Numerical Ship Hydrodynamics - An Assessment of The

Lecture Notes in Applied and Computational Mechanics 94
Takanori Hino · Frederick Stern ·

Lars Larsson · Michel Visonneau ·
Nobuyuki Hirata · Jin Kim Editors
Numerical Ship
Hydrodynamics
An Assessment of the Tokyo 2015
Workshop
Lecture Notes in Applied and Computational
Mechanics
Volume 94
Series Editors
Peter Wriggers, Institut für Kontinuumsmechanik, Leibniz Universität Hannover,
Hannover, Niedersachsen, Germany
Peter Eberhard, Institute of Engineering and Computational Mechanics, University
of Stuttgart, Stuttgart, Germany
This series aims to report new developments in applied and computational
mechanics—quickly, informally and at a high level. This includes the fields of fluid,
solid and structural mechanics, dynamics and control, and related disciplines. The
applied methods can be of analytical, numerical and computational nature. The
series scope includes monographs, professional books, selected contributions from
specialized conferences or workshops, edited volumes, as well as outstanding
advanced textbooks.
Indexed by EI-Compendex, SCOPUS, Zentralblatt Math, Ulrich’s, Current
Mathematical Publications, Mathematical Reviews and MetaPress.
More information about this series at http://www.springer.com/series/4623

Takanori Hino Frederick Stern
• •
Lars Larsson Michel Visonneau

• •
Nobuyuki Hirata Jin Kim

•
Editors
Numerical Ship
Hydrodynamics
An Assessment of the Tokyo 2015 Workshop
123
Editors
Takanori Hino Frederick Stern
Faculty of Engineering IIHR – Hydroscience & Engineering
Yokohama National University University of Iowa
Yokohama, Japan Iowa City, IA, USA
Lars Larsson Michel Visonneau

Mechanics and Maritime Sciences Fluid Mechanics Laboratory
Chalmers University of Technology Ecole Centrale de Nantes
Gothenburg, Sweden Nantes, France
Nobuyuki Hirata Jin Kim

National Maritime Research Institute Korea Research Institute of Ships and Oc
Mitaka, Tokyo, Japan Daejeon, Korea (Republic of)
ISSN 1613-7736 ISSN 1860-0816 (electronic)

Lecture Notes in Applied and Computational Mechanics
ISBN 978-3-030-47571-0 ISBN 978-3-030-47572-7 (eBook)
https://doi.org/10.1007/978-3-030-47572-7
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Switzerland AG 2021
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with regard
to jurisdictional claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Since 1980, workshops on CFD in Ship Hydrodynamics have been held regularly.
The main purpose of these workshops is to assess the state of the art in CFD for
hydrodynamic applications. Active researchers in the field worldwide are invited to
provide computed results for a number of well-specified test cases, and the orga-
nizers collect and present the results such that comparisons between different
methods can be made easily. Detailed information about each method is also
reported via a questionnaire provided by the organizers. All results are discussed at
a meeting, and a final assessment of the workshop is made by the organizers.
The Tokyo 2015 Workshop attracted 36 groups from all over the world, and
different types of computations were carried out for three hulls. It was the largest
of the workshops in the series so far. All computed results were compiled in a
volume, called Proceedings II, and distributed at the meeting, which was held in
Tokyo in December 2015. The volume also includes short papers describing the
computational methods and the results in more detail.
In the present book, in-depth evaluations of all computed results are presented.
For some of the test cases, additional computations by the organizers are presented
on topics of particular interest found at the meeting. All experimental data are
reported, as well as a comprehensive set of new data is obtained after the
workshop. The book has been written by the organizers and their co-workers.
Supplementary materials are available for free at https://www.springer.com/gp/
book/9783030475710 the book constitutes the final documentation of the Tokyo
2015 Workshop and gives a state-of-the-art assessment of the CFD capabilities
within the area of Ship Hydrodynamics.
Yokohama, Japan Takanori Hino

Mitaka, Japan Nobuyuki Hirata
Iowa City, USA Frederick Stern
Gothenburg, Sweden Lars Larsson
Nantes, France Michel Visonneau
Daejeon, South Korea Jin Kim
February 2020
v
Contents
Introduction, Conclusions and Recommendations . . . . . . . . . . . . . . . . . 1

Takanori Hino, Nobuyuki Hirata, Frederick Stern, Lars Larsson,
Michel Visonneau, and Jin Kim
Experimental Data for JBC Resistance, Sinkage, Trim,
Self-Propulsion Factors, Longitudinal Wave Cut and Detailed Flow
with and without an Energy Saving Circular Duct . . . . . . . . . . . . . . . . 23
Nobuyuki Hirata, Hiroshi Kobayashi, Takanori Hino, Yasuyuki Toda,
Moustafa Abdel-Maksoud, and Frederick Stern
Experimental Data for KCS Resistance, Sinkage, Trim,
and Self-propulsion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Jin Kim
Experimental Data for KCS Added Resistance and ONRT Free
Running Course Keeping/Speed Loss in Head and Oblique Waves . . . . 61
Yugo Sanada, Claus Simonsen, Janne Otzen, Hamid Sadat-Hosseini,
Yasuyuki Toda, and Frederick Stern
Evaluation of Resistance, Sinkage, Trim and Wave Pattern
Predictions for JBC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Lars Larsson
Analysis of the Local Flow around JBC . . . . . . . . . . . . . . . . . . . . . . . . . 159
Michel Visonneau
Evaluation of Self-propulsion and Energy Saving Device Performance
Predictions for JBC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
Takanori Hino
vii
viii Contents
Evaluation of Resistance, Sinkage, Trim and Self-propulsion

Predictions for KCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
Jin Kim
Assessment of CFD for KCS Added Resistance and for ONRT Course
Keeping/Speed Loss in Regular Head and Oblique Waves . . . . . . . . . . . 333
Frederick Stern, Hamid Sadat-Hosseini, Timur Dogan, Matteo Diez,
Dong Hwan Kim, Sungtek Park, and Yugo Sanada
Contributors
Moustafa Abdel-Maksoud Hamburg University of Technology, Hamburg,

Germany
Matteo Diez CNR-INM, National Research Council-Institute of Marine
Engineering, Rome, Italy
Timur Dogan University of Iowa and IIHR-Hydroscience & Engineering, Iowa
City, IA, USA
Takanori Hino Yokohama National University, Yokohama, Japan
Nobuyuki Hirata National Maritime Research Institute, Mitaka, Japan
Dong Hwan Kim Korea Research Institute of Ocean Science & Technology,
Daejeon, South Korea
Jin Kim Korea Research Institute of Ships and Ocean Engineering (KRISO),
Daejeon, South Korea
Hiroshi Kobayashi National Maritime Research Institute, Mitaka, Japan
Lars Larsson Chalmers University of Technology, Gothenburg, Sweden
Janne Otzen FORCE Technology, Lyngby, Denmark
Sungtek Park University of Iowa and IIHR-Hydroscience & Engineering, Iowa
City, IA, USA
Hamid Sadat-Hosseini University of North Texas, Denton, TX, USA
Yugo Sanada University of Iowa and IIHR-Hydroscience & Engineering, Iowa
City, IA, USA
Claus Simonsen FORCE Technology, Lyngby, Denmark
ix
x Contributors
Frederick Stern University of Iowa and Iowa Institute of Hydraulic Research

(IIHR), Iowa City, IA, USA
Yasuyuki Toda Osaka University, Suita, Japan
Michel Visonneau CNRS/Centrale Nantes, Nantes, France
Introduction, Conclusions
and Recommendations
Takanori Hino, Nobuyuki Hirata, Frederick Stern, Lars Larsson,

Michel Visonneau, and Jin Kim
Abstract The Tokyo 2015 Workshop on CFD in Hydrodynamics was the seventh
in a series started in 1980. The purpose of the Workshops is to regularly assess the
state of the art in Numerical Hydrodynamics and to provide guidelines for further
developments in the area. The 2015 Workshop offered 16 test cases for three ship
hulls. A total of 36 participating groups of CFD specialists submitted their computed
results during the fall of 2015. The results were compiled by the organizers and
discussed at a meeting in Tokyo in December 2015. In this chapter the background
and development of the Workshops since the start are presented. The three hulls
used in the 2015 Workshop are introduced and the computations requested from the
participants are specified. Based on a questionnaire sent to all participants the details
of their CFD methods are listed, and finally the general conclusions from each chapter
and recommendations for future Workshops are presented. The detailed results of
the computations are discussed in subsequent Chapters.
T. Hino (B)
Yokohama National University, Yokohama, Japan
e-mail: [email protected]
N. Hirata
National Maritime Research Institute, Mitaka, Japan
F. Stern
University of Iowa and Iowa Institute of Hydraulic Research (IIHR), Iowa City, IA, USA
L. Larsson
Chalmers University of Technology, Gothenburg, Sweden
M. Visonneau
CNRS/Centrale Nantes, Nantes, France
J. Kim
Korea Research Institute of Ships & Ocean Engineering (KRISO), Daejeon, South Korea
© The Editor(s) (if applicable) and The Author(s), under exclusive license 1
to Springer Nature Switzerland AG 2021
T. Hino et al. (eds.), Numerical Ship Hydrodynamics, Lecture Notes in Applied
and Computational Mechanics 94, https://doi.org/10.1007/978-3-030-47572-7_1
2 T. Hino et al.
1 Background
The history of CFD Workshops dates to 1980. That year, the first Workshop was
held in Gothenburg, Sweden (Larsson 1981). The objective of the workshop was
the assessment of up-to-date numerical methods for ship hydrodynamics to aid code
development and guide industry.
Several test cases were set for viscous flows around ship hulls and the participants
were requested to submit their numerical results. Through the comparison with exper-
imental results both for integral variables such as resistance or self-propulsion factors
and local flow quantities such as wake or pressure distributions, the state-of-the-art
of ship CFD methods was evaluated. The objective and the style of the workshop
remained the same in the following workshops, which were held approximately every
five years between Gothenburg and Tokyo, i.e. at Gothenburg in 1990 (Larsson et al.
1991), (Larsson et al. 2002, 2003) and (Larsson et al. 2014) and at Tokyo in 1994
(Kodama et al. 1994) and (Hino 2005).
In the 35 years from the first workshop, computer technology has progressed
exponentially, and available computing resources have increased several orders of
magnitude. Numerical flow analysis methods have also evolved from boundary-layer-
theory-based methods to Navier-Stokes methods and from resistance prediction to
unsteady complex flow simulations with waves, propellers and ship motions.
In 2015, again at Tokyo, the seventh CFD workshop in the series was organized.
The purpose of the workshop was the same as at the preceding workshops and it was
to assess state-of-the-art of the current CFD codes for ship hydrodynamics. In the
present workshop, three ship hulls, two of which are new, were selected and a total of
16 test cases were specified by the organizers. One of the new ship hulls is the Japan
Bulk Carrier (JBC) which is a cape-size bulk carrier equipped with a stern duct as
an energy saving device. The other new hull is the ONR Tumblehome model 5613
(ONRT) which is a preliminary design of a modern surface combatant. The third hull
is the KRISO Container Ship (KCS) which is a 3,600 TEU container ship and has
been used in the preceding workshops. Like in all previous workshops, participants
were asked to provide computed results, information about the method used and a
paper summarizing the computations. 36 groups submitted their computed results
for one or more cases.
The present chapter gives a background to the Workshop, specifications of the hulls
and test cases followed by the summary of the numerical methods the participants
used. General conclusions and recommendations for future work are also included in
this chapter. Chapter 2 is a report of the experimental data for JBC, including resis-
tance and self-propulsion tests, local flow measurement using stereo particle image
velocimetry (SPIV) with and without a stern duct and wave profiles. Chapter 3 is a
summary of the resistance and self-propulsion tests of the KCS. Chapter 4 is devoted
to the experimental data for added resistance in waves of the KCS and the ONRT
hulls. Chapters 5–9 are the evaluation reports of the submitted data by the organizers.
Chapters 5 and 7 are the evaluations of the resistance and self-propulsion simula-
tions, respectively, for the JBC hull. Chapter 6 evaluates the local flow predictions for
Introduction, Conclusions and Recommendations 3
the JBC. Chapter 8 is the evaluations of resistance and self-propulsion simulations

for the KCS. Chapter 9 is the evaluations of added resistance and course keeping
simulations for the KCS and the ONRT. It should be noted that these reports are
based solely on results submitted for evaluation at the Workshop. No results from
submitted papers or from computations carried out afterwards are included except the
additional analysis of the SPIV measurement in Chapter 2 (Figs. 17, 19-21, 35–38)
and the vortex flow analysis in Chap. 6 (Figs. 50–53, 70–79).
Similarly with the previous Gothenburg 2010 Workshop, this book replaces
Proceedings, Part 1 of the Workshop. The detailed results and papers by all partic-
ipants are presented in the Proceedings, Part 2, distributed at the Workshop. These
Proceedings can be downloaded from the web site SpringerExtras (for address, see
the book cover) together with much other supplementary information, such as the
spreadsheet with detailed information about all methods, discussions at the Workshop
and some photos.
2 Hulls
The following three hulls were used in the Workshop:

(1) The JBC, Japan Bulk Carrier
(2) The KCS, KRISO Container Ship
(3) The ONRT, ONR Tumblehome model 5613.
The JBC (Japan Bulk Carrier) is a cape-size bulk carrier equipped with a stern duct
as an energy saving device. National Maritime Research Institute (NMRI), Yokohama
National University and the Ship Building Research Centre of Japan (SRC) were
jointly involved in the design of a ship hull, a duct and a rudder. The hull design and
measurements were conducted with the support of ClassNK as part of the ClassNK
joint R&D for Industry Program (Hino et al. 2016). Towing tank experiments were
carried out at NMRI (Hino et al. 2016) and Osaka University (Jufuku et al. 2015),
including resistance tests, self-propulsion tests and SPIV measurements of stern flow
fields. Additionally, LDV/SPIV measurement data in the wind tunnel are available
from Technical University of Hamburg (Shevchuk et al. 2020).
The KCS (KRISO Container Ship) was conceived to provide data for both explica-
tion of flow physics and CFD validation for a modern container ship with a bulbous
bow, and towing tank tests were carried out at Korea Research Institute of Ships
and Ocean Engineering (KRISO) in Daejeon (Van et al. 1998, Kim et al. 2001).
Self-propulsion tests were carried out at the Ship Research Institute (now NMRI) in
Tokyo and are reported in the proceedings of Tokyo 2015 Workshop (Hino 2005).
Later, resistance tests were also reported by NMRI (See Larsson et al. 2014). Data for
pitch, heave, and added resistance are also available from Force/DMI measurements
(Simonsen et al. 2008).
The ONRT (ONR Tumblehome model 5613) is a preliminary design of a modern
surface combatant, which is publicly accessible for fundamental research. The 1/49
4 T. Hino et al.
Fig. 1 The three ships used in the workshop (top: JBC; middle: KCS; bottom: ONRT)
scaled ship model is appended with skeg and bilge keels. The model has a wave
piercing hull design with 10° tumblehome sides and a transom stern. The model also
has rudders, shafts and propellers with propeller shaft brackets, but the superstructure
is not attached. Free-running tests including course keeping, zig-zag and turning in
calm water and in regular waves were performed at IIHR Hydraulics Wave Basin
Facility (Sanada et al. 2013, 2019).
Side views of the three hulls are seen in Fig. 1 and the main particulars are given
in Table 1. No full-scale ships exist for all three hulls.
The Cartesian coordinate system adopted in the workshop has its origin at the
forward perpendicular and still water level, x is backwards, y to starboard and z
vertically upwards, as shown in Fig. 2.
3 Test Cases
Several types of computations were requested, namely:

(1) Resistance, sinkage and trim either at zero sinkage and trim (denoted FX0 ) or
dynamic sinkage and trim (FRzθ )
(2) Local flow at fixed condition, either at FX0 or FRzθ . In some cases FX0 is
simulated by pre-ballasting the hull to obtain zero sinkage and trim in the free
condition at the correct Froude number (FR0 ).
(3) Self-propulsion at FX0 , FRzθ or FR0
(4) Captive tow in waves either at FRzθ or with free surge, heave, roll and pitch
(FRxzθ )
(5) Free running with all degrees of freedom (FRall ).
All test cases for the three hulls are listed in Table 2. The measured data are
provided by the organizations listed. A total of 16 cases are specified and the
participants were free to select which cases to compute.
Table 1 Main particulars of the three ships

Main particulars (Full scale) JBC KCS ONRT
Length between L PP (m) 280.0 230.0 154.0 (L WL )
perpendiculars
Maximum beam of B (m) 45.0 32.2 18.78
waterline
Draft T (m) 16.5 10.8 5.494
Displacement volume ∇ (m3 ) 178369.9 52030.0 8507 ton
Wetted area of a bare SW (m2 ) 19556.1 9424.0
hull
Wetted surface area of a S R (m2 ) 164.3 115.0
rudder
Wetted surface area of S ESD (m2 ) 77.8 – –
ESD
Block coefficient (C B ) ∇/(L PP ·B·T ) 0.8580 0.6505 0.535 (L WL )
Propeller center, long. x/L PP 0.9846 0.9825 0.9267 (L WL )
location(from FP)
Propeller center, vert. −z/L PP 0.0404214 0.02913 0.03565 (L WL )
location (below WL)
Propeller center, lateral y/L PP 0 0 ±0.02661 (L WL )
location (from CL)
LCB (%L PP ), fwd+ – 2.5475 −1.48 79.52 m from FP
Vertical center of KG (m) 13.29 14.35 7.63
gravity(from keel)
Metacentric height GM (m) 5.30 0.60 2.07
Moment of inertia K xx /B 0.40 0.40 0.444
Moment of inertia K yy /L PP , K zz /L PP 0.25 0.25 0.25 (L WL )
Speed U (knots) 14.5 24.0 15.1
Froude number Fr 0.142 0.26 0.20
Fig. 2 Cartesian coordinate system

6 T. Hino et al.
Table 2 Test cases

Case Hull Condition Attitude Validation Data provider
variables
1.1a JBC w/o ESD Towed in calm FRzθ Resistance, NMRI
water sinkage and trim
1.2a JBC with ESD Towed in calm FRzθ Resistance, NMRI
1.3a JBC w/o ESD Towed in calm FRzθ Mean velocities, NMRI
water turbulence and
wave patterns
1.3b JBC w/o ESD Double model in FX0 Mean velocities TUHH
wind tunnel
1.4a JBC with ESD Towed in calm FRzθ Mean velocities NMRI
water
1.5a JBC w/o ESD Self-propelled in FRzθ Thrust, torque, NMRI
calm water RPM or force
balance and
resistance
1.6a JBC with ESD Self-propelled in FRzθ Thrust, torque, NMRI
calm water RPM or force
balance and
resistance
1.7a JBC w/o ESD Self-propelled in FRzθ Mean velocities NMRI
calm water
1.8a JBC with ESD Self-propelled in FRzθ Mean velocities NMRI
calm water
2.1 KCS Towed in calm FRzθ Resistance, KRISO/NMRI
2.5 KCS Self-propelled in FR0 Thrust, torque and NMRI
calm water propulsive
coefficients
2.7 KCS Self-propelled in FR0 Mean velocities NMRI
calm water
2.10 KCS Captive towed in FRzθ Resistance, heave FORCE
head waves and pitch
2.11 KCS Captive towed in FRxzφθ Resistance, surge, IIHR
beam, follow and heave, roll and
oblique waves pitch
3.9 ONRT Free running in FRall Thrust, torque, IIHR
calm water propulsive
coefficients,
sinkage and trim
head waves rpm, motions and
trajectory
(continued)
Table 2 (continued)
Case Hull Condition Attitude Validation Data provider
variables
beam, follow and rpm, motions and
oblique waves trajectory
4 Participants and Methods
The workshop participants together with the main features of their methods are listed
in Table 3. In the first column the acronym of the participating group is given. This is
used in combination with the code name of column three to identify each submission.
The cases computed are given in column two. In the remaining columns the features
of each method are given.
Types of the codes used and their numbers are in-house codes (12), open-source
codes (11) and commercial codes (13). In the previous Workshop, they were in-house
(14), open-source (3) and commercial (16). The number of open-source code users
has increased considerably.
For turbulence models, the majority of methods use two-equation models, k-ω
SST or k-ε. There are also some one-equation models, either Spalart-Allmaras or
Menter. The anisotropic models are either of the algebraic stress or Reynolds stress
type. Note that there are also some LES/DES methods.
Most of the participants use no-slip wall boundary conditions, but there are also
several methods, particularly for open-source and commercial codes, with wall func-
tions, both with and without pressure gradient corrections. The Volume of Fluid
(VOF) technique is the most popular one for the free-surface modeling, but there
are also several level set (LS) methods. There is only one entry with surface fitting
approach. The propeller is represented either as an actual rotating propeller or through
a body force approximation.
Most simulations were performed using finite volume discretization but there are
a few methods which adopt the finite difference or finite element method. For spatial
accuracy, 2nd order accurate schemes were used in most codes and limited studies
used 3rd or 4th order schemes. Time accuracy is either 1st or 2nd order. The most
common grids were unstructured. Multi-block structured or overlapping structured
grids were also used to some extent. Most methods are pressure based but there are
also several solving the equations directly or with an artificial compressibility.
A complete specification of each method and application is given in the supple-
mentary material on SpringerExtras based on the replies to a questionnaire answered
by all participating groups.
8
Table 3 Workshop participants and methods

Organization Cases Code Turbulence model Wall Free Propeller Grid Spatial Spatial Time Vel-press
submitted model surface type disc. accuracy accuracy coupling
ABS 1.1a-8a OpenFOAM SST W VOF A OS, FV 2 2 PR
3.9 3.12 U
3.13
Chalmers/FLOWTECH 1.1a-8a SHIPFLOW5.2 EASM N BL OS FV M AC
CNR-INSEAN 1.3a-8a Xnavis SA N LS BX,A OS FV 2 2 AC
CSSRC 1.5a 1.6a FLUENT RNG-KE W VOF A MS FV 2 2 PR
3.9 3.12 +
MU
CTO 2.1 2.5 STAR-CCM+ 2E VOF U FV 2 1 PR
Damen 2.1 FINE/Marine SST N VOF U FV 2 2 PR
ECN/CNRS 1.1a-8a ISIS-CFD SST, EASM W VOF BX, A U FV 2 2 PR
1.3b 3.9
3.12 3.13
FOI 1.3b OpenFOAM LES N, W N/A U FV 2 2 PR
FORCE 2.1 StarCCM+ k-ε/k-ω SST N VOF MU FV 2 2 PR
HHI 1.1a 1.2a HiFoam Realizable-KE N VOF A U FV 2 1 PR
1.5a 1.6a
HHI 1.1a 1.2a STAR-CCM+ RS W VOF A U FV 2 1 PR
1.5a 1.6a
2.1 2.5
3.9 3.12
HSVA 1.1a-8a FreSCo+ 2E W, VOF BV, BX U FV 2 2 PR
2.10 WO
(continued)
T. Hino et al.
Table 3 (continued)
IIHR 2.10 2.11 CFDShip-Iowa Hybrid N LS BX OS FD 4 2 PR
3.9 3.12
3.13
IIHR 1.5a-8a REX Hybrid N LS A OS FD/FV 4 2 PR
3.9 3.12
3.13
KRISO 1.1a-8a WAVIS Realizable-KE,EARSM LS BS MS FV 3 1 PR
2.1-7
2.10
MARIC 1.3a-6a ISIS-CFD SST N VOF BX U FV 2 2 PR
MARIN 2.5 PARNASSOS Menters 1E model with N Fitting BS MS FD 2 D
Introduction, Conclusions and Recommendations
Dacles-Mariani
correction
MARIN 1.1a-8a ReFRESCO SST N VOF A U FV 2 1 PR
2.10
MHI 1.1a-4a Fluent v14.5 RNG-KE, SST N VOF U FV 2 PR
MIJAC 1.1a 1.2a OpenFOAM SST N N/A BX U FV 2 N/A PR
1.5a 1.6a v2.3.0
NMRI 1.1a-8a NAGISA EASM N, LS BX OS FV 3 1 A
WO
NUMECA 2.1 2.10 ISIS-CFD SST WO VOF U FV 2 2 PR
PNU 1.3a 1.4a FLUENTv15 2E W VOF A U FV 2 2 PR
1.7a 1.8a
2.1 2.5
(continued)
9
Table 3 (continued)
10
SHIME-CFD 1.1a 1.2a OpenFOAM SST N N/A MU FV 2 2 PR
1.5a 1.6a
SJTU 1.1a 1.2a naoe-FOAM-SJTU 2E N VOF BX, A MU FV M 2 PR
1.5a 1.6a
2.1 3.9
SRC 1.3b FrontFlow/blue LES N VOF U FE 2 2 C
Southampton 1.3a 1.4a OpenFOAM SST N N/A BP MS FV 2 N/A PR
1.7a 1.8a
Southampton 1.3a 1.4a Star CCM+ SST N N/A BP MS FV 2 N/A PR
1.7a
UDE 2.1 ISIS-CFD SST N VOF U FV 2 2 PR
UDE 2.1 OpenFOAM 2E N, W VOF U FV 2 2 PR
ISMT/UDE 1.5a 1.6a STAR-CCM+ SST N VOF A U FV 2 2 PR
UM 2.7 OF23x SST WO VOF A U FV 2 1 PR
LeMoS/Uni Rostock 1.3a 1.7a OpenFOAM SST/LES N N/A A S FV 2 2 PR
UNIZAG-FSB 2.1 navalFoam SST N VOF U FV M 1 PR
UNIZAG-FSB 2.10 2.11 swenseFoam SST N LS U FV M 2, 1 PR
YNU 1.1a-8a SURFv7 EASM N N/A BX OS FV 2 1, 2 AC
A–Actual propeller, BL–Body force propeller (Lifting line), BP–Body Force Propeller (Prescribed), BS–Body force propeller (lifting surface), BV–Body force
propeller (Vortex lattice), BX–Body force propeller (Other), AC–Artificial compressibility; C–Compressibility with low Mach number assumption, D–Direct
method; FD–Finite difference; FV–Finite volume; MS–Multiblock structured; MU–Multiblock unstructured; N–No slip; OS–Overlapping structured; PR–Pressure
correction; S–Single block structured; U–Unstructured; W/WO–Wall functions with/without pressure gradient correction
T. Hino et al.
5 Conclusions
The analysis of the experimental and computational data from the Tokyo 2015 Work-
shop on CFD in Hydrodynamics are given in the following chapters. The summary
of the main conclusions of each chapter is shown below. The detailed discussions are
included in each chapter. Some chapters provide the extensive conclusions as well.
Chapter 2 Experimental Data for JBC Resistance, Sinkage, Trim,

Self-Propulsion Factors, Longitudinal Wave Cut and Detailed Flow
with and without an Energy Saving Circular Duct
In this Chapter, the measured data for Japan Bulk Carrier (JBC) with an energy
saving circular duct are presented. The measurement is conducted in three facilities,
i.e., the towing tanks of National Maritime Research Institute (NMRI) and Osaka
University (OU) and the wind tunnel of Technical University of Hamburg (TUHH).
• Resistance, sinkage, trim, self-propulsion factors with and without the duct are
acquired by the tank tests at NMRI and OU. Wave field data is measured at the
NMRI towing tank.
• The detailed flows fields in seven stations in a stern region are measured by
using SPIV in the tanks of NMRI and OU. The data are acquired in towed
and self-propelled conditions without and with the duct. In addition, LDV/SPIV
measurement is conducted at the TUHH wind tunnel.
• For the SPIV measurement at NMRI, the error estimates are carried out using
the uniform flow test results and the statistical convergence analysis of the actual
measurement. It turned out that uncertainty of the velocity measurement is approx-
imately 2–3% of the uniform flow and the uncertainty of TKE measurement is
approximately 1% of the square of the uniform flow magnitude. Uncertainty of
v velocity is largest and this may be attributed to the problem with the reflection
mirror setting.
• For the mean velocity components, the data of three facilities are generally in
good accordance in spite of some problems such as the laser reflection in TUHH
measurement or the reflection mirror problem in NMRI measurement.
• On the other hand, it is found that there is a large difference in the measured TKE
levels between NMRI and OU/TUHH. Examination of the statistical convergence
of NMRI measurement shows that the difference between NMRI and other facil-
ities does not seem to come from the statistical convergence, though apparently
the more frames are needed for the accurate estimation of turbulent quantities.
The effect of Reynolds number difference is also investigated for the local TKE
distributions along the horizontal lines at the shaft height near the propeller plane.
The distributions in the bare-hull towing condition are compared in the various
scaling. It appears that the TKE distributions look more wake-like rather than
boundary-layer-like from the comparisons with the flat plate data. However, it is
rather difficult to specify the exact reason for differences of the extremely higher
TKE of NMRI data. Further investigations, such as additional measurements in
12 T. Hino et al.
different facilities and/or different measuring systems, are desirable for obtaining
the reasonable distributions of TKE.
• Finally, the recommendations to the future workshops at present are as follows:
For the resistance and self-propulsion, sinkage and trim and the wave profiles,
the measured data from NMRI can be considered to be appropriate. For the local
flow data, the OU data seems to have no serious deficiencies and thus can be
recommended as the reliable measured data of mean velocities. Further works
will be needed to establish reliable turbulence data of this JBC case.
Chapter 3 Experimental Data for KCS Resistance, Sinkage, Trim,

and Self-propulsion
• Additional experimental data for KCS calm water resistance obtained after the
workshop G2010 are reported. All KCS calm water models are listed at Table 1 in
Chap. 4. In the present chapter, all experimental data are obtained with 7.3 m model
from the KRISO and NMRI towing tanks. These experimental data have been
widely used as a standard benchmarking case for CFD validation after G2010.
However, turbulence quantities are still missing in the set of experimental data
and we also feel inconvenienced by the lack of local flow measurement for self-
propulsion including a rudder. It is desirable to have more data to make up the
full set of experimental data, especially for the standard size KCS model ship.
Chapter 4 Experimental Data for KCS Added Resistance and ONRT Free
Running Course Keeping/Speed Loss in Head and Oblique Waves
• Evaluation is performed of the data used for T2015 test cases for the KCS captive
added resistance σAR and ONRT free running course keeping/speed loss in head
and oblique waves.
• For KCS calm water resistance, the individual facility N-order level testing uncer-
tainty is UXi = 1%D and the multiple facility standard deviation based on three
institutes using 2 model sizes (L = 7.3 and 6.1 m) is SD = 0.74%D, such that the
individual facility MxN-order level testing uncertainty is UDi = 1.75%D. UXi was
not reported for sinkage and trim; however, the SD = 4 and 7%D, respectively. D
corresponds to either the individual or mean facility, as implied by the context of
the discussion.
• For KCS head waves, the analysis was based on three institutes with model sizes
L = 6.1, 3.2 and 2.7 m. For the 6.1 m model, FORCE provided UXi = 8, 4
and 4% D, respectively, for σAR and first harmonic heave z1 /ζ1 and pitch θ1 /ζ1 k
amplitudes, whereas for the 2.7 m model, FORCE/IIHR provided UXi = 7/18,
8/2 and 9/5% D, respectively. In consideration of the differences in model sizes
and rigid vs. surge free mounts the agreement between the facilities is reasonable:
for the primary variables, SD = 3, 20 and 10% D for z1 /ζ1 and θ1 /ζ1 k amplitudes
and σAR , respectively; and for secondary variable SD = 66% D for first harmonic
resistance CT1 .
• For KCS oblique waves, the data is only available from IIHR for tests in 2015 and
2016. The agreement is good, i.e., SD values are comparable to the corresponding
values for head waves: for primary variables, SD = 6 and 63%D for z1 /ζ1 and
θ1 /ζ1 k amplitudes and σAR , respectively; and for secondary variable SD = 13%D
for CT1 .
• For ONRT self-propulsion and head and oblique waves, the data is only available
from IIHR and only preliminary uncertainty analysis is available.
• The evaluation showed reasonably reliable data for both KCS and ONRT.
However, clearly it is desirable to have data from more facilities including uncer-
tainty analysis for assessment of facility biases, which will provide more robust
data and uncertainty analysis for CFD validation.
Chapter 5 Evaluation of Resistance, Sinkage, Trim and Wave Pattern

