
helicopter

lead-lag dampers = (see blade dampers)

lead-lag motion = horizontal motion of blade sections caused by either
rotation about the vertical hinge (i.e., nearly parallel
to the axis of rotation) at the blade root or elastic
bending of the blade
longitudinal cyclic = component of cyclic control used to pitch the heli-
pitch copter
Mach scaling = special type of model scaling wherein the velocities
are scaled to full-scale values
matrix-related terms:
matrix = set of numbers or elements arranged in a rect-
angular array of m rows and n columns. The array
is then called an m × n matrix. The matrix defines a
linear transformation. In this text the array will be
represented by the following symbols:
 
a11 a12 · · · a1n
 a ··· a2n 
A = [A] =  21
aij · · · 
(A.1)
···
am1 · · · amn
(If no confusion arises, it is sometimes convenient to

drop the brackets and indices altogether.)
row matrix = matrix of order 1 × n; notation:
A = [A] = [a1 a2 . . . an ] (A.2)
column matrix = matrix of order m × 1; notation:

 
 a1 

 a2 

A = {A} = .. (A.3)

 

 . 
am
square matrix = matrix of order n × n. Important: Only with square

matrices can we ascribe meaning to the concept
of determinant. The value of the determinant is
designated by the symbol |A|.
diagonal matrix = square matrix of which the elements other than those
in the principal diagonal are zero; notation:
 
a11 0 · · · 0
 0 a22 0 
A = [\A\] = 
aii · · · 
(A.4)
···
0 ··· ann
unit matrix = diagonal matrix of which the diagonal elements are

equal to unity; notation: [I]

APPENDIX A 661
null matrix = any matrix whose elements are all zero; notation: [0]
matrix minor = In any square matrix the minor of the element Aij ,
Mij is the remaining determinant if the ith column
and jth row are deleted.
cofactor = cofactor of the element Aij , is the minor Mij multi-
plied by (−1)i + j
singular matrix = matrix is said to be singular if its determinant is zero
coefficient matrix = Given a set of linear equations in the unknowns {x},
the coefficients of {x} form the coefficient matrix.
augmented matrix = coefficient matrix to which the column of constants
has been added
rank of a matrix = order of the largest square array in the matrix of
which the determinant does not vanish
transposed matrix = matrix AT = [Aji ] is called the transpose of any given
matrix A(=[Aij ]); matrix that is equal to its transpose
is said to be symmetrical.
adjoint matrix = If A is a square matrix, the matrix obtained from A
by replacing each element by its cofactor and then
transposing the resulting matrix is called the adjoint
of A.
mechanical = for sinusoidally excited mechanical systems, the
admittance ratio of response velocity to excitation force;
mechanical admittance (mobility) is the reciprocal
or inverse of mechanical impedance
mechanical = for sinusoidally excited mechanical systems, the
impedance ratio of excitation force to response velocity;
impedance is the reciprocal of mechanical admit-
tance (mobility)
mobility = for sinusoidally excited systems, the ratio of response
quantity to force quantity; the term is used gener-
ally to describe such a ratio wherein the response
is taken to be displacement, velocity, or accelera-
tion and used specifically wherein the response is
velocity (see mechanical admittance). Whether used
generally or specifically, impedance is the reciprocal
(inverse) of mobility as long as the response quanti-
ties are consistently defined. Impedance and mobility
can be scalar or matrix quantities.
node point = point in a dynamic system, oscillating at one of its
natural frequencies, where the response is zero
orthotropic = having two dissimilar planes of symmetry at right
angles to each other
pitch = angle between a blade chord line and a plane perpen-
dicular to the rotor mast
pitch-flap coupling = automatic kinematic pitch change caused by flapping
motion, as results from the delta-three hinge
pitch-lag coupling = automatic kinematic pitch change caused by lead-lag
motion, as results from the alpha-two hinge

plane of rotation = plane perpendicular to the rotor shaft

ply group = in a composite structure, a group of n contiguous
uniaxial plies that all have the same fiber orientation
and fiber composition
preconing = constant angle built into the blade roots of hingeless
rotor blades to place the blades in a fixed coning
position independent of lift or centrifugal forces
pylon = portion of fuselage that supports the drive shaft to a
rotor
pylon damping = dampers attached to the helicopter airframe to damp
motion of the pylon relative to the ground; typically
installed in landing-gear struts
quasi-isotropic = composite laminate that behaves structurally like an
composite laminate isotropic (metallic) material.
quasi-steady = use of steady characteristics, such as aerodynamic
coefficients, in a dynamic context without recogni-
tion of rate and/or acceleration effects
resonance condition = response condition in a dynamic system wherein a
small sinusoidal excitation force oscillating at or very
near to the system natural frequency will cause the
system to respond at amplitudes several orders of
magnitude greater than would be obtained if the force
were of constant magnitude
retrograde whirl = whirl of a body in direction opposite to the actual
spin rotation of the body
reverse-flow region = area on the retreating side of the rotor where the
component of free stream velocity approaching the
trailing edge of the blade exceeds the tangential
velocity of rotation
root vortex = local whirling motion of air particles exiting in the
chordwise direction from the most inboard aerody-
namic portion of the rotor blade, wherein the whirling
is about an axis normal to the blade
settling with power = (see vortex-ring state)
shaft critical speed = rotation speed of a shaft that is equal to one of the
lateral bending frequencies
shed vortex = local whirling motion of air particles exiting from the
trailing edge of the rotor blade, wherein the whirling
is about an axis parallel to the blade
stall = reduction of lift (or change in pitching moment or
increase in drag) associated with separation of air-
flow from the surface of the blade or other component
subcritical operation = any operational condition wherein the rotor speed
(frequency) is lower than any significant natural
frequencies of the system
supercritical = any operational condition wherein the rotor speed
operation (frequency) is higher that any significant natural
frequencies of the system

APPENDIX A 663
swash plate = portion of a rotor control system that converts control

inputs from the pilot (in the nonrotating coordinate
system) to collective and cyclic pitch angles to the
separate blades
teetering rotor = two-bladed rotor with structure to carry moments
from one blade to the other, mounted on a single
horizontal hinge that allows flapping that is always
equal and opposite on the two blades
thickness ratio = ratio of maximum thickness to airfoil chord
thrust = rotor force perpendicular to the tip-path plane
tip loss = loss of lift at the blade tip caused by three-
dimensional induced effects
tip-path plane = plane containing the path of the rotor blade tips
tip speed = tangential velocity of the rotor blade in hover
tip speed ratio = ratio of aircraft forward flight speed to tip speed
tip vortex = local whirling motion of air particles exiting in the
chordwise direction from the tip of the rotor blade,
wherein the whirling is about an axis normal to the
blade in a rotational sense from the bottom to the top
surfaces
torque = moment about the rotor shaft in the top view as a
result of the total drag of the rotor blades
transmissibility = ratio of the amplitude of vibrational output response
(often accelerations) to the amplitude of input
response
vortex-ring state = unstable aerodynamic condition of a vertically
descending rotor wherein a portion of the airflow
through the rotor disk recirculates and forms a
vortical structure at the blade tips
whirl = circular motion of the spin axis apart from the spin
of the body itself
Abbreviations and Acronyms

DAVI = Dynamic Antiresonant Vibration Isolator
FADEC = Full Authority Digital Engine Control
FFT = Fast Fourier Transform
FPM = Force-Phasing Matrices
FRF = Frequency Response Function
HHC = Higher Harmonic Control
MDF = Multiple Degrees of Freedom
PENDAB = Blade-appended Pendulum Vibration Absorber
PTP = Peak to Peak
SDF = Single Degree of Freedom
TVA = Tuned Vibration Absorber
VTOL = Vertical Takeoff and Landing (aircraft)

Appendix B:
Charts for Blade Frequency Estimation
As described earlier in Chapter 3, Yntema developed a collection of charts for

rapidly estimating the natural frequencies of rotating beams for two types of mass
and stiffness distributions: 1) concurrent linear distributions of both mass and
stiffness, and 2) constant mass and stiffness distributions, but with a concentrated
tip mass. This appendix presents a reproduction of the original charts from NACA
TN 3459 (see Figs. B.1–B. 10). The charts are organized according to configuration
(hinged and cantilever inboard constraint) and with regard to mass distribution
(linear variation of properties and concentrated mass at the end of the beam).
665
“appendixb(new)” — 2005/10/20 — page 665 — #1

Fig. B.1 Bending frequency coefficients an for hinged beams with linear mass and
stiffness distributions.

APPENDIX B 667
Fig. B.2 Zero offset Southwell coefficient K0n for hinged beams with linear mass and

Fig. B.3 Offset correction factors for Southwell coefficients K1n for hinged beams
with linear mass and stiffness distributions.

APPENDIX B 669
Fig. B.3 (continued) Offset correction factors for Southwell coefficients K1n for
hinged beams with linear mass and stiffness distributions.

Fig. B.4 Bending frequency coefficients an for cantilever beams with linear mass and

APPENDIX B 671
Fig. B.5 Zero offset Southwell coefficient K0n for cantilever beams with linear mass
and stiffness distributions.

Fig. B.6 Offset correction factors for Southwell coefficients K1n for cantilever beams
with linear mass and stiffness distributions.

APPENDIX B 673
Fig. B.7 Bending frequency coefficients for nonrotating uniform hinged beams with
a mass at the tip, r = tip mass/beam mass.

Fig. B.8 Zero offset Southwell coefficients for a uniform hinged beam with a mass at
the tip, r = tip mass/beam mass.

APPENDIX B 675
Fig. B.9 Bending frequency coefficients for nonrotating uniform cantilever beams
with a mass at the tip, r = tip mass/beam mass.

Fig. B.10 Zero offset Southwell coefficients for a uniform cantilever beam with a
mass at the tip, r = tip mass/beam mass.

Appendix C:
Generalized Frequency-Domain
Substructure Synthesis
The concept of analytical testing can be generalized to enable the systematic

synthesis of several substructures into a single composite structure. The paramount
assumption made is that the responses, as described using mobilities, are taken at
a discrete frequency. The methodology enables the direct utilization of shake-test-
derived mobilities along with analytical finite element modeling (FEM) results.
Furthermore, the methodology enables a reduction of model size to include only
those degrees of freedom of interest. Additionally, the methodology enables the
use of general complex-valued responses for each substructure, thereby providing
a rigorous method for synthesizing substructures with diverse forms and values of
damping.
Partitioning of Mobility Matrices

Later developments require a generalization of the subsystem mobility matrix
partitioning, and an explanation of the notation used for this purpose is first pre-
sented. The system mobility matrix can generally be partitioned according to sets
of either internal or interface degrees of freedoms in the following form:
   a a a a  f 


qA 

YAA YAI YAJ : YAM 
 A 
 
   a   
 qI   YIA YII YIJ
a a YIM   fI 
a
= 
 (C.1)
 : 

   :
 : : : 
 :

 
  
  
qM a
YMA a
YMI a
YMJ : YMMa fM
where the following sets are defined:
A ≡ set of retained internal degrees of freedom in structure A (number of
degrees of freedom in set A = nA )
I ≡ first set of interface degrees of freedom (number of degrees of freedom
in set I = nI )
M ≡ mth set of interface degrees of freedom (number of degrees of freedom
in set M = nM )
The indices in Eq. (C.1) have three functions: They identify the size, set, and
system matrix attributes for each of the component (both substructure and interface)
677
“appendixc(new)” — 2005/10/25 — page 677 — #1

mobilities. The superscripts identify the substructures from which the partition
originates (i.e., the substructure into which we are looking when we define the
mobility). Accordingly,
a
YAA ≡ nA × nA , AA partition of the mobility matrix from substructure a as
defined by set A
a
YIA ≡ nI × nA , IA crosspartition of interface/internal mobilities as defined
by sets I and A
YIIa ≡ nI × nI , II partition of the mobilities, as defined by set I
qA , qI ≡ subsets of the internal and interface degrees of freedom
fA , fI ≡ subsets of internal and interface forces
Simple Coupling of Substructures

The simplest coupling possible is that of two substructures involving only one
interface point, as symbolically represented in Fig. C.1.
A and B are two substructures each having their respective sets of internal
degrees of freedom and one set of interface degrees of freedom in common. The
equations describing the coupling of the mobilities for the two subsystems are
given by a generalized form of the basic analytical testing equation (see Chapter 7),
Eq. (C.2), wherein the coupled mobilities ab are calculated from the uncoupled
systems A and B. The derivation of Eq. (C.2) and all equations to follow is based
on implicit displacement and force continuity considerations at the connection
nodes. The actual derivation of Eq. (C.2) is an extension of the derivation given in
Chapter 7 wherein the form of matrix notation previously outlined is implied.
 ab  
YAA YAI YAB Y a Y a 0
   AA AI 
 YIA YII YIB  =   Y a Y a 0


   IA II 
YBA YBI YBB 0 0 YBBb
   T
a
YAI a
YAI
  −1  
   a 
−  YIIa  YIIa + YIIb  YII  (C.2)
   
−YBIb −YBI b
Fig. C.1 Symbolic coupling of two substructures.

APPENDIX C 679
Note that in this equation the interface degrees of freedom are retained with system
A (and not B). The interface degrees of freedom could equally as well have been
carried with B (and not A), and an equivalent expression with terms rearranged
would be achieved.
By a redefinition of subscripts where Ã now represents the union of A and
I (Ã = A ∪ I), Eq. (C.2) can also be written in more compact form:
ab T
YÃÃ YÃB Ya 0 Ya −1 Ya
ÃÃ ÃI ÃI
= − YIIa + YIIb (C.3)
YBÃ YBB 0 b
YBB −YBI
b −YBI
b
where the special Ã partitioning for the substructure carrying the interface is to be
noted. The important characteristic of these two relationships is that the coupling
process, wherein the total coupled mobilities are found, utilizes (additionally) only
interface information from the two unconnected subsystems. Also, one should
note that the algebra involved is quite efficient since the inverse, representing
the most time-consuming calculation, is only of order nI × nI , and is generally
much smaller than the system matrices themselves. These equations form the
fundamental foundation for the generalization of the coupling process for any
number of substructures. This generalization is presented next.
Generalized Mobility Coupling of N Subsystems

With the equations previously defined in mind, one can now extend the theory to
the general case of N subsystems. In general, N structures can be coupled through
no more than N(N − 1)/2 interface sets, each of which may have any number
of motion coordinates. In practice this would be accomplished by modeling each
natural substructure of the (rotorcraft) airframe separately and then combining
them through the synthesis technique outlined in this development. A potential
coupling application could be as indicated in Fig. C.2.
As represented in this figure, physical fuselage subassemblies (i.e., sub-
structures) are to be synthesized to form the expression for the complete dynamic
response (i.e., mobilities). The subassemblies indicated are normally “bolted”
together, which makes independent modeling convenient. The fuselage fore-
section (A) is connected to the tailboom (B) through the interface set (I). The
turbine (C) is connected to the tailboom through the interface set (J), to the
Fig. C.2 Representative coupling of rotorcraft fuselage substructures.

foresection through interface set (M), and to the transmission (D) through interface
set (K). And last, the transmission is connected to the turbine through interface
set (K) and to the fuselage through interface set (L). After separating out the sets
of common interfaces for each substructure, the coupling process of Fig. C.2 can
be pictured symbolically in Fig. C.3.
By labeling the sets of interfaces, the coupling process is now fully described.
The utilization of graph theory for the assembly process at hand will assure proper
sign selection of common coordinates. In Fig. C.3 a “tail” is identified by a (+), and
an “arrow” is identified by a (−). Although this convention represents an arbitrary
selection of arrows for common connection points, it will make the coupling of
the subsystems consistent. In accordance with Fig. C.3 a Boolean mapping matrix
[M] can be constructed:
  
1 0 0 1 1 A

−1 1 0 0 0 B
[M] = 
 0
 (C.4)
−1
Bodies
−1 1 0 C

0 0 −1 −1 0 D
I J K L M

Interface sets
This matrix defines which “bodies” are connected and through which “interface
sets” the connections are to be accomplished. It also reflects the arbitrary sign
convention chosen for Fig. C.3. For example, the first column of Eq. (C.4) can be
constructed in the following manner: According to Fig. C.3, the interface set I is
given a +1 on A and consequently a −1 on B. This corresponds directly to the
first column values given in Eq. (C.4).
In order to complete the coupling process, another Boolean map is needed. This
matrix organizes the interface degrees of freedom and their interrelations. Such a
matrix is formed from the rows of the previously defined [M] mapping matrix:
[N (i) ] = [Mi ]T [Mi ] (C.5)
Fig. C.3 Symbolic mapping of substructures.

APPENDIX C 681
where [Mi ] is the ith row of the [M] matrix. Consequently there exist as many
of these matrices as there are subcomponents to assemble. Consider, for instance,
substructure C, the third substructure in the coupling sequence:
 
0 0 0 0 0 I
0 1 −1 0 1 J
 
[N (3) ] = [Mi ]T [Mi ] = 
0 −1 1 0 −1 K (C.6)
(i=3) 0 0 0 0 0 L
0 1 −1 0 1 M
I J K L M
where [M3 ] is the third row of [M], the mapping matrix defined by Eq. (C.4).
The individual terms in this matrix correspond to 1) the partition of the point
and the cross mobilities that are present for a particular substructure and 2) the
sign convention for the same. Similar matrices can be derived for the A, B and D
substructures.
With all of these tools for the assemblage process in mind, the generalized
mobility coupling equation for N subassemblies can then be written as

[Y ] = [Yαα ] − [M] ⊗ [Yαγ ]
N
−1 T
· [N ] ⊗ Yγ γ
(i) (i)
[M] ⊗ [Yαγ ] (C.7)
i=1
where
α ≡ A, B, . . . , K, substructure identification in addition to set definition of
internal degrees of freedom
γ ≡ I, J, . . . , L, (interface set definition)
Y ≡ total synthesized mobility matrix
Yαα ≡ diagonal mobility matrix of the uncoupled subsystems as defined by
sets α
M ≡ mapping matrix, as defined by Eq. (C.4)
Yαγ ≡ mobility matrix for the cross partitionings between the internal and
interface degrees of freedom, as defined by sets α and γ
Y(i)
γγ ≡ mobility matrix for the cross and point relations between the interface
degrees of freedom, as defined by set γ for substructure i
N(i) ≡ ith auxiliary mapping matrix, formed from the ith row of [M] as defined
in Eq. (C.5)
N = number of substructures

⊗ ≡ symbology for a pseudo-Hadamard matrix multiplication, wherein

instead of an element by element matrix product, the indicated
premultiplications by the Boolean matrices M and N(i) are taken to
mean scalar multiplications of all of the elements of each of the indi-
cated matrix partitions by one of the respective Boolean values (either
0, −1, or +1); the appropriate Boolean value is taken from the element
of the M or N(i) matrix in accordance with the partition indexing of
the matrix partition
In accordance with Eqs. (C.2) and (C.3) it can be seen that the generalized
relationship of Eq. (C.7) is valid for 1) total synthesis of all degrees of freedom (i.e.,
calculation of all the degrees of freedom including interface degrees of freedoms)
and 2) condensation of the interface degrees of freedom (with interface degrees
of freedom not included in the coupled result). If a total relationship involving a
synthesis of all degrees of freedom is wanted, then an arbitrary choice of which
substructure should carry the interface sets has to be made. The common interface
sets of two connected substructures can only be retained with one system and not
both [see Eqs. (C.2) and (C.3)]. If this is chosen in a consistent manner, Eq. (C.7)
will effectively synthesize all degrees of freedom of all bodies.
The steps involved in the mobility coupling of N subsystems according to
Eq. (C.7) can then be summarized as follows:
1) Configure a symbolic map of the subassemblies and interfaces involved.
2) Construct the [M] and [N (i) ] mapping matrices.
3) For synthesizing all degrees of freedom choose the substructure that will
carry the interfaces.
4) Construct the set of internal and interface degrees of freedom according
to [M].
5) All portions of Eq. (C.7) are then available, and the coupling process can be
completed.
Impedance Connections Between Substructures

The methodology can readily be expanded to include stiffeners and dampers at
the connection points. The impedance between the connecting interfaces can be
incorporated in the following manner. Define the connecting force at interface γ
as follows:
{ fγ } = [Zγ ]{qγ } (C.8)
where
qγ ≡ difference in displacements between two substructures for
the common interface set γ
Zγ ≡ diagonal impedance matrix for the interface set γ
f r ≡ the nodal force vector for connections γ
The same graph theory convention as that given before is followed, with the
modification that
{qγ } = {qγ+ − qγ− } (C.9)

APPENDIX C 683
where {qγ+ } and {aγ− } are displacements in correspondence with the sign convention
chosen in the map of Eq. (C.4).
A new general equation, similar to Eq. (C.7) for the synthesis of N substructures,
can then be written:
N −1

(i) T
[Y ] = [Yαα ] − [M] ⊗ [Yαγ ] · [N ] ⊗ [Y γ γ ]
(i)
[M] ⊗ [Yαγ ]
i=1
(C.10)
where all terms, except [Ȳγ(i)γ ], remain the same as those in Eq. (C.7). [Ȳγ(i)γ ] is the
mobility matrix of the interface relations with the modification that the diagonal
terms have the connection impedance added in for the interface set in question.
For example, for interface I with an “impedance connection” between common
interface set I for substructures A and B, the diagonal term would read:
· · · [YIIa + YIIb + ZI−1 ]−1 · · · (C.11)
This abbreviated development serves as a basic description of the mobility

substructure synthesis with the inclusion of “impedance connections.” This
methodology is equally applicable to other more general interface mechanisms,
including those demonstrating frequency-dependent interface characteristics, and
all use the formulation defined by Eq. (C.7).

Appendix D:
Basic Equations of Motion for Ground
Resonance and Air Resonance
The simplified (linear) equations of motion presented in this appendix represent

a modal approach to both of these closely allied rotor instabilities. They are
intended as an approximate, but reasonably representative and unified, analysis
of the two phenomena and should thereby maximize insights into their respective
physics. Although the equations are not intended for general analysis applications
in support of actual helicopter design efforts, they nevertheless serve three useful
functions. They provide 1) a vehicle for identifying the principal coupling terms
inherent in these phenomena, 2) cost-effective “first cuts” at the stability character-
istics, and 3) a basis for expansion of the equations to other rotor-pylon coupled
phenomena in the form of dynamic equation “building blocks.” The assumptions
leading to the equations of motion are first presented; the equations themselves
are then presented with abbreviated mathematical development that derives from
the principles put forth in the text.
Assumptions
The principal assumption made in this analytical formulation is that both the
ground- and air-resonance phenomena can be adequately modeled using linear
constant coefficient differential equations of motion. That ground resonance can
be so modeled has been adequately verified experimentally [see Bielawa, 1962],
and the use of linear analyses is thus well-established. However, the assumption
that air resonance can also be modeled using only linear differential equations is
less secure, and certain nonlinearities must be addressed and linearized for inclu-
sion when appropriate. The predominance of linearity is recognized, however, on
the basis that air resonance is basically a phenomenon closely akin to ground res-
onance. Like ground resonance, air resonance involves first-order blade elastic
in-plane bending effects (reasonably and adequately describable using linear
dynamics) as a principal dynamic ingredient. Finally, although not yet experi-
mentally verified, the linear analysis presented herein has been shown to be capable
of predicting “air resonance-like” instabilities.
Another principal assumption is that, air resonance can be modeled with the use
of rotor mode descriptions of the blade in both flapwise and inplane bending. This
assumption is made on the basis that the rotor is isotropic (i.e., having three or more
blades) and is thereby subject to the same basic approach followed in the work of
685
“appendixd(new)” — 2005/11/17 — page 685 — #1

Coleman and Feingold for ground resonance. To make the formulation sufficiently
general and yet particularly applicable to hingeless and bearingless rotor types,
the equations are developed using a modal formulation for the blade motions. The
equations of motion are therefore formulated using a Galerkin technique based on
the following detailed assumptions:
1) The transverse deflection of the blade is elastomechanically represented by
a single (uncoupled, but rotating) spanwise variable normal bending mode each in
the flatwise and edgewise directions. These modes are “uncoupled” in the sense
that they are calculated assuming that there is zero inertia or elastic coupling caused
by pitch, twist, or precone angles.
2) The blades respond quasi-statically in torsion. The principal nonlinear
elastomechanical coupling between the flatwise and edgewise modes is retained
and linearized. This linearization is achieved using appropriate trimmed flight
conditions.
3) The blade has a constant value of built-in coning angle βB maintained by
the aerodynamic lifting loads.
4) The rotor hub is attached to a general pylon substructure that does not rotate
at the rotor speed. This pylon substructure is defined in terms of the five rigid-body
degrees of freedom of the hub: longitudinal, lateral, and vertical translations, and
pitch and roll rotations. For explicit ground- and/or air-resonance applications the
pylon substructure is taken to be the fuselage, which is assumed to be a rigid body.
5) The effects of control angles, as determined by appropriate trimmed flight
conditions, are retained in the aerodynamic formulation.
6) The hub loads are equilibrated in the hub-fixed, nonrotating coordinate sys-
tem. Thus, longitudinal and lateral (gravity) forces arise in the formulation as a
result of rotations of the fuselage in pitch and roll, respectively.
7) For free-flight conditions the equation governing the vertical translation
pitch-angle degrees of freedom will contain load factor augmenting terms propor-
tional to velocity and pitch rate. For constrained flight (wind-tunnel) conditions
these terms are omitted.
8) Basic quasi-steady aerodynamic theory is used to define the unsteady air-
loads on the blade. Low angle-of-attack (unstalled) conditions are assumed: The
lift coefficient cl is represented by a lift-curve slope a(= clα ) times an angle of
attack α, and the drag coefficient cd is represented by a uniform constant drag
coefficient cd0 together with even-powered variations in angle of attack δn α n . The
moment coefficient cm is assumed to be zero.
Kinematic Descriptions
Pertinent Degrees of Freedom
For both the ground resonance and air resonance phenomena, the basic mathe-
matical description is provided by 11 differential equations that respectively model
the responses in the hub x, y, and z translations, hub roll and pitch rotations, and
rotor mode descriptions for the blade elastic responses: blade cyclic in-plane (edge-
wise) bending in the x and y directions, blade cyclic flapwise (flatwise) bending in
the roll and pitch directions, and finally, cosine and sine components of the torsion

APPENDIX D 687
responses. These degrees of freedom are defined mathematically by the following

state vector:
{Z} = [ẋ, ẏ, ż, θxF , θyF , x , y , θxR , θyR , φx , φy ]T (D.1)
Elastomechanical Portions of the Equations

Dynamic terms caused by rotor mass elements. The dynamic terms in
both the blade and fuselage equations of motion arising from the rotor-blade mass
elements are obtained using D’Alembert’s principle. For this basic methodology
the acceleration relative to an inertial reference frame of an arbitrary rotor (blade)
mass element is required. However, the components of this acceleration must
be measured in two different moving reference frames: 1) that defined by axes
attached to the translating and rotating (but not rotating at rotor speed ) hub (the
I, J, and K unit vectors) and 2) that defined by axes attached to the rotating and
preconed (by the angle βB ) blade, which, in turn, is attached to the moving hub
(the i, j, and k unit vectors). These two sets of moving coordinate systems are
depicted in Fig. D.1.
With the small-angle assumption on the precone angle βB , the two moving
coordinate systems can then be related to each other by the following coordinate
transformation, expressed in matrix form:

I cos ψ − sin ψ −βB cos ψ i i
J = sin ψ cos ψ −βB sin ψ j = [U] j (D.2)
K βB 0 1 k k
Fig. D.1 Depiction of the sets of unit vectors in the two moving coordinate systems.

The position vector defining the blade mass element is given by
r = ri + y5 j + z5 k (D.3)
where the elastic in-plane and out-of-plane deformations are, respectively, given
by the first bending modal responses:
y5 = Rγv (r)[qv0 + y cos ψ − x sin ψ] (D.4a)

z5 = Rγw (r)[qw0 − θyR cos ψ + θxR sin ψ] (D.4b)
where qv0 and qw0 are the steady in-plane and out-of-plane modal deformations,
respectively, as determined by the trimmed flight condition. The derivation of the
dynamic terms follows from a straightforward application of Coriolis theorem.
The theorem accurately gives the components of inertial acceleration as measured
in each of the two moving coordinate systems. Upon denoting acceleration with
respect to inertial space as [ p2I ] and simple differentiation with respect to time as
[p], the Coriolis theorem can be used to write the inertial acceleration vector for
the position vector r for either of the two moving coordinate systems (ν = 1, 2):
p2I r = pv0 + ωivv × v0 + p2 r + pωivv × r + 2ωivv × pr

+ ωivv × (ωivv × r) (D.5)
where ωivv is the angular velocity of the vth coordinate system (v = 1 for the
hub-centered coordinate system and v = 2 for the rotating, coned blade axis coor-
dinate system) relative to an inertial coordinate system. The angular velocity vector
of the hub-centered axis system (relative to the inertial frame) is expressed as
ωiv1 = θ̇xF I + θ̇yF J + (0)K (D.6a)
On the other hand, the angular velocity vector of the rotating, coned-blade axis
system is expressed as
ωiv2 = [θ̇xF cos ψ + θ̇yF sin ψ + βB ]i

+ [−θ̇xF sin ψ + θ̇yF cos ψ] j
+ [−βB (θ̇xF cos ψ + θ̇yF sin ψ) + ]k (D.6b)
Additionally in the former case of the hub-centered coordinate system, the hub
velocity v0 is composed of a finite steady-state value in the negative longitudinal
x direction V and perturbational values in each of the three component axes:
v0 = (−V + ẋ)I + ẏJ + ż K (D.7)
After evaluation and combination of the various terms in the preceding equation,
the inertial acceleration vector can then be expressed using the rotating, preconed

APPENDIX D 689
coordinate system unit vectors, as given in the following:
p2I r = {βB (z̈ + κV θ̇yF ) − r2

+ [ẍ − 2Rγv (˙y − x ) + 2rβB θ̇xF
− βB 2 Rγw θyR ] cos ψ
+ [ÿ + 2Rγv (˙x + y ) + 2rβB θ̇yF
+ βB 2 Rγw θxR ] sin ψ}i

+ [ÿ + Rγv (ÿ − 2˙x − 22 y ) − rβB θ̈xF
+ 2βB Rγw (θ̇yR − θxR )] cos ψ
− [ẍ + Rγv (¨x + 2˙y − 22 x ) + rβB θ̈yF
+ 2βB Rγw (θ̇xR + θyR )] sin ψ} j
+ {z̈ + κV θ̇yF + [−βB ẍ − Rγw (θ̈yR − 2θ̇xR

− 2 θyR ) − r(θ̈yF − 2θ̇xF )
+ 2βB Rγv (˙y − x )] cos ψ
+ [−βB ÿ + Rγw (θ̈xR + 2θ̇yR − 2 θxR )
+ r(θ̈xF + 2θ̇yF )

− 2βB Rγv (˙x + y )] sin ψ k (D.8)
where the κ factor is equal to 1 or 0, depending on whether the pylon (fuselage) is,
respectively, in a free-flight or gimballed (constrained) configuration. The preced-
ing expression for inertial acceleration can then be represented in the following
abbreviated form:

i
p2I r = ARx ARy ARz j (D.9)
k
where the ARx , ARy , and ARz vector components represent the components of the
inertial acceleration vector, as measured in the number 2 (coned and rotating)
moving coordinate system.
The dynamic terms in the rotor degree-of-freedom equations resulting from
rotor mass elements can then be obtained from the preceding equation using
D’Alembert’s principle together with an application of the Galerkin technique.
A separate application of the Galerkin technique is made for each of the appro-
priate Fourier rotor mode components of the blade elastic modal variables. To
obtain the dynamic inertia terms caused by rotor blade mass that are required
for the nonrotating degree-of-freedom equations, it is much more appropriate and
convenient to take inertial acceleration vector components in the directions of the

unit vectors in the number 1 (unconed and nonrotating) moving coordinate system:

I
p2I r = [ARx ARy ARz ] [U]−1 J (D.10a)
K

I
pI r = [ANRx ANRy ANRz ] J
2
(D.10b)
K
The three components of inertia loads and the rolling and pitching components
of inertia moment at the hub caused by rotor mass (as expressed in the nonrotating
coordinate system) can then be expressed in terms of the ANRα coefficients resulting
from the operations defined by Eqs. (D.2), (D.8), and (D.10a).
Dynamic terms caused by rotor stiffness and damping. One great advan-
tage of the use of a modal description for the blade bending equations is the ease
with which the stiffness and (viscous equivalent) structural damping can be intro-
duced into the blade bending equations. These generalized stiffness and damping
terms for the edgewise bending excitation can be written in terms of the generalized
mass as follows:
R
(s)
+ (d)
= R2 m γv2 dr [ωv2 + 2ζv ωv0 ]
˙
0
= S49 [ωv2 + 2ζv ωv0 ]

˙ (D.11)
where ωv is the “rotating” natural frequency of the first edgewise mode at the
(nominal) rotor speed . Note that the natural frequencies of all of the blade
modes vary in a monotonically increasing manner with rotor speed. Thus, strictly
speaking, this use of the blade modal frequency requires that the variation of
modal frequency with rotor speed be known for appropriate input to the analysis.
An alternate method for including the variation of the natural frequency with rotor
speed is to use the Southwell hyperbolic approximation of frequency with rotor
speed, wherein a simple variation with rotor speed about some nominal condition
is assumed. Thus, the variation of ωv with about some nominal value ωvnom
(relative to the nominal rotor speed nom ) can be closely approximated by
ωv2 = ωv2nom + Kv (2 − 2nom ) (D.12)

where Kv is the Southwell coefficient for the first edgewise bending mode as
described in Chapter 3. Note that in Eq. (D.11) the viscous equivalent struc-
tural damping term [=S49 (2ζv ωv0 )]˙ is written in terms of the natural frequency at
zero rotor speed ωv0 , that is, the “nonrotating” natural frequency. This is required
because the structural damping would not be expected to increase with rotor speed
(which would be the case if the blade rotating natural frequency were used to form
this approximation to the blade structural damping). Also, Eq. (D.11) indirectly
defines and shows the use of a modal integration coefficient S49 . This coefficient
and others to be defined for the other degrees of freedom are summarized in a
subsequent subsection.

APPENDIX D 691
In a similar manner the stiffness and structural damping terms in the flapwise
bending equations can be introduced:
R
m γw2 dr [ωw2 θR + 2ζw ωw0 θ̇R ]
(s) (d)
θR + θR = R2
0
= S10 [ωw2 θR + 2ζw ωw0 θ̇R ] (D.13)
where, similar to the development just given,
ωw2 = ωw2 nom + Kw (2 − 2nom ) (D.14)
The Kw factor is now the Southwell coefficient for the first flatwise bending mode,
and the viscous equivalent structural damping is formulated the same way.
Torsion equation. The appropriate starting point for formulating the quasi-
static torsion response is Eq. (3.114), written in abbreviated form as
GJθe + Tka2 θe + · · ·

= [−qx5 − · · ·] + [(EIE − EIF )ve we ] (D.15)
The blade-torsion-mode response is defined using a rotor-mode description:
θe (r) = γθ (r) qθ (t) = γθ (r)[φx cos ψ + φy sin ψ] (D.16)
In keeping with the intent to maintain a “simplified” development, the blade is

again assumed to be of constant chord and without built-in twist. The linear bend-
ing to torsion coupling resulting from chordwise mass center offsets is assumed
to be negligible (consistent with customary good blade design). Essentially, the
chordwise mass centers and elastic axis locations are both taken to be coinci-
dent with the blade quarter-chord point along the blade. Also, as is done with the
bending equations, a simplified hyperbolic approximation to the torsion natural
frequency variation with rotor speed is assumed:
ωθ2 = ωθ2nom + Kθ 2 − 2nom (D.17)
After the usual application of Galerkin’s method, the simplified equation for
blade torsion can then be written as
S77 θ̈ blade root + S78 (q̈θ + 2ζθ ωθ q̇θ + ωθ2 qθ ) + 2 (S80 cos 2θ0 − S78 )qθ
(a)
= φ − S90 (qw0 + qv0 θR ) (D.18)
where the angular acceleration at the blade root arises from the angular motion
(a)
of the hub in pitch and roll. The aerodynamic torsional excitation φ can be
expressed as a combination of cosine and sine components, using the hub angular
rates and the rotor-mode description for the torsional response. This aerodynamic
excitation is developed in detail in a subsequent subsection.

Dynamic terms resulting from fuselage inertia. The derivation of the

dynamic terms in the fuselage equations caused by fuselage mass and inertia can
be obtained in more than one way. The approach followed here is again to use a
Newtonian approach with an application of D’Alembert’s principle. In this case,
however, the acceleration of the fuselage mass center relative to the inertial ref-
erence frame is required. In addition to the preceding development, the position
vector of the fuselage mass center is now required:
rFcg = xcg I + (0) J − h1 K (D.19)
Because this position vector is constant in the moving body-fixed coordinate
system, the inertial acceleration vector then becomes
p21 rFcg = [ẍ − h1 θ̈yR ]I + [ÿ − h1 θ̈xF ] J + [z̈ − κV θ̇yF − xcg θ̈yF ]K (D.20)
The inertial forces and moments at the rotor hub as a result of the inertial
accelerations at the fuselage center of mass are correspondingly again obtained
using D’Alembert’s principle.
Dynamic terms resulting from rotor and fuselage gravity loads. One of
the ramifications of choosing a body-fixed coordinate system is that the orientation
of the axes relative to the horizon is variable. Hence, components of the gravity
force in the x and y directions will vary in accordance with the fuselage pitch and
roll angles. Because the gravity force of the total aircraft can be expressed using
the rotor thrust T , the hub force terms caused by gravity can be written as
(g) (g) (g)
FH = FHR + FHR = g(mR + mF )[θy I − θx J − K]
= T [θy I − θx J − K] (D.21)
where, by considerations of equilibrium, T is the total rotor thrust, and where
only the first two components are considered because they are perturbational. In a
manner similar to that followed for the inertial loads, the gravitational hub moments
caused by gravity loads can then be similarly defined.
Dynamic Equilibrium Equations

Hub longitudinal force Fx :
(mFx + mR )ẍ + cFx ẋ + kFx x + (bβB S1 − mFx h1 )θ̈yF
b b
+ βB S16 θ̈yR + S48 ¨x − T θyF = Fx(a) + Fx(h) (D.22)
2 2
Hub lateral force Fy :
(mFy + mR )ÿ + cFy ẏ + kFy y − (bβB S1 − mFy h1 )θ̈xF
b b
− βB S16 θ̈xR + S48 ÿ − T θxF = Fy(a) + Fy(h) (D.23)
2 2

APPENDIX D 693
Hub vertical force Fz :
(mFz + mR )z̈ − mFz xcg θ̈yF + κ(mFz + mR )V θ̇yF = Fz(a) + Fz(h) (D.24)
Hub roll moment MxF :
−(bβB S1 − mFy h1 )ÿ

b
+ IφF + mFy h12 + S2 (1 + βB2 ) θ̈xF
2
b
+ S12 θ̈xR + (1 − κ) cFθx θ̇xF
2

b b b
+ 2 S2 θ̇yF + 2 S12 θ̇yR − βB S46 ÿ
2 2 2
+ [mFy gh1 + (1 − κ)kFθx ]θxF = Mx(a)
F
+ Mx(h)
F
(D.25)
Hub pitch moment MyF :

(bβB S1 − mFx h1 )ẍ + IθF + mFx (h12 + xcg
2
)
b b
+ S2 (1 + βB2 ) θ̈yF + S12 θ̈yR
2 2

b b
− 2 S2 θ̇xF + (1 − κ) cFθy θ̇yF − 2 S12 θ̇xR
2 2

b
+ βB S46 ¨x + [mFx gh1 + (1 − κ)kFθy ]θyF
2
− mFz xcg (z̈ + κV θ̇yF ) = My(a)

F
+ My(h)
F
(D.26)
Rotor longitudinal edgewise excitation x :
S48 ẍ + βB S46 θ̈yF + S49 [¨x + 2ζv ωv0 ˙x + (ωv2 − 2 )x ]
+ 2S49 (˙y + ζv ωv0 y ) + 2βB S25 (θ̇xR + θyR ) = (a)

x (D.27)
Rotor lateral edgewise excitation y :
S48 ÿ − βB S46 θ̈xF + S49 [ÿ + 2ζv ωv0 ˙y + (ωv2 − 2 )y ]
− 2S49 (˙x + ζv ωv0 x ) + 2βB S25 (θ̇yR − θxR ) = (a)

y (D.28)

Rotor rollwise flatwise excitation θxR :
− βB S16 ÿ + S12 (θ̈xF + 2θ̇yF ) − 2βB S25 (˙x + y )

+ 2S10 (θ̇yR + ζw ωw0 θyR )
(a)
+ S10 [θ̈xR + 2ζw ωw0 θ̇xR + (ωw2 − 2 )θxR ] = θxR (D.29)
Rotor pitchwise flatwise excitation θyR :
βB S16 ẍ + S12 (θ̈yF + 2θ̇xF ) − 2βB S25 (˙y + x )

− 2S10 (θ̇xR + ζw ωw0 θxR )
(a)
+ S10 [θ̈yR + 2ζw ωw0 θ̇yR + (ωw2 − 2 )θyR ] = θyR (D.30)
Rotor cosine torsion response φx :
S77 (θ̈xF + θ̇yF )+ S78 φ̈x + 2ζθ ωθ φ̇x +(ωθ2 − 2 )φx + 2 (S80 cos 2θ0 − S78 )φx
(a)
+ 2S78 (φ̇y + ζθ ωθ φy ) + S90 (qw0 y − qv0 θyR ) = φx (D.31)
Rotor sine torsion response φy :
S77 (θ̈yF + θ̇xF )+ S78 φ̈y + 2ζθ ωθ φ̇y +(ωθ2 − 2 )φy + 2 (S80 cos 2θ0 − S78 )φy
(a)
−2S78 (φ̇x + ζθ ωθ φx ) + S90 (−qw0 εx + qv0 θxR ) = φy (D.32)
The ( )(h) terms represent arbitrary explicit forces and/or moments applied
directly to the hub from some other source. For some applications these ( )(h) terms
can be expressed in terms of existing degrees of freedom, or they can be used more
generally to define rotor aeroelastic operators, that is, aeromechanical transfer
functions. Note that the preceding equation set is intended for the dual purpose of
modeling both ground resonance and air resonance characteristics. For simplified
(minimal sophistication) ground-resonance applications, only Eqs. (D.22), (D.23),
(D.27), and (D.28) need to be used, with the ( ) terms suppressed. A more
rigorous modeling of ground resonance is affordable by including the fuselage
and rotor degrees of freedom. In any given application of the preceding equations,
the control over which equations and degrees of freedom are to be included in the
equation set can be accomplished using the formalized constraint matrix described
in a subsection to follow.
For air resonance applications all of the equations are generally used, but with
the ( ) terms suppressed. Additionally, the ( ) terms serve a dual
purpose; in the case of a more rigorous ground resonance modeling, wherein the
roll and pitch degrees of freedom must be included, they allow a more direct
modeling of the landing-gear stiffnesses and damping characteristics. In the case