Predictions for JBC
The Workshop submissions for JBC contained four cases: the hull with or without an
ESD both in towing and self-propulsion. Altogether 88 predictions were provided. In
this chapter, resistance, sinkage and trim, as well as wave pattern results are evaluated.
Conclusions are as follows:
• Like at the previous workshops the mean signed comparison error for resistance is
very small, around 1%, which is also the accuracy reported for the experimental
data. However, more interesting is the mean absolute comparison error, since
this represents the average deviation from the data. This is around 2% for the
towed cases and 3% for self-propulsion. The scatter is represented by the standard
deviation and is about 2% and 4% respectively for the two conditions. In the 2005
and 2010 Workshops the standard deviations were 6% and 1%. There was thus a
significant reduction in scatter between 2005 and 2010, but no further reduction
in 2015.
• A resistance reduction around 1% is predicted with the ESD fitted. This is
confirmed by the measurements.
• For the comparison error to be smaller than 4% the grids have to contain more
than 10 M cells. In 2010 the corresponding limit was 3 M cells. Thus, more grid
cells were required for the same accuracy in 2015.
• There is hardly any influence of the grid type on the predicted resistance. If any,
there is a slight advantage of the structured grids, compared to hex-dominated and
completely unstructured grids.
• Like in 2010 the two-equation turbulence models predict better results than the
more advanced models. The k-ε and k-ω models have mean signed comparison
errors less than 1% and mean absolute errors just below 2%. EASM has gained
in popularity and its mean absolute error is now 2%.
• Resistance predictions using wall-functions are more accurate than those with
wall resolved flows. Both the absolute error and the standard deviation are about
twice as large with wall resolution.
• Grid convergence is demonstrated for 64% of the unstructured grids and 77% of
the structured grids. The achieved order of accuracy in the grid refinement studies
14 T. Hino et al.
is in many cases far from the theoretical accuracy. Most methods predict a too
high order of accuracy, 2.3 times the theoretical one, on the average. For those
with a too low order of accuracy the mean value is 0.7 times the nominal one.
• As stated above, the mean absolute comparison error in all resistance predictions
is 2%, which is considerably smaller than the mean validation uncertainty, 5%,
composed of the mean numerical uncertainty of 4.9% and the stated experimental
accuracy of 1%. The methods may thus be considered validated in a mean sense.
• The comparison errors for sinkage were very large in 2010, but are now reduced
considerably. The mean signed error, absolute error and standard deviation are
now -1.9%, 4.7% and 5.6% respectively.
• For trim the corresponding errors are also reduced considerably and are now 0.6%,
3.1% and 3.5%, respectively.
• Very accurate predictions are reported for the wave profile along the hull, while
less correspondence with measurements is demonstrated for two wave cuts. There
is a consistent phase shift in all results, which may indicate an error in the data.
A small over prediction of the bow wave height is also noted in all results.
Chapter 6 Analysis of the Local Flow around JBC

In this chapter, we have reviewed computations of the flow around the Japanese
Bulk Carrier (JBC) provided by the contributors to the Tokyo T2015 workshop from
the standpoint of the local flow analysis. Five flow configurations were studied: (i)
the naked hull without propeller nor duct (case 1–3a), (ii) the double body model
(case 1–3b), (iii) the hull with a duct but without propeller (case 1–4), (iv) the hull
with a propeller and without duct (case 1–7) and finally, (v) the hull with a ducted
propeller (case 1–8). This chapter aimed at understanding the physics of the flow from
a local point of view with the help of the flow measurements performed by NMRI at
three stations S2, S4 and S7. But several additional objectives were pursued during
this analysis. The first one, in the continuity of the G2010 Gothenburg workshop
concerned the verification and validation of stern flows over the JBC hull. This hull
is characterised by a relatively high block coefficient, which means that the flow is
more complex in the stern region. An intense bilge vortex was detected and some
questions arose about the steadiness of the flow in the stern region. It was therefore
decided to focus the study on the bilge vortex core and try to compare the various
contributions in this specific region. The second objective was to understand the
mechanisms which might improve the propulsive efficiency when a ship is equipped
with a ducted propeller. To separate physical effects, several configurations with and
without duct combined with the presence or absence of a propeller were studied with
the help of the measurements made by NMRI at stations S2, S4 and S7. This gave us
the opportunity (i) to compare the various turbulence closures in presence of a duct
or a propeller, (ii) to assess the reliability of the various propeller representations
implemented by the contributors, and (iii) to check if the observed influence of
the duct on the local flow is accurately captured by the computations. The main
conclusions that stem from all these studies are summarized as follows
• For the case 1–3a, since the typical grids which are used by the contributors (5-
10 M points) are fine enough, the influence of the grid discretisation is moderate
for RANSE, which means that the numerical error appears to be under control.
It was feasible therefore to focus the study on the influence of the turbulence
closures and see if it was possible to draw general conclusions form all the contri-
butions. As noticed during the Gothenburg 2010 workshop for the analysis of the
local flow around the KVLCC2, the major influence comes from the turbulence
closure. One noticed that the linear isotropic closures significantly under-predict
the longitudinal vorticity while full RSM closures slightly over-predict it. Non-
linear anisotropic closures (EARSM) seemed to offer a good compromise from
the standpoint of the local flow although they slightly underpredict the vorticity at
the key station S2. Contributions employing hybrid RANS-LES were also taken
into account. The results appeared promising although IDDES seemed to over-
predict the vorticity again. No spectacular advantage was noticed compared to the
best RANS models at the measurement stations.
• This study was completed by a local core vortex analysis which provided a first
interesting attempt to carry out a more local analysis of the time-averaged bilge
vortex. Globally, the agreement between most of the computations and the rather
coarse experiments was satisfactory in terms of global trends. The longitudinal
vorticity was somewhat underestimated and the longitudinal velocity distribution
was fairly reproduced. These experiments pointed out very large differences on the
turbulence kinetic energy and this was the most striking (and unexpected) result
of this comparison. Only the hybrid RANS/LES closures were able to reproduce
this characteristic, thanks to the contribution of the resolved turbulence kinetic
energy. Whether the NMRI tke measurements are reliable or not to provide an
accurate turbulence kinetic energy at this location is still a matter of intense debate
which is partly addressed in Chap. 2 of this book. Further much more detailed
measurements will be necessary to draw safer conclusions.
• In the case 1–3b, the turbulence closure study led to very similar conclusions;
i.e. that the major influence on the computed distribution of isowakes in the stern
region comes from the turbulence closure. Linear isotropic closures under-predict
the longitudinal vorticity at station S2 while full RSM closures tend to overpredict
it at the same station. But what was remarkable and quite unique was that two
LES computations were presented and especially, one from SRC on an extremely
fine grid (4 billion cells) which was in good agreement with measurements and
predicted the level of turbulence kinetic energy measured by NMRI in the core of
the bilge vortex (at least three times higher than what was modeled by EASM).
Unfortunately, it was not possible to exploit further the results of the (almost)
wall-resolved LES simulation, due to the unavailability of this flow database.
• For the case 1–4, it was again noticed that the turbulence models have a critical role
to play once the grid is fine enough. The main modifications of the flow created
by the presence of the duct were correctly reproduced by linear isotropic models
which provided the correct qualitative trend. A better quantitative agreement was
reached by anisotropic closures although the inner part of the vortex was not
16 T. Hino et al.
perfectly reproduced as illustrated by the occurrence of a reversed flow region

which was over-estimated by both RSTM and EARSM turbulence closures.
• For the test case 1–7, it was observed that the computations based on the simulation
of the actual propeller were the only ones able to represent all the characteristics
of the flow in the wake of propeller, specifically at station S7. Various codes using
different meshes and different turbulence models yielded globally comparable
results in terms of isowake distribution when the real rotating propeller is taken
into account in the simulation. However, alternate coupling strategies based on
simplified propeller models appeared to produce a local flow which was quite
satisfactory for some of them, even if they were not able to capture all the charac-
teristics of the wake flow. Hybrid RANS/LES model used by IIHR/Rex yielded
very good results but not necessarily way superior for the averaged flow field to
what is obtained with usual k −ω-like models to justify the very high cost asso-
ciated with this time-accurate approach. However, this hybrid RANS/LES model
was the only one able to capture most of the details of the flow revealed by the
phase-averaged measurements.
• For the test case 1–8, the computations confirmed the fact that when a duct is
mounted before the propeller, the flow in the wake of the propeller is more homoge-
neous. This influence was correctly captured by most of the contributors, whatever
the turbulence model or propeller representation used. As expected, the isowake
distribution behind the ducted propeller was better represented when the actual
rotating propeller was accounted for. However, some simplified propeller models
worked reasonably well and were able to indicate the right trend concerning the
influence of the duct. RANS models provided a reasonably accurate prediction
of the time-averaged flow behind the propeller. For phase-averaged quantities,
some URANS-based computations were in fair agreement with the experiments
while some others did not exhibit any unsteadiness in the wake of the propeller,
probably due to the combined influence of a locally too coarse grid and a too
diffusive turbulence model. From that standpoint, the hybrid RANS/LES model
appeared again more consistent with the physics of the flow.
Chapter 7 Evaluation of Self-propulsion and Energy Saving Device

Performance Predictions for JBC
Conclusions from the analysis of submitted data for JBC self-propulsion cases are
summarized as follows:
• Grid uncertainty U G in general is about 4–5% S for K T and K Q .
• Averages of absolute comparison errors of K T and K Q are 3–4% D for a bare hull
case and 2–3% D with ESD case. ESD case is slightly smaller than a bare hull
case. Actual propeller models give slightly better results than body force models
though the difference is smaller in case of a bare hull.
• Averages of absolute comparison errors of self-propulsion factors, 1 − t and 1 −
w are 7–10% D in a bare hull case, which are considerably larger than the errors
of thrust and torque. This is due to the increased number of quantities involved to
estimate the factors. Errors become smaller with the presence of a duct, maybe
due to the smoothening of stern flow fields by a duct.
• Average absolute errors of model scale delivered power (DP) is in the range of 5–
6%, about the same level of other quantities. However, the estimated DP reduction
rates due to a duct scatter from 0.88 to 1.0, whereas the experimental value is
0.94. The current accuracy of CFD estimation of ESD performance seems to be
reasonably well in terms of the magnitude, although this may not be sufficient for
design/performance prediction of an energy saving duct which aims the reduction
rate of a few percent.
Chapter 8 Evaluation of Resistance, Sinkage, Trim and Self-propulsion

Predictions for KCS
• For the predictions of resistance, sinkage and trim, the number of grid points are
varied from 1.1 M to 6.6 M. The submitted results do not show large dependency
of grid points. It means that CFD users who have efficient grid making technique
can have an acceptable simulation results for low block coefficient hull form such
as KCS.
• Most participants use two equation turbulence model and only few participants
use an advanced turbulence model. Based on these submissions, there is no
visible improvement in accuracy of the resistance prediction for more advanced
turbulence model than the two equation models.
• The comparison error for resistance prediction at the design speed (Fr = 0.26)
is −0.2% of the data and the standard deviation is 1.5% of the data value. The
mean comparison error for all 6 speeds is 0.43% and the mean standard deviation
is 2.48%. The increased error and standard deviation are caused by the results for
low Fr simulation submissions. In the meanwhile, the absolute error |E| is also
analyzed. The mean absolute error |E| is 2%, which is slightly larger than 1.64%
the results of Gothenburg 2010.
• The comparison errors and standard deviation of the sinkage and trim are much
larger than for the resistance. The errors come from the results in a speed range
below Fr = 0.2. Two possible reasons can be considered. This range generates
very short waves and so it requires better grid distribution to resolve the waves.
Other problems are likely to be due to the difficulties of measuring two quantities
at low speed. The data might have larger uncertainties than for high speed.
• Self-propulsion submissions are slightly improved than the results of Gothenburg
2010 Workshop. The mean comparison errors of K T and K Q are 0.5% and −
3.5% respectively (-0.6% and −4.6% for Gothenburg 2010 workshop). The stan-
dard deviations for of K T and K Q are 2.7% and 2.4% respectively, which are
reduced the results of 7.2% and 6.1% respectively for Gothenburg 2010 work-
shop. However, K Q is still over predicted than predictions of the resistance. Self-
propulsion parameters are slightly better predicted by body force models (FD
fixed) but local flow characteristics are comparatively well predicted by actual
propeller rotating simulation (rps fixed).
18 T. Hino et al.
• For the future workshops, new experimental data for KCS or other container hull
would be recommended to simulate not only captive but free running 6DoF calm
water self-propulsion condition.
Chapter 9 Assessment of CFD for KCS Added Resistance and for ONRT
Course Keeping/Speed Loss in Regular Head and Oblique Waves
• CFD is assessed for added resistance for KCS (captive test cases 2.10 and 2.11)
and course keeping/speed loss for ONRT (free running test cases 3.9/3.12/3.13)
in head and oblique waves. The number of submissions were 10, 2, and 8 for test
cases 2.10, 2.11, and 3.9/3.12/3.13, respectively.
• The assessment approach uses both solution and N-version validation. The former
considers whether the absolute error |E i | = |D−S i | is less, equal or greater than
the validation uncertainty, which is the root sum square of the numerical and
experimental uncertainties, i.e., |E i | ≤ UVi = U S2Ni + U D2 . The latter considers
whether the absolute error is less,equal or greater than the state-of-the-art SoAi
uncertainty, i.e., |E i | ≤ U So Ai = UV2i + P|E
2
i|
where P|Ei | = kσ|E| is the uncer-
tainty due to the scatter in the solution absolute error. Errors and uncertainties are
normalized using both the data value D and its dynamic range DR. The analysis
uses k = 2.
• The captive resistance CT and free running self-propulsion propeller revolutions
RPS |E| <2% D with UD less than but comparable P|Ei| < 3% D such that only
3/1 solutions were validated but 8/4 codes/solutions were N-version validated for
CT /RPS.
• The head waves captive and free running heave and pitch |E| is less than 8%
DR with UD less than 5% DR and P|Ei| less than 13% DR such that about 5 for
captive and 2 for free running solutions were validated and about 7 for captive
and 5 for free running codes/solutions were N-version validated. The errors for
added resistance and speed loss were less than 13% DR with UD less than 7% DR
and P|Ei| = 25 and 13% DR such that about 4 for captive and 1 for free running
solutions were validated and 7 for captive 4 for free running codes/solutions were
N-version validated. For captive head waves, the errors and scatter are smaller
than those for potential flow.
• The oblique waves captive and free running motion errors |E| are less than 10%
DR except for roll with UD large 23% DR for captive pitch and other wise small
<2% DR. The errors for added resistance and speed loss were <8% DR with UD
< 9% DR. The largest errors were for roll, which had errors for captive of 12%
DR and for free running of 20% DR.
• None of the KCS and ONRT test cases were able to achieve a programmatic
requirement of 5% D and 5% DR for calm water and waves, respectively. Note
that not all the experimental uncertainties are able to meet this requirement either
such that both experiments and CFD need to reduce their uncertainties and in
addition CFD its errors. Nonetheless, in view the comparable CFD capability for
ONRT free running vs. KCS captive conditions, the prognosis for CFD capability
is excellent.
• Future CFD assessment should extend the current test cases for consideration of
added power in regular and irregular waves. Verification should be required, and
more limited/most important validation variables should be confirmed and used
for the V&V and SoA assessment.
6 Future Workshops
The Tokyo 2015 Workshop was the seventh in a series started in 1980. Since 1990
the workshops have been held essentially every 5 years. This is a period long enough
to allow significant progress to take place, but short enough to enable an evaluation
of the new developments without undue delay. Therefore, the 5-year period should
be maintained in principle. Following the selection of the organizer and venue of the
next workshop by the previous Steering Committee, the new Steering Committee
for the CFD Workshops in Ship Hydrodynamics was formed in April 2016. Its
missions are to define the test cases and to coordinate campaigns to obtain more
data and to support the local Organizing Committee in the organization of each
workshop. The members of the Steering Committee are the organizer of the next
workshop (presently Dr. Serge Toxopeus, MARIN), area representatives of the USA
(presently Prof. Frederick Stern, IIHR), Europe (presently Dr. Michel Visonneau,
ECN/CNRS) and Asia (presently Prof. Takanori Hino, YNU, Dr. Jin Kim, KRISO
and Prof. Decheng Wan, SJTU). The present Steering Committee is chaired by Prof.
Stern.
The date and venue of the next workshop are tentatively set to Fall of 2021 at the
Maritime Research Institute Netherland (MARIN), Wageningen, Netherland. Test
cases are now under development, but the current plan is as follows. Some test
cases of the previous workshop will remain such as JBC or other hull for turbulence
modeling assessment, but there will be new cases of breaking bow waves for air-
water interface flow modeling assessment, free running added power in regular and
irregular head and oblique waves for motions and hull-propeller-rudder interaction
modeling assessment and full-scale simulations for scale effects and “real-world”
vs. “laboratory” modeling assessment.
7 Closing Remarks
The workshops clearly document the progression of prediction capability for viscous
ship hydrodynamics from integral and differential boundary layer to current CFD
methods with test case complexity progressing from double body resistance and local
flow to captive calm water resistance, sinkage and trim, self-propulsion and local flow
to captive added resistance and ship motions in head and oblique waves to free running
20 T. Hino et al.
self-propulsion and speed loss in head and oblique waves. The number of partici-
pants/submissions has increased as has the use, i.e., number of research-institute
vs. commercial and open-source codes submissions. The assessment methods have
also progressed from consideration of errors only to quantitative verification and
validation.
The original workshop objectives were to provide guidance for code developers
which were mostly research institutes that were also members of ITTC in support
of institute prediction capability and to assess the SoA similarly as ITTC committee
reports did for experimental ship hydrodynamics. However, with increased submis-
sions using commercial and open source codes the assessment should include guide-
lines here again with similarity to the more recent ITTC focus on its Quality Manual;
however, such an effort would require resources beyond those currently available
without financial support, which is also needed to support the procurement of reli-
able benchmark validation data and to support the host costs. The organization of the
CFD workshops has also progressed from several interested persons to the present
CFD Workshop Organizing Committee with representatives from US, Europe and
Asia. The documentation of the CFD Workshops progressed from institute reports
to proceedings to journal articles to most recently books. The planning of the CFD
Workshop extends over the period between workshops which has been 5 or 6 years.
The advent of ever more complex test cases and increased number of submissions
(especially using commercial and open-source codes rather than research-institute
codes) puts greater burden on the Organizing Committee for procurement of bench-
mark data, preparation of submission instructions, data reduction for CFD assess-
ment, more emphasis on SoA assessment and the need for guidelines, all of which
affect the motivation of the Organizing Committee members.
In view of the progress of both CFD and the CFD Workshops, the following
recommendations are made:
1. Experiments including uncertainty analysis should be conducted by multiple
facilities for each test case to establish reliable benchmark validation data,
including facility biases and scale effects.
2. Fewer and only challenging test cases should be used for the assessment.
3. Fewer and only primary validation variables should be used for the assessment.
4. All test cases should uniformly use the established verification and validation
methods.
5. Improved submission instructions are needed that enable automated data
reduction that the organizers can more easily use for the assessments.
6. If financial support can be obtained, the Organizing Committee should be
formalized; ideally with ability for certification of CFD performance capability.
7. The Organizing Committee should liaison with and leverage the resources of the
ITTC but should continue as voluntary and retain its independence so as not to
be influenced by commercial interests.
Acknowledgements The workshop was organized by the International Steering Committee of six
members, i.e. the authors of this book and also by the local organizers at National Maritime Research
Institute, Japan: Dr. Hiroshi Kobayashi, Dr. Yusuke Tahara, Dr. Kunihide Ohashi and Dr. Naoyuki
Onodera who significantly contributed to the success of the workshop. Finally, the great efforts
by all workshop participants in the preparation and delivery of all computed results are gratefully
acknowledged.
References
Hino, T. (Ed.) (2005). Proceedings of CFD workshop Tokyo 2005, NMRI report.
Hino, T., Hirata, N., Ohashi, K., Toda, Y., Zhu, T., Makino, K., Takai, M., Nishigaki, M., Kimura,
K., Anda, M., & Shingo, S. (2016). Hull form design and flow measurements of a bulk carrier
with an energy-saving device for CFD validations. In: Proc. 13th International Symposium on
Practical Design of Ships and Other Floating Structures (PRADS’ 2016), Copenhagen, Denmark.
Jufuku, N., Hori, M., Itou, S., Toda, Y., & Hinatsu, M. (2015). SPIV stern flow measurement around
operating propeller for with and without duct condition of japan bulk carrier (in Japanese).
Conference Proceedings of the Japan Society of Naval Architects and Ocean Engineers, 21,
309–312.
Kim, W. J., Van, D. H., & Kim, D. H. (2001). Measurement of flows around modern commercial
ship models. Experiment in Fluids, 31, 567–578.
Kodama, Y., Takeshi, H., Hinatsu, M., Hino, T., Uto, S., Hirata, N., et al. (1994). Proceedings, CFD
Workshop. Tokyo, Japan: Ship Research Institute.
Larsson, L. (Ed.). (1981). SSPA-ITTC workshop on ship boundary layers. Gothenburg, Sweden:
SSPA Report 90.
Larsson, L., Patel, V. C., & Dyne, G. (Eds.). (1991). SSPA-CTH-IIHR workshop on viscous flow.
Gothenburg, Sweden: Flowtech Research Report 2, Flowtech Int. AB.
Larsson, L., Stern, F., & Bertram, V. (Eds.). (2002). Gothenburg 2000-A workshop on numerical
hydrodynamics. Gothenburg, Sweden: Department of Naval Architecture and Ocean Engineering,
Chalmers University of Technolog.
Larsson, L., Stern, F., & Bertram, V. (2003). Benchmarking of computational fluid dynamics for
ship flow: the Gothenburg 2000 Workshop. Journal Ship Research, 47, 63–81.
Larsson, L., Stern, F., & Visonneau, M. (Eds.) (2014). Numerical ship hydrodynamics—An
assessment of the Gothenburg 2010 Workshop. Springer. https://doi.org/10.1007/978-94-007-
7189-5
Sanada, Y., Tanimoto, K., Takagi, K., Toda, Y., & Stern, F. (2013). Trajectories and local flow field
measurements around ONR tumblehome in maneuvering motion. Ocean Engineering, 72(2013),
45–65.
Sanada, Y., Elshiekh, H., Toda, Y., & Stern, F. (2019). Effects of waves on course keeping and
manoeuvring for surface combatant ONR tumblehome. Journal Marin Science Technology, 24(3),
948–967.
Shevchuk, I., Sahab, A., Stern, F., & Abdel-Maksoud, M. (2020). Experimental and numerical
studies of the flow around the JBC hull form at straight ahead condition and 8° drift angle. In
Proc. the 33rd Symposium on Naval Hydrodynamics. Osaka, Japan.
Simonsen, C., Otzen, J., & Stern, F. (2008). EFD and CFD for KCS heaving and pitching in regular
head waves. In Proc. 27th Symp. Naval Hydrodynamics. Seoul, Korea.
Van, S. H., Kim, W. J., Yim, G. T., Kim, D. H., & Lee, C. J. (1998). Experimental investigation of
the flow characteristics around practical hull forms. In: Proceedings 3rd Osaka Colloquium on
Advanced CFD Applications to Ship Flow and Hull Form Design. Osaka, Japan.
Experimental Data for JBC Resistance,
Sinkage, Trim, Self-Propulsion Factors,
Longitudinal Wave Cut and Detailed
Flow with and without an Energy Saving
Circular Duct
Nobuyuki Hirata, Hiroshi Kobayashi, Takanori Hino, Yasuyuki Toda,

Moustafa Abdel-Maksoud, and Frederick Stern
Abstract In this Chapter, measured data of resistance, sinkage, trim, self-propulsion

factors, longitudinal wave cut and detailed flow are summarized for Japan Bulk
Carrier (JBC) with and without an energy saving circular duct. JBC is a newly
designed ship for CFD validation of a ship with an energy saving device (ESD).
Resistance and self-propulsion tests are conducted in towing tanks of National
Maritime Research Institute (NMRI) and Osaka University (OU) using model ships
of different sizes. Detailed local flow data are acquired using SPIV (Stereo Particle
Image Velocimetry) for several cross sections at towing tanks of NMRI and OU. The
local flow data are also measured by SPIV in a wind tunnel at Hamburg University
of Technology (TUHH).
1 Introduction
New test cases associated with hydrodynamics of flows around a ship equipped with
an energy saving device (ESD) are adopted in T2015 Workshop. A ship hull called
Japan Bulk Carrier (JBC) which is a Cape-size bulk carrier was designed together
with a circular duct in front of a propeller as an ESD.
N. Hirata · H. Kobayashi
National Maritime Research Institute, Mitaka, Japan
T. Hino (B)
Y. Toda
Osaka University, Suita, Japan
M. Abdel-Maksoud
Hamburg University of Technology, Hamburg, Germany
F. Stern
University of Iowa and Iowa Institute of Hydraulic Research (IIHR), Iowa, IA, USA
24 N. Hirata et al.
Validation data for the test cases were obtained by model tests in the multiple
facilities including National Maritime Research Institute (NMRI), Osaka Univer-
sity (OU) and Hamburg University of Technology (TUHH). Resistance and self-
propulsion tests were conducted at the towing tanks of NMRI and OU. In addition,
local flow measurements are carried out at two towing tanks on NMRI and OU, and
also at the wind tunnel of TUHH.
In this chapter, description of the ship hull, the propeller and the energy saving
duct is given first. It is followed by the summary of towing tank tests. The next section
is devoted to the local velocity measurements. Accuracy estimations are given for
NMRI measurement and credibility of SPIV (Stereo Particle Image Velocimetry) data
of three facilities is discussed. The next section gives the wave height measurement
results at NMRI towing tank. Conclusions of the chapter are summarised in the last
section.
2 JBC (Japan Bulk Carrier)
2.1 Ship Hull
Japan Bulk Carrier (JBC) is a Cape-size bulk carrier designed for the validation of
CFD analysis of a ship with an energy saving device (ESD). A ship type of a bulk
carrier is selected since it is one of the major cargo vessels in international shipping
and since a blunt ship hull is considered to be appropriate for the examination of an
effect of ESDs. A circular duct placed ahead of a propeller is adopted as an ESD.
The ship hull and the duct have been newly designed in the collaborative research
project in Japan which were organized by universities, research organizations and
shipyards in Japan under the sponsorship of ClassNK (Hino et al. 2016).
The principal particulars of a ship are determined following the representative
values of current vessels as shown in Table√ 1. The design speed is set to 14.5 knots
which corresponds to Fn (L P P ) = U0 / gL P P = 0.142. Based on the comparative
study (Hino et al. 2016) of resistance and wake distributions at a propeller plane by
numerical simulations at the design Froude number Fn = 0.142 and the model scale
Reynolds number Rn (L P P ) = 7.245 × 106 , the final hull form is designed as shown
in Fig. 1.
2.2 Propeller
A propeller used is a conventional five-bladed MAU type propeller. The particulars

of the model propeller for a model ship of L P P = 7.000 m are shown in Table 2.
Experimental Data for JBC Resistance … 25
Table 1 Principal particulars of JBC in actual ship scale

Length between perpendiculars L P P (m) 280.0
Length of waterline L W L (m) 285.0
Maximum beam of waterline BW L (m) 45.0
Depth D (m) 25.0
Draft T (m) 16.5
Displacement volume ∇ (m3) 178369.9
Wetted surface area w/o ESD S_w/oESD (m2) 19556.1
Wetted surface area with ESD S_withESD (m2) 19633.9
Block coefficient (C B ) ∇/(L P P BW L T ) 0.8580
Midship section coefficient (C M ) 0.9981
LCB (%L P P ), fwd+ 2.5475
Vertical Center of Gravity (from keel)*1 KG (m) 13.29
Metacentric height*1 GM (m) 5.3
Moment of Inertia*1 K x x /BW L 0.40
Moment of Inertia*1 K yy /BW L , K zz /BW L 0.25
Propeller center, longitudinal location (from FP) x/L P P 0.9864
Propeller center, vertical location (below waterline) −z/L P P 0.0404214
Propeller rotation direction (view from stern) Clockwise
Design speed (knots) 14.5
*1: These values are determined after the workshop. Please note that these values were not available
in computations for the workshop
Fig. 1 Body plan of JBC

26 N. Hirata et al.
Table 2 Principal particulars of a model propeller

Diameter (mm) 203.0
Boss ratio 0.1800
Pitch ratio 0.7500
Expanded area ratio 0.5000
Max blade width ratio 0.2262
Blade thickness ratio 0.050
Angle of rake 5.00 deg
Number of blades 5
Blade section MAU
Direction of rotation Clockwise
2.3 Energy Saving Device
A circular duct installed ahead of a propeller is selected as an energy saving device,

considering that a computational setup is easy due to its simple geometry. A duct
is designed based on a parametric study using CFD simulations aiming to better
propulsive efficiency. A series of computations are made for various configurations
of ducts with diameters at trailing edge ranging from 0.70D P to 1.10D P where D P
is a diameter of a propeller, opening angles ranging from 5 deg. to 20 deg. and several
vertical positions at model scale Reynolds number Rn (L P P ) = 7.245 × 106 . Final
selection is made between two candidates based on self-propulsion test as shown
below. Principal particulars of the designed duct are listed in Table 3. The duct is
attached to the stern-tube using a vertical strut as shown in Fig. 2 in such a way that
the leading edge of the duct is located at S.S. 1/4.
The geometry data of the hull, the propeller and the duct can be found at the
Workshop web site (https://t2015.nmri.go.jp/).
Table 3 Principal particulars of a duct

Section NACA4420
Diameter at trailing edge 0.55D P ∗1
Chord length 0.3D P ∗1
Opening angle 20 deg.
*1 DP : propeller diameter
Fig. 2 Energy saving duct layout
2.4 Models
Three different models are used by National Maritime Research Institute, Japan
(NMRI), Osaka University (OU) and Hamburg University of Technology (TUHH).
Table 4 shows the principal particulars of each model used in three facilities. Two
models of NMRI and OU are used in their tank tests (resistance and self-propulsion)
and SPIV flow measurement. A model by TUHH is used to measure flow field by
LDV and PIV in its wind tunnel. Note that the TUHH model is a double model shape.
A rudder is not installed in all measurements to avoid interference with measuring
Table 4 Principal particulars of NMRI, OU and TUHH models

NMRI OU TUHH
Length between perpendiculars L PP (m) 7.000 3.200 3.513
Length of waterline L W L (m) 7.125 3.257 3.576
Maximum beam of waterline BWL (m) 1.1250 0.5143 0.5646
Depth D (m) 0.6250 0.2857 −
Draft d (m) 0.4125 0.1886 0.2070
Wetted surface area w/o ESD S_w/oESD (m2 ) 12.223 2.554 3.078
Wetted surface area with ESD S_withESD (m2 ) 12.271 2.564 3/091
Block coefficient CB 0.8580
Midship section coefficient CM 0.9981
LCB (%L pp ), fwd+ 2.55
28 N. Hirata et al.
instruments. Figures 3 and 4 are photographs of the NMRI model and the OU model,
respectively. Figure 5 shows the TUHH model in the wind tunnel.
Fig. 3 NMRI model
Fig. 4 OU model
Fig. 5 TUHH model
Model propellers used in self-propulsion tests and SPIV measurements at NMRI

and OU are the MAU propeller shown in Table 3 with the diameters being 0.203 m
for NMRI and 0.0928 m for OU, respectively. Propeller open water characteristics
of the NMRI model propeller are shown in Fig. 6.
Fig. 6 Propeller open water characteristics of the NMRI model propeller

30 N. Hirata et al.
3 Resistance and Self-propulsion Tests
Resistance and self-propulsion tests are carried out in towing tanks of NMRI (Hino
et al. 2016) and OU (Jufuku et al. 2015). The dimension (length, width and water
depth) of tanks are 400 × 18 × 8 m for NMRI and 100 × 8 × 4.35 m for OU. The
tests are conducted in the conditions with and without a duct while a rudder is not
installed throughout the measurements.
The results at the design Froude number F n = 0.142 are shown in Table 5 for both
NMRI and OU. The form factors 1 + k are obtained using ITTC 1957 correlation line.
Note that resistance coefficients of OU are based on the same wetted area (the value
with the duct) for both configurations. The self-propulsion tests in both facilities are
carried out at the ship point, where the roughness allowance C F is set to 0.12 ×
10-3 .
NMRI data is obtained at the Reynolds number Rn = 7.569 × 106 for both with
and without the duct. However, test cases 1.1a and 1.2a of T2015 Workshop adopted
Rn = 7.46 × 106 , the same as the SPIV measurement. This makes it possible to use
the same computations between cases for integral values and for local flow data. On
the other hand, the C T , K T , K Q or propeller revolution values given as the validation
data at the Workshop are slightly different (approximately 0.3% in case of C T ) from
the values with specified Reynolds number. Nevertheless, this difference is within
the nominal uncertainty of the resistance tests (1% D).
Table 5 Results of resistance and self-propulsion tests at design speed

NMRI OU
ESD Without Duct With Duct Without Duct With Duct
V m (m/s) 1.179 1.179 0.795 0.795
F n (L pp ) 0.142 0.142 0.142 0.142
Rn (L pp ) 7.569 × 106 7.569 × 106 2.17 × 106 2.17 × 106
CT 4.29 × 10−3 4.26 × 10−3 5.27 × 10−3 5.28 × 10−3
1+k 1.314 1.305 1.26 1.26
Cw 1.50 × 10−4 1.52 × 10−4 2.447 × 10−4 2.547 × 10−4
sinkage/L PP (%) −8.57 × 10−2 −8.46 × 10−2 −7.65 × 10−2 −7.45 × 10−2
trim/L PP (%) −1.80 × 10−1 −1.82 × 10−1 −1.75 × 10−1 −1.75 × 10−1
n(1/s) 7.810 7.516 11.8 11.8
KT 0.217 0.233 0.217 0.230
10K Q 0.279 0.295 − −
1−t 0.804 0.810 0.748 0.751
1 − wT 0.552 0.481 0.400 0.346
ηR 1.015 1.009 − −
Sinkage is positive upward and trim is positive bow up
Fig. 7 Total resistance coefficient (C T ) of NMRI model
Figures 7 through 10 show all the measured results of resistance tests at NMRI.
Figure 7 shows total resistance coefficient C T = 0.5ρU RT
2 curves of the cases with
0S
and without the duct. C T with the duct is smaller than that without
the ductat every
Froude number, while wave making resistance coefficient C W = 0.5ρU RW
2 curves
0S
shown in Fig. 8 are not so different between with and without the duct. Trim and
sinkage are shown in Figs. 9 and 10, respectively. Trim τ and sinkage σ are defined
as τ = (da − d f )/L pp and σ = −(da + d f )/2L pp , where d f and da are dipping at
FP and AP, respectively. Differences of trim and sinkage between with and without
duct are very small.
Figures 11 through 15 are the results of self-propulsion tests at NMRI in which
a range of Froude numbers from 0.12 to 0.16 are covered. The thrust deduction
coefficient 1 − t in Fig. 11 and relative rotative efficiency η R in Fig. 13 show small
differences between with and without the duct. 1 − t is larger with the duct than
without the duct, while η R shows the opposite trend. 1 − wT with wT being wake
Fig. 8 Wave making resistance coefficient (C W ) of NMRI model

32 N. Hirata et al.
Fig. 9 Trim of NMRI model
Fig. 10 Sinkage of NMRI model
Fig. 11 Thrust deduction coefficient (1 −t) of NMRI model
fraction is largely improved by the presence of the duct as shown in Fig. 12. Both
K T in Fig. 14 and K Q in Fig. 15 increase with the duct.
Fig. 12 Thrust wake coefficient (1 − wT ) of NMRI model
Fig. 13 Relative rotative efficiency (η R ) of NMRI model
Fig. 14 Propeller thrust coefficient (K T ) of NMRI model
4 Detailed Flow Measurements
4.1 Overview
For the validation of computed results, not only the integrated values such as C T ,
the attitude of the hull and self-propulsion factors but also detailed flow field data
are desired. To this end, Stereo Particle Image Velocimetry (SPIV) measurements
34 N. Hirata et al.
Fig. 15 Propeller torque coefficient (K Q ) of NMRI model
are carried out at the towing tanks of NMRI (Hino et al. 2016) and OU (Jufuku
et al. 2015). Also, LDV/PIV measurement is conducted at the TUHH wind tunnel
(Shevchuk et al. 2020).
The models used in NMRI, OU and TUHH measurements are specified in Table 4.
Measuring sections are common in three facilities. Seven cross sections with constant
x/L P P ahead and behind a duct ranging from S.S. 1/2 to AP are adopted as shown
in Fig. 16 for NMRI measurements. Section 1 (S1) is at S.S. 1/2 (x/L P P = 0.950),
Sect. 2 (S2) at S.S. 3/8 (x/L P P = 0.9625), Sect. 3 (S3) is at the duct mid-chord
(x/L P P = 0.9788), Sect. 4 (S4) is between the duct and the propeller (x/L P P =
0.9843), Sect. 5 (S5) is at the propeller plane (x/L P P = 0.9864), Sect. 6 (S6) is behind
the propeller boss (x/L P P = 0.9923) and Sect. 7 is at AP (x/L P P = 1.0000).
The measurement in NMRI is carried in its middle towing tank (length × width
× depth = 150 × 7.5 × 3.5 m) using 7.0 m model, while that in OU is done in its
towing tank (length × width × depth = 100×8×4.35 m) using 3.2 m model. Froude
number is Fn = 0.142 in both NMRI and OU measurements. Reynolds numbers are
7.46 × 106 and 2.17 × 106 in NMRI and OU, respectively.
LDV/SPIV measurement is conducted in the TUHH wind tunnel. The TUHH
low-speed wind tunnel is 40 m long with the test section (length × width × height
= 5.5 × 3 × 2 m). The double model of length 3.513 (m) is used and the blockage
coefficient is 0.04. The wind velocity U0 = 11.8 m/s which corresponds to Reynolds
number of 2.74 × 106 for LDV measurement and U = 10.0 m/s and Reynolds
number of 2.42 × 106 for SPIV measurement.
4.2 Accuracy Estimation of SPIV Measurement at NMRI
The uniform flow measurements without a ship model using SPIV system are
conducted at NMRI for clarification of the accuracy. Figure 17 shows the measured
data of uniform flow at U 0 = 1.0 m/s which consists of three-components of velocity
u, v and w and turbulent kinetic energy TKE. The accuracy of the averaged velocities
can be estimated to be 2−3% of U 0 for u, 3−4% of U 0 for v and 1% of U 0 for w.
Fig. 16 Measuring sections of velocity fields
The reason why the accuracy of v is somewhat large is that the reflection mirror in
the PIV system is not adjusted correctly. For the accuracy of TKE, 0.3−0.5% of U 20
can be estimated. Note that all the data is acquired using 250 images.
To assess the accuracy of NMRI measurement further, the statistical conver-
gence of PIV data is examined. Measurement at the S4 section without a propeller
and without a duct is taken as an example. Figure 18 is the contours of the axial
velocity u and TKE k, respectively, measured at NMRI. Figure 19 is the cumula-
tive moving errors of velocity components and TKE with respect to the number of
images. Figure 20 is the histogram of three velocity components. Figure 21 shows
the statistical convergence errors of the mean velocity components and TKE. The
sample point is at y/L P P = −0.009814 and z/L P P = −0.03926 also shown as a
red circle in Fig. 18. The velocity components and the TKE values in Fig. 19 do not
converge completely since the number of images is limited to about 1000, though
the data variations are small. The histograms in Fig. 20 show that the fluctuations of
velocity components are large. This may be also due to the small number of images.
The statistical convergence of the mean velocity components in Fig. 21 is defined
using the sample standard deviation sx as
36 N. Hirata et al.
U V
W TKE
TKE
W
Fig. 17 SPIV measurement of uniform flow (U 0 = 1.0 m/s) at NMRI
Fig. 18 Contours of the axial velocity u (left) and TKE k (right) in S4 (x/L P P = 0.9843) station
of NMRI measurement
Fig. 19 Cumulative moving averages of velocity components and TKE in NMRI measurement
Fig. 20 Histogram of velocity components in NMRI measurement
tn;α/2 sx
E SC (%) = √ × 100
N xref
where N is the sample size and tn;α/2 is the Student-t variable of n = N −1 degrees of
freedom for a 100(1 − α) percent level of confidence and xref is the reference velocity
equal to the uniform flow magnitude, U0 (Yoon et al. 2015). For the convergence of
TKE, the χ 2 -distribution is assumed and the upper and lower limits of the confidence
interval is defined based on the sample variance sx2 as
38 N. Hirata et al.
Fig. 21 Statistical convergence errors of mean velocity components (left) and TKE (right) at in
NMRI measurement at and (y/L pp , z/L pp ) = (−0.009814, −0.03926)

n sx2
U
E SC (%) = − × 100
χn;1−α/2
2
sr2e f

n sx2
L
E SC (%) = 1− × 100
χn;α/2
2
sr2e f
for a 100(1 − α) percent level of confidence. The χn;1−α/22

and χn;α/2
2
are the χ 2
2
variables and sref is the reference velocity-squared value for which the square of the
uniform velocity magnitude U02 is taken (Yoon et al. 2015). Note that 95% level of
confidence (α = 0.05) is set for all variables. When the sample size N is approaching
to 1000, the statistical convergence of the mean velocity component is around 1% of
U0 and that of TKE is below 1% of U02 . Figure 22 depicts the contours of the statistical
U
convergence errors E SC for the mean velocity components and E SC for TKE in the
S4 section. The convergence error of the mean velocity u, v and w are proportional to
their standard
√ deviations
√ which
√ correspond to the square-root of the normal Reynolds
stresses, u u , v v and w w . The patterns, therefore, are similar to that of TKE
in Fig. 18. The error levels of u, v and w are estimated as 1.2, 1.4 and 0.6% of U0 for
the average and 1.8, 2.0 and 0.9% of U0 for the maximum, respectively. The error of
TKE is 1.0% of U02 for the average and 1.6% of U02 for the maximum. The errors of
v are slightly larger than those of u, which is also observed in Fig. 22. Though the
reason is not clear, it may be related to the mirror alignment problem.
Total uncertainty of PIV measurement is estimated using the analysis above. The
bias errors E b are estimated from the uniform flow test and the precision errors E p
from the statistical convergence errors and the uncertainty is defined as E b2 + E 2p .
Results are shown in Table 6 and the average uncertainty of the mean velocity is
1−3% of U0 and that of TKE is 1.0% of U02 . The maximum uncertainty is slightly
larger than the average.
Fig. 22 Contours of the statistical convergence errors mean velocity u (top left), v (top right),
w (bottom left) and TKE k (bottom right) in S4 (x/L pp = 0.9843) station of NMRI measurement
Table 6 Uncertainty of PIV measurement

Bias error Precision error Precision error Uncertainty Uncertainty
(average) (max.) (average) (max.)
u (%U 0 ) 2.0 1.2 1.8 2.3 2.7
v (%U 0 ) 3.0 1.4 2.0 3.3 3.6
w (%U 0 ) 1.0 0.6 0.9 1.2 1.3
TKE (%U 20 ) 0.3 1.0 1.6 1.0 1.6
4.3 Credibility of SPIV Measurement at NMRI, OU

and TUHH
In order evaluate the credibility of measurement, the measured data of three facilities
are compared. The test case 1.3 in the workshop, i.e. the towed condition without a
propeller and without an ESD is chosen since this configuration is simplest.
In the S4 section (x/L P P = 0.9843), the contours of the axial velocity u, the
horizontal velocity v and the vertical velocity w are shown in Figs. 23, 24 and 25,
40 N. Hirata et al.
Fig. 23 Contours of measured axial velocity (u) at S4 section. Left: NMRI, middle: TUHH and
right: OU
Fig. 24 Contours of measured horizontal velocity (v) at S4 section. Left: NMRI, middle: TUHH
and right: OU
Fig. 25 Contours of measured vertical velocity (w) at S4 section. Left: NMRI, middle: TUHH and
right: OU
respectively. Comparisons of the crossflow vectors are shown in Fig. 26 and contours
of the x-vorticity ωx and TKE k are shown in Figs. 27 and 28, respectively.
All the velocity contours are similar to each other and the so-called hook shapes
can be observed. While NMRI data is almost symmetric in y direction, the measured
areas of TUHH and OU are too narrow to exhibit symmetry.
Lateral locations of the longitudinal vorticies of OU and TUHH are almost same.
That of NMRI, on the other hand, is a little closer to a symmetry plane. This is due
Fig. 26 Comparisons of cross flow vectors (v, w) at S4 section. Left: NMRI/OU, middle:
NMRI/TUHH and right: OU/TUHH
Fig. 27 Contours of measured vorticity (ωx ) at S4 section. Left: NMRI, middle: TUHH and right:
OU
Fig. 28 Contours of measured turbulent kinetic energy (k) at S4 section. Left: NMRI, middle:
TUHH and right: OU
to the difference of Reynolds numbers. Reynolds numbers of OU and TUHH are

almost the same, 2.17 × 106 and 2.42 × 106 , respectivley, and that of NMRI is larger,
7.46 × 106 . It is known that the boundary layer is thicker in low Reynolds number
flows.
In the narrow region near the center of measuring area (−0.01 < y/ L P P < 0.01,
−0.05 < z/ L P P < -0.03), axial and horizontal velocities u and v agree well between
NMRI and OU. However, the discrepancies between NMRI and OU become large in
42 N. Hirata et al.
| y/ L P P | > 0.015 and the crossflow vectors of NMRI data in Fig. 26 are inclined to
the positive y direction. This is due to the problem of the reflection mirror in NMRI’s
SPIV system as described above. The uniform flow measurement in Fig. 17 indicates
v is overestimated by about 5% of uniform velocity U 0 = 1.00 m/s near both side
edges of the measuring area. In TUHH case, all flow components under the propeller
shaft show asymmetric distributions which is due to the laser reflection effects. The
contours of x-vorticity ωx in Fig. 27 are similar between NMRI and OU. TUHH data
is different from the other two, particularly under the propeller shaft, which is again
attributed to the laser reflection. The distributions of TKE in Fig. 28 show that the
peak value of NMRI data is much larger than that of OU, although the patterns are
similar. Figures 29 through 34 are the same plots in the S7 section as Figs. 23 through
28 in the S4 section. General trends are the same in case of the S4 section.
The large differences are found in the TKE distributions between NMRI and OU
as shown in Figs. 28 and 34. The convergence of TKE data is shown in Fig. 19 for
NMRI case and it exhibits the reasonable convergence with 1000 images. TKE of
OU data is acquired using 3000 images and it is believed the convergence is similar
or better than NMRI data. Therefore, the difference of TKE values between NMRI
and OU seems to come from the reason other than the statistical convergence though
more images may be needed for the accurate measurement of turbulent quantities.
Fig. 29 Contours of measured axial velocity (u) at S7 section. Left: NMRI, middle: TUHH and
right: OU
Fig. 30 Contours of measured horizontal velocity (v) at S7 section. Left: NMRI, middle: TUHH
and right: OU
Fig. 31 Contours of measured vertical velocity (w) at S7 section. Left: NMRI, middle: TUHH and
right: OU
Fig. 32 Comparisons of cross flow vectors (v, w) at S7 section. Left: NMRI/OU, middle:
NMRI/TUHH and right: OU/TUHH
Fig. 33 Contours of measured vorticity (ωx ) at S7 section. Left: NMRI, middle: TUHH and right:
OU
In order to assess the Reynolds number effect, three TKE data are compared by
the scaled variables. TKE distributions along the horizontal line at the shaft height
shown in Fig. 35 are extracted
√ from NMRI, OU and TUHH measurement. The
distributions are plotted as TKE/u τ ∼ y + in left of Fig. 35. Also plotted is the
same scaled data from the propeller plane of KVLCC2. The data is taken from the
wind tunnel measurement by a hot-wire (Lee et al. 2003) where Reynolds number is
Rn = 4.6 × 106 . The wall distance y is measured from the shaft edge and the friction
44 N. Hirata et al.
Fig. 34 Contours of measured turbulent kinetic energy (k) at S7 section. Left: NMRI, middle:
TUHH and right: OU
NMRI OU TUHH KVLCC2
Fig. 35 Left: Distributions of scaled TKE along a horizontal line at the shaft height in S4 section
of JBC and the propeller plane of KVLCC2. Right: Distributions of scaled streamwise normal
Reynolds stress u u of a flat plate boundary layer (Longo et al. 1998). Bottom: Data-extracted lines
velocity u τ is estimated from the local c f of a turbulent boundary layer of a flat plate
using the formula of Prandtl- Schlichting as below

c f = (2.0 log10 (Rex ) − 0.65)−2.3 , u τ /U = c f /2
where Rex is set equal to the Reynolds number of each case. Actual data of u τ /U0
in each case are 0.0367, 0.0405, 0.0401 and 0.0381 for NMRI, OU, TUHH and
KVLCC2, respectively. The plots show the similarity in the distribution patterns of
all cases. Note that the irregular data of NMRI measurement are omitted. The peak
locations and peak values of OU and TUHH are almost the same since the Reynolds
numbers of two cases are similar. The peak value of NMRI is larger than those of
OU and TUHH. KVLCC2 data with the moderate Reynolds number is in-between
OU/TUHH and NMRI for the peak location and value. Right of Fig. 35 is the scaled
streamwise normal Reynolds stress u u in the boundary layer (Longo et al. 1998).
The distributions show the different distribution pattern than the JBC case, though
the general tendency of the turbulence increase with Reynolds number seems to be
the same as in the JBC case.
Figure 36 shows the maximum TKE values of k + (= k/u 2τ ) in the profiles of the
present horizontal lines. Again, the peak values of TKE increase with the Reynolds
number. The NMRI data has the highest peak value but it is rather difficult to tell if
this follows the trend of the other data or not since the date points is quite few.
Figure 37 shows the wake-scale distributions of u velocity. The wake half width b
is determined by manual fitting of the velocity distributions. Due to the presence of
the vortex core, the velocity distributions have the minimum away from the center.
The TKE distributions in the wake-scale are shown in left of Fig. 38. Right of Fig. 38
is the similar plot for the wake of a flat plate (Longo et al. 1998). The distributions of
the JBC case look closer to the wake profiles rather than the boundary layer profiles
in Fig. 35.
In summary, it is rather difficult to assess the quantitative Reynolds number effect
from the data currently available. Although the tendencies seem to be reasonable,
the extremely higher peak of NMRI data compared with other cases cannot be
fully justified. Further investigations, such as additional measurements in different
facilities and/or different measuring systems, are desirable for obtaining the proper
distributions of TKE.
Fig. 36 Relation of Rn and the maximum TKE k +

46 N. Hirata et al.
Fig. 37 Wake-scale distribution of u velocity along a horizontal line at the shaft height in S4 section
shown in Fig. 35
Fig. 38 Left: Wake-scale distribution of TKE along a horizontal line at the shaft height in S4
section shown in Fig. 35. Right: Distributions of turbulent kinetic energy in the wake of a flat plate
(Longo et al. 1998)
On the other hand. for the mean velocity components, the data of three facilities
are generally in good accordance. However, TUHH measurement has the problem
of laser reflection under the propeller shaft and NMRI data has some errors (about
3−4% of the uniform axial velocity) in v near the side edges of measuring area
though the errors can be negligible in the center. In total, the OU data seems to have
no serious deficiencies and thus this can be considered as the reliable measured data.
Table 7 Test cases of JBC

Test case Conditions
local flow fields
Cases1.3a and 1.3b Towed in calm water condition w/o
rudder
w/o propeller and w/o ESD
Case1.4a Towed in calm water condition w/o
rudder
w/o propeller and with ESD
Case1.7a Self-propelled at the ship point in calm
water
w/o rudder, with propeller and w/o ESD
Case1.8a Self-propelled at the ship point in calm
water
w/o rudder, with propeller and with ESD
4.4 Local Flow Data of NMRI, OU and TUHH Used in Test

Cases
In the Workshop, the test cases for local flow fields are set up as Table 7 based on
the measurement described above. SPIV data of the measuring sections S2, S4 and
S7 shown in Fig. 16 are picked up for all the test cases. Cases 1.3 and 1.4 are towed
condition without and with ESD, respectively. Similarly, Cases 1.7 and 1.8 are the
self-propelled condition without and with ESD. Note that the NMRI data were used
for all the cases with the identification “a” and the part of the TUHH data of LDV
measurement was used in Test Case 1.3b. The OU data and SPIV data of TUHH were
not used in the Workshop. All the data both used and not used in the Workshop are
listed in Table 8. In Cases 1.7 and 1.8 with a rotating propeller, ‘averaged’ indicates
the mean flow data and ‘prop000’ to ‘prop048’ indicate the phase averaged data with
the blade angles of 0 to 48 degrees, respectively. Actual figures of the data used in the
Workshop data are shown in Chap. 6 together with the submitted numerical results.
5 Wave Height Measurement
Wave height distributions around JBC advancing at the design speed Fn = 0.142
are measured in the large towing tank of NMRI. The ship is towed in a trim-free
condition without a propeller and a duct. The wave profile on the hull is acquired
from the photographs of the hull surface on which the ordinate and the abscissa are
marked. The longitudinal wave cuts are measured using the capacitance type wave
gauge at y/L P P = 0.1043 and 0.1900 where y is the lateral distance from the center
line of the hull. Figure 39 shows the wave profile on the hull. Figures 40 and 41 show
the longitudinal wave cuts at y/L P P = 0.1043 and 0.1900, respectively.
48 N. Hirata et al.
Table 8 List of local veocity data

Testcase Section Flow variables NMRI TUHH/LDV TUHH/SPIV OU
Case1.3 S2 u, v, w O *
ωx , k O *
S4 u, v, w O O * *
ωx , k O * *
S7 u, v, w O O * *
ωx , k O * *
Case1.4 S2 u, v, w O *
ωx , k O *
S4 u, v, w O *
ωx , k O *
S7 u, v, w O *
ωx , k O *
Case1.7 Averaged S2 u, v, w O *
Case1.7 prop000 S4 u, v, w O *
O: data used in the Workshop *: data not used in the Workshop
6 Conclusions
In this Chapter, the measured data for Japan Bulk Carrier (JBC) with an energy saving
circular duct are presented. The measurement is conducted in three facilities, i.e., the
towing tanks of National Maritime Research Institute (NMRI) and Osaka University
(OU) and the wind tunnel of Hamburg University of Technology (TUHH).
Fig. 39 Wave profile on the hull
Fig. 40 Longitudinal wave cut at y/L P P = 0.1043
Fig. 41 Longitudinal wave cut at y/L P P = 0.1900
Resistance, sinkage, trim, self-propulsion factors with and without the duct are
acquired by the tank tests at NMRI and OU. Wave field data is measured at the NMRI
towing tank.
The detailed flows fields in seven stations in a stern region are measured by
using SPIV in the tanks of NMRI and OU. The data are acquired in towed and self-
propelled conditions without and with the duct. In addition, LDV/SPIV measurement
is conducted at the TUHH wind tunnel.
50 N. Hirata et al.
For the SPIV measurement at NMRI, the error estimates are carried out using
the uniform flow test results and the statistical convergence analysis of the actual
measurement. It turned out that uncertainty of the velocity measurement is approx-
imately 2−3% of the uniform flow and the uncertainty of TKE measurement is
approximately 1% of the square of the uniform flow magnitude. Uncertainty of v
velocity is largest and this may be attributed to the problem with the reflection mirror
setting.
For the mean velocity components, the data of three facilities are generally in
good accordance in spite of some problems such as the laser reflection in TUHH
measurement or the reflection mirror problem in NMRI measurement.
On the other hand, it is found that there is a large difference in the measured TKE
levels between NMRI and OU/TUHH. Examination of the statistical convergence
of NMRI measurement shows that the difference between NMRI and other facilities
does not seem to come from the statistical convergence, though apparently the more
frames are needed for the accurate estimation of turbulent quantities. The effect of
Reynolds number difference is also investigated for the local TKE distributions along
the horizontal lines at the shaft height near the propeller plane. The distributions in
the bare-hull towing condition are compared in the various scaling. It appears that
the TKE distributions look more wake-like rather than boundary-layer-like from
the comparisons with the flat plate data. However, it is rather difficult to specify
the exact reason for differences of the extremely higher TKE of NMRI data. Further
investigations, such as additional measurements in different facilities and/or different
measuring systems, are desirable for obtaining the reasonable distributions of TKE.
Finally, the recommendations to the future workshops at present are as follows:
For the resistance and self-propulsion, sinkage and trim and the wave profiles, the
measured data from NMRI can be considered to be appropriate. For the local flow
data, the OU data seems to have no serious deficiencies and thus can be recommended
as the reliable measured data of mean velocities. Further works will be needed to
establish reliable turbulence data of this JBC case.
References
Hino, T., Hirata, N., Ohashi, K., Toda, Y., Zhu, T., Makino, K., et al. (2016). Hull form design
and flow measurements of a bulk carrier with an energy-saving device for CFD validations. In
Proc. 13th International Symposium on Practical Design of Ships and Other Floating Structures.
Denmark: Copenhagen.
Jufuku, N., Hori, M., Itou, S., Toda, Y., & Hinatsu, M. (2015). SPIV stern flow measurement around
operating propeller for with and without duct condition of Japan bulk carrier (in Japanese). In
Conference Proceedings of the Japan Society of Naval Architects and Ocean Engineers (Vol. 21,
pp. 309 − 312).
Shevchuk, I., Sahab, A., Stern, F., & Abdel-Maksoud, M. (2020). Experimental and numerical
studies of the flow around the JBC hull form at straight ahead condition and 8° drift angle. In
Proc. the 33rd Symposium on Naval Hydrodynamics. Japan: Osaka.
Yoon, H., Longo, J., Toda, Y., & Stern., F. (2015). Benchmark CFD validation data for surface
combatant 5415 in PMM maneuvers – Part II: Phase-averaged stereoscopic PIV flow field
measurements. Ocean Engineering, 109, 735–750.
Longo, L., Huang, H. P., & Stern, F. (1998). Solid/free-surface juncture boundary layer and wake.
Experiments in Fluids, 25, 283 − 297.
Lee, S. -J., Kim, H. -R., Kim, W. -J., & Van, S. –H. (2003). Wind tunnel tests on flow characteristics
of the KRISO 3,600 TEU containership and 300 K VLCC double-deck ship models. Journal of
Ship Research, 47(1), 24 − 38.
Experimental Data for KCS Resistance,
Sinkage, Trim, and Self-propulsion
Jin Kim
Abstract This chapter is dedicated to the experimental results of resistance, sinkage,

trim, and self-propulsion tests of the KCS hull form. The KCS was conceived to
provide data for both explication of flow physics and CFD validation for a modern
container ship with a bulbous bow. The Korea Research Institute of Ships and Ocean
engineering (KRISO) performed towing tank experiments to obtain resistance, mean
flow data and free surface waves. Self-propulsion tests were carried out at the National
Maritime Research Institute (NMRI) and later, additional resistance tests were also
reported in Gothenburg 2010 Workshop on CFD in ship hydrodynamics. The main
information of the KCS hull and its propeller is introduced in Sect. 1.1 and then the
results of resistance and hull attitudes measurement are followed by self-propulsion
results and some local flow measurements.
1 KCS (KRISO Container Ship)
1.1 Ship Model
The ship hull used in Tokyo 2015 Workshop is a 3,600 TEU container carrier called
KRISO Container Ship (KCS) (Van et al. 1998) which has been designed for the vali-
dation of CFD predictions. This hull form was selected also in the Gothenburg 2010
Workshop on CFD in ship hydrodynamics (Larsson et al. 2010; Larsson et al. 2013).
KCS has a low block coefficient and the design speed is 24kn (Kim et al. 2001). The
model ship was made with a scale ratio of 1/31.6 (i.e. Lpp = 7.2786 m). A conven-
tional five-bladed propeller with the NACA66 section is adopted. The geometries of
the hull and the propeller are shown in the Figs. 1 and 2, respectively. The principal
particulars of the ship and the propeller are listed in Tables 1 and 2.
J. Kim (B)
Korea Research Institute of Ships and Ocean Engineering (KRISO), Daejeon, South Korea
54 J. Kim
Fig. 1 Geometry of the KCS (bow and stern view)
Fig. 2 Geometry of the