APPENDIX D 695
of air resonance, they can provide the modeling of the restraint moment about
some gimbal point (as with wind-tunnel installations) for the cases that are so
constrained.
The various (inertia) integration constants appearing in Eqs. (D.23–D.33) are
all functions of the spanwise mass distribution m (having the units of lb-s2 /in2 )
and are defined as follows:
R
S1 = m r dr (D.33a)
0
R
S2 = m r 2 dr (D.33b)
0
R
S10 = R2 m γw2 dr (D.33c)
0
R
S12 = R m rγw dr (D.33d)
0
R
S16 = R m γw dr (D.33e)
0
R
S25 = R2 m γv γw dr (D.33f)
0
R
S46 = R m rγv dr (D.33g)
0
R
S48 = R m γv dr (D.33h)
0
R
S49 = R2 m γv2 dr (D.33i)
0
R
S77 = R 2 2
m γθ k̄m2 + k̄m1
2
dr (D.33j)
0
R
S78 = R2 m γθ2 k̄m2
2
+ k̄m1
2
dr (D.33k)
0
R
S80 = R2 2
m γθ2 k̄m2 − k̄m1
2
dr (D.33l)
0
R
−2
S90 = R {γθ γw γv (EIE − EIF )} dr (D.33m)
0
For scaling purposes the preceding integration constants can be presented in

terms of suitable dimensionalizing factors and nondimensional constants. Thus,
because the mode shapes are already in nondimensional form, we can effect the
desired nondimensionalization by first nondimensionalizing the mass distribution
by some arbitrarily selected reference value m0 and then the spanwise variable r by

the blade radius R. Using this approach, the preceding (dimensional) integration
constants can be expressed as follows:
R 1
m
S1 = m r dr = m0 R 2
r̄ dr̄ = m0 R2 S̄1 (D.34a)
0 0 m0
Similarly,
S2 = m0 R3 S̄2 (D.34b)
S10 = m0 R3 S̄10 (D.34c)
S12 = m0 R3 S̄12 (D.34d)
S16 = m0 R S̄16 2
(D.34e)
S25 = m0 R3 S̄25 (D.34f)
S46 = m0 R S̄46 3
(D.34g)
S48 = m0 R2 S̄48 (D.34h)
S49 = m0 R3 S̄49 (D.34i)
S77 = m0 R S̄77 3
(D.34j)
S78 = m0 R3 S̄78 (D.34k)
S80 = m0 R S̄80 3
(D.34l)
S90 = R−1 S̄90 (D.34m)
Note that S̄90 still has the units of bending stiffness (i.e., lb-in2 ).
Aerodynamic Portions of the Rotor Equations

Quasi-Steady Aerodynamics
The aerodynamic excitations, as indicated by the ( )(a) superscripted terms on
the right-hand side of Eqs. (D.22–D.32) are formed using simple quasi-steady
aerodynamic theory. To this end, the static lift-curve slope a, drag coefficient,
consisting of both a constant term cd0 plus one or more terms with even powers
of angle of attack, that is, δn α n , built-in precone angle βB , collective angle θ0 ,
and inflow λ (as appropriate to either hovering or forward flight) are included in
the formulations. The more realistic effects of twist, air mass dynamics, reversed
flow, lift deficiency, cyclic pitch (arising from trim considerations), and nonuni-
form inflow are omitted, consistent with the intended limited use of the equations.
All of the aerodynamic portions of the equations of motion are formulated using

APPENDIX D 697
the following expressions for the tangential and perpendicular components of air-
foil sectional velocity and the airfoil pitch angle and pitch-angle rate θ and θ,θ̇ ,
respectively:
Tangential section air velocity component UT :
UT = R{(r̄ + µ sin ψ) + ( ẏ cos ψ − ẋ sin ψ)/R

+ γv [(˙y − x ) cos ψ − (˙x + y ) sin ψ]} (D.35)
Perpendicular section air velocity component UP :
UP = RŪP

= R φ + µ[(1 − κ)θyF − cos ψ[βB + γw (θxR sin ψ − θyR cos ψ)]]
1
+ βB (ẋ cos ψ + ẏ sin ψ)/R + r̄(θ̇yF cos ψ − θ̇xF sin ψ) − θyR cos ψ

+ γw [(θ̇yR − θxR ) cos ψ − (θ̇xR + θyR ) sin ψ] (D.36)
Note that in Eq. (D.36) and elsewhere in this appendix, the prime symbol
denotes differentiation with respect to nondimensional span r̄.
θ = φx cos ψ + φy sin ψ (D.37)

θ̇ = [θ̇xF + γθ (φ̇x + φy )] cos ψ
+ [θ̇yF + γθ (φ̇y − φx )] sin ψ (D.38)
The rotor inflow is represented by the inflow angle φ based on the inflow
and advance ratios, both of which have distinct radial distributions depending
on whether the rotor is in hovering or forward flight. The “0” subscript denotes
that dynamic (perturbational) variables are omitted:

ŪP0 −1 −λ(r̄)
φ = arctan = tan (D.39)
ŪT0 r̄ + µ sin ψ
where the inflow ration λ is given by



 σa 32θ0 r̄

 1+ − 1 , hovering flight



 16 σa


−µ tan αTPP + [1 + r̄(kx cos ψ + ky sin ψ)]
λ(r̄) = (D.40)



 2

 µ2 C2 µ2


 ×
 + T − , forward flight
2 4 2

The kx and ky coefficients are functions of both advance ratio and inflow ratio and
are defined in detail in Chapter 10.
The perturbational in-plane and out-of-plane force and pitching-moment airload
distributions from which all of the required aerodynamic coefficients are obtained
are given by
(a)
d[δfy ] 1
= δfy(a)
dr R

ρac2 R2 cd δn
= ŪT0 −2 + φ θ0 + n α (n−1) δ ŪT
2 a a

δn (n−1) δn (n−1)
+ θ0 + 2φ − n α δ ŪP + ŪT0 φ − n α δθ
a a
ρac2 R2 ¯ (a)
= δ fy (D.41)
2
d[δfz(a) ] 1 ρac2 R2
= δfz(a) = ŪT0 {[2θ0 + φ]δ ŪT
dr R 2
ρac2 R2 ¯ (a)
+ δ ŪP + ŪT0 δθ } = δ fz (D.42)
2
d[δmx(a) ] 1 ρπ c3 R
= δmx(a) = {−ŪT0 δ θ̇ }
dr R 8
ρπc3 R (a)
= δ m̄x (D.43)
8
The angle of attack α in Eq. (D.41) is the sum of the pitch angle θ0 and the inflow
angle φ.
Aerodynamic Coupling Matrices

The use of simplified quasi-steady modeling of the aerodynamic terms is more
or less standard. Rather than expanding out the aerodynamic exitations, as given
by the right-hand sides of Eqs. (D.22–D.32), an implicit approach is followed.
The aerodynamic excitations identified in these equations amount eventually to
supplemental damping and stiffness matrices. Thus, rather than use separate coef-
ficients to construct the matrices, it is actually more efficient and less error prone
to calculate the matrices directly. Each of the aerodynamic excitations in Eqs.
(D.22–D.32) represents rows in the supplemental damping and stiffness matrices.
To this end, a vector of dependent variables upon which all of the perturbational

APPENDIX D 699
aerodynamics can be formulated is identified as follows:

 
 ẋ 

 R 


 ẏ 


 R 


 


 ż 


 R 


 


 θ̇ x 



F



 θ̇yF 


 


 θ 



yF 


 (ε̇ + ε ) 


 x y 

{δξi } = (ε̇y − ε x ) (D.44)
 
(θ̇xR + θyR )

 


 
(θ̇yR − θxR )
 


 


 θ xR 


 


 θyR 


 
 (φ̇x + φy ) 
 


 
 (φ̇ − φ ) 
 


 y x 


 


 φ x 

 
φy
All of the aerodynamic excitation terms in Eqs. (D.22–D.32) then can be put
in the form of perturbations of each of the elements in the δξi vector using the
perturbational airloads and the expressions for UT , UP θ̇ , θ:
! 1 2π
(a)
∂ f¯z
(a)
∂Fx 1
Fx =
(a)
δξi = bKa c̄ −βB cos ψ
∂ξi 2π 0 0 ∂ ŪT

∂ f¯ y
(a)
∂ f¯ z ∂ ŪP ∂ f¯ z ∂θ
(a) (a)
∂ ŪT 
+ + − sin ψ
∂ξi ∂ ŪP ∂ξi ∂θ ∂ξi ∂ ŪT
 &
∂ f¯ y ∂ ŪP ∂ f¯ y ∂θ 
(a) (a)
∂ ŪT  dψ dr̄ δξi
+ + (D.45)
∂ξi ∂ ŪP ∂ξi ∂θ ∂ξi 
!
1 1 2π ∂ f¯ z ∂ ŪT
(a)
Fy(a) = bKa c̄ −βB sin ψ
2π 0 0 ∂ ŪT ∂ξi

∂ f¯ y ∂ ŪT
(a)
∂ f¯ z ∂ ŪP ∂ f¯ z ∂θ
(a) (a)
+ + + cos ψ 
∂ ŪP ∂ξi ∂θ ∂ξi ∂ ŪT ∂ξi
 &
∂ f¯ y ∂ ŪP ∂ f¯ y ∂θ 
(a) (a)
+ +  dψ dr̄ δξi (D.46)
∂ ŪP ∂ξi ∂θ ∂ξi 

!
1 1 2π ∂ f¯ z ∂ ŪT
(a)
∂ f¯ z ∂ ŪP
(a)
Fz(a) = bKa c̄ +
2π 0 0 ∂ ŪT ∂ξi ∂ ŪP ∂ξi
&
∂ f¯ z ∂θ
(a)
+ dψ dr̄ δξi (D.47)
∂θ ∂ξi
! 1 2π (a)
1 ∂ f¯ z ∂ ŪT
Mx(a)
F
= bKa R c̄r̄ sin ψ
2π 0 0 ∂ ŪT ∂ξi
&
∂ f¯ z ∂ ŪP ∂ f¯ z ∂θ
(a) (a)
+ + dψ dr̄ δξi (D.48)
∂ ŪP ∂ξi ∂θ ∂ξi
! 1 2π (a)
1 ∂ f¯ z ∂ ŪT
My(a)
F
= −bKa R c̄r̄ cos ψ
2π 0 0 ∂ ŪT ∂ξi
&
∂ f¯ z ∂ ŪP ∂ f¯ z ∂θ
(a) (a)
+ + dψ dr̄ δξi (D.49)
! 
∂ f¯ y ∂ ŪT
(a)
1 1 2π
(a)
x = −Ka R c̄γv sin ψ 
π 0 0 ∂ ŪT ∂ξi
 &
∂ f¯ y ∂θ
(a) (a)
∂ f¯y ∂ ŪP
+ +  dψ dr̄ δξi (D.50)
! (a)
1 1 2π ∂ f¯y ∂ ŪT
(a)
y = Ka R c̄γv cos ψ
π 0 0 ∂ ŪT ∂ξi
(a) (a)
&
∂ f¯y ∂ ŪP ∂ f¯y ∂θ
+ + dψ dr̄ δξi (D.51)
!
(a)
(a) 1 1 2π ∂ f¯z ∂ ŪT
θxR = Ka R c̄γw sin ψ
π 0 0 ∂ ŪT ∂ξi
&
(a) (a)
∂ f¯z ∂ ŪP ∂ f¯z ∂θ
+ + dψ dr̄ δξi (D.52)
!
(a)
(a) 1 1 2π ∂ f¯z ∂ ŪT
θyR = −Ka R c̄γw cos ψ
π 0 0 ∂ ŪT ∂ξi
&
(a) (a)
∂ f¯z ∂ ŪP ∂ f¯z ∂θ
+ + dψ dr̄ δξi (D.53)

APPENDIX D 701
! &
(a)
(a) π 1 1 2π 3 ∂ m̄x ∂ θ̇
φX = Ka R c̄ γw cos ψ dψ dr̄ δξi (D.54)
4a π 0 0 ∂ θ̇ ∂ξi
! &
(a)
(a) π 1 1 2π 3 ∂ m̄x ∂ θ̇
φY = Ka R c̄ γw sin ψ dψ dr̄ δξi (D.55)
4a π 0 0 ∂ θ̇ ∂ξi
where the aerodynamic effectivity constant Ka is defined as

1
Ka = ρaR4 (D.56)
2
Steady Bending Deflections

To use the nonlinear torsion term S90 , the steady first edgewise and flatwise mode
deflections qv0 and qw0 must be evaluated. An approximation to these responses
is afforded by using the steady (nonperturbaional) portions of the airloads defined
earlier along with the development described in Chapter 3 for inertially coupled
responses. Upon relaxing the assumption of zero pitch angle, the coupled equations
for qv0 and qw0 become
S49 (ωv2 + 2 sin2 θ0 )qv0 + S25 2 sin θ0 cos θ0 qw0

= (a)
v0 − (gS48 + βB 2 S46 ) sin θ0 (D.57)
S10 (ωw2 − sin θ0 )qw0 + S25 sin θ0 cos θ0 qv0
2 2 2
= (a)
w0 − (gS16 + βB 2 S12 ) cos θ0 (D.58)
where the steady aerodynamic generalized modal loads are determined by any
standard blade element theory [see Leishman and Prouty]:
R
dFy0 dFz0
v0 = R cos θ0 +
(a)
γv (r) sin θ0 dr (D.59)
0 dr dr
R
dFy0 dFz0
w0 = R γw (r) − sin θ0 +
(a)
cos θ0 dr (D.60)
0 dr dr
Equations (D.59) and (D.60) require the steady in-plane and out-of-plane airload
distributions, which in turn require a suitable trim calculation to determine the
aircraft equilibrium, and more immediate, the inflow ratio and blade (control)
pitch angles. The development and presentation of a suitable trim calculation is
beyond the intent of this text, and the reader is referred to Leishmann, Johnson,
and Prouty for excellent treatments of the equations for rotor trim.
Constraints on System Degrees of Freedom

General Modeling of Constraints
With the inclusion of the aerodynamic terms on the left-hand side of the equal
sign, the basic set of dynamic equations, Eqs. (D.22–D.32), can be written in the

following general matrix form:

 
 Lξ(h) 
[M] {Z̈} + [C] {Ż} + [K] {Z} = − − − (D.61)
 
0
where the total vector of degrees of freedom {Z} is defined by Eq. (D.1), and
{Lξ(h) } is the five-element vector of additional excitation loads at the hub. For
some potential applications of these equations, it might arise that some one or
several of the degrees of freedom can either be considered to be identically zero
or are to be arbitrarily constrained to zero (perhaps to investigate their importance
to an instability). In either case, it would be desirable to be able to simplify the
equations to the resulting smaller size in a convenient, accurate manner. One
method for doing this is to define the constraint on the degrees of freedom using
a constraint matrix. Thus, after neglecting {Lξ(h) } for the moment, we define a
reduced set of independent degrees of freedom as denoted by {Z̃}, which is then
related to the vector of 11 original degrees of freedom {Z} as a linear combination.
Such a relationship can be described in a straightforward manner using a matrix
equation, as follows:
{Z} = [T ] {Z̃} (D.62)
where the [T ] matrix will be 11 × n matrix, where n will generally be <11. The
constraining out of selected degrees of freedom is then achieved by first substituting
Eq. (D.62) into Eq. (D.61) and then premultiplying the resulting matrix equation
by [T ]T . This then produces a requisite square matrix formulation of the dynamic
equations:
[T ]T [[M] [T ] p2 + [C] [T ] p + [K] [T ]] {Z̃} = {0} (D.63)
This equation is then in a form suitable for an application of any standard eigen-
value solution.
Displacement Constraint Matrix Caused by Gimballing

The constraining of the helicopter airframe using a gimbal mount (typically
about the total vehicle mass center) is a typical method used to constrain a helicopter
model for the (approximate) testing of air resonance, either on a hover test stand
or in a wind tunnel. The following matrix equation relates the unconstrained (hub)
degrees of freedom to those constrained by virtue of a gimbal about a point. This
gimbal point is located a distance zfoc below the hub. Note that, although the plunge
degree of freedom could be retained and would represent, in effect, the vertical
motion of the gimbal, for present purposes this degree of freedom will also be

APPENDIX D 703
constrained to zero:
   

xF /R 0 zfoc /R

 
 −zfoc /R

yF /R  0   θ
  xF
zF /R =  0 0  (D.64a)

    θyF
 θxF 
   1
 0 

 

θyF 0 1

θ
= [Tfoc ] xF (D.64b)
θyF
Incremental Hub Loads as Dynamic Modules

Generalized Rotor-Airframe Coupling
In Eqs. (D.22–D.26) the right-hand sides of the equations contain the incre-
(h)
mental loads (force or moment) of the form Lξ (where L = F or M, as
appropriate, and ξ is x, y, z, θxF , or θyF ). Because the equations represent state-
ments of dynamic equilibrium at the rotor hub, these incremental loads can be used
to define a more general coupling of the rotor to the remainder of the rotorcraft,
be it a complete helicopter or only an airframe component of a more complicated
rotorcraft. For example, if the explicitly defined helicopter fuselage contributions
to the force and moment equations were deleted, then these incremental loads
could represent the totality of the reaction loads of the fuselage structure at the
hub. Alternatively, they could represent the loads exerted by one of the wings of
a tilt-rotor flight vehicle on the rotor attached to it. Because these loads constitute
internal loads of the total vehicle, they can also be thought of as the negative of the
loads the rotor exerts on the fuselage or rotorcraft substructure (such as a wing).
More specifically, any one of these five rotor force and moment equations (say,
the mth, associated with one of the ξ subscripts) can be written in symbolically
functional form as the totality of the rotor-related terms [Eqs. (D.22–D.26), but
with the explicitly fuselage terms deleted] on one side of the equation and the
incremental hub loads on the other side:
Fm {[acceleration terms ∼ Z̈i ], [damping terms ∼ Żi ]

and [stiffness terms ∼ Zi ]}rotor = Lξ(h) (D.65)
where Zi are the rotor aeroelastic degrees of freedom defined in Eq. (D.1). These
incremental hub loads Lξ(h) essentially define rotor aeroelastic operators that can
be included in the dynamic equations of motions of other aeroelastic systems as
concentrated forces or moments, as appropriate. Thus, as shown in Fig. D.2, these
hub loads can then be applied to the nonrotor substructure in a direction opposite
to the sense they are defined for the rotor.

Fig. D.2 Schematic representation of rotor-airframe coupling using incremental

hub loads.
Application to a Rotor on Wing Configuration

The use of these rotor aeroelastic operators can be demonstrated more specifi-
cally in the case of a rotor-wing combination (i.e., tilt-rotor and/or tilt-wing
rotorcraft configuration), in the hovering mode, as shown in Fig. D.3.
Fig. D.3 Coupling in a rotor-wing combination using incremental hub

loads—hovering flight mode configuration.

APPENDIX D 705
In the following material, equations of motion are considered for the hovering
mode configuration of a wing with a rotor-nacelle substructure attached to it at a
spanwise station sR and with the rotor vertically located relative to this attachment
point by the distance hR . With the implementation of a Galerkin formulation for
the elastic motion (bending and torsion) of the wing, we can write the equations
of motion for these wing degrees of freedom as integrals of the wing loadings
Lw and Lθ , respectively, multiplied by the bending and torsion mode shapes ηw
and ηθ over the semispan of the wing sW . Note that the Lw and Lθ loadings
represent dynamic loads resulting only from the explicitly wing motion degrees of
freedom qWw and qWθ , that is, the wing bending and torsion generalized (modal)
variables, respectively. Stated another way, these loadings represent the dynamic
loads on the wing in the absence of the rotor and nacelle. The actual inclusion of
the incremental hub loads in this integral formulation can then be accomplished
in two basic steps. First, the hub position variables x, y, z, θxF , and θyF must be
expressed in terms of the wing bending and torsion motion variables qWW and qWθ :
   

 x/R 0 ηθ (sR )hR /R

 
 −η (sR )hR /R 

y/R   w 0  q
  Ww
z/R =  ηw (SR )/R 0  (D.66)

 
  (s )  qWθ

 xF 
θ 
  η 0 

 
w R
θyF 0 ηθ (sR )

qWW
= [TR/W ]hov (D.67)
qWθ
Note that for the forward-flight condition a similar transformation matrix can be
defined.
Then, the incremental hub loads (as seen by the wing) are treated mathematically
as negative delta functions in the wing integral (Galerkin) formulations. Thus, the
resulting wing bending equation would be of the following form:
sW
ηw Lw ds + ηw (sR ) −Fz(h) + ηw (sR ) −Mx(h)
F
0
− ηw (sR )hR −Fy(h) = 0 (D.68)
In an analogous manner the wing torsion equation then becomes

sW
ηθ Lθ ds + ηθ (sR ) hR −Fx(h) + ηθ (sR ) −My(h)
F
=0 (D.69)
0
Equations (D.68) and (D.69) can be combined with the [TR/W ]hov matrix defined
by Eqs. (D.66) and (D.67) to yield a unified formulation of the coupled equations
of motion that contain the rotor-wing coupling and are expressed only in terms
of the wing bending degrees of freedom. This is accomplished by first expressing

Eqs. (D.68) and (D.69) in the following manner:

 (h) 
 S  
 Fx R 

 

W

 

 η L ds  
 F (h)
R 

 0
w w
 

y 

 − − −− 
  − [TR/W ]hov Fz R = {0}
T (h)
(D.70)
 S  
 

 W  
 (h) 

ηθ Lθ ds 
 M xF 

 

0 
 
(h) 
MyF
Inspection of the basic equations of motion, Eqs. (D.22–D.32), shows that

the incremental load vector in Eq. (D.70) is also equal to the first five of these
equations, (D.22–D.26), but with the aerodynamic terms (. . .)(a) transferred to the
left-hand side of the equations. Thus, the incremental load vector is seen to consist
of linear combinations of the hub degrees of freedom, that is, the left-hand side of
Eq. (D.61) and their time derivatives along with the blade cyclic rotor mode degrees
of freedom x , y , θxR , θyR , φx , and φy and their time derivatives. The hub degrees
of freedom and their time derivatives are then removed from these incremental
hub load expressions using the substitution afforded by Eq. (D.67). This operation
leaves the matrix equations for the wing dynamics, Eqs. (D.68) and (D.69), in
terms only of the wing degrees of freedom qWw and qWθ and the blade cyclic
rotor mode degrees of freedom. The remaining equations needed to complete the
eigenvalue problem are the six remaining basic equations, Eqs. (D.27–D.32), but
with the hub degrees of freedom again appropriately converted to the wing degrees
of freedom using Eq. (D.67). The final equation set is thereby reduced to an 8 × 8
formulation.
Passive Elastomechanical Rotor-Pylon Coupling

An alternate way of using these Lξ(h) terms to achieve rotor-airframe coupling
would be to express the internal loads in terms of a combination of passive mechan-
ical coupling elements such as springs and dampers together with the appropriate
relative motion. In this instance distinct hub and wing attachment degrees of free-
dom would be required and the incremental hub load terms, Lξ(h) , appearing in
both the rotor and wing equations of motion would instead be expressed using the
mathematical description of the connecting force and/or moment element(s). Thus,
such an approach to the rotor-pylon coupling problem would result in additional
degrees of freedom and additional equations of motion.