KP505 (KCS propeller)
Table 1 Principal particulars of KCS

Lpp [m] Lwl [m] B [m] D [m] T [m] CB
Full scale 230.0 232.5 32.2 19.0 10.8 0.6505
Model 7.2786 7.3577 1.0190 0.6013 0.3418
Table 2 Principal particulars of the model propeller

Dp[m] Hub ratio Pitch type EAR No. of blades Blade section
0.25 0.18 FPP 0.50 5 NACA66
1.2 Test for Resistance, Sinkage, and Trim
The global force and the hull attitudes were measured in the towing tank of the
KRISO (Van et al. 1998). This test corresponds to case 2.1. The rudder was fitted,
and the measurement was carried out for the 6 speeds shown in Table 3. The obtained
values of C T , sinkage, and trim angle is listed in Table 4 where Sf and Sa mean
sinkage at FP and AP, respectively, and σ means the mean sinkage. Figure 3 shows
Experimental Data for KCS Resistance, Sinkage, Trim … 55
Table 3 Test conditions for case 2.1

No. 1 2 3 4 5 6
Speeds [m/s] 0.915 1.281 1.647 1.922 2.196 2.379
Fr 0.108 0.152 0.195 0.227 0.260 0.282
Re 5.23 × 106 7.33 × 106 9.42 × 106 1.10 × 107 1.26 × 107 1.36 × 107
Table 4 The measured values of CT , sinkage, and Trim angle

Fr C T × 103 Sf × 102 [m] Sa × 102 [m] σ × 102 [m] τ°
0.108 3.796 −0.195 0.015 −0.090 −0.017
0.152 3.641 −0.613 0.063 −0.275 −0.053
0.195 3.475 −1.213 0.015 −0.599 −0.097
0.227 3.467 −1.750 −0.138 −0.944 −0.127
0.260 3.711 −2.465 −0.322 −1.394 −0.169
0.282 4.501 −2.710 −0.695 −1.702 −0.159
Fig. 3 The measured total resistance coefficients
the measured total resistance coefficients. The data uncertainty, U D , of C T is approx-

imately 1.0%D. Figures 4 and 5 show the measured sinkage and trim angle results.
The data uncertainty of the sinkage and trim was not reported at Van et al. (1998).
We assumed 1.0%D, but these values are reasonable only in the high Fn region. The
data uncertainty in the low Fn region might be much higher than 1.0%D. In Chap. 8,
all participants show large comparison errors in the low Fn region.
56 J. Kim
Fig. 4 The measured mean sinkage
Fig. 5 The measured trim angle

1.3 Test for Self-propulsion
The self-propulsion test was carried out at the NMRI (Hino 2005). This test corre-
sponds to case 2.3 and case 2.7. The hull is free to heave and pitch but ballast weights
are arranged to give zero speed attitude at the design speed and no rudder is fitted. The
towing speed is set corresponding to Fr is 0.26. The self-propulsion test is conducted
at the ship point, which means that one should adjust the rate of rotation of the
propeller to get a force equilibrium taking into account the additional towing force
(Skin Friction Correction, SFC). The revolution rate of the propeller model is set to
9.5 rps at the ship point and the SFC is 30.25 N at that time. The three-dimensional
velocity field is measured around the stern by a spherical type eight-hole Pitot tube
(Fujisawa et al. 2000). For the surface pressure measurements on the stern 144 pres-
sure taps are employed (Tsukada et al. 2000). The measured self-propulsion factors
are summarized in Table 5. The measurement of the local velocity field is carried
out downstream of the propeller at x/Lpp = 0.991 and the corresponding results
are presented in Fig. 6. The axial velocity contours are characterized by two regions
inside and outside of the propeller disk. A moon crescent-like region of high velocity
up to U = 1.1 caused by the rotating propeller is shown inside. A more detailed value
on the transversal evolution of the three components of the velocity is also measured
at x/Lpp = 0.991, z/Lpp = −0.03 as shown in Fig. 7. Measurements of the surface
pressure distribution for the validation of CFD results are conducted on the port side
of the hull (ref. Fig. 8).
Table 5 The measured values obtained from the self-propulsion test

CT × 103 KT KQ n(rps) (for given SFC) RT -T(N) (for given n)
3.966 0.170 0.0288 9.5 30.25
Fig. 6 Measured axial velocity contours and cross flow vectors at x/Lpp = 0.9911 (KCS)
58 J. Kim
Fig. 7 Velocity profiles downstream of the propeller
Fig. 8 Hull surface pressure coefficient contours with working propeller
2 Conclusions
Additional experimental data for KCS calm water resistance obtained after the work-
shop G2010 are reported. All KCS calm water models are listed at Table 1 in Chap. 4.
In the present chapter, all experimental data are obtained with 7.3 m model from the
KRISO and NMRI towing tanks. These experimental data have been widely used
as a standard benchmarking case for CFD validation after G2010. However, turbu-
lence quantities are still missing in the set of experimental data and we also feel
inconvenienced by the lack of local flow measurement for self-propulsion including
a rudder. It is desirable to have more data to make up the full set of experimental
data, especially for the standard size KCS model ship.
References
Fujisawa, J., Ukon, Y., Kume, K., & Takeshi, H. (2000). Local velocity field measurements around
the KCS model (SRI M. S. No. 631) in the SRI 400 m towing tank. NMRI report 2000.
Hino, T. (ed.) (2005). Proceedings of CFD Workshop Tokyo 2005. NMRI report 2005.
Kim, W. J., Van, S. H., & Kim, D. H. (2001). Measurement of flows around modern commercial
ship models. Experiment in Fluids, 31, 567–578.
Larsson, L., Stern, F., & Visonneau, M. (2010). Proceedings of Gothenburg 2010 a workshop on
numerical ship hydrodynamics. Gothenburg, Sweden.
Larsson, L., Stern, F., & Visonneau, M. (ed.) (2013). Numerical ship hydrodynamic –An assessment
of the Gothenburg 2010 Workshop. Springer.
Tsukada, Y., Hori, T., Ukon, Y., Kume, K., & Takeshi, H. (2000). Surface pressure measurements
on the KCS model (SRI M. S. No. 631) in the SRI 400 m towing tank. NMRI report 2000.
Van, S.H., Kim, W.J., Yim, G.T., Kim, D.H., & Lee, C.J. (1998). Experimental investigation of
the flow characteristics around practical hull forms. In: Proceedings 3rd Osaka Colloquium on
Advanced CFD Applications to Ship Flow and Hull Form Design. Osaka, Japan.
Experimental Data for KCS Added
Resistance and ONRT Free Running
Course Keeping/Speed Loss in Head
and Oblique Waves
Yugo Sanada, Claus Simonsen, Janne Otzen, Hamid Sadat-Hosseini,

Yasuyuki Toda, and Frederick Stern
Abstract Evaluation is performed of the data used for T2015 test cases for the KCS
captive added resistance σAR and ONRT free running course keeping/speed loss
in head and oblique waves. For KCS calm water resistance, the individual facility
N-order level testing uncertainty is UXi = 1%D and the multiple facility standard
deviation based on three institutes using 2 model sizes (L = 7.3 and 6.1 m) is SD
= 0.74%D, such that the individual facility MxN-order level testing uncertainty is
UDi = 1.75%D. UXi was not reported for sinkage and trim; however, the SD = 4 and
7%D, respectively. For KCS head waves, the analysis was based on three institutes
with model sizes L = 6.1, 3.2 and 2.7 m. For the 6.1 m model, FORCE provided
UXi = 8, 4 and 4%D, respectively, for σAR and first harmonic heave z1 /ζ1 and pitch
θ1 /ζ1 k amplitudes, whereas for the 2.7 m model, FORCE/IIHR provided UXi =
7/18, 8/2 and 9/5%D, respectively. In consideration of the differences in model sizes
and rigid vs. surge free mounts the agreement between the facilities is reasonable:
for the primary variables, SD = 3, 20 and 10%D for z1 /ζ1 and θ1 /ζ1 k amplitudes
and σAR , respectively; and for secondary variable SD = 66%D for first harmonic
resistance CT1 . For KCS oblique waves, the data is only available from IIHR for
tests in 2015 and 2016. The agreement is good, i.e., SD values are comparable to
the corresponding values for head waves: for primary variables, SD = 6 and 63%D
for z1 /ζ1 and θ1 /ζ1 k amplitudes and σAR , respectively; and for secondary variable SD
= 13%D for CT1 . For ONRT self-propulsion and head and oblique waves, the data
is only available from IIHR and only preliminary uncertainty analysis is available.
The evaluation showed reasonably reliable data for both KCS and ONRT. However,
Y. Sanada · F. Stern (B)

University of Iowa and IIHR-Hydroscience & Engineering, Iowa City, IA, USA
C. Simonsen · J. Otzen
FORCE Technology, Lyngby, Denmark
H. Sadat-Hosseini
University of North Texas, Denton, TX, USA
Y. Toda
Osaka University, Suita, Japan
62 Y. Sanada et al.
clearly it is desirable to have data from more facilities including uncertainty analysis
for assessment of facility biases, which will provide more robust data and uncertainty
analysis for CFD validation.
Nomenclature
A Desired wave amplitude (m)

bR Bias error
B Breadth (m)
BWL Maximum beam at the waterline (m)
CB Block coefficient
CF Frictional resistance coefficient
◦
C F15 Nominal frictional resistance coefficient
CR Residual resistance coefficient
CT Total resistance coefficient
◦
C T15 Nominal total resistance coefficient (Converted from measured tempera-
ture to nominal temperature 15 °C)
CTi Ith harmonic of total resistance coefficient
D Depth (m)
DR Dynamic range
f nz Natural heave frequency (Hz)
f nθ Natural pitch frequency (Hz)
f nφ Natural roll frequency (Hz)
FORCE Technology, Denmark
Fr Froude number
GM Metacentric Height (m)
H Wave height (m)
HS Significant wave height (m)
ITTC International Towing Tank Conference
k Wave number
1+k Form factor
KG Vertical center of gravity (m)
kxx Radius of gyration in roll (m)
kyy Radius of gyration in pitch (m)
KRISO Korea Research Institute of Ships and Ocean Engineering
KCS KRISO Container Ship
L Ship length (m)
LCB Longitudinal center of buoyancy
LPP Length between perpendiculars (m)
LWL Length at the waterline (m)
n Propeller revolutions (rps)
NMRI National Maritime Research Institute
NRIFE National Research Institute of Fisheries Engineering
Experimental Data for KCS Added Resistance … 63
ONRT ONR Tumblehome

OU Osaka University
Re Reynolds number
SR Random error
S Wetted surface area (m2 )
SD Standard deviation
SR Wetted suface area of rudder (m2 )
SW Wetted surface area of hull (m2 )
t Time (s)
T Draft (m)
Te Wave encounter period (s)
TP Peak wave period (s)
u Surge velocity (m/s)
ui I-th harmonic surge velocity amplitude (m/s)
uR Total uncertainty
U Ship speed (m/s)
UDi Individual facility experimental MxN uncertainty
UXi Individual facility experimental uncertainty
U95,R 95% confidence experimental uncertainty
v Sway velocity (m/s)
vi I-th harmonic sway velocity amplitude (m/s)
V Ship speed (m/s)
x Surge (m)
X X force (N)
Zi I-th harmonic heave amplitude (m)
δ Rudder angle (deg)
FP Draft changes at FP (m)
AP Draft changes at AP (m)
ζ Wave elevation (m)
ζi I-th harmonic wave amplitude
θ Pitch angle (rad)
θi I-th harmonic pitch amplitude (rad)
λ Wave length (m)
ρ Density of water (kg/m3 )
σAR Added resistance
σ/L Sinkage
τ Trim(deg)
φ Roll angle (rad)
φi I-th harmonic roll amplitude (rad)
χ,μ Wave encounter angle (deg)
ψ Yaw angle (rad|deg)
ψi I-th harmonic yaw amplitude (rad)
∇ Displacement (m3 )
64 Y. Sanada et al.
1 Introduction
The experimental data used for T2015 test cases for KCS added resistance in regular
head (2.10) and oblique (2.11) waves and ONRT free running course keeping/speed
loss in regular head (3.12) and oblique (3.13) waves is evaluated, including in the
former case calm water resistance, sinkage and trim and in the latter case calm water
self-propulsion (3.9). The Froude numbers Fr for KCS and ONRT are 0.26 and 0.2,
respectively. H/λ = 1/60 and 1/50 for the KCS and ONRT test cases, respectively. The
evaluation compares data and uncertainty analysis from multiple facilities and model
sizes when available. Only FORCE and IIHR provided experimental uncertainty
estimates following ITTC (2008) and ASME (2013), respectively.
The individual facility data value is Xi and when available its individual facility
N-order level testing uncertainty estimate is UXi . The multiple facility MxN order
M
level testing average value X = M1 X i and precision limit are PXi = 2σXi , where
i=1
M
2
σ X2 i = 1
M−1
Xi − X
is the standard deviation SD such that the individual
i=1
facility MxN-order level testing uncertainty estimate is UDi = U Xi
2
+ PXi
2
. The
discussions use both Xi and X , which are often simply referred to as D and similarly
for UXi and UDi as UD for convenience. Note that herein SD was calculated using the
population definition with factor 1/M vs/the sample definition with factor 1/(M-1)
such that the values are under conservative by a factor of 3/2.
The validation variables for calm water KCS are resistance, sinkage and trim and
for ONRT self-propulsion are propeller revolutions, sinkage and trim. Resistance
and propeller revolutions are considered primary and motions secondary variables.
For KCS head waves, added/0th harmonic resistance, and 1st harmonic heave and
pitch are considered primary and 1st harmonic resistance as secondary variables;
with additionally 1st harmonic roll as primary variable for oblique waves. Similarly,
for ONRT free running in head and oblique waves primary variables are speed loss
(0th harmonic axial velocity) and first harmonic motions which additionally include
yaw are the primary variables and other 0th and 2nd harmonic variables as secondary.
2 KCS Calm Water
Table 1 lists the KCS full and model scale dimensions, mass properties, natural
frequencies and form factors. The full-scale and NMRI/KRISO 7.3 m model data is
from Zou and Larsson (2014). The FORCE 4.3 m model data is from Simonsen et al.
(2013). The FORCE 6.1 m model data and uncertainty values are from Simonsen et al.
(2014) and Otzen (2015a). The FORCE 2.7 m model data and uncertainty values are
from Otzen (2015b). The IIHR 2.7 m model data is from both 2015 (Sadat-Hosseini
Table 1 Principle particulars for full scale, KRISO, NMRI, FORCE, and IIHR KCS models
Full Scale NMRI KRISO FORCE1 FORCE2 FORCE3 IIHR OU
Length LPP (m) 230 7.2786 7.2786 4.367 6.070 ± 0.001 2.700 ± 0.001 2.700 ± 0.001 3.200 ± 0.001
between
perpendiculars
Length of LWL (m) 232.482 7.3571 7.3571 4.414 6.135 ± 0.001 2.729 ± 0.001 2.729 ± 0.001 3.235 ± 0.001
waterline
Maximum BWL (m) 32.2 1.019 1.019 0.611 0.850 ± 0.001 0.378 ± 0.001 0.378 ± 0.001 0.4480 ± 0.0001
beam of
waterline
Depth D (m) 19 0.601 0.601 – – – 0.223 0.2643 ± 0.0001
Draft T (m) 10.8 0.3418 0.3418 0.205 0.2850 ± 0.0006 0.1268 ± 0.0006 0.1268 ± 0.0006 0.1503 ± 0.0001
Displacement ∇/(m3 ) 52061.738 1.649 1.649 0.356 0.9571 ± 0.002 0.0842 ± 0.0005 0.0842 ± 0.0005 0.140
volume
Experimental Data for KCS Added Resistance …
Wetted surface SW (m2 ) 9500.728 9.4379 9.4379 3.436 6.6177 ± 0.02 1.309 ± 0.008 1.309 ± 0.008 1.839
area w/o
rudder
Wetted surface SR (m2 ) 115 0.0741 0.0741 – 0.08 0.016 0.016 0.022
area of rudder
Block ∇/(LPP BWL T) 0.6505 0.6505 0.6505 0.651 0.6505 0.643 0.643 0.651
coefficient
(CB)
Midship 0.9849 0.9849 0.9849 0.985 0.9849 0.983 0.983 0.9849
section
coefficient
(CM)
(continued)
65
Table 1 (continued)
66
Full Scale NMRI KRISO FORCE1 FORCE2 FORCE3 IIHR OU

LCB (%LPP), −1.48 −1.48 −1.48 −1.48 −1.48 −1.48 −1.48 −1.48
fwd+
Vertical Center KG (m) 14.324 0.378 0.168 ± 0.02
of Gravity
(from keel)
Metacentric GM (m) 0.623 0.016 0.0067 ± 0.0001
height
Moment of kxx /B 0.4 0.4 0.4 0.4 0.4 0.4 0.39 ± 0.02
Inertia
Moment of kyy /LPP 0.25 0.25 0.25 0.25 0.252 0.25 0.25 ± 0.01 0.25
Inertia
Natural f nz (Hz) 0.121 0.121 0.681 0.880 0.746 1.119 1.119
Frequency of
Heave
(Estimated)
Natural f nφ (Hz) 0.031 0.031 0.168 0.217 0.185 0.276 0.283
Frequency of
Roll
(Estimated)
Natural f nθ (Hz) 0.118 3.739 0.665 0.859 0.722 1.092 1.092
Frequency of
Pitch
(Estimated)
Form factor 1+k 1.082 1.083 1.087 1.088 1.283
Y. Sanada et al.
et al. 2015a) and 2016 and uncertainty values are from Stocker (2016). The OU 3.2 m
model data is from Toda (2018).
Adjustment of ship mass properties such as metacentric height (GM), radius of
gyration in roll (kxx ) and pitch (kyy ) for the IIHR model was done as follows. (1)
Ballasting weights were added to adjust the water line. The weights were distributed
in a way that there was no pitch and roll. (2) Inclining tests were conducted in water.
One of the ballasting weights that was placed around the ship center of gravity in the
1st step was selected as the moving weight and was moved transversally to induce
a roll angle. If the calculated GM based on the roll angle is not within ±2% of the
desired GM, the distribution of ballasting weights in the vertical direction would be
changed and the test would be repeated. (3) The ship was removed from the water
and the swing tests were conducted. kxx and kyy can be obtained by measuring the
swing period. In the swing tests for kxx , the kxx was adjusted to the desired value by
moving weights away from the centerline symmetrically. For the kyy adjustment, the
weights were moved away from the center of gravity in the longitudinal direction.
The swing tests were repeated until the obtained kxx and kyy are within ±2% of the
desired value. (4) Free roll decay experiments were conducted to evaluate the natural
roll period in water.
IIHR made improvements to the experimental setup between 2015 and 2016. The
improvements were using a reference wave gauge mounted to the surge free mount vs.
sub-carriage and use of a laser tracking system and accelerometers for the resistance
and x data reduction. The OU 3.2 m model data is from Sanada et al. (2018a). In
summary, calm water resistance C T , sinkage σ/LPP and trim τ experimental data are
available from five facilities (KRISO, NMRI, FORCE, IIHR and OU) using five
different model sizes (7.3, 4.4, 6.1, 2.7 and 3.2 m) and different types of mounts.
The resistance, sinkage and trim data-reduction equations are:
◦ ◦
C T15 = C R + C F15 (1 + k) (1)
X
CT = (2)
1/2ρ SV 2
C F = .075/(log10 Re − 2)2 (3)
C R = C T − C F (1 + k) (4)
CT Fr 4
=m + (1 + k) (5)
CF CF
σ
= (F P + A P)/L P P (6)
LPP
τ = (F P − A P)/L P P (7)

68 Y. Sanada et al.
where ρ is the water density, S is the wetted surface area excluding wetted surface
area of rudder (S = SW ), V is the ship velocity, L PP is the ship length, ΔFP and
ΔAP are the draft changes at the fore and aft perpendiculars, k is the form factor, X
is the resistance in calm water, m is the slope of the line based on Prohaska method,
◦
C T15 corrects the resistance coefficient for the same temperature at 15 °C, CF is
the frictional resistance coefficient, and CR partially removes scale/Re effects. Form
factor k was evaluated using Eq. (5), i.e., Prohaska method.
◦
Tables 2 and 3 and Fig. 1 provide C T15 without and with form factor. Figure 1
15◦
shows C T increases from large to small model sizes and values are often higher
with form factor. All the available data is tabulated for all Fr. Most data is for Fr
= 0.26. The standard deviation SD is presented using both %D and %DR where D
is the average data value and DR is the dynamic range and both with and without
the 2.7 and 3.2 m model data. The scatter in the results is much larger including the
2.7 and 3.2 m models, but reasonably small for the larger models both without and
with form factor, as shown in Fig. 1. For Fr = 0.26, the scatter in the results for
models excluding the 2.7 m model is < 7% and 3%D without and with form factor,
respectively. Including the 2.7 m model increases the scatter by factors of about 2
and 5 for without and with form factor, respectively. The differences without vs.
with form factor are small if all models are included but large if the smaller models
are excluded. Tables 4 and 5 and Fig. 2 report residual resistance CR without and
◦
with form factor and they show similar results to those for C T15 , but somewhat larger
◦
%D and %DR as CR < C T15 . In this case, the SD differences without vs. with the
form factor are large both including and excluding the smaller models. Thus, the SD
results without and with form factor do not show a consistent trend.
Experimental uncertainty UXi estimates for calm water tests are provided from
only two institutes. For CT , FORCE provided UXi = 1 and 4%D for 6.1 and 2.7 m
models, respectively. IIHR provided UXi = 2.5%D for the 2.7 m model. For CFD
validation studies in Chap. 9 for test case 2.10, the SD for CR with form factor and
without the 2.7 m model was normalized by D for FORCE and multiplied by 2 and
was used as the MxN-order level testing precision limit PXi for CT [Chap. 9, Eq. (29)],
i.e., 4.7 × (0.602/3.863) × 2 = 1.46%D.
Tables 6 and 7 and Fig. 3 show the sinkage and trim provided from different
institutes. Figure 3 shows magnitudes of both sinkage and trim increase for increasing
◦
Fr. The trends for the sinkage and trim scatter are like those for C T15 and CR i.e.
including the 2.7 and 3.2 m models increases the scatter. For Fr = 0.26, the scatter
excluding the 2.7 and 3.2 m model is 4 and 7%D for sinkage and trim, respectively,
◦
i.e. larger than that for C T15 with form factor.
Only IIHR provided UXi estimates of 1 and 66%D for sinkage and trim, respec-
tively. For CFD validation studies in Chap. 9 for test case 2.10, SD = 4 and 7%D for
the larger models was used for sinkage and trim, respectively, normalized by D for
FORCE, i.e., for sinkage 3.8 × (−0.2014/−0.2074) × 2 = 7.38%D and for trim 7 ×
(−0.168/−0.165) × 2 = 14.25%D were used as the as the MxN-order level testing
precision limit [Chap. 9, Eq. (29)]. UXi was not considered for CFD validation since
it was not available for the large models. Note that FORCE and IIHR both used added
◦
Table 2 Comparison of total resistance coefficient calculated without using Prohaska’s method, C T15 , for KCS in calm water
Institute NMRI KRISO FORCE1 FORCE2 IIHR (Aug 2016) OU (Oct 2016) All Facilities Without 2.7 m and OU model
Fr CT15◦ × 103 Ave (D) SD%D {SD%DR} Ave (D) SD%D {SD%DR}
0.0867 3.841 3.841 0.0 3.841
0.1084 3.740 3.729 3.916 3.795 2.3 3.795 2.3
0.1300 3.657 3.657 0.0 3.657
0.1517 3.580 3.578 3.767 3.641 2.4 3.641 2.4
0.1734 3.503 3.678 3.591 2.4 3.591 2.4
0.1950 3.438 3.417 3.605 3.487 2.4 3.487 2.4
0.2059 3.415 3.415 0.0 3.415
0.2167 3.404 3.593 3.498 2.7 3.498 2.7

0.2276 3.409 3.413 3.411 0.0 3.411 0.0
0.2384 3.440 3.639 3.540 2.8 3.540 2.8
0.2492 3.505 3.505 0.0 3.505
0.2601 3.650 3.659 4.31 3.860 5.203 4.637 4.220 13.4 3.870 6.9
0.2709 3.981 3.981 0.0 3.981
0.2817 4.459 4.447 4.561 4.489 1.1 4.489 1.1
Ave (Total) 3.645 3.707 3.827 3.719 2.1.0{14.6*} 3.726 2.6{17.6 +}
*DR = 0.539
+DR = 0.539
69
70
◦
Table 3 Comparison of total resistance coefficient calculated using Prohaska’s method, C T15 , for KCS in calm water
Institute NMRI KRISO FORCE2 IIHR (Aug 2016) OU (Oct 2016) All Facilities Without 2.7 m and OU model
Fr CT15◦ × 103 Ave (D) SD%D {SD%DR} Ave (D) SD%D {SD%DR}
0.0867 3.827 3.840 0.4 3.840 0.4
0.1084 3.727 3.721 3.919 3.789 2.4 3.789 2.4
0.1300 3.645 3.648 3.647 0.0 3.647 0.0
0.1517 3.568 3.573 3.770 3.637 2.6 3.637 2.6
0.1734 3.492 3.487 3.681 3.554 2.5 3.554 2.5
0.1950 3.428 3.412 3.608 3.482 2.6 3.482 2.6
0.2059 3.404 3.392 3.398 0.2 3.398 0.2
0.2167 3.393 3.389 3.596 3.460 2.8 3.460 2.8
0.2276 3.399 3.406 3.402 0.1 3.402 0.1
0.2384 3.430 3.441 3.642 3.504 2.8 3.504 2.8
0.2492 3.495 3.514 3.504 0.3 3.504 0.3
0.2601 3.640 3.652 3.863 4.888 4.676 4.144 12.8 3.718 2.8
0.2709 3.972 3.948 3.960 0.3 3.960 0.3
0.2817 4.449 4.443 4.564 4.485 1.2 4.485 1.2
Ave (Total) 3.634 3.634 3.830 4.888 3.7 2.2{15.1*} 3.670 1.6{10.7 +}
*DR = 0.544
+DR = 0.544
Y. Sanada et al.
7.0 6.0
NMRI NMRI(1+k=1.082)
6.5 KRISO KRISO(1+k=1.083)
1 2
FORCE 5.5 FORCE (1+k=1.087)
2
FORCE IIHR(1+k=1.088)
6.0 IIHR OU(1+k=1.283)
OU 5.0 Mean
5.5 Mean
3
3
CT x10
CT x10
5.0 4.5
15°
15°
4.5
4.0
4.0
3.5
3.5
3.0 3.0
0.05 0.1 0.15 0.2 0.25 0.3 0.05 0.1 0.15 0.2 0.25 0.3
Fr Fr
(a) All Facilities, without Prohaska method (b) All Facilities, using Prohaska method
5.0 5.0
KRISO KRISO(1+k=1.083)
2
FORCE1 FORCE (1+k=1.087)
4.5 2 4.5 Mean
FORCE
Mean
3
3
CT x10
CT x10
4.0 4.0
15°
15°
3.5 3.5
3.0 3.0
0.05 0.1 0.15 0.2 0.25 0.3 0.05 0.1 0.15 0.2 0.25 0.3
Fr Fr
(c) Excluding 2.7m model, without Prohaska method (d) Excluding 2.7m model, using Prohaska method
◦
Fig. 1 Comparison of CT15 × 103 calculated with and without Prohaska’s method, at NMRI,
KRISO, FORCE, IIHR, OU and mean of the facilities. Standard deviation between facilities is
included with the mean line
resistance mounts designed for large motions and not conventional calm water test
setups. Further FORCE used mount with fixed surge, while IIHR used free surge.
The IIHR and OU surge free mounts for calm water resistance, sinkage and trim
measurements were designed primarily for ship motions studies, and possibly small
model size are reasons for its large scale-effect/facility bias. IIHR and OU sinkage
and trim were measured by a 2DOF mount (sinkage and trim free) installed at the
hull center of gravity. The displacement measurement at the bow (FP) and stern (AP)
performed in the ordinary calm water resistance test is more accurate in measuring
small displacements than this method. IIHR and OU should re-take the measurements
using rigid mounts in their towing tanks to provide better calm water reference data.
All FORCE calm water measurements are also performed with added resistance
mounts designed for large motions, but with fixed surge and not with conventional
72
Table 4 Comparison of residuary resistance coefficient calculated without using Prohaska’s method, CR, for KCS in calm water
Fr CR × 103 Ave (D) SD%D {SD%DR} Ave (D) SD%D {SD%DR}
0.0867 0.403 0.403 0.0 0.403
0.1084 0.440 0.427 0.448 0.438 1.9 0.438 1.9
0.1300 0.464 0.464 0.0 0.464
0.1517 0.473 0.472 0.505 0.483 3.2 0.483 3.2
0.1734 0.469 0.495 0.482 2.7 0.482 2.7
0.1950 0.465 0.444 0.485 0.465 3.6 0.465 3.6
0.2059 0.470 0.470 0.0 0.470
0.2167 0.484 0.532 0.508 4.7 0.508 4.7
0.2276 0.514 0.516 0.515 0.2 0.515
0.2384 0.568 0.629 0.599 5.1 0.599 5.1
0.2492 0.654 0.654 0.0 0.654
0.2601 0.820 0.828 1.07 0.897 1.177 1.111 0.984 14.4 0.905 11.3
0.2709 1.170 1.170 0.0 1.170
0.2817 1.666 1.655 1.635 1.652 0.8 1.652 0.8
Ave (Total) 0.647 0.724 0.703 0.663 2.6{2.8*} 0.658 4.2{4.4 +}
*DR = 0.625
+DR = 0.625
Y. Sanada et al.
Table 5 Comparison of residuary resistance coefficient calculated using Prohaska’s method, CR, for KCS in calm water
Institute NMRI KRISO FORCE2 IIHR (Aug 2016) OU (Oct 2016) All Facilities Without 2.7 m and OU model
Fr CR × 103 Ave (D) SD%D {SD%DR} Ave (D) SD%D {SD%DR}
0.0867 0.107 0.118 9.9 0.118 9.9
0.1084 0.156 0.148 0.150 0.151 2.4 0.151 2.4
0.1300 0.190 0.190 0.190 0.1 0.190 0.1
0.1517 0.207 0.209 0.224 0.213 3.6 0.213 3.6
0.1734 0.209 0.201 0.220 0.210 3.7 0.210 3.7
0.1950 0.211 0.192 0.216 0.206 5.1 0.206 5.1
0.2059 0.217 0.202 0.210 3.6 0.210 3.6
0.2167 0.234 0.227 0.268 0.243 7.3 0.243 7.3