Appendix E:
Composite Materials—Basics
Fundamental Differences Between Isotropic Metals

and Uniaxial Composites
The whole subject of composite materials is beyond the scope of the present text,
and only some very basic key concepts are introduced in this appendix. Most of the
material in this appendix is condensed from Tsai and Hahn. The interested reader
is directed to this work and other referenced texts for a more in-depth treatment of
the subject. Furthermore, only those topics that relate directly to composite blade
construction are treated herein. Also, discussion is limited to only “symmetric”
composites.
Except for exotic metal alloys most metals are completely isotropic in that
they have the same properties in any direction. That is, their basic properties are
invariant in all three dimensions. Moreover, the basic properties for metals are only
two in number: the Young’s modulus E, which governs tensile stress and strain,
and the shear modulus G, which governs shear stress and strain. Alternatively,
instead of the shear modulus, Poisson’s ratio ν could be used. The fundamental
starting point in examining the elastic properties of materials is Hooke’s law,
which relates stresses to strains. For the general case of a solid, there are six stress
components σ (three tensile and three shear) and six strain components ε (also
three tensile and three shear) that are related by the following compliance matrix
equation:
    

εxx  a11 a12 a13  σxx 
ε 
  a σ 
 
 yy  a22 a23 0 
  yy 
 
   21 
εzz a a32 a33  σzz
=  31  (E.1)

 ε 
  a44 0  σyz 
 
yz

 εxz   
 σ  
 
 
0 a55
 xz 
 
εxy 0 a66 σxy
[It should be kept in mind that although strains are quantities that can be mea-
sured, stresses are really mathematical abstractions (albeit good ones) that can
only be “measured” by measuring strains and then using them through Hooke’s
law to obtain stresses.]
707
“appendix-e(new)” — 2005/11/16 — page 707 — #1

For isotropic solids Eq. (E.1) simplifies to
    

εxx  1 −ν −ν 
 σxx 
ε 
 σ  
 yy 
 −ν −ν  
  yy 

1 0
  1 
εzz −ν −ν 1  σzz
=   (E.2)

 ε 
  2(1 + ν) 0  σyz 
 
yz E

 εxz 
  2(1 + ν) 
 σ  

   0
 xz 
 
εxy 0 2(1 + ν) σxy
where the quantity E/2(1 + ν) is actually the shear modulus G. To compare

isotropic metals to composites, we reduce Eq. (E.2) to the two-dimensional plane-
stress configuration wherein stresses and strains with subscripts zz, yz, and xz
are omitted, and which would correspond to a thin plate or shell-like solid of
thickness h. The following simplified matrix equation (with a simplification on
the subscripts and designating shear stress and strain with the symbols τ and γ ,
respectively) then results:
 
1 ν
 − 0
 E E  σx
εx  ν 
1
εy = 
− E 0 σy (E.3a)
γxy  E  τxy
 1
0 0
G
The inverse matrix representation of Eq. (E.3a), generally defining the stiffness
characteristics, is given as
E νE 
0
 (1 − ν 2 ) (1 − ν 2 ) 
σx   εx
 
σy =  νE E  εy (E.3b)
 0
τxy  (1 − ν 2 ) (1 − ν 2 )  γxy
0 0 G
Figure E.1a shows how the stresses and strains are oriented on a two-
dimensional element of isotropic material. In the case of uniaxial composite
materials, where there is a distinct orientation of the fibers, the stress-strain
relationship now must use four basic properties, Ex , Ey , Es , and νx (or, alter-
natively, νy ), and take the fiber orientation into account. Figure E.1b identifies
the corresponding stress-strain field for the uniaxial composite two-dimensional
element.

APPENDIX E 709
Fig. E.1 Stress fields for a two-dimensional plane-stress elements—a) isotropic

material, b) uniaxial composite material.

The resulting “on-axis” stress-strain (compliance) relationship for the compo-

site material element is then given by
 
1 νy
− 0
  Ex Ey 
 σx
εx  νx 1 
εy =  −
 E 0  σy
 (E.4)
εs  x Ey  σs
 1
0 0
Es
where a symmetry condition exists to give
νx Ex
= (E.5)
νy Ey
The inverted (stiffness) form of Eq. (E.4), similar to Eq. (E.3b), is not as simple
to write explicitly and is typically written in implicit form as
    
σx  Qxx Qxy 0 εx 
σy = Qyx Qyy 0  εy (E.6)
   
σs 0 0 Qss εs
Off-Axis Stiffness Relationships

In the case of isotropic materials, the stresses at a different orientation can
be easily performed. In the case with uniaxial composites, more care must be
exercised. Figure E.2 poses the problem of knowing and working with strains and
stresses of a uniaxial composite that are oriented relative to the fiber alignment
by an angle θ. The indicated approach is to first resolve the strains to the on-axis
orientations in order to obtain the on-axis stresses. Then the stresses are resolved
back to the original orientation.
The three resolution steps can be combined to give the final stress-strain
relationship for the off-axis orientation.
 
Q11 Q12 Q16 ε
σ1
 .  1
σ2 =  .. Q22 Q26  ε2 (E.7)
σ6 (sym.) · · · Q66 ε 6
where
   
 Q11  m4 n4 2m2 n2 4m2 n2

 
  n4  

 Q22 
  m4 2m2 n2 4m2 n2  Qxx 

   
Q    2 2
m n m 2 n2 m 4 + n4 −4m n2 2 
 Qyy 
 
= 
12
(E.8)

 Q66  
 2 2
m n m 2 n2 −2m2 n2 (m2 − n2 )2 
 Qxy 


 
   
 


 
Q16   m3 n −mn3 mn3 − m3 n 2(mn3 − m3 n) Qss

 

Q26 mn3 −m3 n m3 n − mn3 2(m3 n − mn3 )

APPENDIX E 711
Fig. E.2 Determination of the off-axis stiffnesses (after Tsai and Hahn).
and where m = cos θ , n = sin θ , and Qxx , Qyy , Qxy , and Qss are as defined by
Eq. (E.6).
Multidirectional Laminates
A practical composite structure will use uniaxial composite plies and ply groups
(plies all sharing the same properties and fiber orientations) with a variety of arbi-
trary fiber orientations and sequences. Such a built-up configuration is denoted a
laminate, which can be either symmetrical about some midplane or unsymmetrical.
Laminate Codes
To identify the laminate configuration, a stacking sequence must be used. Fig-
ure E.3a shows a typical stacking sequence for a symmetric laminate. The code
used to define the stacking sequence for a multidirectional composite identifies not
only the order of the orientations (in an ascending order from the bottom ply), but
also the number of plies within each ply group, written as a subscript. Thus, the
appropriate code for the laminate shown in Fig. E.3a, written with the symmetric
notation, would be
[03 /902 / − 45/453 ]S (E.9)

Fig. E.3 Typical stacking sequence details—a) symmetric laminate, b) symmetric

sandwich laminate.
Alternatively, the stacking sequence could be codified using the total notation,
with the subscript T :
[03 /902 / − 45/456 / − 45/902 /03 ]T (E.10)
Because the flexural stiffness and strength characteristics of a laminate are
greatly enhanced by placing the tensile-carrying fibers as far from the neutral
plane as possible, the use of sandwich configurations has become quite practical.
Figure E.3b shows a typical stacking configuration with the inclusion of a
nonstructural core between two load carrying laminates.
The core material is usually referred to as a cellular solid and can be any of
a wide variety of metallic, polymer, or ceramic (foams and/or two-dimensional
honeycomb). Gibson and Ashby present a comprehensive exposition of such mate-
rials. For the most part such materials are used in composite structures in a
nonstructural role, principally for holding the load-carrying laminates in place.
With regard to codifying the ply sequencing with a core insert, the configuration
shown in Fig. E.3b is representative and would be codified as
[03 /902 / − 45/453 /zc ]S (E.11)

APPENDIX E 713
In-Plane Stress-Strain Relationships for Symmetric Laminates

We can consider the built-up multidirectional laminate to be a more compre-
hensive form of the off-axis uniaxial ply or ply group. As shown in Fig. E.4, the
laminate will still be a plate-like structure with a thickness h that is small relative
to the other two dimensions of the structure.
One assumption we must work with is that although the stresses might vary
widely throughout the laminate, the strains will be continuous across ply bound-
aries. Thus, it is customary to regard the stresses at any of the interfaces as
“averaged.” Thus, for the σ1 , σ2 , and σ6 stress components that exist separately
for each ply within the laminate, the same stress components would be averaged
values for the laminate plate. Figure E.4 shows this averaging for the σ2 stress in
some detail. All three of the averaged components are defined as follows:
h/2
1
σ̄1 = σ1 dz (E.12a)
h −h/2
h/2
1
σ̄2 = σ2 dz (E.12b)
h −h/2
h/2
1
σ̄6 = σ6 dz (E.12c)
h −h/2
where the individual σ1 , σ2 , and σ6 stress components for each of the plies are
orientated to the same reference direction using Eqs. (E.7) and (E.8). Also, using
Fig. E.4 In-plane stiffness of a typical symmetric laminate plate, showing corre-
sponding components of the actual ply stresses and the average across the laminate.

the averaged stresses, we can further define stress resultants:
N1 = hσ̄1 (E.13a)
N2 = hσ̄2 (E.13b)
N6 = hσ̄6 (E.13c)
The stress resultants thus have the units of (force)/(length).
Quasi-Isotropic Laminates
For some applications it might be required that the composite laminate behave
exactly as a metallic (isotropic) material. This can be accomplished if the following
conditions are met:
A11 = A22 (E.14a)

A16 = A26 = 0 (E.14b)
1
A66 = [A11 − A12 ] (E.14c)
2
One method for achieving isotropy would be to use many random directions
for the fibers, as would be the case with chopped or matted fibers in the lay-up.
A more controlled way would be to select specific ply orientations that would
mathematically achieve the requirements specified by Eqs. (E.14). Two such ply
orientations exist: the Pi/3 and Pi/4 symmetric lay ups. In the case of the Pi/3 lay
up, the plies on each side of the mid plane must conform to equal numbers of plies
with orientations in the 0-, 60-, and −60-deg directions. That is, a laminate with
a designation of
[0/60/ − 60]S
For a symmetric laminate this defines a minimum of six plies. Similarly, for
the Pi/4 lay up we must have equal numbers of plies with orientations in the 0-,
90-, 45-, and −45-deg directions, for a minimum of eight plies. For this case the
laminate designation is
[0/90/45/ − 45]S
For the quasi-isotropic laminate, the equivalent isotropic engineering constants

become
A12
ν = νx = νy = (E.15a)
A11
G = Es = A66 /h (E.15b)
E = 2(1 + ν)G (E.15c)

APPENDIX E 715
Flexural Stiffness of Symmetric Laminates

If instead of subjecting the composite plate to in-plane stresses, we introduce
moments about the three axes, so as to produce deformations of the plane ele-
ment out of its original plane, we have an alternate mode of structural behavior.
As shown in Fig. E.5, the “2” sides of the plate element are subjected to equal
opposing distributed moments (/unit length) M2 , resulting in an antisymmetric
strain displacement across the faces parallel to the “1” (i.e., local x) axis.
Because the strain displacements must be continuous across the faces, the
corresponding stresses across the thickness of the plate must be discontinuous to
accommodate the variations in ply orientation and properties. In the most general
case where there are moments applied at the edges of the plate, there will be
out-of-plane flexing, describable by some function w, where
w = w(x, y) (E.16)
and where the “1” and “2” axes are taken to be the local x and y axes respectively.
Three applied moments will produce flexing of the plate, the M2 moments shown
in Fig. E.5, similar opposing moment M1 about the “2” axis, and M6 , co-linear
twisting moments about the “1” axis. These moments can be expressed in terms
Fig. E.5 Flexing and stress-strain conditions on the sides of a symmetric laminate
plate caused by the sustained moments on the 2 axis faces.

of the three basic stress components in the plate:

h/2
M1 = σ1 z dz (E.17a)
−h/2
h/2
M2 = σ2 z dz (E.17b)
−h/2
h/2
M6 = σ6 z dz (E.17c)
−h/2
Because σ6 is a shear stress, M6 will take the form of moments applied along
the “1” axis, whereas M1 and M2 are applied at the edges of the plate parallel
to the “2” and “1” axis, respectively. To complete the stress-strain description,
the strains must be defined in terms of the displacements. Thus, in terms of the
out-of-plane displacement w, the displacements in the “1” and “2” directions, u
and v, respectively, are expressed as
∂w
u = −z (E.18a)
∂x
∂w
v = −z (E.18b)
∂y
Finally, the three components of strain can be written as
∂u ∂ 2w
ε1 = = −z 2 (E.19a)
∂x ∂x
∂v ∂ 2w
ε2 = = −z 2 (E.19b)
∂y ∂y
∂u ∂v ∂ 2w
ε6 = + = −2z (E.19c)
∂y ∂x ∂x∂y
The second derivatives in the expressions for the ε1 and ε2 strains can be related
to curvatures k1 and k2 , respectively, and the ε6 strain can be related to a twisting
rate k6 (curvature):
∂ w 2
k1 ∼
=− 2 (E.20a)
∂x
∂ 2w
k2 ∼
=− 2 (E.20b)
∂y

∂ 2w ∂ ∂w ∂φ1
k6 ∼
= −2 = −2 = −2 (E.20c)
∂x∂y ∂x ∂y ∂x

APPENDIX E 717
where φ1 is the angle of twist about the “1” axis. Thus, the strains can be written as
 
ε1 (z) k1
ε2 (z) = z k2 (E.21)
  k6
ε6 (z)
The final expression for the moment-curvature relationship can then be

written as
    
 M1  D11 D12 D16 k1 
M2 =  D22 D26  k2 (E.22)
   
M6 (symm.) D66 k6
where

D11 = Q11 z2 dz (E.23a)

D12 = Q12 z2 dz (E.23b)

D16 = Q16 z2 dz (E.23c)

D22 = Q22 z2 dz (E.23d)

D26 = Q26 z2 dz (E.23e)

D66 = Q66 z2 dz (E.23f)
and where the Q11 thru Q66 terms are defined by Eq. (E.8).
References
Gibson, L. J., and Ashby, M. F., Cellular Solids—Structure and

Properties, 2nd ed. Cambridge Univ. Press, New York, 1997.
Jones, R. M., Mechanics of Composite Materials, McGraw–Hill, New
York, 1975.
Tsai, S. W., and Hahn, H. T., Introduction to Composite Materials,

References
Books
Antoniou, A., Digital Filters: Analysis and Design, McGraw–Hill Book Company,
New York, 1979.
Astrom, K. J., and Wittenmark, B., Computer Controlled Systems, Theory and Design,
(Information and System Sciences Series), Prentice–Hall, Englewood Cliffs, NJ, 1984.
Bathe, K.-J., Finite Element Procedures in Engineering Analysis, Prentice–Hall,
Bisplinghoff, R. L., Ashley, H., and Halfman, R. L., Aeroelasticity, Addison–Wesley,
Cambridge, MA, 1955.
Bramwell, A. R. S., Helicopter Dynamics, Wiley, New York, 1976.
Broch, J. T., Mechanical Vibration and Shock Measurements, Larsen & Son, Denmark
(for Brüel and Kjaer), 1984.
Churchill, R. V., Operational Mathematics, McGraw–Hill, New York, 1958.
Coyle, S., Cyclic & Collective: Second Edition of the Art and Science of Flying
Helicopters, Helobooks, a Division of Mojave Books Limited, Mojave, CA, 2004.
Den Hartog, J. P., Mechanical Vibrations, McGraw–Hill, New York, 1940.
Doughtie, V. L., and James, W. H., Elements of Mechanism, Wiley, New York, 1954.
Dowell, E. H., Clark, R. C., Cox, D., Curtiss, H. C. Jr., Edwards, J. W., Hall, K. C.,
Peters, D. A., Scanlan, R., Simiu, E., Sisto, F., and Strganac, T. W., A Modern Course in
Aeroelasticity, Fourth Edition, Kluwer Academic Publishers, Norwell, MA, 2004.
Ehrich, F. F., Handbook of Rotordynamics, McGraw–Hill, New York, 1992.
Ehrich, F. F., “Self-Excited Vibration,” Shock and Vibration Handbook, Third Edition,
edited by C. M. Harris, McGraw–Hill, New York, 1988, Chap. 5.
Ewins, D. J., Modal Testing: Theory and Practice, Research Studies Press, Herts,
England, U.K. 1984.
Frazer, R. A., Duncan, W. T., and Collar, A. R., Elementary Matrices, Cambridge
Univ. Press, Cambridge, England, 1950.
Fung, Y. C., An Introduction to the Theory of Aeroelasticity, Wiley, New York, 1955.
Gessow, A., and Myers, G. C., Aerodynamics of the Helicopter, Ungar, New York,
1967.
Gibson, L. J., and Ashby, M. F., Cellular Solids - Structure and Properties, 2nd ed.,
Cambridge Univ. Press, New York, 1997.
Greensite, A. L., Elements of Modern Control Theory, Spartan Books, New York,
1970.
Haftka, R. T., Gurdal, Z., and Kamat, M. P., Elements of Structural Optimization,
Kluwer Academic Publishers, Dordrecht, the Netherlands, 1990.
Hildebrand, F. B., Methods of Applied Mathematics, Prentice–Hall, New York, 1952.
Hildebrand, F. B., Introduction to Numerical Analysis, McGraw–Hill, NewYork, 1956.
719
“bib_new” — 2005/11/16 — page 719 — #1

Huelsman, L. P., Active and Passive Analog Filter Design, An Introduction, McGraw–
Hurty, W. C., and Rubinstein, M. F., Dynamics of Structures, Prentice–Hall,
Johnson, D. E., Introduction to Filter Theory, Prentice–Hall Inc., Englewood Cliffs,
NJ, 1976.
Johnson, W., Helicopter Theory, Princeton Univ. Press, Princeton, NJ, 1980.
Jones, D. I. G., “Application of Damping Treatments,” Shock and Vibration Handbook,
Third Edition, edited by C. M. Harris, McGraw–Hill, New York, 1988, Chapter 37.
Jones, R. M., Mechanics of Composite Materials, McGraw–Hill, New York, 1975.
Lamb, H., Hydrodynamics, Dover, New York, 1932.
Leishman, J. G., Principles of Helicopter Aerodynamics, rev. ed., Cambridge Univ.
Press, New York, 2005.
Myklestad, N. O., Vibration Analysis, McGraw–Hill, New York, 1944.
NATO Advisory Group for Aeronautical R & D (AGARD), Structures and Materials
Panel, Manual on Aeroelasticity, edited by W. P. Jones, London, 1960.
Nikolsky, A. A., Helicopter Analysis, Wiley, New York, 1951.
Payne, P. R., Helicopter Dynamics and Aerodynamics, Macmillan, New York, 1959.
Press, W. H., Flannery, B. P., Teukolsky, S. A., and Vetterling, W. T., Numerical
Recipies in FORTRAN, the Art of Scientific Computing, 2nd ed., Cambridge Univ. Press,
New York, 1992.
Prouty, R. W., Practical Helicopter Aerodynamics, P. J. S. Publications, Peoria,
IL, 1982.
Ralston, A., and Wilf, H. S. (eds.), Mathematical Methods for Digital Computers,
Wiley, New York, 1960.
Randall, R. B., Frequency Analysis, Larsen & Son, Denmark (for Brüel and Kjaer),
1984.
Rao, S. S., Mechanical Vibrations, Addison–Wesley, Reading, MA, 1995.
Reed, F. E., “Dynamic Vibration Absorbers and Auxiliary Mass Dampers,” Shock
and Vibration Handbook, 3rd ed., edited by C. M. Harris, McGraw–Hill, New York, 1988,
Chap. 6.
Saaty, T. L., and Bram, J., Nonlinear Mathematics, McGraw–Hill, New York, 1964.
Savant, C. J., Jr., Basic Feedback Control System Design, McGraw–Hill, New York,
1958.
Scanlan, R. H., and Rosenbaum, R., Introduction to the Study of Aircraft Vibration
and Flutter, Macmillan, New York, 1951.
Scarborough, J. B., The Gyroscope, Theory and Applications, Interscience, New York,
1958.
Schwamb, P., Merrill, A. L., and James, W. H., Elements of Mechanisms, Wiley,
New York, 1938.
Smith, B. T., Boyle, J. M., Dongarra, J. J., Garbow, B. S., Ikebe, Y., Klema, V. C.,
and Moler, C. B., Matrix Eigensystem Routines - EISPACK Guide, 2nd ed., Vol. 6, Lecture
Notes in Computer Science, Springer-Verlag, New York, 1976.
Stewart, H. L., Pneumatics and Hydraulics, Audel, (Div. of Bobbs-Merrill).
Indianapolis, IN, 1966.
Timoshenko, S., Vibration Problems in Engineering, Van Nostrand, New York, 1937.
Tong, K. N., Theory of Mechanical Vibrations, Wiley, New York, 1960.
“bib_new” — 2005/11/16 — page 720 — #2

REFERENCES 721
Tsai, S. W., and Hahn, H. T., Introduction to Composite Materials, Technomic