0.2276 0.266 0.270 0.268 0.8 0.268 0.8
0.2384 0.322 0.330 0.369 0.340 6.1 0.340 6.1
0.2492 0.410 0.427 0.418 2.0 0.418 2.0
0.2601 0.577 0.587 0.642 0.862 0.152 0.564 40.8 0.602 4.7
0.2709 0.930 0.903 0.917 1.5 0.917 1.5
0.2817 1.428 1.419 1.383 1.410 1.4 1.410 1.4
Ave (Total) 0.390 0.388 0.434 0.390 6.3{3.8*} 0.393 3.7{2.3 +}
*DR = 0.646
+DR = 0.646
73
74 Y. Sanada et al.
2.0
NMRI
NMRI(1+k=1.082)
KRISO
2.0 1 KRISO(1+k=1.083)
FORCE 1.5 2
2 FORCE (1+k=1.087)
FORCE
IIHR(1+k=1.088)
1.5 IIHR
OU(1+k=1.283)
OU 1.0 Mean
3
3
Mean
CRx10
CRx10
1.0
0.5
0.5
0.0
0.0
-0.5
0.05 0.1 0.15 0.2 0.25 0.3 0.05 0.1 0.15 0.2 0.25 0.3
Fr Fr
(a) All Facilities, without Prohaska method (b) All Facilities, using Prohaska method
2.0 2.0
KRISO KRISO(1+k=1.083)
FORCE1 FORCE2(1+k=1.087)
1.5 2 1.5
FORCE Mean
Mean
3
3
CRx10
CRx10
1.0 1.0
0.5 0.5
0.0 0.0
0.05 0.1 0.15 0.2 0.25 0.3 0.05 0.1 0.15 0.2 0.25 0.3
Fr Fr
(c) Excluding 2.7m model, without Prohaska method (d) Excluding 2.7m model, using Prohaska method
Fig. 2 Comparison of CR calculated with and without Prohaska’s method, at NMRI, FORCE, IIHR,
OU and mean of the facilities. Standard deviation between facilities is included with the mean line
calm water test setup. Ideally comparison of calm water resistance across facilities
should be made based on conventional resistance test setups, but that was not possible,
since data was taken in connection with added resistance test campaigns.
For KCS calm water resistance based on three institutes using 2 model sizes
with L ≥ 6 m, the individual facility N-order level testing uncertainty UXi = 1%D
(only reported one facility) and the MxN-order level testing standard deviation SD
= 0.74%D, such that the best estimate multiple facility uncertainty is UDi = [U2Xi +
(2SD)2 ]1/2 = 1.75%D. UXi was not reported for sinkage and trim; however, SD = 4
and 7%D such that UDi = 2SD = 7.38 and 14.15%D, respectively.
Table 6 Comparison of sinkage for KCS in calm water
Institute NMRI KRISO FORCE1 FORCE2 IIHR (Aug 2016) OU (Oct 2016) All Facilities Without 2.7 m and OU
model
Fr σ/LPP × 102 Ave (D) SD%D Ave (D) SD%D
{SD%DR} {SD%DR}
0.0867 −0.0216 −0.0105 105.7 −0.0105 105.7
0.1084 −0.0316 −0.0124 −0.0382 −0.0274 40.0 −0.0274 40.0
0.1300 −0.0447 −0.0219 −0.0333 34.2 −0.0333 34.2
0.1517 −0.0612 −0.0378 −0.0646 −0.0545 21.8 −0.0545 21.8
0.1734 −0.0807 −0.0608 −0.0902 −0.0772 15.9 −0.0772 15.9
0.1950 −0.1030 −0.0823 −0.1089 −0.0980 11.7 −0.0980 11.7
0.2059 −0.1154 −0.1026 −0.1090 5.9 −0.1090 5.9

0.2167 −0.1288 −0.1185 −0.1315 −0.1263 4.4 −0.1263 4.4
0.2276 −0.1434 −0.1297 −0.1365 5.0 −0.1365 5.0
0.2384 −0.1594 −0.1529 −0.1736 −0.1619 5.4 −0.1619 5.4
0.2492 −0.1771 −0.1716 −0.1743 1.6 −0.1743 1.6
0.2601 −0.1967 −0.1915 −0.2100 −0.2074 −0.1704 −0.1747 −0.1918 7.8 −0.2014 3.8
0.2709 −0.2184 −0.2138 −0.2161 1.1 −0.2161 1.1
0.2817 −0.2423 −0.2339 −0.2572 −0.2445 3.9 −0.2445 3.9
Ave (Total) −0.1231 −0.1092 −0.1340 −0.1187 18.9{19.2*} −0.1193 18.6{19.0 +
}
*DR = 0.117
+DR = 0.117
75
76
Table 7 Comparison of trim for KCS in calm water

Fr τ [deg] Ave (D) SD%D {SD%DR} Ave (D) SD%D {SD%DR}
0.0867 −0.011 −0.009 28.8 −0.009 28.8
0.1084 −0.019 −0.017 −0.029 −0.022 25.0 −0.022 25.0
0.1300 −0.031 −0.040 −0.035 13.0 −0.035 13.0
0.1517 −0.045 −0.053 −0.056 −0.051 9.3 −0.051 9.3
0.1734 −0.062 −0.077 −0.082 −0.073 11.7 −0.073 11.7
0.1950 −0.082 −0.097 −0.103 −0.094 9.2 −0.094 9.2
0.2059 −0.095 −0.105 −0.100 5.2 −0.100 5.2
0.2167 −0.108 −0.113 −0.128 −0.116 7.3 −0.116 7.3
0.2276 −0.122 −0.127 −0.124 2.1 −0.124 2.1
0.2384 −0.135 −0.144 −0.168 −0.149 9.3 −0.149 9.3
0.2492 −0.146 −0.158 −0.152 4.0 −0.152 4.0
0.2601 −0.152 −0.169 −0.185 −0.165 −0.022 0.006 −0.114 66.7 −0.168 7.0
0.2709 −0.150 −0.171 −0.160 6.8 −0.160 6.8
0.2817 −0.132 −0.159 −0.221 −0.171 21.9 −0.171 21.9
Ave (Total) −0.092 −0.102 −0.119 −0.098 15.7{19.0*} −0.102 11.5{14.4 +}
*DR = 0.081
+DR = 0.081
Y. Sanada et al.
0.05 0.10
0.00 0.05
-0.05 0.00
2
-0.10 -0.05
[deg]
/Lx10
NMRI
KRISO NMRI
-0.15 FORCE
1 -0.10 KRISO
1
2 FORCE
FORCE 2
-0.20 -0.15 FORCE
IIHR IIHR
OU OU
-0.25 Mean -0.20 Mean
-0.30 -0.25
0.05 0.1 0.15 0.2 0.25 0.3 0.05 0.1 0.15 0.2 0.25 0.3
Fr Fr
(a) σ/L, All Facilities (b) τ, All Facilities
0.05 0.10
NMRI
0.00 0.05 KRISO
1
FORCE
2
-0.05 0.00 FORCE
Mean
2
-0.10 -0.05
[deg]
/Lx10
-0.15 -0.10
NMRI
-0.20 KRISO -0.15
1
FORCE
2
-0.25 FORCE -0.20
Mean
-0.30 -0.25
0.05 0.10 0.15 0.20 0.25 0.30 0.05 0.1 0.15 0.2 0.25 0.3
Fr Fr
(c) σ/L, Excluding 2.7m model (d) τ, Excluding 2.7m model
Fig. 3 Comparison of σ/L and τ, at NMRI, KRISO, FORCE, IIHR, OU and mean of the facilities.
Standard deviation between facilities is included with the mean line
3 KCS Head Waves
Regular head wave experimental data are available from three facilities using three
model sizes (6.1, 2.7 and 3.2 m). The FORCE 6.1 m model data and uncertainty
values are from Simonsen et al. (2014) and Otzen (2015a). The FORCE 2.7 m model
data and uncertainty values are from Otzen (2015b). The IIHR 2.7 m model data is
from both 2015 (Sadat-Hosseini et al. 2015a) and 2016 and uncertainty values are
from Stocker (2016). IIHR made improvements to the experimental setup between
2015 and 2016, as discussed previously. Unfortunately, the FORCE 4.3 m model
data from Simonsen et al. (2013) is not available. Data is also recently available
78 Y. Sanada et al.
from Osaka University (Toda 2018), although not used at T2015 useful for assessing
the scale effects/facility biases.
Table 8 provides SD analysis for calm water resistance, sinkage, trim and added
resistance variables in head waves for Fr = 0.26. Here again the SD trends are
scattered and generally larger than for the larger models only, especially for CR with
the Prohaska method and trim.
Figures 4, 5, 6, 7 and 8 compare time histories from FORCE 6.1 m, IIHR 2.7 m,
OU 3.2 m models for wavelengths of λ/L = 0.65,0.85, 1.15, 1.37, and 1.95 for
wave elevation (ζ/A), resistance (CT ), surge (x/ζ1 ), heave (z/ζ1 ), and pitch (θ/ζ1 k).
Motions are non-dimensional by either wave amplitude (ζ1 ) or wave slope (ζ1 k).
Unlike FORCE and OU, IIHR used a roll free condition since roll measurement was
important for other headings for test case 2.11. Results show differences between
FORCE, IIHR and OU data for CT , heave, and pitch for short wave length conditions.
Note that FORCE experienced structural resonance in the rigid mount close to the
heave and pitch resonance condition. This is most clearly seen in CT for λ/L equal
to 0.85 and 1.15 in Figs. 5 and 6.
Figure 9 compares AR and 0th, 1st and 2nd harmonic amplitudes for ζ1 /A, CT ,
z/ζ1 , θ/ζ1 k; and Fig. 10 compares their 1st and 2nd harmonic phases. For the IIHR
data the natural surge damping effects of the surge free mount is removed by a least
square method (Stocker 2016) and the harmonic analysis window size for all λ/L in
head waves is 5Te. FORCE, IIHR and OU data agree well for AR, CT0 , z0 /ζ1 and
1st harmonic amplitudes, except CT1 (OU is outlier). The θ0 /ζ1 k and 2nd harmonic
amplitudes for CT (2nd harmonic amplitudes for heave and pitch are small) and CT1
1st and all 2nd harmonic phases show differences between the three measurements.
All three measurements show similar trends vs. wavelength for all 0th harmonic
amplitudes and heave and pitch 1st harmonic phases.
Tables 9, 10, 11, 12 and 13 provide SD analysis for added resistance σAR =
X −X calm
ρgA2 B2 /L
(X is the mean resistance in waves, A = ζ1 is the wave amplitude, and B
is the ship beam) and 1st and 2nd harmonic amplitudes for ζ/A, CT , z/ζ1 , and θ/ζ1 k,
respectively. Table 14, 15 and 16 provide SD analysis for 0th harmonic amplitudes for
CT , z/ζ1 and θ/ζ1 k, respectively; and Tables 17, 18, 19 and 20 provide SD analysis for
1st and 2nd harmonic phases for ζ/A, CT , z/ζ1 and θ/ζ1 k, respectively. The scatter for
the 1st harmonic amplitudes is mostly small, i.e., 4, 66, 3 and 20%D, respectively, for
ζ/A, CT , z/ζ1 , θ/ζ1 k. The scatter for 1st harmonic amplitudes are often larger for short
waves conditions. The scatter for 0th harmonic for CT (SD = 5%D) and z/ζ1 (SD
= 16%D) are relatively small. The scatter for the σAR (SD = 10%D), 0th harmonic
θ/ζ1 k (SD = 163%D), 1st harmonic phases, and 2nd harmonic amplitudes and phases
are mostly large.
FORCE provided UXi for σAR , heave and pitch, i.e., 8, 4 and 4%D, respectively, for
the 6.1 m model. FORCE/IIHR provided UXi for AR, heave and pitch, i.e., 7/18, 8/2
and 9/5%D, respectively, for the 2.7 m model. For CFD validation studies in Chap. 9
test case 2.10, SD was not considered, and the UXi values provided by FORCE for
the 6.1 m model were used, including estimates for the uncertainty due to statistical
convergence.
Table 8 Comparison of total resistance coefficient, residuary resistance coefficient, sinkage and trim in calm water and added resistance variables in head waves
(Fr = 0.26)
Institute LPP [m] Without Prohaska With Prohaska σ/LPP × 102 τ [deg] Added resistance in head wavesa
CT15◦ × 103 CR × 103 15◦
CT × 103 CR × 103 σAR ζ1 /A CT1 z1 /ζ1 θ1 /ζ1 k
FORCE 6.07 3.860 0.897 3.863 0.642 −0.2074 −0.165 4.850 1.010 9.790 0.613 0.598
IIHR 2.7 5.203 1.177 4.888 0.862 −0.1704 −0.022 5.180 1.070 9.300 0.610 0.606
OU 3.2 4.637 1.111 4.676 0.152 −0.1747 0.006 5.580 0.983 0.500 0.595 0.536
Ave. (D) 4.567 1.062 4.476 0.552 −0.1842 −0.060 5.200 1.020 6.530 0.606 0.580
SD%D 12.1 11.2 9.9 53.8 9.0 124.3 10.2 4.3 66.4 2.7 20.3
DR%D 14.7 13.2 11.5 64.3 10.1 131.9 14.0 8.5 142.3 3.0 12.1
a Average of all 5 different λ/L
79
80 Y. Sanada et al.
2
/L = 0.65
1
/A [-]
-1 FORCE(6.0702 m)
IIHR(2.7 m)
OU(3.2 m)
-2
0 2 4 6 8 10
t/Te
0.01
CT [-]
-0.01
0 2 4 6 8 10
t/Te
0.4
0.2
[-]
0
x/ 1
-0.2
-0.4
0 2 4 6 8 10
t/Te
1
0.5
[-]
0
z/ 1
-0.5
-1
0 2 4 6 8 10
t/Te
0.4
0.2
/ 1k [-]
-0.2
-0.4
0 2 4 6 8 10
t/Te
Fig. 4 Time series for force and motions of KCS in regular head waves at λ/L = 0.65 for FORCE2
IIHR and OU
2
/L = 0.85
1
/A [-]
-1 FORCE(6.0702 m)
IIHR(2.7 m)
OU(3.2 m)
-2 0 2 4 6 8 10
t/Te
0.02
0.01
CT [-]
-0.01
-0.02
0 2 4 6 8 10
t/Te
0.4
0.2
[-]
0
x/ 1
-0.2
-0.4
0 2 4 6 8 10
t/Te
1
0.5
[-]
0
z/ 1
-0.5
-1
0 2 4 6 8 10
t/Te
1
0.5
/ 1k [-]
-0.5
-1
0 2 4 6 8 10
t/Te
Fig. 5 Time series for force and motions of KCS in regular head waves at λ/L = 0.85 for FORCE2 ,
IIHR and OU
82 Y. Sanada et al.
2
/L = 1.15
1
/A [-]
-1 FORCE(6.0702 m)
IIHR(2.7 m)
OU(3.2 m)
-2
0 2 4 6 8 10
t/Te
0.04
0.02
CT [-]
-0.02
-0.04
0 2 4 6 8 10
t/Te
2
1
[-]
0
x/ 1
-1
-2
0 2 4 6 8 10
t/Te
1
[-]
0
z/ 1
-1
-2
0 2 4 6 8 10
t/Te
2
1
/ 1k [-]
-1
-2
0 2 4 6 8 10
t/Te
Fig. 6 Time series for force and motions of KCS in regular head waves at λ/L = 1.15 for
FORCE2 ,IIHR and OU
2
/L = 1.37
1
/A [-]
-1 FORCE(6.0702 m)
IIHR(2.7 m)
OU(3.2 m)
-2
0 2 4 6 8 10
t/Te
0.04
0.02
CT [-]
-0.02
-0.04
0 2 4 6 8 10
t/Te
0.4
0.2
[-]
0
x/ 1
-0.2
-0.4
0 2 4 6 8 10
t/Te
1
[-]
0
z/ 1
-1
0 2 4 6 8 10
t/Te
1
/ 1k [-]
-1
0 2 4 6 8 10
t/Te
Fig. 7 Time series for force and motions of KCS in regular head waves at λ/L = 1.37 for
FORCE2 ,IIHR and OU
84 Y. Sanada et al.
2
/L = 1.95
1
/A [-]
-1 FORCE(6.0702 m)
IIHR(2.7 m)
OU(3.2 m)
-2
0 2 4 6 8 10
t/Te
0.04
0.02
CT [-]
-0.02
-0.04
0 2 4 6 8 10
t/Te
0.4
0.2
[-]
0
x/ 1
-0.2
-0.4
0 2 4 6 8 10
t/Te
1
[-]
0
z/ 1
-1
0 2 4 6 8 10
t/Te
1
/ 1k [-]
-1
0 2 4 6 8 10
t/Te
Fig. 8 Time series for force and motions of KCS in regular head waves at λ/L = 1.95 for FORCE2
IIHR and OU
Fig. 9 Harmonic amplitudes of resistance and motions for KCS in regular head waves
85
86
Fig. 10 Harmonic phases of resistance and motions for KCS in regular head waves
Y. Sanada et al.
Table 9 SD analysis on added resistance for KCS in head waves
σAR
λ/L FORCE (6.0702 m) IIHR (2.7 m, Aug 2016) OU (3.2 m, Oct 2016) AVERAGE (D) SD SD%D
0.65 2.06E + 00 1.96E + 00 2.80E + 00 2.27E + 00 3.75E-01 16.5
0.85 4.17E + 00 5.72E + 00 5.59E + 00 5.16E + 00 7.00E-01 13.6

1.15 9.92E + 00 1.04E + 01 1.04E + 01 1.03E + 01 2.34E-01 2.3
1.37 6.22E + 00 6.37E + 00 7.21E + 00 6.60E + 00 4.34E-01 6.6
1.95 1.87E + 00 1.43E + 00 1.89E + 00 1.73E + 00 2.13E-01 12.3
Average (Total) 4.85E + 00 5.18E + 00 5.58E + 00 5.20E + 00 3.91E-01 10.24
87
88
Table 10 SD analysis on 1st and 2nd harmonic amplitude of ζ/A for KCS in head waves
ζ1 /A
0.65 9.74E-01 1.08E + 00 1.05E + 00 1.03E + 00 4.35E-02 5.0
0.85 8.95E-01 1.09E + 00 9.26E-01 9.72E-01 8.71E-02 10.2
1.15 1.12E + 00 1.06E + 00 9.74E-01 1.05E + 00 5.97E-02 2.8
1.37 1.07E + 00 1.10E + 00 9.50E-01 1.04E + 00 6.60E-02 1.3
1.95 9.85E-01 1.03E + 00 1.02E + 00 1.01E + 00 1.90E-02 2.2
Average (Total) 1.01E + 00 1.07E + 00 9.83E-01 1.02E + 00 5.51E-02 4.31
ζ2 /A
0.65 7.18E-03 2.60E-02 3.82E-02 2.38E-02 1.27E-02 53.6
0.85 1.48E-02 4.01E-02 1.72E-02 2.40E-02 1.14E-02 47.4
1.15 4.55E-03 3.58E-02 3.41E-02 2.48E-02 1.43E-02 57.8
1.37 2.89E-02 3.32E-02 3.32E-02 3.17E-02 2.01E-03 6.3
1.95 2.67E-02 6.74E-02 2.28E-02 3.89E-02 2.02E-02 51.8
Average (Total) 1.64E-02 4.05E-02 2.91E-02 2.87E-02 1.21E-02 43.40
Y. Sanada et al.
Table 11 SD analysis on 1st and 2nd harmonic amplitudes of CT for KCS in head waves
CT1
0.65 3.11E-03 1.65E-03 2.36E-04 1.67E-03 1.17E-03 70.5
0.85 4.11E-03 2.91E-03 2.76E-04 2.43E-03 1.60E-03 65.9
1.15 4.00E-03 3.49E-03 2.79E-04 2.59E-03 1.65E-03 63.6
1.37 1.28E-02 1.26E-02 5.63E-04 8.66E-03 5.72E-03 66.1
1.95 2.50E-02 2.58E-02 1.15E-03 1.73E-02 1.14E-02 66.1

CT2
0.65 4.63E-04 4.51E-04 1.36E-05 3.09E-04 2.09E-04 67.6
0.85 6.45E-04 2.34E-04 2.43E-05 3.01E-04 2.58E-04 85.6
1.15 2.79E-03 1.02E-03 7.48E-05 1.30E-03 1.12E-03 86.8
1.37 2.23E-03 1.86E-03 5.79E-05 1.38E-03 9.49E-04 68.6
1.95 3.85E-04 7.25E-04 1.59E-04 4.23E-04 2.32E-04 55.0
89
90
Table 12 SD analysis on 1st and 2nd harmonic amplitudes of z/ζ1 for KCS in head waves
Z1 /ζ1
0.65 1.16E-01 1.27E-01 1.23E-01 1.22E-01 4.31E-03 3.5
0.85 2.46E-01 2.55E-01 2.21E-01 2.40E-01 1.43E-02 6.0
1.15 9.01E-01 8.95E-01 8.82E-01 8.92E-01 8.01E-03 0.9
1.37 8.73E-01 8.53E-01 8.62E-01 8.63E-01 7.87E-03 0.9
1.95 9.29E-01 9.18E-01 8.87E-01 9.11E-01 1.80E-02 2.0
Z2 /ζ1
0.65 9.21E-04 3.98E-03 3.22E-03 2.71E-03 1.30E-03 48.0
0.85 2.68E-03 2.73E-03 4.96E-03 3.46E-03 1.06E-03 30.7
1.15 1.82E-02 1.05E-02 1.22E-02 1.36E-02 3.31E-03 24.3
1.37 1.32E-02 5.28E-03 5.01E-03 7.82E-03 3.79E-03 48.5
1.95 2.45E-02 9.86E-03 1.04E-02 1.49E-02 6.76E-03 45.3
Y. Sanada et al.
Table 13 SD analysis on 1st and 2nd harmonic amplitudes of θ/ζ1 k for KCS in head waves
θ1 /ζ1 k
λ/L FORCE (6.0702 m) IIHR (2.7 m,Aug 2016) OU (3.2 m, Oct 2016) AVERAGE (D) SD SD%D
0.65 1.34E-02 1.12E-02 9.23E-04 8.52E-03 5.45E-03 63.9
0.85 1.46E-01 2.41E-01 1.71E-01 1.86E-01 4.06E-02 21.8
1.15 7.51E-01 7.49E-01 6.73E-01 7.24E-01 3.65E-02 5.0
1.37 9.66E-01 9.63E-01 8.83E-01 9.38E-01 3.84E-02 4.1
1.95 1.12E + 00 1.07E + 00 9.53E-01 1.04E + 00 6.81E-02 6.5

θ2 /ζ1 k
0.65 7.36E-04 9.40E-04 5.98E-04 7.58E-04 1.40E-04 18.5
0.85 1.26E-03 7.46E-04 1.96E-03 1.32E-03 4.97E-04 37.6
1.15 1.73E-02 1.11E-02 1.10E-02 1.31E-02 2.93E-03 22.3
1.37 1.56E-02 1.60E-02 1.47E-02 1.54E-02 5.23E-04 3.4
1.95 4.37E-02 2.37E-02 3.56E-02 3.43E-02 8.24E-03 24.0
91
92
Table 14 SD analysis on 0th harmonic amplitudes of CT for KCS in head waves

CT0
0.65 4.05E-03 4.58E-03 4.70E-03 4.45E-03 2.81E-04 6.3
0.85 4.63E-03 5.37E-03 5.12E-03 5.04E-03 3.09E-04 6.1
1.15 7.10E-03 7.77E-03 7.68E-03 7.52E-03 2.97E-04 3.9
1.37 6.97E-03 7.66E-03 7.59E-03 7.41E-03 3.11E-04 4.2
1.95 5.50E-03 5.32E-03 5.68E-03 5.50E-03 1.47E-04 2.7
Y. Sanada et al.
Table 15 SD analysis on 0th harmonic amplitudes of z/A for KCS in head waves
Z0 /ζ1
0.65 −4.10E-01 −3.05E-01 −2.33E-01 −3.16E-01 7.26E-02 16.6
0.85 −3.20E-01 −2.06E-01 −2.89E-01 −2.72E-01 4.82E-02 21.0

1.15 −1.41E-01 −1.16E-01 −1.29E-01 −1.28E-01 1.01E-02 9.6
1.37 −1.25E-01 −8.35E-02 −9.26E-02 −1.00E-01 1.80E-02 20.9
1.95 −1.05E-01 −8.47E-02 −1.10E-01 −9.99E-02 1.10E-02 10.1
Average (Total) −2.20E-01 −1.59E-01 −1.71E-01 −1.83E-01 3.20E-02 15.64
93
94
Table 16 SD analysis on 0th harmonic amplitudes of θ/Ak for KCS in head waves
θ0 /ζ1 k
0.65 −5.50E-02 −3.21E-03 1.84E-01 4.20E-02 1.03E-01 244.6
0.85 −6.60E-02 2.44E-02 2.38E-01 6.56E-02 1.28E-01 194.5
1.15 −1.57E-03 6.70E-02 2.35E-01 1.00E-01 9.93E-02 99.2
1.37 −3.74E-03 6.22E-02 2.84E-01 1.14E-01 1.23E-01 107.8
1.95 −2.47E-02 −1.37E-03 2.30E-01 6.79E-02 1.15E-01 169.1
Average (Total) −3.02E-02 −2.98E-02 −2.34E-01 −7.80E-02 −1.14E-01 −163.06
Y. Sanada et al.
Table 17 SD analysis on 2nd harmonic phases of ζ/A for KCS in head waves
2nd Phase ζ2 /A
0.65 −2.36E + 00 −6.76E-01 3.84E-01 −8.85E-01 1.13E + 00 127.8
0.85 3.07E + 00 −5.09E-01 2.09E + 00 1.55E + 00 1.51E + 00 97.5

1.15 1.78E + 00 1.41E + 00 −2.63E-01 9.74E-01 8.88E-01 91.1
1.37 −2.52E + 00 −1.08E + 00 9.65E-02 −1.17E + 00 1.07E + 00 91.5
1.95 −2.93E + 00 2.59E-01 −9.81E-01 −1.22E + 00 1.31E + 00 107.8
Average (Total) −5.94E-01 −1.20E-01 2.64E-01 −1.50E-01 1.18E + 00 103.15
95
96
Table 18 SD analysis on 1st and 2nd harmonic phases of CT for KCS in head waves
1st Phase CT1
0.65 −9.76E-01 3.99E-01 −2.30E + 00 −9.58E-01 1.10E + 00 114.9
0.85 −8.16E-01 3.38E-01 −1.79E + 00 −7.57E-01 8.71E-01 115.0
1.15 −1.69E + 00 −1.20E + 00 −2.96E + 00 −1.95E + 00 7.43E-01 38.1
1.37 −1.34E + 00 −8.65E-01 −2.61E + 00 −1.60E + 00 7.36E-01 45.9
1.95 −3.69E-01 1.27E-02 −2.15E + 00 −8.37E-01 9.44E-01 112.8
Average (Total) −1.04E + 00 −2.63E-01 −2.36E + 00 −1.22E + 00 8.79E-01 85.36
2nd Phase CT2
0.65 −4.55E-01 −1.36E + 00 −1.31E + 00 −1.04E + 00 4.16E-01 39.9
0.85 −5.64E-02 1.48E + 00 −1.14E + 00 9.52E-02 1.07E + 00 1127.4
1.15 1.59E + 00 2.16E + 00 3.88E-01 1.38E + 00 7.37E-01 53.5
1.37 3.07E + 00 3.65E + 00 2.45E + 00 3.05E + 00 4.90E-01 16.0
1.95 3.41E-01 −1.79E-01 −1.88E + 00 −5.74E-01 9.50E-01 165.6
Average (Total) 8.97E-01 1.15E + 00 −2.98E-01 5.82E-01 7.33E-01 280.50
Y. Sanada et al.
Table 19 SD analysis on 1st and 2nd harmonic phases of z/A for KCS in head waves
1st Phase Z1 /A
0.65 1.59E + 00 3.04E + 00 1.76E + 00 2.13E + 00 6.47E-01 30.3
0.85 1.41E + 00 2.54E + 00 1.69E + 00 1.88E + 00 4.78E-01 25.4
1.15 −3.22E + 00 −2.60E + 00 −3.09E + 00 −2.97E + 00 2.68E-01 −9.0
1.37 −2.39E + 00 −1.91E + 00 −2.30E + 00 −2.20E + 00 2.11E-01 −9.6
1.95 −1.72E + 00 −1.39E + 00 −1.63E + 00 −1.58E + 00 1.43E-01 −9.0
Average (Total) −8.67E-01 −6.34E-02 −7.15E-01 −5.48E-01 3.49E-01 5.62

2nd Phase Z2 /A
0.65 −1.03E + 00 1.53E + 00 2.92E + 00 1.14E + 00 1.64E + 00 143.4
0.85 −2.55E + 00 −6.90E-01 −2.48E + 00 −1.91E + 00 8.62E-01 −45.2
1.15 −1.75E-02 1.46E + 00 1.73E-01 5.38E-01 6.56E-01 121.9
1.37 5.64E-01 2.10E + 00 6.06E-01 1.09E + 00 7.12E-01 65.5
1.95 1.52E + 00 2.18E-01 1.57E + 00 1.10E + 00 6.25E-01 56.7
Average (Total) −3.03E-01 9.22E-01 5.57E-01 3.92E-01 8.98E-01 68.45
97
98
Table 20 SD analysis on 1st and 2nd harmonic phases of θ/Ak for KCS in head waves
1st Phase θ1 /Ak
0.65 9.95E-01 1.93E + 00 7.55E-01 1.23E + 00 5.05E-01 41.2
0.85 1.80E + 00 2.93E + 00 1.88E + 00 2.21E + 00 5.15E-01 23.4
1.15 −2.62E + 00 −2.06E + 00 −2.54E + 00 −2.41E + 00 2.47E-01 10.3
1.37 −1.70E + 00 −1.28E + 00 −1.62E + 00 −1.53E + 00 1.84E-01 12.0
1.95 −5.90E-01 −2.58E-01 −5.07E-01 −4.51E-01 1.41E-01 31.2
Average (Total) −4.23E-01 2.52E-01 −4.06E-01 −1.92E-01 3.19E-01 23.61
2nd Phase θ2 /Ak
0.65 −3.68E + 00 −1.98E + 00 −4.66E-01 −2.04E + 00 1.31E + 00 64.2
0.85 −2.14E + 00 −2.90E + 00 −2.69E + 00 −2.58E + 00 3.19E-01 12.4
1.15 7.07E-02 9.24E-01 3.25E-02 3.43E-01 4.12E-01 120.2
1.37 1.03E + 00 1.48E + 00 5.48E-01 1.02E + 00 3.78E-01 37.2
1.95 2.47E + 00 2.08E + 00 2.82E + 00 2.46E + 00 3.04E-01 12.4
Average (Total) −4.49E-01 −8.02E-02 5.02E-02 −1.60E-01 5.45E-01 49.27
Y. Sanada et al.
In consideration differences in model sizes and rigid (FORCE) vs. surge free
(IIHR and OU) mounts the agreement is reasonable: for primary variables, SD = 3,
20 and 10%D for heave and pitch amplitudes and added resistance, respectively; and
for secondary variable SD = 66%D for CT1 .
4 KCS Oblique Waves λ/L = 1
Regular oblique wave experimental data are available from IIHR using one model
size 2.7 m and from both 2015 (Sadat-Hosseini et al. 2015a) and 2016 and uncertainty
values from Stocker (2016). Data is available for headings χ = 0, 45, 90, 135 and
180 degrees with 0.5 ≤ λ/L ≤ 2. The λ/L = 1 data was used for test case 2.11.
Figures 11, 12, 13, 14 and 15 compare the time histories from 2015 and 2016 for
ζ/A, CT , x/A, z/A, θ/Ak and φ/Ak for χ = 0, 45, 90, 135 and 180 degrees and λ/L
= 1. Figure 16 compares σAR , 1st and 2nd harmonic amplitude for ζ/A, 0th, 1st and
2nd harmonic amplitudes for CT , x/A, z/A, φ/Ak and θ/Ak; and Fig. 17 compares
1st and 2nd harmonic phases for CT , x/A, z/A, φ/Ak and θ/Ak. Harmonic analysis
window size for χ = 0,45 degrees and χ = 90,135,180 degrees are 5Te and 3Te,
respectively. Results show reasonable agreement between the time histories of the
two sets of data.
Tables 21 and 22 provide SD analysis for σAR and 1st and 2nd harmonic amplitude
for ζ/A. Tables 23, 24, 25, 26 and 27 provide SD analysis for 0th, 1st and 2nd harmonic
amplitudes for CT , x/A, z/A, φ/Ak and θ/Ak, respectively. Tables 28, 29, 30, 31, 32
and 33 provide SD analysis for 1st and 2nd harmonic phases for CT , x/A, z/A, φ/Ak
and θ/Ak, respectively. The scatter for the 1st harmonic amplitudes is relatively small
4, 13, 25, 6, 17, 6%D, respectively, for ζ/A, CT , x/A, z/A, φ/Ak, θ/Ak. The scatter for
0th harmonic for CT (SD = 1%D) and 1st harmonic phases for ζ/A (SD = 0.1%D)
and z/A (SD = 4%D) are also small. The scatter for σAR (SD = 63%D) and other
harmonic amplitudes and phases is mostly large.
The agreement between the two series of tests is reasonably good, i.e., SD values
are comparable to the corresponding values for test case 2.10: for primary variables,
SD = 6, 6 and 63%D for heave and pitch amplitudes and added resistance, respec-
tively; and for secondary variable SD = 13%D for CT1 . T2015 test case 2.11 used
the 2016 data.
5 KCS Oblique Waves All λ/L
Table 34 lists all IIHR KCS oblique wave data for all headings and wavelengths.
Figures 18, 19, 20, 21 and 22 provide time histories and Figs. 23–24 harmonic
amplitudes and phases. The harmonic analysis window size for oblique heading case
are 2Te only for λ/L = 2.0 and μ = 135 degree and 3Te to 5Te for other λ/L and χ.
The figures show the trends are very different for different wave headings.
100 Y. Sanada et al.
Fig. 11 Time series for force and motions of KCS in regular oblique waves at λ/L = 1.00, χ = 0°
for IIHR Aug-15 and Aug-16
Fig. 12 Time series for force and motions of KCS in regular oblique waves at λ/L = 1.00, χ =
45° for IIHR Aug-15 and Aug-16
Fig. 16 Harmonic amplitudes of resistance and motions for KCS in regular oblique waves at Fr = 0.26
105
106
Fig. 17 Harmonic phases of resistance and motions for KCS in regular oblique waves at Fr = 0.26
Y. Sanada et al.
Table 21 SD analysis on added resistance for KCS in oblique waves (λ/L = 1)
σAR
χ [deg] λ/L IIHR (2.7 m, Aug 2015) IIHR (2.7 m, Aug 2016) AVERAGE (D) SD SD%D
0 1.00 9.73 8.28 9.01 7.24E-01 8.0
45 1.00 7.58 6.12 6.85 7.30E-01 10.7

90 1.00 4.07 2.06 3.07 1.00E + 00 32.7
135 1.00 0.46 −0.17 0.14 3.16E-01 217.6
180 1.00 1.34 0.50 0.92 4.18E-01 45.3
0.00 4.64 3.36 4.00 0.64 62.86
107
108
Table 22 SD analysis on 1st and 2nd harmonic amplitudes of ζ/A for KCS in oblique waves (λ/L = 1)
ζ1 /A
0 1.0 9.86E-01 1.04E + 00 1.01E + 00 2.91E-02 2.9
45 1.0 9.56E-01 1.06E + 00 1.01E + 00 5.12E-02 5.1
90 1.0 9.93E-01 1.04E + 00 1.02E + 00 2.52E-02 2.5
135 1.0 1.03E + 00 1.11E + 00 1.07E + 00 3.82E-02 3.6
180 1.0 9.18E-01 1.03E + 00 9.72E-01 5.43E-02 5.6
Average (Total) 9.77E-01 1.06E + 00 1.02E + 00 3.96E-02 3.9
ζ2 /A
0 1.0 9.18E-03 7.56E-03 8.37E-03 8.07E-04 9.6
45 1.0 2.45E-02 2.05E-02 2.25E-02 1.99E-03 8.8
90 1.0 5.22E-02 4.30E-02 4.76E-02 4.63E-03 9.7
135 1.0 3.10E-02 9.50E-02 6.30E-02 3.20E-02 50.8
180 1.0 1.94E-02 8.26E-02 5.10E-02 3.16E-02 62.0
Average (Total) 2.73E-02 4.97E-02 3.85E-02 1.42E-02 28.2
Y. Sanada et al.
Table 23 SD analysis on 0th, 1st and 2nd harmonic amplitudes of CT for KCS in oblique waves (λ/L = 1)
CT0
0 1.0 6.29E-03 6.43E-03 6.36E-03 6.76E-05 1.1
45 1.0 5.91E-03 6.01E-03 5.96E-03 5.13E-05 0.9
90 1.0 4.81E-03 4.91E-03 4.86E-03 5.17E-05 1.1
135 1.0 4.22E-03 4.38E-03 4.30E-03 8.43E-05 2.0
180 1.0 4.56E-03 4.52E-03 4.54E-03 2.04E-05 0.5
CT1
0 1.0 3.01E-03 2.45E-03 2.73E-03 2.79E-04 10.2
45 1.0 5.87E-03 6.50E-03 6.18E-03 3.18E-04 5.1

90 1.0 2.06E-03 2.73E-03 2.40E-03 3.35E-04 14.0
135 1.0 3.71E-03 7.09E-03 5.40E-03 1.69E-03 31.3
180 1.0 1.09E-03 1.18E-03 1.13E-03 4.66E-05 4.1
CT2
0 1.0 4.25E-04 5.78E-04 5.02E-04 7.68E-05 15.3
45 1.0 6.23E-03 3.78E-03 5.01E-03 1.22E-03 24.5
90 1.0 4.69E-03 2.02E-03 3.36E-03 1.33E-03 39.7
135 1.0 2.12E-03 1.38E-03 1.75E-03 3.66E-04 20.9
180 1.0 3.05E-04 5.57E-04 4.31E-04 1.26E-04 29.2
109
110
Table 24 SD analysis on 0th, 1st and 2nd harmonic amplitudes of x/A for KCS in oblique waves (λ/L = 1)
x0 /A
0 1.0 −1.39E-02 −1.13E-02 −1.26E-02 1.34E-03 −10.6
45 1.0 −1.23E-02 −3.45E-02 −2.34E-02 1.11E-02 −47.3
90 1.0 −4.15E-04 1.87E-02 9.16E-03 9.58E-03 104.5
135 1.0 1.45E-02 3.85E-02 2.65E-02 1.20E-02 45.4
180 1.0 1.75E-03 −2.54E-03 −3.94E-04 2.15E-03 −544.5
Average (Total) −2.09E-03 1.80E-03 −1.43E-04 7.23E-03 −90.5
x1 /A
0 1.0 3.90E-02 2.95E-02 3.43E-02 4.78E-03 13.9
45 1.0 8.85E-02 9.89E-02 9.37E-02 5.19E-03 5.5
90 1.0 3.05E-02 5.33E-02 4.19E-02 1.14E-02 27.2
135 1.0 3.30E-01 7.12E-01 5.21E-01 1.91E-01 36.7
180 1.0 1.97E-01 4.71E-01 3.34E-01 1.37E-01 41.1
x2 /A
0 1.0 1.22E-03 1.89E-03 1.56E-03 3.36E-04 21.6
45 1.0 3.79E-02 2.12E-02 2.95E-02 8.38E-03 28.4
90 1.0 4.37E-02 1.63E-02 3.00E-02 1.37E-02 45.6
135 1.0 6.64E-02 6.14E-02 6.39E-02 2.48E-03 3.9
180 1.0 1.91E-02 4.20E-02 3.05E-02 1.14E-02 37.4
Y. Sanada et al.
Table 25 SD analysis on 0th, 1st and 2nd harmonic amplitudes of z/A for KCS in oblique waves (λ/L = 1)
z0 /A
0 1.0 −1.70E-01 −1.61E-01 −1.66E-01 4.76E-03 −2.9
45 1.0 −1.58E-01 −1.30E-01 −1.44E-01 1.41E-02 −9.8
90 1.0 5.41E-02 −1.49E-01 −4.73E-02 1.01E-01 −214.3
135 1.0 5.08E-02 −1.34E-01 −4.17E-02 9.26E-02 −221.9
180 1.0 −2.53E-01 −2.37E-01 −2.45E-01 7.85E-03 −3.2
Average (Total) −9.52E-02 −1.62E-01 −1.29E-01 4.42E-02 −90.4
z1 /A
0 1.0 5.79E-01 6.08E-01 5.94E-01 1.45E-02 2.4
45 1.0 1.06E + 00 1.16E + 00 1.11E + 00 4.84E-02 4.4

90 1.0 1.04E + 00 1.13E + 00 1.09E + 00 4.52E-02 4.2
135 1.0 3.70E-01 4.04E-01 3.87E-01 1.68E-02 4.3
180 1.0 4.54E-02 5.95E-02 5.24E-02 7.04E-03 13.4
z2 /A
0 1.0 1.72E-02 1.98E-02 1.85E-02 1.31E-03 7.1
45 1.0 9.45E-03 7.85E-03 8.65E-03 8.01E-04 9.3
90 1.0 2.14E-02 3.42E-02 2.78E-02 6.40E-03 23.0
135 1.0 1.81E-02 1.03E-03 9.59E-03 8.56E-03 89.3
180 1.0 2.52E-02 2.36E-02 2.44E-02 7.94E-04 3.3
111
112
Table 26 SD analysis on 0th, 1st and 2nd harmonic amplitudes of φ/Ak for KCS in oblique waves (λ/L = 1)
φ0 /Ak
0 1.0 7.84E-02 3.24E-01 2.01E-01 1.23E-01 61.0
45 1.0 2.83E-01 −1.32E-01 7.54E-02 2.07E-01 275.0
90 1.0 −2.79E-01 5.88E-01 1.55E-01 4.34E-01 280.0
135 1.0 −3.71E-01 5.13E-01 7.11E-02 4.42E-01 621.5
180 1.0 1.02E-01 −2.57E-01 −7.72E-02 1.80E-01 −232.7
Average (Total) −3.72E-02 2.07E-01 8.50E-02 2.77E-01 200.9
φ1 /Ak
0 1.0 1.24E-02 1.19E-02 1.22E-02 2.49E-04 2.0
45 1.0 5.99E-01 5.58E-01 5.78E-01 2.07E-02 3.6
90 1.0 7.59E-01 3.78E-01 5.69E-01 1.90E-01 33.5
135 1.0 2.39E + 00 2.79E + 00 2.59E + 00 2.00E-01 7.7
180 1.0 1.06E-01 2.47E-01 1.77E-01 7.03E-02 39.8
φ2 /Ak
0 1.0 2.74E-03 5.82E-03 4.28E-03 1.54E-03 36.0
45 1.0 1.89E-01 1.48E-01 1.68E-01 2.08E-02 12.3
90 1.0 1.47E-01 6.71E-02 1.07E-01 4.00E-02 37.4
135 1.0 8.21E-02 9.36E-02 8.79E-02 5.76E-03 6.6
180 1.0 3.49E-02 4.61E-02 4.05E-02 5.60E-03 13.8
Y. Sanada et al.
Table 27 SD analysis on 0th, 1st and 2nd harmonic amplitudes of θ/Ak for KCS in oblique waves (λ/L = 1)
θ0 /Ak
0 1.0 9.12E-02 4.42E-02 6.77E-02 2.35E-02 34.7
45 1.0 1.81E-02 −1.35E-02 2.30E-03 1.58E-02 687.3
90 1.0 2.05E-02 −2.33E-02 −1.38E-03 2.19E-02 −1587.6
135 1.0 2.42E-02 −1.04E-02 6.88E-03 1.73E-02 251.1
180 1.0 6.84E-02 1.55E-02 4.19E-02 2.64E-02 63.1
Average (Total) 4.45E-02 2.50E-03 2.35E-02 2.10E-02 −110.3
θ1 /Ak
0 1.0 4.64E-01 4.92E-01 4.78E-01 1.40E-02 2.9
45 1.0 6.05E-01 6.85E-01 6.45E-01 3.98E-02 6.2

90 1.0 3.53E-02 4.76E-02 4.14E-02 6.17E-03 14.9
135 1.0 3.66E-01 3.65E-01 3.65E-01 4.57E-04 0.1
180 1.0 1.28E-01 1.38E-01 1.33E-01 4.56E-03 3.4
θ2 /Ak
0 1.0 1.05E-02 1.30E-02 1.18E-02 1.27E-03 10.8
45 1.0 2.64E-02 9.74E-03 1.81E-02 8.32E-03 46.1
90 1.0 1.19E-02 1.96E-02 1.57E-02 3.88E-03 24.7
135 1.0 3.95E-02 2.81E-02 3.38E-02 5.68E-03 16.8
180 1.0 1.79E-02 1.91E-02 1.85E-02 5.96E-04 3.2
113
114
Table 28 SD analysis on 1st and 2nd harmonic phases for CT of KCS in oblique waves (λ/L = 1)
1st Phase CT1
0 1.0 −7.44E-02 1.65E-01 4.55E-02 1.20E-01 263.4
45 1.0 1.02E-02 −5.09E-01 −2.50E-01 2.60E-01 −104.1
90 1.0 −1.08E + 00 −1.41E + 00 −1.24E + 00 1.63E-01 −13.1
135 1.0 1.69E + 00 1.49E + 00 1.59E + 00 9.83E-02 6.2
180 1.0 −2.98E + 00 −3.03E + 00 −3.01E + 00 2.53E-02 −0.8
Average (Total) −4.88E-01 −6.58E-01 −5.73E-01 1.33E-01 30.3
2nd Phase CT2
0 1.0 2.30E + 00 2.17E + 00 2.23E + 00 6.67E-02 3.0
45 1.0 2.86E + 00 2.57E + 00 2.72E + 00 1.44E-01 5.3
90 1.0 −2.31E + 00 −3.56E + 00 −2.94E + 00 6.22E-01 −21.2
135 1.0 −1.27E + 00 −1.81E + 00 −1.54E + 00 2.69E-01 −17.5
180 1.0 −1.17E + 00 −2.77E + 00 −1.97E + 00 8.02E-01 −40.7
Average (Total) 8.32E-02 8.32E-02 −6.79E-01 −2.98E-01 3.81E-01
Y. Sanada et al.
Table 29 SD analysis on 1st and 2nd harmonic phases for x/A of KCS in oblique waves (λ/L = 1)
1st Phase x1 /A
0 1.0 2.98E + 00 3.21E + 00 3.10E + 00 1.15E-01 3.7
45 1.0 3.11E + 00 2.57E + 00 2.84E + 00 2.68E-01 9.4
90 1.0 1.92E + 00 3.01E + 00 2.46E + 00 5.42E-01 22.0
135 1.0 −1.54E + 00 −1.55E + 00 −1.55E + 00 2.45E-04 0.0
180 1.0 −4.99E-01 −1.25E-01 −3.12E-01 1.87E-01 −60.1
Average (Total) 1.19E + 00 1.19E + 00 1.42E + 00 1.31E + 00 2.23E-01

2nd Phase x2 /A
0 1.0 −2.52E-02 −3.97E-01 −2.11E-01 1.86E-01 −88.1
45 1.0 −2.03E-01 −3.59E-01 −2.81E-01 7.81E-02 −27.8
90 1.0 2.73E-01 2.18E-01 2.45E-01 2.77E-02 11.3
135 1.0 1.81E + 00 1.58E + 00 1.70E + 00 1.15E-01 6.8
180 1.0 −4.50E-01 1.73E-01 −1.38E-01 3.11E-01 −225.2
Average (Total) 2.81E-01 2.43E-01 2.62E-01 1.44E-01 −64.6
115
116
Table 30 SD analysis on 1st and 2nd harmonic phases for z/A of KCS in oblique waves (λ/L = 1)
1st Phase z1 /A
0 1.0 2.69E + 00 2.95E + 00 2.82E + 00 1.32E-01 4.7
45 1.0 −1.98E + 00 −2.21E + 00 −2.09E + 00 1.16E-01 −5.5
90 1.0 1.29E-01 1.62E-01 1.46E-01 1.66E-02 11.4
135 1.0 2.98E + 00 3.07E + 00 3.03E + 00 4.35E-02 1.4
180 1.0 2.64E + 00 3.01E + 00 2.83E + 00 1.82E-01 6.5
Average (Total) 1.29E + 00 1.40E + 00 1.34E + 00 9.81E-02 3.69
2nd Phase z2 /A
0 1.0 −2.58E-01 7.45E-02 −9.16E-02 1.66E-01 −181.3
45 1.0 −5.99E-01 −4.33E-01 −5.16E-01 8.30E-02 −16.1
90 1.0 −2.43E + 00 −1.60E + 00 −2.01E + 00 4.14E-01 −20.6
135 1.0 1.52E + 00 −1.56E + 00 −2.14E-02 1.54E + 00 −7222.9
180 1.0 −1.78E + 00 −1.23E + 00 −1.51E + 00 2.73E-01 −18.1
Y. Sanada et al.
Table 31 SD analysis on 1st and 2nd harmonic phases for φ/Ak of KCS in oblique waves (λ/L = 1)
1st Phase φ1 /Ak
0 1.0 −1.39E + 00 4.15E-01 −4.87E-01 9.02E-01 −185.1
45 1.0 9.56E-01 4.39E-01 6.98E-01 2.59E-01 37.1
90 1.0 −1.75E + 00 −9.94E-01 −1.37E + 00 3.76E-01 −27.4
135 1.0 −1.75E + 00 −1.76E + 00 −1.75E + 00 9.97E-04 −0.1
180 1.0 −9.29E-01 −2.94E + 00 −1.93E + 00 1.00E + 00 −52.0

2nd Phase φ2 /Ak
0 1.0 7.49E-02 −2.88E + 00 −1.40E + 00 1.48E + 00 −105.3
45 1.0 −5.58E-02 −1.20E-01 −8.80E-02 3.22E-02 −36.6
90 1.0 −1.97E + 00 −1.98E + 00 −1.97E + 00 3.91E-03 −0.2
135 1.0 3.31E-01 1.17E + 00 7.51E-01 4.20E-01 55.9
180 1.0 −1.18E + 00 −1.08E + 00 −1.13E + 00 5.10E-02 −4.5
117
118
Table 32 SD analysis on 1st and 2nd harmonic phases for θ/Ak of KCS in oblique waves (λ/L = 1)
1st Phase θ1/ Ak
0 1.0 −3.00E + 00 −2.74E + 00 −2.87E + 00 1.31E-01 −4.6
45 1.0 −1.18E + 00 −1.44E + 00 −1.31E + 00 1.28E-01 −9.8
90 1.0 −9.41E-01 −7.17E-01 −8.29E-01 1.12E-01 −13.5
135 1.0 1.69E + 00 1.74E + 00 1.71E + 00 2.33E-02 1.4
180 1.0 2.87E + 00 3.18E + 00 3.03E + 00 1.55E-01 5.1
Average (Total) −1.13E-01 −1.13E-01 4.26E-03 −5.44E-02 1.10E-01
2nd Phase θ2 /Ak
0 1.0 −2.94E-01 −1.20E-02 −1.53E-01 1.41E-01 −92.1
45 1.0 2.52E + 00 1.56E + 00 2.04E + 00 4.80E-01 23.5
90 1.0 5.53E-02 1.38E + 00 7.18E-01 6.62E-01 92.3
135 1.0 −1.34E + 00 −1.39E + 00 −1.36E + 00 2.45E-02 −1.8
180 1.0 −3.45E + 00 −2.76E + 00 v3.11E + 00 3.44E-01 −11.1
Average (Total) 7.56E-01 −5.01E-01 −2.44E-01 −3.72E-01 3.30E-01
Y. Sanada et al.
Table 33 SD analysis on 1st and 2nd harmonic phases for ζ/A of KCS in oblique waves (λ/L = 1)
1st Phase ζ1 /A
0 1.0 −6.19E-02 −3.89E-02 −5.04E-02 1.15E-02 −22.8
45 1.0 −9.49E-02 −2.14E-03 −4.85E-02 4.64E-02 −95.6
90 1.0 −7.74E-02 −1.26E-01 −1.02E-01 2.43E-02 −23.9
135 1.0 1.19E-01 3.00E-02 7.45E-02 4.45E-02 59.8
180 1.0 1.38E-01 1.28E-02 7.52E-02 6.24E-02 83.0
Average (Total) 4.50E-03 −2.48E-02 −1.02E-02 3.78E-02 0.1

2nd Phase ζ2 /A
0 1.0 −2.42E + 00 7.47E-02 −1.17E + 00 1.25E + 00 −106.4
45 1.0 −5.40E-01 −6.97E-01 −6.19E-01 7.85E-02 −12.7
90 1.0 −2.85E + 00 2.22E + 00 −3.19E-01 2.53E + 00 −793.8
135 1.0 −2.19E + 00 3.57E-01 −9.16E-01 1.27E + 00 −138.9
180 1.0 1.69E + 00 1.85E + 00 1.77E + 00 7.98E-02 4.5
Average (Total) −1.26E + 00 7.61E-01 −2.51E-01 1.04E + 00 −209.5
119
Table 34 Data lists of oblique wave cases

χ [°] λ/L 15-Aug 16-Aug
45 0.50 0
1.00 0 0
2.00 0
90 0.25 0
0.30 0
0.40 0
0.50 0
1.00 0 0
2.00 0
135 0.25 0
0.30 0
0.40 0
0.50 0
1.00 0 0
2.00 0
180 0.50 0
1.00 0 0
2.00 0
T2015
Figure 25 compares the present and Fujii-Takahashi (1975) added resistance σAR
vs. both wavelength and heading for reference. Fujii-Takahashi (1975) is the only
available added resistance data for oblique headings but note that the data was taken
at Fr = 0.25 using the S-175 container ship model which has different CB and L/B
from KCS. Table 35 lists the principal-particulars comparisons between the KCS
and the S-175 model used in Fujii-Takahashi (1975). The IIHR trends vs. λ/L for
χ = 0 and 90° are consistent with expectation for the effects of fe /fn (for heave
and pitch fn ), i.e., resonance for shorter waves for χ = 90° than χ = 0°. Since χ
= 180° is far from resonance it is not known why σAR is larger for short waves.
The Fujii-Takahashi (1975) results for head waves are in reasonably close agreement
with IIHR, but only roughly similar for the other headings. Figures 26 and 27 are
like Figs. 23–24 but show the harmonic amplitudes and phases vs. heading. The data
not was used at T2015 but was used for CFD assessment by Sadat-Hosseini et al.
(2015a); and provided herein for use by others.
Fig. 18 Time series for force and motions of KCS in regular oblique waves at χ = 0° (IIHR
Aug-15)
Aug-15)
Aug-15)
Aug-15)
Aug-15)
Fig. 23 Harmonic amplitudes of resistance and motions for KCS in regular oblique waves at Fr =
0.26
Fig. 23 (continued)
Fig. 25 Added resistance comparison between IIHR and Fujii-Takahashi (1975)
Table 35 Principal-particulars comparison between KCS and S-175

KCS (IIHR) S-175 (Fujii-Takahashi 1975)
L [m] 2.700 3.500
B [m] 0.378 0.508
T [m] 0.127 0.190
CB 0.643 0.572
B/T 2.981 2.674
L/B 7.143 6.890
6 ONRT Calm Water
IIHR has extensive experimental data and uncertainty analysis for ONRT free running
calm water self-propulsion and maneuvering; and course keeping/speed loss and
maneuvering in regular and irregular waves (Sanada et al. 2013, 2014 and 2018b;
Elsheikh 2014; Sadat-Hosseini et al. 2015b; Bottiglieri 2016), as summarized in
Table 36. Table 37 lists the particulars of the ship model. The ONRT mass properties
adjustment was done similar as for the KCS model as described in Sect. 2 except
for kxx . Instead of kxx adjustment, free roll decay tests were performed to adjust the
natural roll period to match with the reference value provided by NRIFE (Sanada et al.
2013). The uncertainty analysis is preliminary, and improvements are in progress for
Fig. 26 Harmonic amplitudes of resistance and motions for KCS in regular oblique waves at Fr =
0.26
Fig. 26 (continued)
differential filter for surge and sway velocities based on the model position and its
uncertainties.
In addition, the time histories display a time shift between the yaw and rudder
angle in course keeping/speed loss (the delay is 0.15 s at least) due to the processing
time of the free running system. Since the free running system can receive updated
yaw angle data every 100 ms from the tracking system and the free-running system
requires 50 ms at least to update the rudder after receiving the updated yaw angle. To
make an accurate comparison between the time series of the experiments and CFD,
the latency of the yaw acquisition should be included as the CFD condition in the
future. Both the tracking and free running system data rates are 20 Hz, but they have
different internal clocks and it is technically difficult to completely synchronize the
current two systems. Time uncertainty due to jittering of the internal clocks should
be evaluated in the future. Methods are under investigation to remove this issue in the
future; however, it should not have much of an effect on the T2015 V&V and SoA
assessment of the CFD. Additional discussion on this issue is provided in Chap. 9,
Sect. 4.3.3. Also discussed therein is that a coordinate transformation sign error was
discovered post T2015 such that the (u, v) required corrections, which were included
in the Chap. 9 CFD assessments and the UXi values used therein.
Free running calm water self-propulsion and course keeping/speed loss in regular
waves with H/λ = 0.02 and λ/L = 1.0 are used in T2015. Bottiglieri (2016) provides
uncertainty analysis for sinkage, trim, and propeller RPS at Fr = 0.2. UXi for propeller
RPS is very small (0.08%D) whereas sinkage and trim showed large uncertainties.
The CFD validation studies in Chap. 9 used the UXi values provided.
Table 36 ONRT course keeping and maneuvering tests conditions
132
T2015 Case
Test Fr* δ ψ ** [deg] H/λ λ/L χ [deg] Numbers of runs
[deg]
Calm water Course keeping 0.2 N/A 0,180 0,180 3,3
± 45, ± 135 ± 45, ± 135 3/3,3/3
± 90 ± 90 3/3
Zigzag 0.2 10 10 0,180 3,3
20 20 21,3
35 35,90 3,3
Turning circle 0.1 ± 35 N/A 0 5/5
0.2 0, ± 90,180 6/21,3/3,3/3,3/3
−5,−10, −15, −20, −25, −30 0 2,2,2,2,3,3
0.3 ± 35 0 5/5
Head Wave Course keeping 0.2 N/A 0 0.02 0.5 0 3
1 3
1.2 3
Course keeping 0.2 N/A 0 0.04 0.5 0 3
Zigzag 0.2 10 10 0.02 0.5,1.0,1.2 0 3,3,3
20 20 0.5,1.0,1.2 3,3,3
35 35 0.5,1.0,1.2 3,3,3
Turning circle 0.1 ± 35 N/A 0.02 0.5,1.0,1.2 0 5/5,5/5,5/5
0.2 0.5,1.0,1.2 0, ± 90,180 2 × 3 × 3 × 4 = 72
(continued)
Y. Sanada et al.
Table 36 (continued)
T2015 Case
[deg]
−35 0.04 0.5 0 3
0.3 ± 35 0.02 0.5,1.0 0 5/5,5/5
Long approach 0.2 −35 0.02 0.5,1.0,1.2 0 3,3,3
Turning circle
Following waves Course keeping 0.2 N/A 180 0.02 0.5 180 3
1 3
1.2 3
Zigzag 0.2 10 10 0.02 0.5,1.0,1.2 180 3,3,3
20 20 0.5,1.0,1.2 3,3,3
35 35 0.02,0.03 0.5,1.0,1.2 3,3,3/3,3,3

Beam wave Course keeping 0.2 N/A ± 90 0.02 0.5 ± 90 3
1 3
1.2 3
Quartering wave Course keeping 0.2 N/A ± 45, ± 135 0.02 0.5 ± 45, ± 135 3/3,3/3
1 3/3,3/3
1.2 3/3,3/3
(continued)
133
134
T2015 Case
[deg]
Irregular wave Course keeping 0.2 N/A 0 Sea State 4 (Tp = 1.26 0 7
[s], Hs = 0.038 [m])
(Duration:
60[s],Uniform division
by frequency) (3
repeated runs with
same phase, 4 runs with
diferent phases)
0.2 N/A 0 Sea State 4 (Tp = 1.26 0 12
[s], Hs = 0.038 [m])
(Duration 30[s],
Uniform division by
frequency)
Long approach 0.2 −35 0 Sea State 4 (Tp = 1.26 0 7
Turning Circle [s], Hs = 0.038 [m])
(Duration 120
[s],Uniform division by
frequency) (3 repeated
runs with same phase, 4
runs with diferent
phases)
(continued)
Y. Sanada et al.
T2015 Case
[deg]
0.2 −35 0 Sea State 4 (Tp = 1.26 0 7
[s], Hs = 0.038 [m])
(Duration:
120[s],Uniform
division by wave
energy) (3 repeated
runs with same phase, 4
runs with diferent
phases)
Fr* : the nominal Froude number
ψ* : the target yaw angle

135
Table 37 ONRT Particulars

Model Scale
Length (L) 3.147 m
Breadth 0.384 m
Depth 0.266 m
Draft 0.112 m
Displacement 72.6 kg
Metacentric height 0.0422 m
Natural roll period 1.665 s
Rudder area 0.012 m2 × 2
Block coefficient 0.535
Radius of gyration in pitch 0.246 × L
Maximum rudder Angle ±35°
7 ONRT Head and Oblique Wave
Bottiglieri (2016) provides uncertainty analysis for 0th harmonic amplitudes of heave
z/A, roll φ/Ak, pitch θ/Ak, yaw ψ/Ak, axial velocity u/U, and side velocity v/U,
respectively; and uncertainty analysis for 1st and 2nd harmonic phases for ζ/A, z/A,
φ/Ak, θ/Ak, ψ/Ak, u/U, and v/U, respectively. The harmonic analysis window sizes
for μ = 0, 45, 90, 135, 180 degrees are 10Te, 4Te, 5Te, 3Te, 5Te, respectively.
For 0th harmonic amplitudes, the UXi values are small for u/U (1 ~ 3%D) whereas
other values show large uncertainties. The UXi values for 1st harmonic amplitudes
are reasonably small for most variables except for φ/Ak and ψ/Ak for μ = 0 and 180
deg, and u/U for μ = 0 and v/U for μ = 90. 1st harmonic phases mostly have small
uncertainties. The uncertainty values are often larger for 2nd harmonic amplitudes
and phases. For CFD validation studies in Chap. 9 used the UD values provided.
8 Overall Conclusions
The experimental data used for T2015 test cases for KCS added resistance in regular
head (2.10) and oblique (2.11) waves and ONRT free running course keeping in
regular head (3.12) and oblique (3.13) waves evaluation, including in the former
case calm water resistance, sinkage and trim and in the latter case calm water self-
propulsion (3.9), is shown to be reasonably reliable. However, clearly it is desirable
to have data from more facilities including uncertainty analysis for assessment of
facility biases, which will provide more robust data and UDi for CFD validation/SoA
assessment.
Acknowledgements The Office of Naval Research supports the IIHR research under grants admin-
istered by Drs. Thomas Fu, Woei-Min Lin, Ki-Han Kim, and Robert Brizzolara. Mark Stocker and
Michael Bottiglieri master’s theses contributed to the IIHR experiments.
References
ASME. (2013). Test Uncertainty: Instruments and Apparatus, PTC 19.1.