Publishing, Lancaster, PA, 1980.
Tse, F. S., Morse, I. E., and Hinkle, R. T., Mechanical Vibrations, Allyn & Bacon,
Boston, 1964.
Tuma, J. J., Handbook of Numerical Calculations in Engineering, McGraw–Hill,
New York, 1989.
Vanderplaats, G. N., Numerical Optimization Techniques for Engineering Design, with
Applications, McGraw-Hill, New York, 1984.
Whitney, J. M., Structural Analysis of Laminated Anisotropic Plates, Technomic
Publishing, Lancaster, PA, 1987.
Wilkinson, J. J., and Reinsch, C., Handbook for Automatic Computation, Vol. II -
Linear Algebra, Springer–Verlag, New York, 1971.
Wrigley, W., Hollister, W. M., and Denhard, W. G., Gyroscopic Theory, Design and
Instrumentation, MIT Press, Cambridge, MA, 1969.
Young, W. C., Roark’s Formulas for Stress and Strain, McGraw–Hill,
New York, 1989.
Zaveri, K., Modal Analysis of Large Structures—Multiple Exciter Systems, Naerum
Offset, Denmark (for Brüel and Kjaer), 1985.
Internet
Bellis, M., Inventors, The History of Computers, available online at
http://inventors.about.com/library/blcoindex.htm (cited June 2005).
Bellis, M., Inventors of the Modern Computer, FORTRAN - The First Successful High
Level Programming Language, available online at http://inventors.about.com/library/wee-
kly/aa072198.htm (cited June 2005).
Stewart, C. D., Finite Element Analysis/Method Resources, available online at
http://www.steelynx.net/fea/html (cited July 2005).
Wikipedia, the free encyclopedia, Fast Fourier Transform, available online at
http://en.wikipedia.org/wiki/Fast_Fourier_transform (cited June 2005).
Journal Articles
Alwang, J. R., and Skarvan, C. A., “Engine Control Stabilizing Compensation - Test-
ing & Optimization,” Journal of the American Helicopter Society, Vol. 22, No. 3, 1977,
pp. 13–18.
Bailey, C. D., “Exact and Direct Analytical Solutions to Vibrating Systems with
Discontinuities,” Journal of Sound and Vibration, Vol. 44, No. 1, 1976, pp. 15–25.
Beddoes, T. S., “A Synthesis of Unsteady Aerodynamic Effects Including Stall
Hysteresis,” Vertica, Vol. 1, 1976, pp. 113–123.
Bielawa, R. L., “An Improved Technique for Testing Helicopter Rotor-Pylon Aerome-
chanical Stability Using Measured Rotor Dynamic Impedance Characteristics,” Vertica,
Vol. 9, No. 2, 1985, pp. 181–197.
Bielawa, R. L., “Notes Regarding Fundamental Understandings of Rotorcraft Aeroe-
lastic Instability,” Journal of the American Helicopter Society, Vol. 32, No. 4, 1987,
pp. 4–15.
“bib_new” — 2005/11/16 — page 721 — #3

Bielawa, R. L., “Validation of a Method for Air Resonance Testing of Helicopters

at Model Scale Using Active Control of Pylon Dynamic Characteristics,” Journal of the
Bishop, R. E. D., “The Vibration of Rotating Shafts,” Journal of Mechanical
Engineering Science, Vol. 1, No. 1, 1959, pp. 50–65.
Brackbill, C. R., Lesieutre, G. A., Smith, E. C., and Ruhl, L. E., “Characterization and
Modeling of the Low Strain Amplitude and Frequency Dependent Behavior of Elastomeric
Damper Materials,” Journal of the American Helicopter Society, Vol. 45, No. 1, 2000,
pp. 34–43.
Brosens, P. J., and Crandall, S. H., “Whirling of Unsymmetrical Rotors,” Journal of
Applied Mechanics, Vol. 83, Sept. 1961, pp. 355–362.
Burkam, J. E., and Miao, W.-L., “Exploration of Aeroelastic Stability Boundaries
with a Soft-in-Plane Hingeless Rotor Model,” Journal of the American Helicopter Society,
Vol. 17, No. 4, 1972, pp. 27–35.
Cooley, J. W., and Tukey, J. W., “An Algorithm for the Machine Calculation of
Copmplex Fourier Series,” Mathematics of Computation, Vol. 19, 1965, pp. 297–301.
Corliss, L. D., “The Effects of Engine and Height-Control Characteristics on Helicopter
Handling Qualities,” Journal of the American Helicopter Society, Vol. 28, No. 3, 1983,
pp. 56–62.
Crandall, S. H., and Brosens, P. J., “On the Stability of a Rotor with Rotationally
Unsymmetric Inertia and Stiffness Properties,” Journal of Applied Mechanics, Vol. 83,
Dec. 1961, pp. 567–570.
Davis, M. W., and Weller, W. H., “Application of Design Optimization Techniques
to Rotor Dynamic Problems,” Journal of the American Helicopter Society, Vol. 33, No. 3,
1988, pp. 42–50.
Desjardin, R. A., and Hooper, W. E., “Antiresonant Rotor Isolation for Vibration
Reduction,” Journal of the American Helicopter Society, Vol. 25, No. 3, 1980, pp. 46–55.
Donham, R. E., Cardinale, S. V., and Sachs, I. B., “Ground and Air Resonance Char-
acteristics of a Soft In-Plane Rigid Rotor System,” Journal of the American Helicopter
Society, Vol. 14, No. 4, 1969, pp. 33–41.
Ehrich, F., and Childs, D., “Self-Excited Vibration in High-Performance Turboma-
chinery,” Mechanical Engineering, Vol. 106, No. 5, 1984, pp. 66–79.
Fabunmi, J. A., “Developments in Helicopter Ground Vibration Testing,” Journal of
Felker, F. F., Lau, B. H., McLaughlin, S., and Johnson, W., “Nonlinear Behavior of
an Elastomeric Lag Damper Undergoing Dual-Frequency Motion and its Effect on Rotor
Dynamics,” Journal of the American Helicopter Society, Vol. 32, No. 4, 1987, pp. 45–53.
Franklin, J. N., “On the Numerical Solution of Characteristic Equations in Flutter
Analysis,” Journal of the Association for Computing Machinery, Vol. 5, No. 1, 1958,
pp. 45–52.
Friedmann, P. P., and Shanthakumaran, P., “Optimum Design of Rotor Blades for
Vibration Reduction in Forward Flight,” Journal of the American Helicopter Society, Vol. 29,
No. 4, 1984, pp. 70–80.
Friedmann, P. P., Hammond, C. E., and Woo, T.-H., “Efficient Numerical Treatment
of Periodic Systems with Application to Stability Problems,” International Journal for
Numerical Methods in Engineering, Vol. 11, No. 7, 1977, pp. 1117–1136.
“bib_new” — 2005/11/16 — page 722 — #4

REFERENCES 723
Friedmann, P. P., and Straub, F., “Application of the Finite Element Method to Rotary-
Wing Aeroelasticity,” Journal of the American Helicopter Society, Vol. 25, No. 1, 1980,
pp. 36–44.
Gabel, R., and Capurso, V., “Exact Mechanical Instability Boundaries as Determined
from the Coleman Equation,” Journal of the American Helicopter Society, Vol. 7, No. 1,
1962, pp. 17–23.
Gandhi, F., and Chopra, I., “Analysis of Bearingless Main Rotor Aeroelasticity Using
an Improved Time Domain Nonlinear Elastomeric Damper Model,”Journal of the American
Gangwani, S. T., “Calculation of Rotor Wake Induced Empennage Airloads,” Journal
of the American Helicopter Society, Vol. 28, No. 2, 1983, pp. 37–46.
Gangwani, S. T., “Prediction of Dynamic Stall and Unsteady Airloads for Rotor
Blades,” Journal of the American Helicopter Society, Vol. 27, No. 4, 1982, pp. 57–64.
Gaonkar, G. H., Sastry, V. V. S. S., Reddy, T. S. R., Nagabhushanam, J., and Peters,
D. A., “The Use of Actuator-Disc Dynamic Inflow for Helicopter Flap-Lag Stability,”
Journal of the American Helicopter Society, Vol. 28, No. 3, 1983, pp. 79–88.
Gaukroger, D. R., Skingle, C. W., and Heron, K. H., “Numerical Analysis of Vector
Response Loci,” Journal of Sound and Vibration, Vol. 29, No. 3, 1973, pp. 341–353.
Gladwell, G. M., and Stammers, C. W., “On the Stability of an Unsymmetrical Rigid
Rotor Supported in Unsymmetrical Bearings,” Journal of Sound and Vibration, Vol. 3,
No. 3, 1966, pp. 221–232.
Hodges, D. H., “Vibration and Response of Nonuniform Rotating Beams with
Discontinuities,” Journal of the American Helicopter Society, Vol. 24, No. 5, 1979,
pp. 43–50.
Hohenemser, K. H., and Yin, S.-K., “Some Applications of the Method of Multiblade
Coordinates,” Journal of the American Helicopter Society, Vol. 17, No. 3, 1972, pp. 3–12.
Horvay, G., and Yuan, S. W., “Stability of Rotor Blade Flapping Motion When the
Hinges are Tilted: Generalization of the ‘Rectangular Ripple’ Method of Solution,” Journal
of the Aeronautical Sciences, Vol. 14, No. 10, 1947, pp. 583–593.
Houbolt, J. C., and Reed, W. H., “Propeller-Nacelle Whirl Flutter,” Journal of the
Howlett, J. J., Morrison, T., and Zagranski, R. D., “Adaptive Fuel Control for Heli-
copter Applications,” Journal of the American Helicopter Society, Vol. 29, No. 4, 1984,
pp. 43–54.
Ibrahim, S. R., and Mikulcik, E. C., “A Method for the Direct Identification of Vibration
Parameters from the Free Response,” Shock and Vibration Bulletin, No. 47, Pt. 4, 1977,
pp. 183–198.
Isaacs, R., “Airfoil Theory for Flows of Variable Velocity,” Journal of the Aeronautical
Isaacs, R., “Airfoil Theory for Rotary Wing Aircraft,” Journal of the Aeronautical
Jetmundsen, B., Bielawa, R. L., and Flannelly, W. G., “Generalized Frequency Domain
Substructure Synthesis,” Journal of the American Helicopter Society, Vol. 33, No. 1, 1988,
pp. 55–64.
Kemp, N. H., “On the Lift and Circulation of Airfoils in Some Unsteady-Flow
Problems,” Journal of the Aeronautical Sciences, Vol. 19, No. 10, 1952, pp. 713, 714.
“bib_new” — 2005/11/16 — page 723 — #5

Kennedy, C. C., and Pancu, C. D. P., “Use of Vectors in Vibration Measurement and
Analysis,” Journal of the Aeronautical Sciences, Vol. 14, No. 11, 1947, pp. 603–625.
Kuczynski, W. A., Cooper, D. E., Twomey, W. J., and Howlett, J. J., “The Influence of
Engine/Fuel Control Design on Helicopter Dynamics and Handling Qualities,” Journal of
Leishman, J. G. and Beddoes, T. S., “A Semi-Empirical Model for Dynamic Stall,”
Loewy, R. G., “Helicopter Vibrations: A Technological Perspective (The AHS
Alexander A. Nikolsky Honorary Lecture),” Journal of the American Helicopter Society,
Vol. 29, No. 4, 1984, pp. 4–30.
Loewy, R. G., “A Two-Dimensional Approximation to the Unsteady Aerodynam-
ics of Rotary Wings,” Journal of the Aeronautical Sciences, Vol. 24, No. 2, 1957,
pp. 81–92.
Loewy, R. G., Rosen, A., and Mathew, M. B., “Application of the Principal Curva-
ture Transformation to Nonlinear Rotor Blade Analysis,” Vertica, Vol. 11, No. 1/2, 1987,
pp. 263–296.
Luke, Y. L., and Dengler, M. A., “Tables of the Theodorsen Circulation Function
for Generalized Motion,” Journal of the Aeronautical Sciences, Vol. 18, No. 7, 1951,
pp. 478–483.
Lytwyn, R. T., Miao, W.-L., and Woitsch, W., “Airborne and Ground Resonance of
Helicopter Rotors,” Journal of the American Helicopter Society, Vol. 16, No. 2, 1971,
pp. 2–9.
MacFarlane, A. G. J., and Postlethwaite, I., “The General Nyquist Stability Criterion
and Multivariable Root Loci,” International Journal of Control, Vol. 25, 1977, pp. 81–127.
McCarthy, T. R., and Chattopadhyay, A., “A Coupled Rotor/Wing Optimization Pro-
cedure for High Speed Tilt-Rotor Aircraft,” Journal of the American Helicopter Society,
Vol. 41, No. 4, 1996, pp. 360–369.
Miller, R. H., and Ellis, C. W., “Blade Vibration and Flutter,” Journal of the American
Nagy, E. J., “Improved Methods in Ground Vibration Testing,” Journal of the American
Nelson, R. B., “Simplified Calculation of Eigenvector Derivatives,” AIAA Journal,
Vol. 14, No. 9, 1976, pp. 1201–1205.
Ormiston, R. A., and Hodges, D. H., “Linear Flap-Lag Dynamics of Hingeless Heli-
copter Rotor Blades in Hover,” Journal of the American Helicopter Society, Vol. 17, No. 2,
1972, pp. 2–14.
Paul, W. F., “A Self-Excited Rotor Blade Oscillation at High Subsonic Mach Numbers,”
Penrose, R., “On the Generalized Inverse of a Matrix,” Proceedings of the Cambridge
Philosophical Society, 1955, pp. 406–413.
Peters, D. A., “Flap-Lag Stability of Helicopter Rotor Blades in Forward Flight,”
Pitt, D. M., and Peters, D. A., “Theoretical Prediction of Dynamic-Inflow Derivatives,”
Vertica, Vol. 5, No. 1, 1981, pp. 21–34.
Popelka, D. A., Sheffler, M. W., and Bilger, J. “Correlation of Test and Analysis for the
1/5-Scale V-22 Aeroelastic Model,” Journal of the American Helicopter Society, Vol. 32,
No. 2, 1987, pp. 21–33.
“bib_new” — 2005/11/16 — page 724 — #6

REFERENCES 725
Reed, W. H., “Propeller-Rotor Whirl Flutter: A State-of-the-Art Review,” Journal of

Sound and Vibration, Vol. 4, No. 3, 1966, pp. 526–544.
Sears, W. R., “Some Aspects of Non-Stationary Airfoil Theory and Its
Practical Application,” Journal of the Aeronautical Sciences, Vol. 8, No. 3, 1941,
pp. 104–108.
Sela, N. M., and Rosen, A., “The Influence of Alternate Inter-Blade Connections on
Ground Resonance,” Journal of the American Helicopter Society, Vol. 39, No. 3, 1994,
pp. 75–78.
Taylor, R. B., and Teare, P. A., “Helicopter Vibration Reduction with Pendulum
Absorbers,” Journal of the American Helicopter Society, Vol. 20, No. 3, 1975, pp. 9–17.
Tongue, B. H., “Limit Cycle Oscillations of a Nonlinear Rotorcraft Model,” AIAA
Journal, Vol. 22, No. 7, 1984, pp. 967–974.
Urabe, M., “Galerkin’s Procedure for Nonlinear Periodic Systems,” Archives of
Rational Mechanics and Analysis, Vol. 20, 1965, pp. 120–152.
van der Wall, B., and Leishman, J. G., ”The Influence of Variable Flow Velocity on
Unsteady Airfoil Behavior,” Journal of the American Helicopter Society, Vol. 39, No. 4,
1994, pp. 288–297.
Weller, W. H., and Davis, M. W., “Wind Tunnel Tests of Helicopter Blade Designs
Optimized for Minimum Vibration,” Journal of the American Helicopter Society, Vol. 34,
No. 3, 1989, pp. 40–50.
Wernike, K. G., and Gaffey, T. M., “Review and Discussion of The Influence of
Blade Flapping Restraint on the Dynamic Stability of Low Disk Loading Propeller-Rotors,”
Winckler, S. J., “Hygrothermally Curvature Stable Laminates with Tension-Torsion
Coupling,” Journal of the American Helicopter Society, Vol. 30, No. 3, 1985, pp. 56–58.
Wood, E. R., Powers, R. W., Cline, J. H., and Hammond, C. E., “On Developing and
Flight Testing a Higher Harmonic Control System,” Journal of the American Helicopter
Society, Vol. 30, No. 1, 1985, pp. 3–20.
Young, M. I., and Lytwyn, R. T., “The Influence of Blade Flapping Restraint on
the Dynamic Stability of Low Disk Loading Propeller-Rotors,” Journal of the American
Zajac, E. E., “The Kelvin-Tait-Chetaev Theorem and Extensions,” Journal of the
Proceedings
Beddoes, T. S., “Onset of Leading Edge Separation Effects Under Dynamic Conditions
and Low Mach Number,” Proceedings of the 34th Annual National Forum of the American
Helicopter Society, May 1978.
Bielawa, R. L., “Techniques for Stability Analysis and Design Optimization with
Dynamic Constraints of Nonconservative Linear Systems,” Proceedings of the 12th
AIAA/ASME/AHS Structures, Structural Dynamics and Materials Conference, AIAA Paper
71-388, 1971.
Bielawa, R. L., “Dynamic Analysis of Multi-Degree-of-Freedom Systems Using Phas-
ing Matrices,” Proceedings of Specialists’ Meeting on Rotorcraft Dynamics, NASA Ames
Research Center, 1974.
“bib_new” — 2005/11/16 — page 725 — #7

Bielawa, R. L., “Frequency-Domain Methodology for Practical Ground Resonance

Analysis and Test,” Proceedings of the 28th European Rotorcraft Forum, Paper 69, 2002.
Bielawa, R. L., “Practical Analysis and Testing of Rotorcraft Instabilities Using a
Frequency-Domain Based Methodology,” Proceedings of HeliJapan 2002, AHS Interna-
tional Meeting on Advanced Rotorcraft Technology and Life Saving Activities, Paper T222-3,
2002.
Browning, C. E., Husman, G. E., and Whitney, J. M., “Moisture Effects in Epoxy
Matrix Composites,” Composite Masterials: Testing and Design (Fourth Conference),
ASTM STP 617, American Society for Testing and Materials, 1977, pp. 481–496.
Carlson, R. G., Blackwell, R. H., Commerford, G. L., and Mirick, P. H., “Dynamic Stall
Modeling and Correlation with Experimental Data on Airfoils and Rotors,” Proceedings of
Specialists’ Meeting on Rotorcraft Dynamics, NASA Ames Research Center, 1974.
Carta, F. O., and Ham, N. D., “An Analysis of the Stall Flutter Instability of Helicopter
Rotor Blades,” Proceedings of the 23rd Annual National Forum of the American Helicopter
Society, May 1967.
Childs, D. W., “Identification and Avoidance of Instabilities in High-Performance
Turbomachinery,” Von Karman Institute for Fluid Dynamics, lecture, Rhode-Saint-Genese,
Belgium, 1987.
Edwards, J. V., Ashley, H., and Breakwell, J. B. “Unsteady Aerodynamic Modeling
for Arbitrary Motions,” Proceedings of the AIAA Dynamic Specialist Conference, Paper
77–451, 1977.
Flannelly, W. G., “The Dynamic Antiresonant Vibration Isolator,” Proceedings of the
22nd Annual National Forum of the American Helicopter Society, May 1966.
Halwes, D. R., “LIVE—Liquid Inertia Vibration Eliminator,” Proceedings of the 36th
Annual National Forum of the American Helicopter Society, Paper 80-22, May 1980.
Jose, A. I., Leishman, J. G., and Baeder, J. D., “On Calculating Unsteady Airloads
with Time-Varying Free-Stream Mach Numbers,” Proceedings of the 61st Annual National
Forum of the American Helicopter Society, 2005.
Kresselmeier, A., and Steinhauser, R., “Systematic Control Design by Optimizing a
Vector Performance Index,” Proceedings of IFAC Symposioum on Computer Aided Design
of Control Systems, 1979, pp. 113–117.
Lytwyn, R. T., “Aeroelastic Stability Analysis of Hingeless Rotor Helicopters in For-
ward Flight Using Blade and Airframe Normal Modes,” Proceedings of the 36th Annual
National Forum of the American Helicopter Society, Paper 79-23, 1979.
Miao, W.-L., and Huber, H. B., “Rotor Aeroelasticity Coupled with Helicopter
Body Motion,” Proceedings of Specialists’ Meeting on Rotorcraft Dynamics, NASA Ames
Research Center, 1974.
Mindlin, R. D., and Deresíewicz, H., “Timoshenko’s Shear Coefficient for Flexu-
ral Vibrations of Beams,” Proceedings of Second National Congress Applied Mechanics,
ASME, New York, 1954.
Paul, W. F., “Development and Evaluation of the Main Rotor Bifilar Absorber,” Pro-
ceedings of the 25th Annual National Forum of the American Helicopter Society, May
1969.
Peters, D. A., and Izadpanah, A., “Helicopter Trim by Periodic Shooting with Newton-
Raphson Iteration,” Proceedings of the 37th Annual National Forum of the American
Helicopter Society, Paper 81-18, 1981.
Rao, B. M., and Jones, W. P., “Application to Rotary Wings of a Simplified
Aerodynamic Lifting Surface Theory for Unsteady Compressible Flow,” Proceedings of
Specialists’ Meeting on Rotorcraft Dynamics, NASA Ames Research Center, 1974.
“bib_new” — 2005/11/16 — page 726 — #8

REFERENCES 727
Rehfield, L. W., “Design Analysis Methodology for Composite Rotor Blades,” Pro-
ceedings of the Seventh DoD/NASA Conference on Fibrous Composites in Structural Design,
AFWAL-TR-85-3094, U.S. Air Force, 1985, pp. V(a)-1–V(a)-144.
Young, D. K., and Tarzanin, R. J. (Jr.), “Structural Optimization and Mach Scale Test
Validation for a Low Vibration Rotor,” Proceedings of the 47th Annual National Forum of
the American Helicopter Society, May 1991.
Reports
Baker, B. C., “Anti-Aliasing, Analog Filters for Data Acquisition Systems,” Microchip
Application Note AN699, 1999.
Bielawa, R. L., Johnson, S. A., Chi, R. M., and Gangwani, S. T., “Aeroelastic Analysis
for Propellers,” NASA CR-3729, 1983.
Bielawa, R. L., “Simplified Dynamic Equations for Ground and Air Resonance,”
Rensselaer Polytechnic Inst., Rotorcraft Technology Center, Troy, NY, Rept. D-90-1, 1990.
Bielawa, R. L., “An Experimental and Analytical Study of the Mechanical Instability
of Rotors on Multiple-Degree-of-Freedom Supports,” Princeton Univ., Princeton, NJ, Dept.
of Aeronautical Engineering Rept. 612, 1962.
Cansdale, R., Gaukroger, D. L., and Skingle, C. W., “A Technique for Measuring
Impedance of a Spinning Model Rotor,” Royal Aircraft Establishment TR-71092, London,
1971.
Castles, W. Jr., and Gray, R. B., “Empirical Relation Between Induced Velocity, Thrust
and Rate of Descent of a Helicopter Rotor as Determined by Wind-Tunnel Tests on Four
Model Rotors,” NACA TN 2474, 1951.
Coleman, R. P., and Feingold, A. M., “Theory of Self-Excited Mechanical Oscillations
of Helicopter Rotors with Hinged Blades,” NACA Rept. 1351, 1958.
Flannelly, W. G., Fabunmi, J. A., and Nagy, E. J., “Analytical Testing,” NASA CR-
3429, 1981.
Gangwani, S. T., “Synthesized Airfoil Data Method for Predicition of Dynamic Stall
and Unsteady Airloads,” NASA CR-3672, 1983.
Greenberg, J. M., “Airfoil in Sinusoidal Motion in a Pulsating Stream,” NACA TN-
1326, 1947.
Halfman, R. L., Johnson, H. C., and Haley, S. M., “Evaluation of High Angle-of-Attack
Aerodynamic—Derivative Data and Stall-Flutter Prediction Techniques,” NACA TN-2533,
1951.
Ham, H. D., “Helicopter Blade Flutter,” AGARD Rept. 607 (revision of Pt. III, Chap.
10 of the AGARD Manual of Aeroelasticity), 1973.
Hammond, C. E., and Doggett, R., “Demonstration of Subcritical Damping by
Moving-Block/Randomdec Applications,” NASA SP-415, 1976, pp. 59–76.
Hodges, D. H., and Dowell, E. H., “Nonlinear Equations of Motion for the Elastic
Bending and Torsion of Twisted Nonuniform Rotor Blades,” NASA TN D-7818, 1974.
Hodges, R. V., Nixon, M. W., and Rehfield, L. W., “Comparison of Composite Rotor
Blade Models: A Coupled-BeamAnalysis and an MSC/NASTRAN Finite-Element Model,”
NASA TM-89024, 1987.
Houbolt, J. C., and Brooks, G. W., “Differential Equations of Motion for Combined
Flapwise Bending, Chordwise Bending and Torsion of Twisted Non-uniform Rotor Blades,”
NACA Rept. 1346, 1958.
“bib_new” — 2005/11/16 — page 727 — #9