Bottiglieri, M. (2016). Uncertainty Assessment For Free-Running Model Cases At The IIHR Wave
Basin, M.S. Thesis, University of Iowa, August.
Elsheikh, H. (2014). Maneuvering Characteristics in Calm Water and Regular Waves for ONR
Tumblehome, M.S. Thesis, University of Iowa, May.
Fujii, H., Takahashi, T. (1975). Experimental Study on the Resistance Increase of a ship in Regular
Oblique Waves, 14th ITTC, Ottawa.
ITTC, 7.5-02-01-01. (2008). Guide to the Expression of Uncertainty in Experimental Hydrody-
namics, Specialist Committee on Uncertainty Analysis of 25th ITTC.
Otzen, J. F. (2015a). Uncertainty Assessment for KCS Added Resistance in Waves, FORCE
Technology Report 112-31859-1C/ JFO/ 22-May.
Otzen, J. F. (2015b). Uncertainty Assessment for KCS Added Resistance in Waves, FORCE
Technology Report 112-31859-2A/ JFO/ 11-November.
Sadat-Hosseini, H. et al. (2015a). Experiments and Computations for KCS Added Resistance for
Variable Heading, World Maritime Technology Conference, Providence, Rhode Island, USA.
Sadat-Hosseini, H., Sanada, Y., Stern, F. (2015b). Experiments and CFD for ONRT Course Keeping
in Regular Waves, 5th World Maritime Technology Conference November 3−7, Rhode Island
Convention & Omni Hotel Providence, Rhode Island, USA.
Sanada,Y., Tanimoto, K., Takagi, K., Toda, Y., & Stern, F. (2013). Trajectories and Local Flow
Field Measurements around ONR Tumblehome in Maneuvering Motion. Ocean Engineering,
72, 45–65.
Sanada, Y., Elshiekh, H., Toda, Y., & Stern, F. (2014). Effects of Waves on Course Keeping
and Maneuvering for Surface Combatant ONR Tumblehome, 30th Symposium on Naval
Hydrodynamics Hobart, Tasmania, Australia, 2−7 November.
Sanada, Y., Elshiekh, H., Toda, Y., & Stern, F. (2018b). Effects of Waves on Course Keeping and
Manoeuvring for Surface Combatant ONR Tumblehome, online JMST October.
Simonsen, C. D., Otzen, J. F., Joncquez, S., & Stern, F. (2013). EFD and CFD for KCS Heaving
and Pitching in Regular Head Waves. Journal of Marine Science and Technology, 18, 435–459.
Simonsen, C. D., Otzen, J. F., Nielson, C., & Stern F. (2014). CFD Prediction of Added Resistance
of the KCS in Regular Head and Oblique Waves, 30th SNH, Hobart, Australia, 2−7 November.
Stocker, M. (2016). Surge Free Added Resistance Tests In Oblique Wave Headings For The KRISO
Container Ship Model, M.S. Thesis, University of Iowa, August.
Toda, Y. (2018). Private communication.
Yugo, S., Htike, A. K., Ayaka, M., Yasuyuki, T., Janne, F. O. (2018a). Claus Simonsen, Fred-
erick Stern, Added Resistance/Powering and Motion Measurements of KRISO Container Ship
in Regular Head Waves, Proceedings of 8th PAAMES and AMEC 2018, 9−12 October, Busan.
Zou, L., & Larsson, L. (2014). Additional Data for resistance, sinkage and trim. In Larsson et al.
(Ed.), Numerical Ship Hydrodynamics—An Assessment of the Gothenburg 2010 Workshop,
Springer Business Media, Dordrecht.
Evaluation of Resistance, Sinkage, Trim
and Wave Pattern Predictions for JBC
Lars Larsson
Abstract JBC predictions are presented for four cases: towed and self-propulsion for
the bare hull and a hull with an Energy Saving Device (ESD). A statistical evaluation
is made based on the 88 resistance predictions submitted. The comparison error is
defined as the difference between the measured data and the numerically predicted
value. This error is analyzed in different ways. It is seen that the mean signed error is
as small as the measurement accuracy, while the mean absolute error is about twice
as large. This represents the typical error in the prediction. A similar accuracy was
found in the 2010 Workshop. The number of grid cells has increased since 2010
and the required grid size for a given uncertainty of 4% has increased from 3 M
cells to 10 M cells. Reasons for this are discussed. As in the previous workshop
the two-equation turbulence models produce more accurate results than the more
advanced models, although the more and more popular Explicit Algebraic Stress
Model (EASM) is close. A surprising result of the analysis is that methods with
wall-functions give significantly smaller resistance errors than those with a wall-
resolved flow. Grid convergence is discussed and it is shown that the vast majority
of results converge with grid refinement, but that the achieved order of accuracy is
often far from the theoretical one. Sinkage is much better predicted than in 2010 and
the trim results are also improved. Finally, the wave pattern prediction is discussed.
The best methods in 2010 predicted the waves extremely well, in fact better than in
the present workshop. A consistent shift in the predicted wave phase relative to the
data could indicate a measurement error.
L. Larsson (B)
Chalmers University of Technology, Gothenburg, Sweden
140 L. Larsson
1 Resistance
1.1 Statistics
The resistance cases of the Workshop were 1.1 for the bare hull and 1.2 for the
hull equipped with an ESD. Resistance results are however also reported for the
self-propulsion cases 1.5, without, and 1.6 with an ESD. Altogether, there were 88
fine grid resistance solutions reported. This is a large data set, suitable for statistical
evaluation. The particular questions to be answered are:
• How accurately can resistance be predicted for a bluff hull, like the JBC? And
how close are we to experimental data? Is there any improvement compared to
the previous Workshops?
In Table 1 the four cases with reported resistance data are listed together with the
corresponding results for the bluff hulls used in the two previous Workshops: the
KVLCC2 in Gothenburg (Larsson et al. 2014) and the KVLCC2m in Tokyo (Hino
2005). The former is a VLCC designed and tested at KRISO in Korea in the late
1990s (Van et al. 1998) and the latter is a slightly modified version tested at NMRI
in Japan a few years later (Hino 2005).
The table shows the comparison error E defined as E = D − S, where D and S are
the measured and computed values, respectively. In the third column, the mean error
E mean , i.e. the mean value of all reported errors with sign, are presented in percent
of D. Column four gives |E|mean , i.e. the mean of the absolute value of the error. This
is followed by the standard deviation of the error σ, the mean numerical uncertainty
U SN and the reported uncertainty of the measured data U D . Finally, the number of
entries is shown.
Perhaps the most striking results of Table 1 are the small mean errors. These are
around 1%, which is the same as the measurement accuracy in most cases. There is
thus practically no systematic error in the CFD predictions, although there is a scatter
between the different methods. This scatter is represented by the mean absolute error
Table 1 Resistance statistics. Blue: towing, yellow: self-propulsion

Hull Year/Case Emean /E/mean σ USNmean UD No. of
%D %D %D %D %D Entries
KVLCC2m 2005/1.4 0.0 6.2 0.7 13
KVLCC2 2010/1.2a -1.7 1.3 1.0 5
JBC 2015/1.1a 1.0 2.1 2.4 3.8 1.0 22
2015/1.2a 1.0 1.5 2.0 5.1 1.0 20
2015/1.5a -0.1 3.2 4.0 4.6 1.0 23
2015/1.6a 0.7 3.1 4.0 6.0 1.0 23

Evaluation of Resistance, Sinkage, Trim … 141
|E|mean , which represents the average deviation from the measured data. Note that
the numerical uncertainty U SN reported by the participants is considerably larger
than |E|mean . Unfortunately neither of these two quantities are available from the
two previous Workshops, so we cannot draw any conclusions on their development.
However, the standard deviation is available, and there was a significant improvement
between 2005 and 2010, but no further improvement after that. Now we should
compare only the blue fields, representing towed results, since the self-propelled
results in the yellow field are more scattered. Also, we have to keep in mind that
there were only 5 entries for KVLCC2 at Gothenburg 2010. It should be noted that
the mean signed error for all hulls in 2010 was very small, 0.1%D, while the standard
deviation was 2.1%D. We will return to the lack of progress between 2010 and 2015
below.
The resistance statistics may be compared with that of the Fifth AIAA Compu-
tational Fluid Dynamics Drag Prediction Workshop (Vassberg 2016), DPW. The
standard test case was a wing-body configuration, which was tested in two wind
tunnels, where the drag differed by 3.2%. 57 drag predictions were submitted, and
the mean drag differed from the mean of the two wind tunnel results by 2.4%. This
is a larger comparison error than those of Table 1, and a probable explanation is the
larger data uncertainty. The standard deviation was 2.1%, which is similar to the
results for the towed case in Table 1. An interesting difference compared with the
present workshop was the use of extrapolated predictions. All participants had to
carry out a grid dependence study and the drag reported was the continuum solution,
i.e. the solution extrapolated to zero cell size.
Table 2 shows the reduction in resistance due to the ESD. Although an increase in
resistance might be expected when adding the ESD, both computations and measure-
ments exhibit a resistance reduction. For the towed cases the computed and measured
reductions are extremely close, while for the self-propelled case the difference is
slightly larger. However, all reductions are within the reported measurement accuracy
(see Table 1), so their significance is questionable.
An objective of the series of Workshops is the possibility to study the performance
of the different components of the CFD methods. For this reason participants have
to report the details of their methods in a questionnaire. The results are collected in a
spreadsheet, which may be found in the supplementary information on the web site
Springer Extras (for address, se book cover). By linking the performance of a method
to its basic components, guidelines may be obtained for future developments. While
rather detailed comparisons were possible 35 years ago, when the first workshop was
held (Larsson 1981), this is not possible anymore due to the very much increased
complexity of the methods. Like at Gothenburg 2010, we will therefore make the
Table 2 Resistance reduction due to ESD

Computed (mean) Measured
Towing (1.2a-1.1a)% −0.7 −0.6
Self-propulsion (1.6a-1.5a)% −0.2 −1.0
142 L. Larsson
SP
SP
Fig. 1 Gothenburg 2010. Comparison error for all resistance submissions versus grid size. Blue
lines represent ±4%. SP stands for self-propulsion
comparison for the two main components: grid and turbulence model. Here we will
also add the wall boundary condition.
• How important is grid size?
Figure 1 shows the comparison errors for all resistance cases of Gothenburg 2010.
It is seen that for grid sizes above 3 million (red line) virtually all errors are within
±4%. The outliers are three self-propulsion cases. The corresponding plot for JBC
in the present Workshop is presented in Fig. 2.
Apparently, the ±4% level is now achieved at a larger cell count, approximately
10 million (red line). This is a disappointing result, indicating that more grid cells
are needed in 2015 than five years earlier. The reasons for that are not obvious,
but one explanation could be general CFD experience of the participants. Over the
years there has been a shift from only developers in the first workshops to more and
more users in the most recent ones. In the present workshop there were considerably
more users than developers. This reflects the maturity of the codes that are now used
routinely in hydrodynamic development work, but a regular user may need more grid
cells to obtain the same accuracy as an experienced developer. Another reason may
be the increasing speed of computers. The manual labor to create a perfect grid is
considerable, and it may be more efficient to bear with a less perfect, but larger grid
to get a certain accuracy
a. Bare hull, Case 1.1
b. Hull with ESD, Case 1.2
Fig. 2 Present JBC submissions. Comparison error for resistance versus grid size. Blue lines
represent ±4%
144 L. Larsson
a. Bare hull
b. Hull with ESD
Fig. 3 Comparison of JBC submissions, resistance versus grid type
The next grid-related question is:

• Is there any influence of grid type?
Table 3 Resistance comparison error for different turbulence models

Turbulence model E %D |E| %D No. of entries
mean mean
2015 2010 2015 2010 2015 2010

1.1,1.2 All 1.1,1.2 All 1.1,1.2 All
1-E Spalart Allmaras - 3.6 3.6 - 1
1-E Menter - 0.1 0.1 - 3
2-E k-e 0.8 0.3 1.8 2.1 8 37
2-E k-w 0.7 -0.5 1.6 1.8 19 37
EASM 1.6 1.8 2.0 5.9 12 3
RS, ARS -1.3 -8.0 1.3 -8.0 2 2
DES 4.4 0.6 4.4 3.1 1 6
Figure 3 is the same plot as Fig. 2, but with an emphasis on grid type. Square symbols
represent structured grids, diamonds unstructured grids dominated by hexahedral
cells and triangles unstructured cells of various shapes. There seems to be a slight
advantage of structured grids, particularly in Fig. 3b, but even in Fig. 3a the maximum
errors for the structured grids are smaller than for the other two. The largest errors
in both figures come from the completely unstructured grids, but the difference
compared to the hex-dominated ones is small and may well be insignificant.
• What is the influence of the turbulence model?
In Table 3 the comparison error for different turbulence models is presented. Results
from Gothenburg 2010 are also included. These are for all three hulls at that work-
shop. For Tokyo 2015 the results are for cases 1.1 and 1.2. There were no one-equation
(1-E) models in the present Workshop, but many two-equation (2-E) models, like in
2010. The Explicit Algebraic Stress Model (EASM) has gained in popularity and the
submissions have increased from 3 to 12.
Like in 2010, the two-equation models perform better than the more advanced
models. Both the mean signed error and the mean absolute errors are smaller, even
though the EASM model is now rather close. The very few submissions for the
Reynolds stress models (RS, ARS) makes it difficult to draw any firm conclusions
on the accuracy. This holds also for the Detached Eddy Simulation models (DES)
in the present Workshop. It should be stressed that these results are for resistance.
Wake predictions are a completely different matter.
The final component of the methods to be compared is the boundary condition
applied on the hull surface.
• Are wall-resolved solutions better than those using wall-functions?
Table 4 shows a comparison between wall-function (WF), and low Reynolds number
(LRN), wall resolved solutions for cases 1.1 and 1.2. Most methods use LRN
boundary conditions, but it is clear from the table that the wall function approach
yields the smallest errors for these two cases. The mean error, the mean absolute
error and the standard deviation are all smaller for the wall functions.
146 L. Larsson
Table 4 Resistance comparison error for different wall treatments

Case/BC E mean %D |E|mean %D σ %D No. of entries
1.1, WF 0.3 1.2 1.4 8
1.1, LRN 1.4 2.6 2.7 14
1.2, WF 0.3 1.1 1.3 8
1.2, LRN 1.4 1.8 2.2 12
1.2 Uncertainty Estimation
According to the ITTC (ITTC 2017) a solution has been validated if the comparison
error |E| is smaller than the validation uncertainty U val , i.e. |E| ≤ U val , where E = D
− S and U 2val = U 2SN + U 2D . U SN is the numerical uncertainty computed from U 2SN
= U 2G + U 2I and U D is the data uncertainty. U G and U I are the grid and iterative
uncertainties, respectively. Validation is then achieved at the U val level, meaning that
the error is within the “noise level”. If on the other hand |E| U val , validation has
not been achieved and the sign and magnitude of E can be used to improve the CFD
modeling.
Information on the procedure for estimating iterative uncertainty U I was provided
by the participants and most of them used the variation of the quantity of interest
(here total resistance) during the latter part of the convergence history. To obtain U G,
participants were asked to provide solutions for at least three grids. The screening
criterion adopted by the organizers for iterative convergence was |U I /ε12 | < 0.1,
where ε12 is the difference between the solutions on the two finest grids. 25% of the
submissions did not satisfy this criterion and these are not included in the statistics
below. As shown in Zou and Larsson (2014) a larger U I will cause an unacceptably
large error in U G .
The error in a numerical solution may be expressed as a series expansion of the
typical step size h. When the step size tends to zero the first term in the expansion
will become dominant. Thus for sufficiently small h the discretization error δ RE of
a solution1 may be approximated as αhp , where α is a constant and p is the formal
order of accuracy of the method. This error may also be written as S 1 − S 0 where S 1
is the known solution and S 0 is the solution for zero step size. We thus have δ RE =
S 1 − S 0 = αhp . For a given step size h the only unknowns are S 0 and α.
In industrial applications the grids are never fine enough for higher order terms in
the error expansion to be neglected. However, rather than incorporating more terms in
the expression for the error, the same expression is used, but with a variable p. There
are then three unknowns of the problem: S 0 , α and p, which may be determined from
three solutions with systematically varied grids. This is the approach by Xing and
Stern (2010). Another technique is advocated by Eca and Hoekstra (2014), where the
three unknowns are determined by a least squares fit to more than three solutions. The
reason for this more elaborate procedure is to reduce the problem of “scatter” in the
1 The index stands for Richardson Extrapolation, the basis of this error estimation technique.
solutions caused by, for example, not exactly scaled grids and limiters of different
kinds in turbulence models and numerical schemes. The main error estimator in
both procedures is the discretization error δ RE . This is used in more or less complex
expressions to estimate the grid uncertainty U G. 2 The general idea is to base the
uncertainty estimate on the deviation from the theoretical value of p. The larger the
difference the more uncertain the solution.
The Gothenburg 2010 Workshop and the present one offer a unique opportunity
to test the uncertainty estimation procedures, and detailed studies of the 2010 results
are presented in Zou and Larsson (2014). Here we will make a more limited study
of the question:
• How well does systematic grid variation work for uncertainty estimation in
industrial problems?
Most participants followed the instruction to submit solutions for at least three
grids. The largest set was for 1.1.a and 1.2 a and contained 6 grids. There were a few
submissions with 5 and slightly more with 4, but the majority reported 3 grids. The
total number of grid sets with at least 3 grids was 60. This is a good data base for
answering the question above.
First, the convergence of the grid sequencing may be investigated. Figure 4
presents the comparison errors in percent of D for Cases 1.1a and 1.2a. The error
is presented as a function of the normalized step size h/h1 , where h1 represents the
finest grid. Visual inspection shows that most computations converge. The difference
between two successive solutions becomes smaller and smaller, i.e. 0 < R < 1, where
R = (S n+1 − S n )/(S n − S n−1 ) and S n represents the solution on the n:th grid. However,
there are also solutions that diverge (|R| > 1) and with oscillatory convergence (−1
< R < 0).
2 Eça and Hoekstra also use other error estimators.

148 L. Larsson
15.0 ABS (OpenFOAM)
CHALMERS-FLOWTECH (SHIPFLOW)
ECN/CNRS (ISISCFD/LRN-EASM)
10.0
ECN/CNRS (ISISCFD/LRN-SST)
Comparison error E %D
ECN/CNRS (ISISCFD/WF-EASM) #5=AGR

5.0 HHI (HiFoam)
HSVA (FreSCo+)
0.0 KRISO (WAVIS)
MARIN (ReFreSCo)
MIJAC (OpenFOAM)
-5.0
NMRI (NAGISA)
PNU (FLUENTv15)
-10.0 SHIME (OpenFOAM)
SJTU (naoeFOAM)
SOTON (OpenFOAM)
-15.0
0.00 1.00 2.00 3.00 UNI-ROSTOCK (OpenFOAM-SST)
hi /h1 UNI-ROSTOCK (OpenFOAM-HYBRID)
a. Bare hull, Case 1.1a
15.0 ABS (OpenFOAM)
CHALMERS-FLOWTECH (SHIPFLOW)
10.0 ECN/CNRS (ISISCFD/LRN-EASM)
ECN/CNRS (ISISCFD/LRN-SST)
Comparison error E %D
5.0 ECN/CNRS (ISISCFD/WF-EASM) #5=AGR
HSVA (FreSCo+)
0.0 KRISO (WAVIS)
MARIN (ReFreSCo)
-5.0 MIJAC (OpenFOAM)
NMRI (NAGISA)
-10.0 PNU (FLUENTv15)
SHIME (OpenFOAM)
SJTU (naoeFOAM)
0.00 1.00 2.00 3.00
hi /h1
b. Hull with ESD, Case 1.2a
Fig. 4 Resistance versus grid step size
In Table 5 the grid convergence is reported more quantitatively, and a comparison

is made with Gothenburg 2010, all cases. In the present Workshop there were 28
submissions with more than 3 grids for cases 1.1a and 1.2a, compared with 42 for all
cases in 2010. Of the 28 submissions, 38 grid triplets could be formed (since some
contained more than 3 grids). The table is split between structured and unstructured
grids. This is an important distinction, since the theory above assumes systematically
refined grids, i.e. all cells should be scaled down geometrically, a requirement that
is very hard to satisfy for unstructured grids.
There is a significant shift from structured to unstructured grids between the two
Workshops. While in 2010 only 7 out of 40 triplets were unstructured, in 2015 this
was the case for 25 out of 38. This reflects the larger influence of the general purpose
commercial codes, of which practically all use unstructured grids. Another interesting
conclusion from Table 5 is that the unstructured grids converge surprisingly well,
considering the lack of similarity in the grid refinement. Convergence is obtained for
5 out of 7 in 2010 and 16 out of 25 in 2015. 64% were thus convergent at the present
workshop, which is not too far from the structured grids, which converged in 77%
of the cases.
Another aspect of the grid refinement is the achieved value of the exponent p,
i.e. the achieved order of accuracy. The more different from the theoretical order
of accuracy the more uncertain the computed value. In Table 6 statistics are given
for the ratio p/pth , where pth is the theoretical order of accuracy. Ideally, this ratio
should be 1. This is however not the case at all. For most submissions p/pth is larger
than 1 and he average value for these is 2.3 (2.7 in 2010). Those below 1 have an
average of 0.5 (0.7 in 2010). While a smaller order of accuracy than the theoretical
one is explainable by limiters of different kinds, reducing the order in regions with
large gradients, and by less strict boundary conditions, a higher value cannot be
explained by such shortcomings. It must be caused by the influence of higher order
terms and/or scatter, as explained above. The large values are a matter of concern,
since the empirical relations used for the uncertainty estimation are based on smaller
values (maximum 2 for Xing and Stern (2010)).
Table 5 Grid convergence

G2010 all T2015 1.1a, 1.2a
Submissions >2 grids 42 28
Total # triplets ? 38
Structured grids 33 13
– Convergent 32 10
– Divergent 1 3
Unstructured grids 7 25
– Convergent 5 16
– Divergent 2 9
Table 6 Achieved order of accuracy

G2010 all T2015 1.1a, 1.2a
# submissions with p/pth 29 27
# p/pth >1 18 21
– Mean value 2.7 2.3
# p/pth <1 11 6
– Mean value 0.7 0.5
150 L. Larsson
Knowing the grid uncertainty U G and the iterative uncertainty U I , the numerical
uncertainty U SN may be obtained as explained above. Together with the experimental
accuracy U D this yields the validation uncertainty U val . This can now be compared
with the comparison error E. It turns out that |E| < U val in 82% of the reported vali-
dations. In 2010 the evaluation was made by the organizers and both uncertainty
assessment methods were applied to the reported grid variations. For the Xing and
Stern method, |E| < U val in 60% of the submissions, while for the Eca and Hoek-
stra method this was the case for 91% of the submissions. In fact, |E| is normally
considerably smaller than U val . In the present workshop (for JBC) the mean value
of |E| was 2.0%D, while the mean value of U val was 5.0%D. Thus, the codes may be
considered validated in a mean sense!
2 Sinkage and Trim
Table 7 presents a statistical evaluation of the sinkage predictions for JBC, Cases 1.1a
and 1.2a together with results from Gothenburg 2010. Unfortunately, there is only one
submission for the bluff hull, KVLCC2 (for varying Froude numbers). Therefore the
statistics from all hulls are also included. Apparently, the comparison errors are very
large in the 2010 results, 20–30%D. This fact was very surprising and was discussed
at length at that workshop, Larsson and Zou (2014). The conclusion was that it is
extremely difficult to measure accurately the very small sinkages, of the order of 1 mm
for the slow speeds. No uncertainty estimate for the measured data was available,
except for the high speed results of the frigate used in that workshop. The conclusion
was that the large comparison errors are mainly due to errors in the experimental
data. However, a huge improvement is seen in the present computations for JBC in
the table. Neither in this case were any uncertainty estimates of the measured data
provided.
The corresponding trim statistics are presented in Table 8. For the trim the
measured uncertainty was reported in 2010 and it is seen in the table that the computed
mean errors are of the same order as this uncertainty. However, there is a large stan-
dard deviation in the collected results for all cases. Again, the comparison error is
Table 7 Sinkage statistics

Hull Case no. E mean |E|mean σ %D UD No. of entries
%D %D %D
All hulls 2010/1.2b, 2.2b, 2.3b 3.2 −21.0 11.2 – 51
KVLCC2 2010/1.2b −33.3 33.3 – – 1
All Fr
KVLCC2 2010/1.2b −18.2 18.2 – – 1
Fr = 0.142
JBC 2015/1.1a, 1.2a −1.9 4.7 5.6 – 26
Table 8 Trim statistics

Hull Case no. E mean |E|mean σ %D UD No. of entries
%D %D %D
All hulls 2010/1.2b, 2.2b, 2.3b 3.2 8.9 36.9 6 51
KVLCC2 2010/1.2b 7.4 7.4 – 7 1
All Fr
KVLCC2 2010/1.2b 3.2 3.2 – 7 1
Fr = 0.142
JBC 2015/1.1a, 1.2a 0.6 3.1 3.5 – 26
very small in the results of the present Workshop. This may well be a result of a
more accurate trim measurement, but no uncertainty estimation was presented for
the experimental data.
3 Wave Pattern
The wave pattern submissions to the Workshop are listed in Table 9. Three results
were submitted by ECN CNRS with different turbulence models and wall treatments.
The Volume of Fluid (VOF) method is by far the most popular free-surface method
with 9 out of 11 submissions. The remaining two use Level Set. All VOF submissions
use hex-dominated, unstructured grids, while the Level Set entries use structured
grids. The total number of grid points used in the simulations varies considerably,
from 0.7 to 55 million.
Table 9 Wave pattern submissions