Hunt, G. K., “Similarity Requirements for Aeroelastic Models of Helicopter Rotors,”

Aeronautical Research Council, London, CP-1245, 1972.
Johnson, W., “A Comprehensive Analytical Model of Rotorcraft Aerodynamics and
Dynamics, Part I: Analysis Development,” NASA TM-81182, 1980.
Jones, R., Flannelly, W. G., Nagy, E. J., and Fabunmi, J. A., “Experimental Verification
of Force Determination and Ground Flying on a Full-Scale Helicopter,” U. S. Army Applied
Technology Lab., USAAVRADCOM-TR-81-D-11, Fort Eustis, VA, May 1981.
Jones, R. T., “Operational Treatment of the Non-Uniform Lift Theory in Airplane
Dynamics,” NACA Technical Note 667, 1938.
Jordan, P. F., “Aerodynamic Flutter Coefficients for Subsonic, Sonic and Super-
sonic Flow (Linear Two-Dimensional Theory),” Royal Aircraft Establishment Reports and
Memoranda No. 2932, London, April 1953.
Kvaternik, R. G., “Studies in Tilt Rotor VTOL Aircraft Aeroelasticity,” Ph.D. Disser-
tation, Dept. of Solid Mechanics, Structures and Mechanical Design, Case Western Reserve
Univ., Cleveland, OH, 1973.
Lawrence, C., and Kielb, R. E., “Nonlinear Displacement Analysis of Advanced
Propeller Structures Using NASTRAN,” NASA TM-83737, 1984.
Lawrence, C., Aiello, R. A., Ernst, M. A., and McGee, O. G., “A NASTRAN Primer
for the Analysis of Rotating Flexible Blades,” NASA TM-89861, 1987.
McGuire, D. P., “The Application of Elastomeric Lead-Lag Dampers to Helicopter
Rotors,” Lord Library No. LL2133, 1976.
McGuire, E. P., “Fluidlastic Dampers and Isolators for Vibration Control in Heli-
copters,” Lord Library No. LL-6502, 2001.
Rehfield, L. W., Atilgan, A. R., and Hodges, D. H., “Structural Modeling for Multicell
Composite Rotor Blades,” AIAA Paper No. 88-2250, 1988.
“Requirements for Rotorcraft Vibration Specifications, Modeling and Testing,” U. S.
Army AVSCOM, Aeronautical Design Standards, ADS-27, St. Louis, MO, Oct. 1985.
Rodden, W. P., and Johnson, E. H., “MSC/NASTRAN—Aeroelastic Analysis,” Users
Guide, Ver. 68, MacNeal-Schwendler Corp., Los Angeles, CA, 1994.
St. Hilaire, A. O., Carta, F. O., and Jepson, W. D., “Influence of Sweep on the
Aerodynamic Loading on an Oscillating NACA 0012 Airfoil,” NASA CR-3092, 1979.
Scheiman, J., and Kelly, H. L., “Comparison of Flight-Measured Helicopter Rotor
Blade Chordwise Pressure Distributions with Static Two-Dimension Airfoil Characteris-
tics,” NASA TN D-3936, 1967.
“Standard Guide for Dynamic Testing of Vulcanized Rubber and Rubber-Like
Materials Using Vibratory Methods,”ASTM D5992-96, 1996 (Reapproved 2001).
Stetson, K. A., and Elkins, J. N., “Optical System for Dynamic Analysis of Rotating
Structures,” Air Force Aero Propulsion Laboratory Technical Report, AFAPL-TR-51, 1977.
Vanerplaats, G. N., “CONMIN - a Fortran Program for Constrained Function
Minimization, User’s Manual,” NASA TMX-62282, 1987.
Wilkerson, J. B., “An Analytical Study of the Mechanical Stability of Two-Bladed
Rigid Rotor Systems,” Naval Ship Research and Development Center Report 3351, Jan.
1970.
Yntema, R. T., “Simplified Procedures and Charts for the Rapid Estimation of Bending
Frequencies of Rotating Beams,” NACA TN-3459, June 1955.
“bib_new” — 2005/11/16 — page 728 — #10

Index
absorber mass response, 263 practical implementation scheme,

accelerances, 249 647–648
acceleration, 293 supplementary responses, 646
decoupling, 539 aeroelastic responses, 318
potential, 500 aeroelastic systems, 331
accelerometers, 281–282, 565 aeromechanical instability, 654
calibration factors, 282–283 aeromechanical transfer functions, 694
acoustoelasticity methodology, 654 AH-1G helicopter, 249
active vibration control, 654 air-bag configuration, 278–279
actuator–hub linkage, 281 air mass dynamics, 345, 433
Adams method, 537 basic concepts and assumptions,
adaptive control algorithm, 257 433–434
advance ratio, 344 in aeroelastic analyses, 441–442
advancing blade compressibility air mass equations, alternate forms,
phenomena, 545 440–441
advancing blade concept rotor system, 146 air resonance, 424–433
advancing whirl, 397 aerodynamic portions of rotor equations
aerodynamic angle, time histories, 473 of motion, 696–701
aerodynamic coupling matrices, 698–701 basic considerations and dynamic
aerodynamic damping, 109 equations, 424–426
aerodynamic effectivity constant, 701 blade frequency characteristics, 425–426
aerodynamic environment, 239 effects of torsional dynamics and
aerodynamic excitations, 696, 698–699 precone angle, 430–431
aerodynamic nonlinearities, 529 elastomechanical portions of equations
aerodynamic performance testing, of motion, 687–692
574–575 equations of motion, 405–406, 685–706
aerodynamic related integrals, 355 pertinent degrees of freedom,
aerodynamic torsional excitation, 691 425, 686–687
aeroelastic analysis, 297 air resonance instability
air mass dynamics in, 441–442 mathematical modeling, 426
aeroelastic conformability, 244–245 modeling, 427
aeroelastic instability, 654 pertinent parameters and scaling
aeroelastic optimization, 642–648 considerations, 428–431
application to rotorcraft structural air resonance model, 431–432, 562
dynamics and aeroelasticity, air resonance stability
644–646 analytical predictions, 431
constraints, 645–646 general guidelines, 432–433
cost function, 644–645 mathematical formulation using
design parameters, 644, 646 momentum considerations, 434–436
iteration process, 646 air-spring suspension system, 280
729
“index” — 2005/11/17 — page 729 — #1

730 INDEX
airfoil coefficients, 347 artificial damping, 306

airfoil section autogenous mass, 311
geometrical features, 391–392 autospectra, 287–288
properties, 347 auxiliary equation, 304–305
response to step gust velocity, 207 auxiliary mass, 259–260
airframe with only damper attachment, 264–265
natural frequencies, 275 averaged power quantities, 287
see also rotor-airframe axial inertia, 338
airframe vibrations, future directions,
654 backward-differencing techniques, 540
airload components backward whirl, 397
noncirculatory and circulatory, 452–454 Bell Model 412 damper, 584–588
virtual mass terms, 452–454 bending
aliasing, 284, 291, 319 basic equations, 67–79
amplitude-dependent describing see also specific types and applications
function, 546 bending frequency coefficients
amplitude limiter/delimiter, 569–570 cantilever beams, 670
amplitude reducers, 251 hinged beams, 666
amplitude response ratio, 264 nonrotating uniform cantilever
amplitude responses, 29 beams, 675
analog filters, 638–639 nonrotating uniform hinged beams, 673
elements, 636 bending moment, 201
analytical techniques, 5–65 distribution, 198, 200
analytical testing, 246–247, 677 bending stiffness (EI), 96
equation, 272 alternate equivalent distributions for
equation derivation, 247–251 nonlinear (stepped) distribution, 101
methodology, 249, 251 distributions over entire length of blade,
predictions, 249 98–101
structural synthesis using, 250–251 equivalencies over short sections of
analytical variable inflow, 344 blade, 97–98
angle of attack, 345–347, 367, 447, 473, equivalent distributions, 97
475, 477, 483, 503, 511, 518–519, idealization of blade, 100
521–522, 534, 542, 698 tip value, 99
numerical solution, 478 bending-torsion flutter, 485–496
angle rotations, 59 basic considerations, 485–486
angular momentum, 127–128 methods of solution, 492–495
alternate form, 128–130 Bernoilli equation, 464
vector representation, 129–130 Bernoulli–Euler beam representation, 160
angular velocity, 128, 136 Bessel filters, 284–285
vector representation, 130 Bessel functions, 472
anisotropic inertias and/or bifilar absorber, 182, 254–255
stiffnesses, 336 Biot–Savart theorem, 496–499
anisotropic rotor, stability characteristics, biquadratic filter, transfer function, 632
342 blade bending, 365–366
antiresonant condition, 260–261 descriptors, 423–424
antiresonant frequency, 267 blade deflection components, 364
antiresonant isolation, 268 blade discretization, 109
aperiodic instability, 423 blade dynamics, modification, 244–246
arbitrary airfoil motion, 470 blade elastic properties, 591–594
articulated rotor blade, 69 blade element momentum theory, 366–367
articulated rotors, 362, 368 blade equations of motion, 652
pitch-lag stability of, 367–371 blade flapping frequency ratio, 398–399
“index” — 2005/11/17 — page 730 — #2

INDEX 731
blade flapping motion, 138 cancellation effects, 243–244

blade flapping response, 363 canonical form, 310
blade flapwise flexibility, 397–398 cantilever boundary condition (hingeless
blade flutter, 574 blades), 77–78
and divergence characteristics, 495 cantilever root elastic deformation, 107
blade frequency, 292–295 cantilevered shaft, disk attached to, 156
and coupling effects of pitch and twist center of gravity of rotating rotor,
rate, 118 388–389
estimation charts, 665–676 center of instability point, 409
blade frequency ratio, 399 centrifugal force, 602
blade higher harmonic control (HHC), CH-53A helicopter, 543
255–258 chaos
advantages and disadvantages, 257–258 concept, 292
blade load using dynamic magnification theory, 554
factors (DMF), 214–216 characteristic amplitude, 299
blade root characteristic damping coefficient, 299
bending moments, 218 characteristic determinant, expansion of,
boundary conditions, 110 300–301
shears, 216–217 characteristic equation, 7, 299–302,
blade section 338–339
circulation, 500 graphical method, 302–303
elastomechanical properties, 592 characteristic multiplier, 322–323
inward radial view, 354 characteristic natural frequency, 338
pitch angle, 112–114 characteristic root, 299–300
properties, 591–623 nomenclature, 301
blade tip chordwise mass moment of inertia, 617
applied torque, 602 chordwise offsets, 362
boundary conditions, 110 chordwise point, 346, 454
blade-to-blade effectiveness ratios, 413 circumferential distortions, 388
blade-to-blade lead-lag dampers, 413–414 closed-loop block diagrams, 307
blade torsion cockpit panel, vibration level, 273–274
dynamics, 104–112 coherence function, 288
equilibrium, basic elastic description, Coleman diagram, 406–409, 412, 422–424
105–109 extended, 427–428
kinematics, 104 collective angle, 361, 371, 378, 432
mode response, 691 combined lift and blade weight, 601
zeroth-order linearization, 107 combined rotor/drive/engine/fuel control
BMR/BO-105 air-resonance model, system, 625–642
431–432, 562 analysis techniques, 626–628
Boolean mapping matrix, 680 control system, 627–628
branch idealization into cascaded elements, drive system and rotor lead-lag
175 dynamics, 628
branch points, 170 dynamic subsystems, 627
branch root deflection, 177, 179 engine/fuel metering system, 628
branched drive systems, 170–173 flow block diagram, 627
Brüel and Kjaer Model 3560-L Pulse Lite frequency spectrum, 626
Pocket Analyzer, 290 intersystem coupling, 628
bungee fuselage suspension system, stabilization of combined torsion drive
278–279 system, 628–629
burst chirp, 291–292 complex characteristic root plane, 300
burst random, 291–292 complex quantities, 311
Butterworth filter, 284–285 complex stiffness, 38
“index” — 2005/11/17 — page 731 — #3

732 INDEX
complex-valued response amplitudes for coupling factors, 373–374, 409

multiple resonances, 42 coupling matrices, 698–701
complex-valued response velocities, 40 coupling strains, 615–616
composite materials, 707–717 couplings, pylon degrees of freedom, 429,
hygrothermal effects, 620–622 559
rotor blades, 592 Cramer’s rule, 50
uniaxial, 707 critical damping, 8
see also specific laminates attenuation characteristics, 9
compressibility corrections, 476 critical damping ratio, 10, 328, 406
unsteady decay parameter, 476 critical drivers, 313–314
compressibility effects, 447, 496 critical speed stability characteristics, 339
computational-fluid-dynamics (CFD), 482, cross-coupling, 382
496 cross-coupling stiffness, 382
concentrated force, 273 cross product, 57
concentrated mass/inertia part, 88 vector product, 55–57
coning, 361, 367 cross spectra, 287–288
coning angle, 358, 363, 371 cubic-spline fits, 328, 330
constant coefficients, cyclically flapping rotor, vectorial
multiple-degrees-of-freedom representation, 140
system, 309–315
constitutive equation matrix, 603
constrained minimization process, 644 D’Alembert’s principle, 69, 689, 692
constrained optimization, 642–644 damped natural frequency, 299
constraint equation, 172 damper characteristics, 587
constraint modeling, 701–702 damper complex spring rate characteristics,
control system damping, 358 584
controller operation, 256 damper damping, 587
convergence characteristics, 302 damper stiffness, 587
convergence criterion, 170 damping
coordinate system transformation, 58–59, and forcing frequency, 13
105–106 attenuation characteristics, 8
Coriolis theorem, 60–61, 128–130, 688 effects, 342–343
Coulomb friction, 382–384 optimum value, 154
coupled frequencies, 372–373 sources, 564
coupled modes, 113 vs mass unbalance, 154
coupling effects, 244, 374 vs speed ratio at peak response, 155
characteristics, 117 zero damped natural frequency, 9
edgewise bending and flatwise damping characteristics
bending, 121 ground resonance instability, 408–413
frequency and mode shape isotropic case, 409
characteristics, 117 damping coefficient, 8, 515
general issues, 30–35 damping measurement, 294–295
nonlinear coupling caused by damping ratio, 153
combinative blade bending, decoupling with use of normal modes,
120–121 27–28
pitch and twist rate, 118 deflection correction functions, 116
rotating beams, 112–121 deflection vector, 248
shear center and mass center offsets, 119 degrees of freedom constraints, 701–702
sources of coupling, 112 density of shaft, 167
substructures, 678–679 derotator instrumentation, 571–574
torsion, 118–121 describing functions, 531–532
see also cross-coupling design parameters, 240
“index” — 2005/11/17 — page 732 — #4

INDEX 733
determinants, 49 dynamic antiresonant vibration isolator

diagonal matrix, 43 (DAVI), 266–269
diametral moment of inertia, 135, 137 dynamic equations, 127–128
differential equation of motion, 7, 137, two-degrees-of-freedom gyro, 131–134
145, 192 dynamic interaction between existing
digital filters, 638–639 structure and dynamic subsystem
elements, 636–637 Modification, 248
digital signal processing (DSP), 286, dynamic loads, 189–216
569 examples, 190–201
discrete Fourier analysis (DFA), 15–19, methods of calculation, 190
550–551 rotating beams, 201–205
discrete frequency, 677 dynamic magnification factors (DMF),
discrete frequency excitation, 291 211–213
disk attached to cantilevered shaft, blade load using, 214–216
156–157 single-degree-of-freedom system, 236
displacement, 293 dynamic moment stall, 513
displacement constraint matrix caused by dynamic RAM device, 284
gimballing, 702 dynamic signal analyzers, 284–286
displacement decoupling, 539 dynamic stall, 508–523
displacement impedance matrix, 224 analytical representation, 517
displacement mobility matrix, 225 characteristics, 511–513
distributed inertia system, 168 modeling parameters, 519
distributed torsional inertia, 176 modeling procedures, 518
disturbance transverse loading, 210 related parameters, 518–519
divergence boundaries, 358
divergence in/stability, 356, 358, 394
DMAP feature, 653–654 Earth-fixed coordinate system, 61
dot product, 54–56 Earth-fixed reference frame, 60
drag coefficient, 510 edgewise bending, 366
drag load distribution, 348 excitation, 690
drive rod, 281 frequency variation, 425
drive system stiffness, 618
branched, 170–173 effectiveness factors, 412
dynamics, 149–188 eigenfunction, 299
helicopter, 173–180 eigensolution techniques, 137–138
schematic representation, 174 eigenvalue analysis, ground resonance
special devices, 180–186 instability, 413–416
see also combined eigenvalue derivatives, 241
rotor/drive/engine/fuel control eigenvalue problem, 22, 91, 241
system eigenvalue solutions, 323
driving forces, 310 non-symmetric matrices, 652
dry friction rubbing, 382–384 eigenvector derivatives, 242–243
dual channel dynamic signal analyzer, 286 EISPACK programs, 652
dual-frequency analysis, alternate elastic attachment element, 281
methodology, 585–588 elastic blade formulation, 371
dual-frequency motion, 586–587 elastic coupling, 112–113, 115
dual-frequency operation, 584–588 elastic deformation functions, 115
Duhamel integral, 470–471, 480 elastic forces, 529
dynamic absorber, 259–262 elastic massless part, 89–92
dynamic amplification, reduction by elastic root pitching motion, 108
detuning, 240 elastic spring, 181
dynamic amplitude, 588 elastic tailoring parameter, 608
“index” — 2005/11/17 — page 733 — #5

734 INDEX
elastic transfer function, 90 explicit functions, 241

elastomechanical distributions, 617 explicit nonlinearities, quasi-linearization
elastomeric damper/pivot device, 580 for, 546–548
elastomeric devices, 579–590 exponential functions, 552
examples, 579–581 exponential solution, 7
miscellaneous applications, 589 external applied loading, 206
elastomeric lead-lag dampers, 584, 588 external damping, 155, 343
elastomeric materials, basic characteristics, external disturbances, 206
581–583 external stiffnesses, 342
elastomeric modulus, 583
elastomeric tune vibration absorber, 589
electromagnetic induction proportional to FADEC systems, 641–642
linear motion, 565 fan chart, 73
element equivalencies, 162–164 far-field effects, 450
elementary forms, 36–37 far-field lifting element, 504
energy flow diagram, ground resonance fast Fourier transforms (FFT), 19,
instability, 415 284–286, 288, 550–551, 569, 652
energy flow paths, 310 fatigue life, 275
force phasing matrices, 313–314 fatigue testing, 295
equality constraints, 643 loading spectrum, 295
equilibration of loads acting on mass number of cycles to be tested, 295
element, 6 feathering damping, 361
equilibrium inclination angle, 137 feathering damping ratio, 362
equilibrium precessional rate, 137 feathering inertia, 361
equilibrium (trim) responses, 317 field transfer matrix, 89–92
equivalency equation, 101 filler material, 604–605
equivalent critical damping ratio, 41 filters
equivalent damping, 530–531, 535 analysis, 629–632
equivalent distributions, 97–101 basic types, 632
equivalent EI distributions, 97 bilinear transformation, 637–638
equivalent linearized function, 531–532 derived digital, 638
equivalent mass distributions, 97 digital filter elements, 635–636
equivalent rotary inertias, 162 digital implementations, 633–641
equivalent solid rigid-body mass, 229 nonrecursive, 635–636
equivalent stiffness, 531–532, 534 realization, 639–641
equivalent stiffness ratios for use with recursive, 635
Yntema charts, 100 warping effect, 638
equivalent viscous damping, 38 finite element analysis, 600, 653
Euler angles, 58 finite element codes, 653
Euler–Bernoulli form, 593 finite element methods, 93–95
Euler’s equations of motion, 129, 132–134, finite element modeling (FEM), 242, 275,
337 677
Euler’s method, 482–483, 540 analysis, 226
excitation, 568–569 forced-responses calculations, 230
excitation equipment frequency-response functions, 230
impulsive, 281 finite elements in time, 553
oscillatory, 280–281 finite spring rate, 110
excitation force vector, 293 first-order backward-differencing, 537
excitation frequency, 29 five-coordinate system, 105–106
excitation signals, 286 fixed coordinate system, 404
exciters, 567 frequency, 142–143
explicit forces, 694 fixed-hub condition, 234
“index” — 2005/11/17 — page 734 — #6

INDEX 735
fixed points, 261, 264 forward flight, 399, 447–448, 496, 705
fixed wing aerodynamics, 316
flutter calculations, 305–306 forward translational flight, 439
vs rotary wing, 1–2 forward whirl, 397, 399
flap-damping coefficient vs frequency Fourier amplitudes, 388
ratio, 468 Fourier analysis, 349, 417
flap-lag stability characteristics in Fourier coefficients, 15–17
hingeless rotors, 378–379 Fourier methods, 14–19
forward flight, 378–379 Fourier series, 17, 145, 349–350, 389
hover condition, 374 Fourier transform, 18–19, 287, 470
nonlinear effects, 379 free-free systems, 31
flapping excitation of rotor blade, 214 frequency and period relationship, 10
flatwise bending stiffness, 618 frequency domain
flatwise mass moment of inertia, 617 analysis, 2
flexibility ratio, 153 formulation, 403
flexible root constraint, 112 methods, 305–309
flexible rotor blade, flutter, 488–492 substructure synthesis, 677–683
flexible shaft elastically deformed in frequency placements, 240
transverse direction, 150 frequency ratio, 262–264
flexible supports with mass vs flap-damping coefficient, 468
eccentricity, 152 vs pitch-damping coefficient, 468
flight acceleration data, 249 frequency response
Floquet stability analysis, 653 characteristics, 285
Floquet theory, 318–319, 323, 332 damped case, 29
Fluidlastic, 588–589 undamped case, 28–29
flutter instability characteristics, frequency response function (FRF), 19,
rigid-blade rotors, 394 226, 230, 275, 286–287, 327, 332,
flywheels, flexible shaft connected, 31 417
focused pylon vibration-alleviation alternate expressions, 287–288
device, 270 calculation, 286–288
folded Abbé prism, 571–572 linear single-degree-of-freedom
force balance for forward whirling or system, 13
whipping shaft, 379 frequency scaling parameter, 428
force boundary conditions, 115 frequency spectrum of rotorcraft dynamic
force displacement for connecting related phenomena, 626
beam, 158 frequency sweep, 170
force integration method, 189–205, 209, frequency sweep excitation, 291
211 frequency-temperature correspondence
force load distribution, 106, 120 principle, 581
force phasing matrices, 297, 310–314, Froude scaling, 560, 563–564
496 fuel control system see combined
energy flow paths, 313–314 rotor/drive/engine/fuel control
mathematical implementation, 312–313 system
force-summation method, 190 functions, tabulations, 168
force transducers, 565 fuselage
forced motion dynamic equations, 151 dynamic characteristics, 226, 275
forced response equation, 192 dynamic modifications, 246–257
forced responses ground resonance analysis, 405
FEM, 230 impedance, 230
single-degree-of-freedom equations, 205 inertia, dynamic terms caused by, 692
forcing frequency and damping, 13 mobility, 224, 230, 232, 234, 239,
FORTRAN, 651 576
“index” — 2005/11/17 — page 735 — #7

736 INDEX
fuselage excitation unstable, 414

nonharmonic sources, 222 ground resonance eigenvector, 415
nonintegral harmonic sources, 222 ground resonance instability, 404–424, 575
nonrotor sources, 220–222 damping characteristics, 408–413
principal sources, 220 eigenvalue analysis, 413–416
fuselage vibrations, 189–237 energy-flow diagram, 415
comparison of measured and analytic isotropic rotors, 411
responses, 227–228 single-degree-of-freedom pylon, 411
general problem of, 189 ground resonance testing
research, 249 frequency-domain stability
see also rotor-fuselage methodology, 420
safety issues, 420–421
gain margin, 330–331 gyroscopic dynamics, Jeffcott rotor,
Galerkin technique, 27, 71–72, 76, 95, 160–162
207–208, 210, 241, 350, 365, 491, gyroscopic element, retention of
686, 689, 691 orientation in space, 134
gear effects, 164 gyroscopic precessional behavior, 394
gear system, branched, 171 gyroscopic terms, stabilizing effect of,
geared shafting configuration, 165 314–315
generalized coordinates, 27 gyroscopics, 127–148, 156–158
geometric scaling, 560 basic dynamic equation for
Gill coefficients, 324 two-degrees-of-freedom gyro,
gimbals, 564 131–134
configuration, 689 rotor blades, 138–144
displacement constraint matrix caused simplified equation, 130–134
by, 702
lateral translation, 33 Hadamard multiplication, 46, 312
tail rotor driven by torsionally flexible Hamilton’s principle, 93–95
drive shaft, 185 Hankel functions, generalized, 471–472
global degree-of-freedom vector, 95 harmonic acceleration method, 538
global matrices, 95 harmonic rotor hub loads, 216–219
global virtual displacement vector, 95 Heaviside expansion, 37, 208
glossary of rotorcraft-related Hermite polynomials, 94
terms, 657–663 high-level computer language, 651–652
gravity loads, dynamic terms hinged root boundary conditions (hinged
caused by, 692 blades), 78
ground resonance, 651 hingeless rotor blade, 69
aerodynamic portions of rotor equations hingeless rotors
of motion, 696–701 flap-lag stability characteristics
aeroelastic considerations, 424 forward-flight condition, 378–379
characteristics, 290 hover condition, 374–378
elastomechanical portions of equations nonlinear effects, 379
of motion, 687–692 pitch-flap-lag stability characteristics,
equations of motion, 405–406, 685–706 374
fuselage analysis, 405 pitch-lag stability characteristics,
isotropic rotor and anisotropic 371–374
fuselage, 407 historical milestones, 651–654
pertinent degrees of freedom, 686–687 Holzer–Myklestad technique, 84–92,
rotor-airframe interaction, 416 109–110, 169–172, 182–183, 193
ground resonance configuration homogeneous solution, 7–10
characteristic multiplier vs phase honeycomb filler, 604–605
angle, 419 Hooke’s joint, 184–186
“index” — 2005/11/17 — page 736 — #8

INDEX 737
hoop stress and strain, 595–596 inertia coupled equations, 115

hovering condition, 33, 496, 704 inertia coupling, 112, 207
hovering rotor inertia inequality, 337
flowfield, 461 inertia loadings, 105, 354, 529
Miller theory, 503–508 inertia product, 129, 618
hub inertia-related integrals, 355
deflections, 365 inertial acceleration, 60
flapping spring, 373 inertial acceleration vector, 692
gimbal configurations, 353 inertial coordinate system accelerations,
lateral force, 406, 692 61
loads, 225, 703–706 inertial coupling
longitudinal force, 406, 692 nonzero twist, 115–117
pitch moment, 693 zero twist, 114–115
replacement fixture, 277 infinite spring rate, 110
responses, 225–226 infinitely anisotropic case, 410
roll moment, 693 inflow mass flow parameter, 441
shear load, 273 inflow modeling, 378
vertical force, 693 inflow ratio, 344
hub-centered axis system, 688 inflow velocity, 460
Hurwitz criterion, 493 influence coefficient matrix, 157
Hurwitz determinant, 303 initial curvature vs rotor speed, 152
hydrodynamic bearings and seals, 381–382 initial shaft curvature, 151
hydrodynamic characteristics of viscous in-plane
fluid, 381 bending, 70
hygrothermal effects in composite articulated rotor, 209–211
materials, 620–622 natural frequencies, 72–73
hyperbolic functional relationship, 73 bending moment and shear of rotating
hysteretic nonlinear response, 547 beam, 204
hysteretic whirl, 380–381 Coriolis force, 363
cyclic rotor modes, 389
Ibrahim time-domain (ITD) method, 549, damping, 406
553 degrees-of-freedom, 363
impedance connections between hub shears, 254
substructures, 682–683 load distribution, 351
impedance matrix, 326 load relief, 205
impedance matrix partitions, 248 offset distance, 205
impedance modeling using discrete response of articulated rotor, 209–211
masses, 228–230 instability
impedance scaling, 558 active/passive, 569–571
impedance synthesis from modal avoiding, 407
characteristics, 226–230 boundary, 356
implicit functions, 241 identification, 651–652
improved rotor isolation system see IRIS rotor see rotor instability
incremental centrifugal force, 273 see also stability
incremental damping, 396 instrumentation
incremental force, 273 shake testing, 281–284
incremental pressure, 434–435 see also specific types and applications
incremental tension, 273 integral function, 116, 501
indicial response method, 480–483 integral harmonics, 217–218, 220–222,
induced velocity, 214–216 466
inductive pickups, 566 integrated hub loads, 239, 435–436
inequality constraints, 643 integrated rotor aerodynamic loads, 234
“index” — 2005/11/17 — page 737 — #9

738 INDEX
integration limits, 321 lift deficiency function, 454, 457, 492,

interblade dampers, 412–413 508
interface load vector, 248–249 lift harmonics, 508
interface mobility change, 272 lifting airfoil section, vortex structure for,
internal damping, 342, 380–381 500
internal elastic moments, 113–114, 120 lifting airload distribution, 348
internal shaft damping, 155 lifting blade airload, 345
internal shear, 194 lifting load distribution, 392–393
internal stiffnesses, 338–342 lifting surfaces, 499
inverse Fourier transform, 19 limit-cycle condition, 547
inverse transformation, 36–37 line integrals, practical discrete
IRIS rotor isolation system, 268–269 implementation, 598–560
Isaacs problem, 479, 481 linear constant coefficient differential
isotropic metals, comparison with uniaxial equations of motion, 685
composites, 707–710 linear differential equations, 2
isotropic rotor case, dynamic description, linear single-degree-of-freedom system,
404–408 5–14
isotropic stiffness, 342 frequency response functions for, 13
iteration process, 169–170, 293, 646 linear spring-mass-damper system, 11
linear stability problem, 297
Jacobian method, 547 linear systems
Jeffcott rotor, 149–154 basic tools, 299–309
gyroscopic dynamics, 160–162 characteristic equations, 299–302
whirl speeds and modes and shaft stability analysis, 297–334
critical speeds, 161 linear transformations using matrix
multiplications, 47–49
Kalman-filtering techniques, 257 linear two-degrees-of-freedom system,
Kelvin–Tait–Chetaev (K–T–C) theorem, 19–35
314–315, 342 characteristic frequencies and modes,
kernel function, 501 22–24
kinetic energy, 34, 175 undamped case, 22–24
Kronecker delta, 218 linear variable differential transformers
(LVDTs), 565
Lagrange’s equation, 20–21, 30, 34–35 linearization
Lanczos method for real eigenvalues, 653 simple, 530–535
Laplace equation, 498–499, 501 torsion equation, 107
Laplace expansion for determinant load cells, 281–282, 565, 574
evaluation, 49 calibration factors, 282–283
Laplace operator, 229 piezoelectric, 566
Laplace space eigenvalue methods, 328 piezoresistive, 566
Laplace transform, 35–37, 208, 231–232, loading coefficients, evaluation of, 193
300, 401, 470, 472, 474–475, 477, Lock number, 359, 369–370, 426–429, 559
484, 487 Loewy function, 459–469
Laplace-variable domain, 329 real and imaginary parts, 465–467
laser/fiber-optic vibrometers, 282 Loewy theory, 462–469
lead-lag damping, 209–211, 581 limitations, 467–469
lead-lag frequency, 368, 370, 373 logarithmic decrement method, 39, 41, 549
lead-lag hub spring rate, 370 loss factors, 294–295, 583, 588
Legendre differential equation, 74 low-amplitude operation, 588–589
Legendre polynomials, 74–75 low-pass filter, 531, 630–632
lift coefficient, 509, 511 basic types, 630–632
lift-curve slope, 399 frequency transformations, 632
“index” — 2005/11/17 — page 738 — #10

INDEX 739
practical loss specifications, 630 minor iterations, 318

typical loss and amplitude response MMB BK-117 helicopter, 252
characteristics, 631–632 mobility coupling of subsystems, 679–682
lubrication, 381 mobility matrix, 249, 275, 326, 677–678
lumped-mass approach for transverse modal acceleration method, 190, 199–200,
vibration analysis, 158–162 203, 205
lumped-mass representation, 193 modal analysis, 286
modal characteristics, impedance synthesis
from, 226–230
M criterion, 331 modal differential equation, 145
Mach number, 347, 385, 399, 484, 519, modal displacement functions, 116
545–546 modal displacement method, 190–193,
Mach scaling, 560, 563–564 197, 199–201, 204
magnetic pickups, 566 modal equations of motion, 491
major iterations, 318 modal integrals, evaluation, 77
mass center, 617 modal orthogonality placements, 240
mass center displacement, 119–120 modal response variable, equation for,
mass coupling parameter, 408 71–72
mass distribution, 617 modal shear coefficients, 202
mass eccentricity, 150, 152 modal variable coefficients, 177
mass elements, 251 modal variables, 27, 175
mass matrix, 180, 539 mode deflection method, 189–201, 209,
mass moments of inertia, 129, 136 211
mass properties, 617 mode shape adjustment factors, 80–81
mass ratio, 262, 264, 361, 408, 429, hinged blades, 82
558–559 hingeless blades, 82
mass unbalance mode shape calculations, 92
vs damping, 154 mode shape vector, 179
vs rotor speed, 151 mode shapes, 80–84, 275, 292–295
matrix algebra representation, 55–56 mode-acceleration method, 194
matrix eigenvalue algorithms, 652–653 model rotor blade configuration cross
matrix eigenvalue method, 172 section, 601
matrix eigenvalue problem, 179, 309–310 model rotor testing, 557–577
matrix eigenvalue statement solution, 170 aerodynamic performance, 563–564
matrix equation of motion, 95 construction considerations, 562–563
matrix product data reduction, 568–571
inverse, 51 instrumentation and test procedures,
transposition, 49 564–574
matrix(ices), 42–52 scaling of pylon, 558–560
addition, 45 scaling of rotor, 557–558
adjoint, 50 test preparation, 568
basic definitions, 42–45 test procedures, 568–571
equivalence, 45 vibration-related, 575–576
inverse, 50–51 see also rotor testing
mathematical operations, 45–49 model rotor whirl tests, 574
multiplication, 46–47, 52, 55–56 modification parameters, 240
partitioning, 51–52 derived, 240
representation, 47–48 detailed, 240
subtraction, 45 modulus amplitude, 583
transpose, 45 moment coefficient, 510, 512, 533
see also specific types and applications moment distributions, 98
Miller theory for hovering rotors, 503–508 moment equations, 440
“index” — 2005/11/17 — page 739 — #11

740 INDEX
moment hysteresis, 515 Newton backward-difference formulas,

moment loading, 106, 120 537
moment of inertia, 167 Newton–Raphson method, 301, 547
momentum equation, 437 Newton’s law, 60–61, 127
motion-induced loading, 194, 205–211, NH-90 helicopter, 414
298 nodal variables, identification of, 94
motion transducers, 565 nodamatic isolation system, 265–266
moving-block technique, 550–552 node points, 31–32, 165, 170
moving coordinate systems, 687 nodes, 93
multiblade coordinates, 387–389 nonanalytical variable inflow, 345
multicell beam construction, 605–616 nondimensional scale factors, 428
multidirectional laminates, 711 nonharmonic sources of fuselage
multiple-degrees-of-freedom excitation, 222
frequency-domain methodology nonintegral harmonics, 219, 222, 467
eigenvalue results with nonlinear aeroelastic phenomena,
frequency-domain results, 418–420 time-history solutions, 542–546
mathematical formulation, 416–421 nonlinear differential equations of motion,
practical implementation issues, 418 530
rotor-nacelle systems, 400–404 nonlinear dynamics, 529
multiple-degrees-of-freedom systems, 293, future directions, 553–554
301, 309 nonlinear moment loading, 120
application of methods, 41 nonlinear systems analysis, 529–556
constant coefficients, 309–315 nonlinear torsion moment, 121
Nyquist criterion, 325–332 nonlinear vibrations, 529
periodic coefficients, 315–325 non-rigid (or distorted) wake, 503
spring-mass system, 85 nonrotating coordinate system, 404, 407
structural damping, 38–39 nonrotor sources of fuselage excitation,
torsional shafting, 169 220–222
multiple resonances non-spin angular momentum, 136
complex-valued response amplitudes non-stationary excitations, 286
for, 42 non-symmetric matrices, eigenvalue
response amplitude curve for, 42 solutions of, 652
multipoint excitation, 292–293 nonuniform blades, mode shapes, 78
Myklestad technique, 202 nonzero rotational speed frequencies,
79–80
n-pass method, 324 normal mode shapes, 70
nacelle see rotor-nacelle system normalized mode shape adjustment
narrowband low-frequency testing, functions
288–291 hinged blades, 83
NASTRAN, 653 hingeless blades, 83
natural frequencies notch filter, 628
calculations, 164–170 frequency responses, 634
rotating beams, 665–676 null matrix, 44
solution for, 79–80, 109–112 numerical integration, 537
uniform reference blade, 80 comparison of methods, 541–542
natural modes, 24–25 transient solutions using, 535–546
near-field effects, 450 numerical iteration, 169
negative dampers, 395 numerical methods for stability estimation,
negative frequency, concept of, 143 548–553
negative springs, 394 nutation, 143
neutral axis point, 618 equivalent characteristics, 143
Newmark method, 537–541 frequency, 135, 138
“index” — 2005/11/17 — page 740 — #12

INDEX 741
homogeneous solution, 135 perturbational flapping and feathering,

two-degrees-of-freedom gyros, 131 general equations of motion,
Nyquist criterion, 307, 325–332, 493–494 354–356
Nyquist frequency methods, 330–331 perturbational in-plane/out-of-plane force
and pitching-moment airload
off-axis orientation, 710 Distributions, 698
offset correction factors for Southwell perturbational plunging and pitching
coefficients, 668–669, 672 motions, 451–452
one-quarter-chord point, 454 phase margin, 330–331
onset of instability (OSI), 382 phototachometers, 566
open-loop block diagrams, 307 piezoelectric crystals, 246, 292, 567
open-loop transfer function, 308–309 piezoresistive devices, 281–282
optimal controls, 257 pilot seat isolation laboratory test, 269
optimization problems, 643 Pimento 4 Channel Vibration Analyzer, 289
optimum damping, 262–266 pitch angle, 115, 364, 367, 370
orthogonality condition, 70–71 pitch-damping, 513–515
undamped case, 24–26 effectivity, 459, 515
orthogonality principle, 73, 76 pitch-damping coefficient vs frequency
oscillatory instability, 356, 361–362 ratio, 468
out-of-plane pitch-flap coupling, 364, 386
bending, 69–70 pitch-flap-lag stability/instability,
degrees-of-freedom, 363 362–379, 574
load distribution, 351 basic considerations and mathematical
motion, 485–486 development, 362–367
shears, 251–252 characteristics of hingeless rotors, 374
overlift, 513 pitch-lag coupling, 363–364
overmoment, 513 pitch-lag stability
articulated rotors, 367–371
Padé formulation, 483–485 boundaries, 369–370
basic methodology, 484–485 characteristics, 368–371
Padé formulation coefficients, physical hingeless rotors, 371–374
constraints, 484 pitching coefficient, 533
partial fractions, 37 pitching-moment, 345–347, 513
particular solutions, 10–14, 134 distribution, 351
partitioning of mobility matrices, pitching rate, 345–347
677–678 Pitt model, 441
passive coupling effect, 244–245 pivot to hub distance, stabilizing effect, 396
passive isolator, 258–259 point transfer matrix, 88
pendulum absorbers (pendabs), 180–184, polar amplitude, 41
251, 253, 272–274 polar moment of inertia, 145, 167
performance trim, 317–318 polar radius of gyration equivalences, 163
period and frequency relationship, 10 polar symmetry, 304
periodic coefficients, 331–332, 548 polynomials
multiple-degrees-of-freedom systems, quotient, 36
315–325 solution of, 301–302
sources of, 315–318 transfer function, 640
periodic function, 16 positive-definite matrices, 315
personal computer, 653 positive feedback loop, 307
perturbation solutions, 318 potential energy, 34, 175
perturbational airloads, 350–352, 366 potential flow, 498
perturbational angular rate, 135 potentiometers, 565
perturbational equations of motion, 137 Prandtl–Glauert correction, 385, 476
“index” — 2005/11/17 — page 741 — #13

742 INDEX
precession, 146 quarter-chord point, 346–347, 454

angular velocity, 137 quasi-isotropic laminates, 714
characteristics, 138–139 quasi-linearization
motion, 131 explicit nonlinearities, 546–548
particular solution, 134 systems with periodicity, 548
rate, 135, 137 quasi-static torsion response, 691
precone angle, 432 quasi-steady
pressure-doublet flow, 498, 500–501 aerodynamics, 207, 343–352, 363,
pressure parameter, 382 696–698
pressure sensors, 566 airloads formulations, 348
prismatic bars, 622–623 angle of attack, 346
product of damping criterion, 408, 410 lift, 508
infinitely anisotropic case, 410 vector diagram, 343–345
limitations, 410 quencher, 311
product of inertia, 129, 618
progressive whirl, 141–143
frequency relationship, 143 radial damping, 382
Prony’s method, 552–553 radius of curvature method, 39–41
proof-of-design validations, 557 random excitation, 291
propeller random noise, 291
advance ratio, 393 ratio of moments of inertia, 138
moment, 490 reactionary gyroscopic moments, 146
stall flutter, 543–545 reactionless modes, 143–144
propeller-nacelle flutter, 651–652 real-valued components, 141
proportional damping, 26 reduced frequency, 449, 581
pseudogyroscopic effect, 156–157 reference diameter shaft, 163–164
pseudoinverse, 272 reference uniform blade, 79–84
pseudostatic moment, 194 regressive whirl, 141–143, 397
pseudostatic response, 213 resolvers, 567
pseudostatic shear, 197 resonance condition, 14, 28, 41
pylon resonance curve width method, 39, 41
angular moment of inertia, 390 response amplitude curve for multiple
constraint, 562–563 resonances, 42
damping in nonaeroelastic test facility, response of structures
575 environment, 2–3
degrees of freedom, 429, 559 evaluation and modification, 3
focused vibration-alleviation response trim, 318
device, 270 returning wake, 459–461, 465–469
frequency criterion, 430, 560 reversed-flow, 348–350
isotropic case, 421–424 signing function, 349
mobility characteristics, 576 Reynolds number, 519
natural frequencies, 430, 558–559 right-hand rule for relationship between
rigid-body motion, 558 rotational quantity and vector
scale factors, 429 Representation, 130
scaling, 558–560 rigid-blade formulation, 367–368
stiffnesses in pitch and roll, 430, 559 rigid-blade rotors, flutter instability
see also rotor-pylon systems; characteristics, 394
single-degree-of-freedom pylons rigid-body
mode, 32
motion of pylon, 558
quadratic factorization, 301–302 oscillations, 108
quadrature condition, 293 torsion, 108
“index” — 2005/11/17 — page 742 — #14

INDEX 743
rigid flapping mode, 75 rotor advance ratio, 316

rigid wake, 502 rotor aeroelastic analysis, 233
ripple method of solution, 320–323 rotor aeroelastic formulations,
rise factor, 72–73, 79, 96, 426 nonlinearities, 317
rise factor variation functions, 80 rotor aeroelastic operators, 694, 703
hinged blades, 81 rotor aeroelastic problem, 530
hingeless blades, 81 rotor aeroelastic response analysis, 402
Ritz/Hamilton method, 174 rotor-airframe
rocket pods, 249 coupling, 703
rotary inertia equivalences, 162 interaction in ground resonance, 416
rotary shafts, whirling motion, 379–380 rotor anisotropy, 316–317
rotary wing rotor azimuth indexing, 566
aerodynamic environment, 447–448 rotor bifilar absorbers, 254–255
flutter, 327 rotor blades
unsteady aerodynamic theory, 448–450 composite construction characteristics,
vs fixed wing, 1–2 592–593
rotating beams, 67–125 construction, 591–592
approximate methods, 95–101 elastic deflections, 67
basic equations for bending, 67–78 flapping excitation, 214
boundary conditions, 69 gyroscopic characteristics, 138–144
coupling effects, 112–121 in hover, 236–237
distributed tension, 69 mode shapes, 73
dynamic loads, 201–205 one-dimensionality, 67
elastomechanical discretization, 86 stiffness properties, 592
flapping motion, 86 rotor disk
free-body diagram of outboard portion, idealized characteristic aerodynamic
202 zones, 320
free-vibration characteristics, 72 whirling motion, 219
geometry defining torsional kinematics, rotor-dynamic instabilities, 379–384
104 rotor flapping spring rate, 559
Holzer–Myklestad method, 86 rotor-fuselage coupling
in-plane bending moment and shear, 204 dynamics, 224, 231–232
mode shapes, 73 general concepts, 225
numerical methods, 84–95 impedances and mobilities, 224–226
torsional mode shapes, 111 instability, 574–575
rotating blade lateral shear, 244–245 relationships, 232
rotating cantilevered beam, loading rotor-fuselage interactions, 222–234
distributions, 67–68 rotor hub, 387
rotating coned-blade axis system, 688 appended bifilar dyamic absorber, 255
rotating frame, 142 rotor impedance, 224, 232–233, 239
rotating to nonrotating data transmission, rotor impedance matrix, 233, 421
567 rotor-induced velocity function, 438–440
rotation angle defining principal axes, 619 rotor inflow, 503
rotation constraints, 608–610, 612 rotor inflow ratio, 215, 316
rotation point, 347 rotor-in-plane bending coupling with shaft
rotational kinematics of mass particle, 128 lateral flexibility, 102
rotational motion rotor instabilities, 335–386
accelerometers, 282 classification, 335
solid body, 127–130 dynamic description, 336–337
rotational quantities, vector representation general equations of motion, 337–338
of, 130 subharmonic, 545–546
rotational stiffness coefficient, 160 unsymmetrical, 335–343
“index” — 2005/11/17 — page 743 — #15

744 INDEX
rotor lateral edgewise excitation, 406, 693 effect of chordwise mass offset, 360
rotor lift, 563 effect of collective angle, 359
rotor lift coefficients, 563 effect of mass ratio (Lock number), 359
rotor longitudinal edgewise excitation, effect of rotor type, 358
406, 693 effect of solidity, 360
rotor mass elements, dynamic terms caused rotor whirl trend studies, 395
by, 687 rotor-wing combination, 704–706
rotor modes, 138–139, 387–389 rotors
rotor-nacelle system center of gravity, 388–389
dynamic components defining flutter, cosine torsion response, 694
402 unsteady aerodynamics and flutter,
dynamic subsystems, 401 447–527
elastomechanical description, 390–391 see also combined
multiple-degrees-of-freedom rotor/drive/engine/fuel control
frequency-domain method, 400–404 system; two-bladed rotor
substructure, 705 Routh array, 303–304, 411
whirl flutter, 389–404 very lightly damped systems, 305
wing flexibility effect, 396–397 Routh criterion, 303, 356, 358, 367, 408,
rotor pitching moment, 436 493
rotor pitchwise flatwise excitation, 694 Routh–Hurwitz criterion, 303–305, 394
rotor-pylon systems Runge–Kutta method, 324–325, 417,
aerodynamic description using 538–539
quasisteady theory, 391–393
comprehensive analyses, 400–404 St. Venant torsion theory, 606
mechanical and aeromechanical scalar formulation, 499
instabilities, 387–446 scalar matrix, 43
passive elastomechanical coupling, 706 scalar multiplication, 46, 54
stability characteristics, 394–396 scalar triple product, 56
rotor rolling moment, 436 scale factors, 429, 560, 564
rotor rollwise flatwise excitation, 694 scaling formulas for aeroelastic quantities,
rotor sine torsion response, 694 561
rotor speed, 316 scaling laws, 557–560
characteristics, 73 scatter factors for aerospace metals, 295
limiting case, 74 Sears function, 455–457
vs initial curvature, 152 second-order backward-differencing, 537
vs mass unbalance, 151 secondary aeroelastic responses, 244
zero, 76 secondary mass, 239
rotor stiffness and damping, dynamic terms secondary section components, 616–617
caused by, 690–670 section centers, 617
rotor stiffness ratio, 154 section moment, pitch rate effects, 347
rotor systems, elastomechanical semianalytical method, 193
properties, 317 semicanonical form, 309
rotor testing see model rotor testing semipositive-definite matrices, 315
rotor thrust, 436 shafts
rotor unsteady airloads theories, configuration with significant distributed
classification, 450–451 inertia, 167
rotor wake, 502–503 critical speeds, 149–162
division into upper and lower parts, 221 density, 167
geometry, 496 with significant mass, 167–168
rotor weaving instability, 352–362, 386 shake testing, 275–292, 677
effect of chordwise aerodynamic offset, advanced measuring devices, 282
361 basic single-point, 275–276
“index” — 2005/11/17 — page 744 — #16

INDEX 745
data acquisition, 283–284 skewing degradation factor, 412

fuselage configuration, 277–278 slip ring assemblies, 567
fuselage suspension, 278–280 small-amplitude limit-cycle oscillations,
instrumentation, 281–284 588
limitations, 276 Söhngen’s inversion formula, 463
simplest form, 277 solidity effect, 362
test procedures, 288–292 source alleviators, 251
test setup, 277–281 Southwell coefficient, 72–74, 96–97,
shakers, 567 690–691
shape function matrix, 93 limiting case
shape functions, 93–94 in-plane modes, 75
shape-memory alloys, 246 out-of-plane modes, 75
shear, 201 uniform beams, 76
shear bearing components, 619 offset correction factors, 668–669, 672
shear center, 606–615, 619 variation with rotor speed, 78
shear coefficient, 193 zero offset, 671, 674, 676
shear distribution, 199 Southwell diagram, 73
shear flows, 606–608, 614–615 Southwell hyperbolic approximation of
rotation constraint, 612 frequency, 690
translation constraint, 612–613 space-curve character, 106
shear modulus, effects of temperature and spanwise
moisture concentration, 620 element
shear strain, 610–611, 615 equilibrium, 67–68
shed vorticity, 507–508 free-body diagram, 68, 88
shed-wake vorticity, 500 massless elastic element, free-body
shift factor, 581–582 diagram, 89
signal analyzers, 288 transfer matrices, 91
simple harmonic motion, 14, 38 variable, 77
simple linearization, 530–535 variable coefficients, 70
simultaneous equations, 50–51 variations in chordwise mass center and
alternate methodology, 51 shear center locations, 120
sine chirp, 291 spectrum analyzer, 284, 286
single-cell beams, 594 speed ratio, 153, 155
theory, 600–601 spherical elastomeric bearing, 579–581
single-degree-of-freedom equations, 205, spin angular momentum, 136
209 spin momentum vector alignment, 131
single-degree-of-freedom flutter in blade spinning tail rotor, 144
torsion, 465 spinning top dynamics, 135–138
single-degree-of-freedom pylons split tip-path plane oscillation, 545
exact stability boundaries, 411 spring-mass-damper system, 33, 311
ground resonance instability regions, configuration, 20
411 two-dimensional airfoil, 458
single-degree-of-freedom systems, 236, spring-mass element, free-body diagram,
301, 307 85
single free-free mode, 165 square matrices, 43
single-frequency motion, 586 stability
single-frequency operation, 584 measuring instrumentation,
single pass method, 324–325 564–567
single-point excitation, 277 see also instability
sinusoidal function, 213 stability analysis
sinusoidal load, particular solution, 11 basic concepts, 297–299
skew-symmetrical matrix, 44 linear systems, 297–334
“index” — 2005/11/17 — page 745 — #17

746 INDEX
stability boundaries, 302–305, 397, 409, structural coupling factors, 371–372

424, 493 structural damping, 37–41
stability characteristics, 366–367 approximations, 37–38
parameter variations, 357–362 descriptors, 275
rotor-pylon systems, 394–396 multiple-degrees-of-freedom systems,
stability criteria, comparison of, 327 38–39
stability critical eigenvalue components, structural damping coefficient, 38–41,
645 306
stability descriptors, 406 structural dynamics, methodology, 2–4
stability diagram structural modification methodology,
rotors with equal elastic and inertia 239–244
inequalities, 341 structural response velocity data,
rotors with inertia inequality and 39–40
gyroscopic coupling, 340 structural synthesis using analytical
rotors with stiffness inequality and testing, 250–251
gyroscope coupling, 340 subcritical operation, 154–155
stability estimation, numerical methods, subharmonic rotor instability, 545–546
548–553 submatrices, 51–52
stability evaluator, 568 substructures
stability indices, 328–330 coupling, 678–679
stability margins, 330–331 impedance connections between,
stability solution, 356 682–683
stabilizing effect of gyroscopic terms, symbolic mapping, 680
314–315 syntheses, 251
stall flutter, 508–523, 532–535, 542 subsystems, mobility coupling, 679–682
stalled airfoil conditions, 448 supercritical operation, 154–156
Stanford Research Systems Model SR785, supercritical rotor, 407
289 suspension device, 277
start-up techniques, 540 symmetric laminates, 712
starting equation, 170 flexural stiffness, 715
static-stall characteristics, 509–510 sandwich, 712
stationary frequencies, 286 stress-strain conditions, 715–717
steady bending deflections, 701 symmetrical matrix, 44
steady load, particular solution, 11
steady-state solution, 151–152 table look-up schemes, 347, 477
step function table of functions, 168
damped spring-mass system, 212 tail excitations, 227–228
undamped spring-mass system, 212 teeter angles, 145
stiffness teetering rotor, 357, 385–386
anisotropy, 342 elastomechanical description,
characteristics, 708 352–353
inequality, 338 tensile modulus, effects of temperature and
matrix, 177, 180, 539, 593–594 moisture concentration, 620
properties of rotor blades, 592 tension-torsion coupling, 620–621
relationships off-axis, 710–717 Theodorsen formulation, modification,
strain gauges, 281, 565 455
stream tube air flow through rotor Theodorsen function, 451–455, 457, 464,
disk, 435 469–476
stress fields for two-dimensional approximate expression, 452
plane-stress elements, 709 basic characteristics, 452
stress-strain relationship, 710, 713, generalized, 471–472
715–717 real and imaginary parts, 453
“index” — 2005/11/17 — page 746 — #18

INDEX 747
thin-shell closed-cell construction beams, transient solutions using numerical

594–605 integration, 535–546
axis locations, 596 transition matrix, 318, 323–325
contour points and geometry of translation constraints, 610–614
segments, 598 translational stiffness coefficient, 160
elements of CS(6)s matrix, 596–598 transmissibility characteristics, 258–259,
third-order backward-differencing, 537 261–262, 264–267
three-dimensional aerodynamics theories, transverse bending, 68–70
496–508 transverse deformation of flexible shaft
basic mathematical structure, 497–498 segment, 159
three-inertia (free-free) problem, 166 transverse load distribution, 69, 205
three-inertia shafting configuration, 166 transverse vibration analysis, lumped-mass
three-quarter-chord point, 346, 454, 459 approach, 158–162
thrust equation, 437, 440 trend information, 240–241, 243
thrusting propeller state, 438 trim conditions, 378
tilt angles, 139–140, 142 trim responses, 317
tilt-rotor concept, 147 trim solutions, 317–318
time-domain simulations, 542 triple products, 56
time-histories, 195–196, 207, 324, trunk inertia equation, 179
476–479, 542–546, 548–549 tuned absorber, 261–262
Timoshenko beam theory, 158 turbulent windmill state, 439
Timoshenko technique, 168 two-bladed rotor, 101–104, 421–424
tip-clearance excitation, 382–383 asymmetrically deflected, 422
torsion dynamic operational conditions,
coupling effects, 118–121 103–104
modal characteristics, 111 dynamic properties, 102
torsional blade segment, 109 equations of motion: isotropic pylon
torsional deflection, 174 case, 421–424
torsional element, mathematical frequency characteristics, 422
description, 175 out-of-phase and in-phase frequency
torsional inertia, 617 variations with lateral stiffness,
torsional mode shapes for rotating beam, 102–103
111 two-component projection vectors, 141
torsional motion, 485–486 two-degrees-of-freedom gyro, dynamic
torsional natural frequencies, 162–180 equations for, 131–134
torsional pendulum absorber, bifilar two-degrees-of-freedom system, 148, 306,
configuration, 182 309
torsional stiffness, 162–164, 606–615, 619 basic configuration, 20
torsional system, effect of gears, 164 equations of motion, 20
torsional vibration Lagrangian approach, 20–21
drive systems, 162 Newtonian approach, 20
mode shapes, 111 spring-mass system, 23–24
trailing blade-tip vortex, 221 summary of results, 494–495
trailing vorticity, 508 two-dimensional aerodynamic damping,
trailing-wake vorticity, 500 surface representation, 516
transducer calibration, 282–283 two-dimensional aerodynamic
transfer function, 286, 307–308 pitch-damping coefficient, 515
biquadratic filter, 632 two-dimensional airfoil
matrices, 416 drag characteristics, 510
transfer matrix, 85, 87, 91–92, 170 lift characteristics, 510–512
equation, 91 moment characteristics, 511, 513
transient loads, 471 spring-mass damper, 458
“index” — 2005/11/17 — page 747 — #19

748 INDEX
two-dimensional arbitrary motion theories, measurement, 256

469–485 minimization, 245–246, 644
two-dimensional frequency-domain vibration absorbers, 589
theories, 451 optimal operation, 29
two-dimensional theory, results of early vibration-suppression devices, 251–270
investigators, 479–480 fuselage-related, 258–270
two-dimensional wing rotor-related, 251–258
basic flutter analysis, 486–488 vibration testing, 275–296
stability solution, 487 commercially available equipment, 288
two-inertia problem, 165–167 noise, 288
simplest form, 276
vibratory environment, 2
umbrella (or collective) mode, 143–144 vibratory load factor, 250
uncoupled modes, 114–115 vibratory moments, 270
undamped natural frequency, 10, 300 vibratory shears, 270
uniform blades, detailed characteristics, vibratory torque amplitude, 182
73–78 virtual inertia, 433
uniform inflow, 344 virtual mass, 433
unit matrix, 43 viscoelastic behavior, effects of frequency
unit vector, 53–55 and temperature, 582
universal joint, 184 viscous fluid, hydrodynamic
unsteady aerodynamic load distribution, characteristics, 381
490 vortex flow, 498
unsteady airfoil data synthesis, 516–519 vortex-ring state, 437
unsteady coefficients, time-domain vortex structure for lifting airfoil section,
functionality, 518 500
unsteady decay parameter, 475–482
unsteady rotor aerodynamic problem, 497 Wagner function, 469–476, 480, 482, 487
wake-induced lift, 508
vectors wake skew angle, 502–503
addition (and subtraction), 53, 57 warping assumption, 601–604
algebra, 54–56 warping effect, filters, 638
angular momentum, 130 warping function
angular velocity, 130 generalized, 603
calculus, 52–59 torsion-related, 603
coplanar, 57 whirl amplification factor, 151–152, 154
definitions, 52–53 whirl amplification function, 153
derivatives, 56 whirl modes, 139–143
differentiation, 56–57 mathematical description, 142
equation, 59–60 whirl ratio, 381–382
formulation, 498–499 whirling motion, 382–384
generalized coordinates, 229 rotary shaft, 379–380
mathematical operations, 53–57 rotor disk, 219
multiplication, 55 wideband frequency testing, 291
products, 49 windmill brake state, 438
rotational quantities, 130 wing bending
triple product, 56, 128 and torsion motion variables, 705
vertical flight conditions, 438 degrees of freedom, 705–706
vertical hub loads, 251 equation, 705
vibration wing flexibility, effect on rotor-nacelle
characteristics, 75 whirl flutter, 396–397
control, 239–274 wing motion degrees of freedom, 705
“index” — 2005/11/17 — page 748 — #20

INDEX 749
wing-root bending moment, 195 uniform beam approximation, 97

wing-rotor system, 147
wing torsion equation, 705 z transform, 635, 638–639
wing unsteady aerodynamic loading, 401 zero air density eigenvalue solution,
207–208
Yntema charts, 95–97 zero offset Southwell coefficient, 667, 671,
equivalent stiffness ratios for use with, 674, 676
100 zero rotational speed frequencies, 79
nonuniform beam approximation, 96 zero rotor speed, 76
“index” — 2005/11/17 — page 749 — #21

Rotary Wing Structural Dynamics and Aeroelasticity

Uploaded by

Copyright:

Available Formats

Rotary Wing Structural Dynamics and Aeroelasticity

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Rotary Wing Structural Dynamics and Aeroelasticity

Uploaded by

Copyright:

Available Formats

Rotary Wing

“prelims” — 2005/11/22 — page i — #1

“prelims” — 2005/11/22 — page iii — #3

Library of Congress Cataloging-in-Publication Data on File

“prelims” — 2005/11/22 — page iv — #4

To my wife, Joan, ever supportive, ever encouraging, and ever patient!

“prelims” — 2005/11/22 — page v — #5

Robert H. Bishop Rakesh K. Kapania

Rickard B. Heslehurst David K. Schmidt

David M. Van Wie

“prelims” — 2005/11/22 — page vii — #7

“prelims” — 2005/11/22 — page ix — #9

Preface to the Second Edition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv

Preface to the First Edition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii

Chapter 2 Basic Analytical Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Chapter 3 Rotating Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

Chapter 4 Gyroscopics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

Chapter 5 Drive System Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

“prelims” — 2005/11/22 — page xi — #11

Chapter 6 Fuselage Vibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189

Chapter 7 Methods for Vibration Control. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

Chapter 8 Vibration Test Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

Chapter 9 Stability Analysis Methods: Linear Systems . . . . . . . . . . . . . . . 297

Chapter 10 Mechanical and Aeromechanical

Chapter 11 Mechanical and Aeromechanical Instabilities of

Chapter 12 Unsteady Aerodynamics and Flutter of Rotors . . . . . . . . . . . . 447

“prelims” — 2005/11/22 — page xii — #12

Chapter 13 Analysis of Nonlinear Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529

Chapter 14 Model Rotor Testing for Aeroelastic Stability . . . . . . . . . . . . . . 557

Chapter 15 Elastomeric Devices for Rotorcraft . . . . . . . . . . . . . . . . . . . . . . . . . . 579

Chapter 16 Blade-Section Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 591

Chapter 17 Cross-over Topics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625

Chapter 18 Concluding Thoughts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651

Appendix A Glossary of Rotorcraft-Related Terms . . . . . . . . . . . . . . . . . . . . . . 657

Appendix B Charts for Blade Frequency Estimation . . . . . . . . . . . . . . . . . . . . 665

Appendix C Generalized Frequency-Domain

“prelims” — 2005/11/22 — page xiii — #13

Appendix D Basic Equations of Motion for Ground Resonance and

Appendix E Composite Materials—Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 707

Supporting Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 751

“prelims” — 2005/11/22 — page xiv — #14

“prelims” — 2005/11/22 — page xv — #15

“prelims” — 2005/11/22 — page xvii — #17

“prelims” — 2005/11/22 — page xviii — #18

Flannelly, and Robert Loewy. The Rensselaer Polytechnic Institute Rotorcraft

“prelims” — 2005/11/22 — page xix — #19

1.1 Rotary Wing vs Fixed Wing

“chapter_1(new)” — 2005/10/18 — page 1 — #1

1.2.1 Knowledge of the Environment

1.2.2 Calculating the Responses

“chapter_1(new)” — 2005/10/18 — page 2 — #2

1.2.3 Evaluation and Modification

1.2.4 Organization of the Text

“chapter_1(new)” — 2005/10/18 — page 3 — #3

“chapter_1(new)” — 2005/10/18 — page 4 — #4

2.1 Linear Single-Degree-of-Freedom System

U (mass, m) = lbf-s2 /ft (2.1a)

U (damper, c) = lbf-s/ft (2.1b)

U (spring, k) = lbf/ft (2.1c)

U [applied force, F(t)] = lbf (2.1d)

F(t) = P0 cos t = Re[P0 eit ] (2.25)

xp = A cos t + B sin t (2.26)

ẋp = (−A sin t + B cos t) (2.27a)

ẍp = −2 (A cos t + B sin t) = −2 xp (2.27b)

[(ωn2 − 2 )A + 2ζ ωn B] cos t

xp (t) = X cos(t − φ) = X[cos φ cos t + sin φ sin t] (2.30)

tan φ = B/A = 2ζ ωn /(ωn2 − 2 ) (2.31c)

1) The sharp rise in amplitude response occurs when ≈ ωn is denoted the

3) As → ∞, all responses attenuate.

F(t) = Fnc cos nt + Fns sin nt n = 1, 2, 3, . . . (2.35)

xp (t) = An cos nt + Bn sin nt (2.36)

Fnc (ωn2 − n2 2 ) − Fns 2ζ ωn n