Code identifier Free surface method Grid type No. of grid points [M]
ECN CNRS-ISISCFD-LRN EASM V U, hex 9.2
ECN CNRS-ISISCFD-LRN SST V U, hex 5.5
ECN CNRS-ISISCFD-WF EASM V U, hex 2.2
HSVA-FreSCo+ V U, hex 14.6?
KRISO-WAVIS LS S 15.5
MARIC-FINEMarine V U, hex 0.7
MARIN-ReFRESCO V U, hex 55
MHI-FLUENTv14 V U, hex 13.9
NMRI-NAGISA LS S 7.8
PNU-FLUENTv15 V U, hex 4.5
SJTU-naoeFOAM V U, hex ?
V VOF; LS Level Set; U, hex Unstructured, hex dominant; S Structured
152 L. Larsson
Fig. 5 Typical wave profile along the hull (HSVA-FresCo+)
All methods predict the wave profile along the hull very well. A typical example is
shown in Fig. 5, where the circles represent measurements and the line computations.
Larger differences are seen in the two wave cut plots to be presented, but first the
best result for the KVLCC2 in 2010 is shown in Fig. 6. Here results are presented for
three wave cuts and there is a remarkable correspondence with the measured data.
Results for the two wave cuts requested in the present workshop are shown in
Fig. 7. These are from the same code as in Fig. 6 (ECN CNRS-ISISCFD) and with the
same number of grid points (5.5 million). Although these new results are not bad, they
are significantly worse than those of Fig. 6. Starting with the innermost cut (Y /Lpp =
−0.1043) there is an over prediction of the wave in front of the bow, particularly at the
crest, and there is a constant phase shift over the entire length of the hull. However,
with the exception of the bow wave, the amplitudes are well predicted back to X/Lpp
= 1.2, where the grid is coarsened. The over prediction of the bow wave and the phase
shift is common to all submissions, indicating that there might be a problem with
the reported data. In fact most submissions are very similar to that of Fig. 7. This
holds for ECN CNRS-ISISCFD-LRN EASM, ECN CNRS-ISISCFD-WF EASM,
HSVA-FreSCo+, KRISO-WAVIS, MARIC-FINEMarine, MARIN-ReFRESCO and
MHI-FLUENTv14, while the waves are more damped in NMRI-NAGISA and PNU-
FLUENTv15. See the Workshop Proceedings for the other plots!
Very similar features are noticed for the outer cut at Y/Lpp = −0.1900. All
submissions exhibit an over predicted bow wave and a constant phase shift over
the length of the hull. At this Y-position, the differences between the methods are
larger, however. MARIN-ReFRESCO is very similar to ECN CNRS-ISISCFD-LRN
SST shown in Fig. 7, while most of the others under predict wave amplitudes to a
smaller or larger extent. HSVA-FresCo+ over predicts the amplitudes and for SJTU-
naoeFOAM there is a problem with the level of the surface both in front of and
behind the hull. In fact, none of the solutions beats the potential flow solution shown
in Fig. 8 for the outer cut. This is along the hull; at the stern and in the wake this
solution over predicts the amplitude, as expected.
Fig. 6 Three wave cuts (Y /Lpp = −0.0964, −0.1581 and −0.2993) from Gothenburg 2010,
KVLCC2
At the Gothenburg 2010 Workshop a diagram for reporting the grid density near
the free surface was introduced. This was requested also at the present Workshop.
It is presented in Fig. 9 for the submission shown in Fig. 7. The horizontal axis
represents a distance either in the X-direction along the symmetry plane/hull surface
or in the Y-direction at midship. The vertical scale to the left shows the number of
grid points per fundamental wave length (ppwl) both in the X-direction (full line)
154 L. Larsson
Fig. 7 Two wave cuts (Y /Lpp = −0.1043 and −0.1900), JBC. ECN CNRS-ISISCFD-LRN SST
Fig. 8 Potential flow solution at Y /Lpp = −0.1900, JBC. CHALMERS-FLOWTECH-SHIPFLOW

Fig. 9 Grid density. Full line: grid density in the X-direction, Dashed line: grid density in the
Y-direction, Dot-dashed line: cell height near the waterline. ECN CNRS-ISISCFD-LRN SST
and the Y-direction (dashed line). To the right, there is a scale for the thickness (z)
of the cells close to the water line (at Z/Lpp = 0.00).
There are peaks in the longitudinal distribution at bow and stern reaching 60–100
ppwl, while along the main part of the hull the density is about 20 ppwl. In the
transverse direction (at midship) the density drops rapidly from a very high value in
the boundary layer, but stays around 20 ppwl out to about Y /Lpp = 0.25. This is the
lateral extension of the wave system at midship. There is thus a high density where
required, but a rapid drop further out. The cell thickness goes down to about 0.0006
Lpp around the undisturbed surface level.
This method (ISIS-CFD) has adaptive grid generation which automatically
concentrates cells in regions of large gradients. Most other submissions exhibit larger
grid densities along the hull, typically 50 ppwl, and the peaks at the hull ends are
normally higher. Also, the density in the transverse direction is mostly larger than in
Fig. 9. The cell height around the waterline varies significantly between the methods
and are in the range 0.0002–0.002. A surprising plot is that of PNU-FLUENTv15,
which exhibits the largest peak longitudinally (800 ppwl), a good density transversely
at midship (around 30 ppwl) and a very small cell height at the surface (0.0002 Lpp).
Still, it has the largest damping of the waves along the hull. Presumably, this is an
effect of the VOF discretization scheme.
4 Conclusions
The Workshop submissions for JBC contained four cases: the hull with or without an
ESD both in towing and self-propulsion. Altogether 88 predictions were provided. In
this chapter, resistance, sinkage and trim, as well as wave pattern results are evaluated.
Conclusions are as follows:
156 L. Larsson
• Like at the previous workshops the mean signed comparison error for resistance is
very small, around 1%, which is also the accuracy reported for the experimental
data. However, more interesting is the mean absolute comparison error, since
this represents the average deviation from the data. This is around 2% for the
towed cases and 3% for self-propulsion. The scatter is represented by the standard
deviation and is about 2% and 4% respectively for the two conditions. In the 2005
and 2010 Workshops the standard deviations were 6 and 1%. There was thus a
significant reduction in scatter between 2005 and 2010, but no further reduction
in 2015.
• A resistance reduction around 1% is predicted with the ESD fitted. This is
confirmed by the measurements.
• For the comparison error to be smaller than 4% the grids have to contain more
than 10 M cells. In 2010 the corresponding limit was 3 M cells. Thus, more grid
cells were required for the same accuracy in 2015.
• There is hardly any influence of the grid type on the predicted resistance. If any,
there is a slight advantage of the structured grids, compared to hex-dominated and
completely unstructured grids.
• Like in 2010 the two-equation turbulence models predict better results than the
more advanced models. The k-ε and k-ω models have mean signed comparison
errors less than 1% and mean absolute errors just below 2%. EASM has gained
in popularity and its mean absolute error is now 2%.
• Resistance predictions using wall-functions are more accurate than those with
wall resolved flows. Both the absolute error and the standard deviation are about
twice as large with wall resolution.
• Grid convergence is demonstrated for 64% of the unstructured grids and 77% of
the structured grids. The achieved order of accuracy in the grid refinement studies
is in many cases far from the theoretical accuracy. Most methods predict a too
high order of accuracy, 2.3 times the theoretical one, on the average. For those
with a too low order of accuracy the mean value is 0.7 times the nominal one.
• As stated above, the mean absolute comparison error in all resistance predictions
is 2%, which is considerably smaller than the mean validation uncertainty, 5%,
composed of the mean numerical uncertainty of 4.9% and the stated experimental
accuracy of 1%. The methods may thus be considered validated in a mean sense.
• The comparison errors for sinkage were very large in 2010, but are now reduced
considerably. The mean signed error, absolute error and standard deviation are
now −1.9%, 4.7% and 5.6% respectively.
• For trim the corresponding errors are also reduced considerably and are now 0.6%,
3.1% and 3.5%, respectively.
• Very accurate predictions are reported for the wave profile along the hull, while
less correspondence with measurements is demonstrated for two wave cuts. There
is a consistent phase shift in all results, which may indicate an error in the data.
A small over prediction of the bow wave height is also noted in all results.
References
Eca, L., & Hoekstra, M. (2014). A procedure for the estimation of the numerical uncertainty of CFD
calculations based on grid refinement studies. Journal of Computational Physics, 262, 104–130.
Hino, T. (Ed.). (2005). Proceedings of CFD Workshop Tokyo 2005. NMRI Report.
ITTC. (2017). ITTC—Recommended Procedures and Guidelines. Uncertainty Analysis in CFD.
Verification and Validation Methodology and Procedures.
Larsson, L. (Ed.). (1981). SSPA-ITTC Workshop on Ship Boundary Layers. SSPA Report 90,
Gothenburg, Sweden.
Larsson, L., Stern, F., & Visonneau, M. (Eds.). (2014). An assessment of the Gothenburg 2010
workshop. Dordrecht: Springer Science and Business Media.
Larsson, L., & Zou, L. (2014). Evaluation of resistance, sinkage and trim, self-propulsion and wave
pattern predictions. In Numerical ship hydrodynamics—An assessment of the Gothenburg 2010
workshop. Dordrecht: Springer Science and Business Media.
Van, S. H., Kim, W. J., Yim, D. H., Kim, G. T., Lee, C. J., & Eom, J. Y. (1998). Flow measurement
around a 300 K VLCC model. In Proceedings of the Annual Spring Meeting, SNAK, Ulsan
(pp. 185–188).
Vassberg, J. C. (2016). Challenges and accomplishments of the AIAA CFD drag prediction work-
shop series. In AVT-246 Progress and Challenges in Validation Testing for Computational Fluid
Dynamics, Avila, Spain.
Xing, T., & Stern, F. (2010). Factors of safety for Richardson extrapolation. Journal of Fluids
Engineering, 132(6).
Zou, L., & Larsson, L. (2014). CFD verification and validation in practice—A study based on
resistance submissions to the Gothenburg 2010 workshop on numerical ship hydrodynamics. In
30th Symposium on Naval Hydrodynamics, Hobart, Australia.
Analysis of the Local Flow around JBC
Michel Visonneau
Abstract This chapter is devoted to an analysis of the local flow field for test cases
1-3, 1-4, 1-7 and 1-8. The general objective of this analysis is to assess the present
computational approaches according to their ability to represent the influence of a
duct on the local flow field in presence or absence of a propeller, thanks to the exis-
tence of local flow measurements mainly performed by NMRI. General conclusions
based on the numerous contributions, are drawn concerning the reliability of turbu-
lence closures in these peculiar flow configurations, the influence of the propeller
representation on the local flow field and the ability of CFD to represent the local
influence of a duct on the flow at the stern region and in the wake of a ship.
1 Introduction
This chapter is devoted to an analysis of the local flow field for test cases 1-3, 1-4, 1-7
and 1-8. The general objective of this analysis is to assess the present computational
approaches according to their ability to represent the influence of a duct on the local
flow field in presence or absence of a propeller, thanks to the existence of local flow
measurements mainly performed by NMRI. During the previous G2010 workshop,
the role played by the turbulence models was systematically assessed for several ships
and it was established that the flow in the stern region of the KVLCC2 was strongly
dependent on the turbulence closures, once a grid fine enough was used. The objective
of this chapter is to conduct a very similar analysis in a more complex framework
involving the interaction of the stern flow with an energy-saving device with or
without a propeller. General conclusions based on the numerous contributions, have
to be drawn concerning the reliability of turbulence closures in these peculiar flow
configurations, the influence of the propeller representation on the local flow field
and the ability of CFD to represent the local influence of a duct on the flow at the
stern region and in the wake of a ship.
M. Visonneau (B)
LHEEA Laboratory, ECN/CNRS, 1, rue de la Noe, 44321 Nantes, Cedex 3, France
160 M. Visonneau
2 Experiments
The Japan Bulk Carrier (JBC) is a Capesize bulk carrier equipped with a stern duct
as an energy saving device (ESD). National Maritime Research Institute (NMRI),
Yokohama National University and Ship Building Research Center of Japan (SRC)
were jointly involved in the design of this ship hull, duct and propeller. Its length
between perpendiculars is Lpp = 280 m. Its service speed is 14.5 knots, leading
to a Froude number Fn = 0.142 and a Reynolds number at model scale of Re =
7.46 106 with Lpp = 7.00 m at model scale. Its main particulars are described in
Table 1. This ship was selected as one of the testcases of the Tokyo2015 workshop
on numerical ship hydrodynamics, the latest edition in a series of workshops which
also includes Gothenburg 2010 (Larsson et al. 2014). Towing tank experiments,
including resistance tests, self-propulsion tests and PIV measurements of stern flow
fields, were performed at NMRI, SRC and Osaka University. Several test cases were
considered, all with free sinkage and trim; test cases 1-3a (resp. 1-4) are for towing
test without (resp. with) ESD, cases 1-7 (resp. 1-8) for self-propulsion tests without
(resp. with) ESD (see Table 2). Global force measurements and local LDV velocity
profiles at three sections named S2, S4 and S7 (i.e. X/Lpp = 0.9625, X/Lpp = 0.9843
and X/Lpp = 1.0000) before and after the propeller and duct were also provided by
the organizers. Figure 1a, b show a view of the stern without and with ESD with the
location of the local measurement sections.
Table 1 JBC—main particulars

Main particulars Full scale
Length between perpendiculars L pp (m) 280.0
Length of waterline LW L (m) 285.0
Maximum beam of waterline BW L (m) 45.0
Depth D (m) 25.0
Draft T (m) 16.5
Displacement volume ∇ (m3 ) 178369.9
Wetted surface area (without ESD) S0−w/oE S D (m2 ) 19556.1
Wetted surface area (with ESD) S0−wE S D (m2 ) 19633.9
Block coefficient ∇/(L pp BW L T ) 0.8580
Midship section coefficient CM 0.9981
LCB (%L P P ), fwd+ 2.5475
Vertical center of gravity (from keel) KG (m) NA
Metacentric height GM (m) NA
Moment of inertia Kx x /B NA
Moment of inertia K yy /Lpp , Kzz /Lpp NA
Propeller center, long. location (from FP) x/Lpp 0.985714
Propeller center, vert. location (below WL) −z/Lpp −0.0404214
Propeller rotation direction (view from stern) Clockwise
Analysis of the Local Flow around JBC 161
Table 2 Definition of the test cases

Test case 1-3a 1-3b 1-4 1-7 1-8
Experiments NMRI TUHH (wind NMRI NMRI NMRI
(towing tank) tunnel) (towing tank) (towing tank) (towing tank)
Propeller No No No Yes Yes
(self-
propulsion)
Energy saving No No Yes No Yes
device
(a) Without ESD (b) With ESD
Fig. 1 Side views of the JBC hull
Moreover, TUHH performed wind-tunnel LDV experiments around a double-

body model of the JBC without rudder nor propeller at a slightly lower Reynolds
number Re = 2.74 106 with Lpp = 3.513 m and U = 11.8 m/s. These experiments
and the computations performed in the same conditions are described in the test case
1-3b.
3 Test Case 1-3a
3.1 Description of Experimental Results
Test case 1-3a concerns the local flow analysis around the JBC without duct nor
propeller towed in the NMRI towing tank, in free trim and sinkage conditions.
Figures 2, 3 and 4 show the experimental iso wakes and iso contours of the two
crossflow velocity components (v, w) at three sections S2, S4 and S7 already defined
before.
From Fig. 2a taken at station S2, one can notice the presence of an intense longi-
tudinal vortex which is characterised by an island of isowakes comprised between
0.2 and 0.4. A second closed isowake contour is visible closer to the wall but at the
boundary of a region where PIV experiments are likely less reliable. One notices
also that it is not possible to get reliable experimental information very close to the
hull, in the boundary layer. The lateral v component (Fig. 2b) is always positive with
closed contours associated to level 0.0 to 0.2 in the core of the main longitudinal
162 M. Visonneau
(a) Iso-u contours (b) Iso-v contours (c) Iso-w contours
Fig. 2 Case 1-3a—experiments at station S2
vortex. For the vertical component w (Fig. 2c), one can notice a well defined line (w
= 0.0) dividing a region close to the wall where w reaches a minimum value of −0.1
and a region outside of the core of the vortex where w reaches the maximum value
of 0.3.
At station S4, Fig. 3a shows the development of the longitudinal vortex with
well developed hook-shaped isowake contours. Two secondary vortices are also
visible at the boundary of the PIV experiments which are maybe related with the
one visible at station S2 (isocontour 0.4) mentioned previously. However, this can be
also an experimental artefact. It is interesting to notice for further comparisons with
computations, the typical shape of the isowake contours (iso u contours between 0.2
and 0.3) in the lower part of the vertical plane of symmetry. Figure 3b shows the
contours of v which are uniformly very small around 0.1–0.2. Figure 3c provides the
same information for the vertical component W at station S4. Here, one should notice
a minimum value for w of −0.3 in the vertical plane of symmetry and a maximum
value of 0.2 outside of the core of the main vortex.
Finally, Fig. 4a shows the isowakes distribution at station S7. An intense longitu-
dinal vortex is still visible with a minimum isowake value of 0.4. It is interesting to
mention that the experimental averaged flow field is not fully symmetric (iso con-
tours 0.5 and 0.4, for instance). This asymmetry might be related to the unsteadiness
of the stern flow or experimental uncertainties close to the vertical plane of symme-
try. Figure 4b shows the contours of v which are again uniformly very small around
0.1–0.2. Figure 4c provides the same information for the vertical component w at
station S7. Here, one notices a minimum value for w of −0.4 in the vertical plane of
symmetry and a maximum value of 0.2 outside of the core of the main vortex.
3.2 Review of Contributions
Eighteen participants provided results for the tescase 1-3a. The main characteristics of
their contributions are listed in Table 3 in terms of turbulence closures, wall treatments
and discretization. Some participants provided several different contributions which
makes possible a comparison of various turbulence closures or wall treatments for
the same grid and same code. Moreover, SOTON performed a comparison of codes
(OpenFoam vs. Star-CCM+) with the same grid and same turbulence closure.
Among the eighteen participants, eleven used unstructured grids while seven
computed on structured grids sometimes locally enriched with overlapping patches
(Chalmers, CNRS/INSEAN). It is well known that the prediction of the nominal
velocity in the stern region depends strongly on the turbulence closure, as soon as a
reasonable dense grid is used. Like during the Gothenburg 2010 workshop, the tur-
bulence models employed by the participants can be classified into three groups: (i)
Unsteady Hybrid RANS-LES model (University of Rostock (URO)), (ii) anisotropic
non-linear statistical turbulence modeling: Reynolds stress transport models (HHI),
Explicit Algebraic Reynolds stress models (Chalmers, ECN/CNRS, NMRI, YNU),
(ii) isotropic linear eddy-viscosity model (ECN/CNRS, HSVA, MARIN, URO, …).
Therefore, in five years, we can observe a slight increased use of more sophisticated
turbulence closures for ship flow computations, probably related with the number
of publications illustrating the crucial role of turbulence modeling for the predic-
tion of local flow in the stern region. All the discretization methods are formally
second-order accurate and based on multi-block structured or unstructured grids
which are all body-fitted. Only Chalmers and CNRS/INSEAN are using an over-
164 M. Visonneau
Table 3 Case 1-3a—main characteristics of the computations

Organization code name Turbulence model Wall model Grid characteristics
ABS k-ω SST Wall function Unstructured grid
OpenFOAM 5 482 000 cells
Chalmers-Flowtech EARSM Wall resolved Structured grid
Shipflow with overlapping
8M cells
CNR-INSEAN Spalart-Allmaras Structured grid
Xnavis with overlapping
7.7M cells
ECN/CNRS SST/EARSM Wall function Unstructured grid
ISIS-CFD aka FINET M /Marine Wall resolved 5.7/9.2M cells
HHI Realizable Wall function Unstructured grid
HiFOAM/StarCCM+ k-/RSTM 1.48M cells
HSVA 2 equations Wall function Unstructured grid
FreSCo+ from 2.1 to 14.6M
cells
KRISO Realizable Wall function Structured grid
Wavis k-/EARSM 15.5M cells
MARIC k-ω SST Wall function Unstructured grid
FINET M /Marine 650 000 cells
MARIN k-ω SST Wall resolved Unstructured grid
ReFRESCO +DM correction 5.5M cells
MHI k- RNG Wall resolved Unstructured grid
Fluentv14 k-ω SST 13.9M cells
MIJAC k-ω SST Wall resolved Unstructured grid
OpenFOAM 8.7M cells
NMRI EARSM Wall resolved Structured grid
Nagisa 10M cells
PNU 2 equations Wall resolved Unstructured grid
Fluentv15 k- RNG Wall function 4.45M cells
SHIME k-ω SST Wall resolved Unstructured grid
OpenFOAM 22.15M cells
SJTU k-ω SST Wall function Unstructured grid
naoeOpenFOAM 2M cells
SOTON k-ω SST Wall resolved Structured grid
OpenFOAM/StarCCM+ 19.4M cells
UniRostock Hybrid RANS-LES Wall resolved Structured grid
OpenFOAM k-ω SST 29.4M cells
YNU EARSM Wall resolved Structured grid
SURFv7 2.5M cells
set approach combining structured grids to reach an optimal grid density in the
stern region.
3.3 Grid Sensitivity
The grids used by the 18 participants can be classified into four categories, based on
their sizes:
• Not So Fine: Ncell < 2M cells (MARIC, HHI, SJTU)
• Fine: 2M cells < Ncell < 10M cells (ABS, CHALMERS, CNR/INSEAN,
ECN/CNRS, HSVA, KRISO, SHIME, MIJAC, NMRI, PNU, YNU)
• Very Fine: 10M cells < Ncell < 50M cells (SOTON, MHI, MARIN, URO)
• Tremendously Fine: Ncell > 50M cells (None)
During the Gothenburg 2010 workshop, the average size of the grids was 5M
cells, excluding the unusually fine grid (305M cells) generated by IIHR for their
DES computations. For the Tokyo 2015 workshop, one notices a trend towards the
use of significantly finer grids since, five years later, the average size is around
8.4M cells, excluding the very fine grid used by URO for their hybrid RANS-LES
computations.
To get an idea of the grid influence, Figs. 5, 6 and 7 show the isowakes at sta-
tions S2 to S7 computed by SHIME on two different grids, keeping unchanged the
code, discretization and turbulence model. One can hardly see any significant differ-
ences between the two results although the number of grid points is multiplied by a
factor 2.5.
Fig. 5 Station S2—computations from SHIME using different grids (left 10M) (right 25M)
166 M. Visonneau
Figures 8, 9 and 10 show a comparison of the same isowakes at the same stations
as previously, computed by two different teams, SOTON and SHIME, using the same
OpenFOAM code, same turbulence model (low Re near wall k-ω SST) and grids of
similar density comprised of 20 to 25M cells. It is reassuring to notice that the results
are quite similar for all stations, which indicates that some conclusions can be drawn
without taking into account specificities related with the user of the computational
approach or local details of the grids.
(a) SOTON/OpenFOAM - 20M (b) SHIME/OpenFOAM - 25M
Fig. 8 Case 1-3a—comparisons at station S2 between two different contributors using the same
code and similar grids
Based on many previous grid sensitivity studies, one can consider that most of
the grids used in this workshop (the Fine category) are such that the turbulence
modelling error should exceed by large the discretization error, which means that it
is meaningful to compare the turbulence closures among computations performed
with grids comprised of more than 2M cells.
168 M. Visonneau
3.4 Wall Treatment
Since more and more computations are now performed with a wall function approach,
it is interesting to compare wall resolved and wall modeled formulations and see
what is the impact of the wall treatment on the development of the flow at the sta-
tions analysed in this workshop. Figures 11, 12 and 13 provide such a comparison
performed with ISIS-CFD on very similar grids (except below y+ = 60) and with
the same anisotropic explicit algebraic Reynolds stress model (EARSM). Globally,
differences exist but are not very significant for the three stations used for compar-
ison even if one can notice some local differences affecting the lower values of the
isowakes. Based on these comparisons, it is therefore difficult to claim that a wall-
function model affects significantly the accuracy of straight ahead resistance viscous
computations.
3.5 Turbulence Closures
The turbulence models used by the contributors for the case 1-3a can be classified
into three groups:
• The isotropic linear closures: INSEAN (Spalart-Allmaras), HSVA (Linear
EARSM), MARIN (k-w SST + DM), MARIC (k-ω SST), URO (k-ω SST), PNU
(k-), SHIME (k-ω SST), CSSRC (k-), ABS (k-ω SST), SOTON (k-ω SST),
ECN-CNRS (k-ω SST), MHI (k-ω SST),
(a) ECN/ISIS-CFD - Wall function (b) ECN/ISIS-CFD - Wall resolved
Fig. 11 Case 1-3a—comparisons at station S2 between wall function and wall resolved formula-
tions
tions
• The anisotropic non-linear models: MHI (RSTM), ECN-CNRS (EARSM), NMRI

(EARSM), CHALMERS (EARSM), YNU (EARSM), KRISO (EARSM),
• The hybrid RANS-LES models: URO (DES).
In the previous workshop held in Gothenburg in 2010 (see Larsson et al. 2014 for
instance), it was noticed that the modelling error induced by the turbulence closures
on the velocity distribution in the stern region is the dominating factor of error,
greatly surpassing the discretisation error as soon as a reasonably fine grid is used.
170 M. Visonneau
tions
In the next subsection, the various turbulence models used by the contributors will
be analyzed and compared to see if one can draw common conclusions on the role
played by the modelling error in such a flow.
3.5.1 Linear Isotropic Turbulence Closures
Figures 14, 15 and 16 show the iso-contours obtained by the one equation Spalart-
Allmaras model implemented in Cnavis by INSEAN. It is clearly unable to provide
the right level of longitudinal vorticity for all the experimental cross-sections. It is
also true at a lesser degree for the linear EARSM variant implemented by HSVA in
Fresco+ (see Figs. 17, 18 and 19). The k-ω SST model appears to behave consistently
better in the results shown by University of Rostock with OpenFoam although still
far from the experiments (see Figs. 20, 21 and 22). In particular, none of these models
is able to predict the characteristic hook shape visible at station S2 with a low level u
= 0.2 present in the measurements. The closed contours u = 0.3 and u = 0.2 visible
in the experiments at station S4 are also not captured by any of these simulations.
3.5.2 Non-linear Anisotropic Turbulence Closures
The second category of RANSE turbulence models comprises closures taking into
account the turbulence anisotropy. Imperfectly when Explicit Algebraic Reynolds
Stress models (EARSM) are used or (potentially) more accurately if Full Reynolds
Stress Transport Models (RSM) are utilised. Figures 23, 24 and 25 show a comparison
of computations performed by ECN/CNRS with the same code (ISIS-CFD) and the
(a) Experiments (b) Spalart-Allmaras
Fig. 14 Case 1-3a—comparisons at station S2 between INSEAN/Cnavis and experiments
same grid with k-ω SST and EARSM turbulence closures. It appears clearly that
EARSM leads to a stern flow with more longitudinal vorticity. Figure 24 shows for
instance at station S4 that EARSM accurately captures the isowake 0.2 while k-
ω SST fails to reach this result. Low iso wakes are of course related with higher
longitudinal vorticity on the core of the main vortex, a result which is associated
with the increased production of vorticity related with the turbulence anisotropy.
172 M. Visonneau
(a) Experiments (b) Linear EARSM
Fig. 17 Case 1-3a—comparisons at station S2 between HSVA/Fresco+ and experiments
Figures 26, 27, 28, 29, 30 and 31 show a comparison of the simulations of
Chalmers University (resp. ECN/CNRS) with the same EARSM closure imple-
mented in ShipFlow (resp. ISIS-CFD). Although these computations are performed
by Chalmers on a slightly finer grid than the one used by ECN/CNRS, one can
observe a very similar result. The lower isowake levels in the core of the main aver-
aged vortex are better captured at every experimental sections and the trend provided
by both codes are almost identical although completely different codes based on
totally different numerical methodologies are compared. However, for both results,
one should notice that the island corresponding to the iso-level u = 0.2 is captured
by none of these simulations. However, the improvement brought by the EARSM
closure for both contributors, underlines the crucial role played by the turbulence
closure as soon as the simulation is carried out on a fine enough grid.
It is also very interesting to assess the predictive capabilities of Full RSM Transport
Turbulence Closures for this typical flow. Full RSM transport equations model is
based on the discretisation of six transport equations for the Reynolds stresses and
174 M. Visonneau
(a) Experiments (b) k-ω SST
Fig. 20 Case 1-3a—comparisons at station S2 between URO/Lemos and experiments
one for the turbulence dissipation or frequency. They are supposed to be superior to
EARSM because they are exact for convection and production but the other terms
like dissipation, diffusion and pressure velocity correlations are crudely modelled.
Moreover, due to the absence of explicit turbulence viscosity, this class of models is
known to be less robust and more sensitive to discretisation errors on distorted grids
and high Reynolds flows. Such a RSTM closure is implemented in Star-CCM+ and
Figs. 32, 33 and 34 show results obtained by HHI. The agreement with experiments
is very good for stations S4 and S7. At station S2, the simulated longitudinal vortex
appears to be slightly too large but for the first time, the isowake u = 0.2 is present
in the computations at station S2.
(a) k-ω SST (b) EARSM
Fig. 23 Case 1-3a—ECN/CNRS—comparisons at station S2 between two turbulence closures

(same grid, same code ISIS-CFD)
3.5.3 Hybrid RANS-LES Turbulence Closures
In the previous subsections, one has observed that the turbulence anisotropy should be
taken into account by the turbulence models to simulate the right level of longitudinal
vorticity in stern flows. EARSM provides a better solution as confirmed indepen-
dently by two contributors using the same model and RSTM leads to an almost
perfect representation of the measurements. All these models are based on statistical
176 M. Visonneau


modelling and are theoretically valid for steady averaged flows since URANSE can
not simulate the intrinsic flow instabilities appearing in any turbulent flows. To refine
further the model and approach the real physics, it is interesting to have recourse to
hybrid RANS-LES unsteady formulations combining RANSE modelling close to
the wall and a somewhat under-resolved LES formulations outside of the boundary
layer. DES variants proposed by Spalart and colleagues belong to such a class of
models. The University of Rostock provides two flow simulations based on a DES
and IDDES hybrid RANS-LES closures. The results for the sections S2 to S7 are
given by Figs. 35, 36, 37, 38, 39 and 40. Both are based on a double body con-
(a) Experiments (b) EARSM
Fig. 26 Case 1-3a—comparisons at station S2 between CHALMERS/Shipflow and experiments
Fig. 27 Case 1-3a—comparisons at station S2 between ECN/ISIS-CFD and experiments
figuration. DES is able to capture the isowake u = 0.2 at station S2 while IDDES
provides a solution with less longitudinal vorticity at this section. However, at sta-
tions S4 and S7, the agreement with measurements is degraded, although slightly
better with IDDES, when compared to the DES solution. Globally, the agreement
on the isowakes is not significantly better for all stations than what can be obtained
with EARSM or full RSTM statistical closures.
178 M. Visonneau
3.6 Influence of the Free-Surface and Wall Treatment
The previous hybrid RANS-LES computations were performed on a double body i.e.
by replacing the free-surface by a plane of symmetry. It is interesting to assess the
influence of the free-surface on the isowakes at the experimental stations. Figures 41,
42 and 43 show computations done by ECN/CNRS with ISIS-CFD with and without
free-surface on grids of very similar density. Apart from some minor modifications
on the shape of lower isowakes, the influence of the free-surface on the vortical flow
at the stern is not significant, which means that the hybrid RANS-LES computa-
tions performed by URO are not jeopardized by the lack of free-surface boundary
condition.
The role played by the wall boundary condition is always very controversial and
it is worthwhile to assess the results obtained by the same code on grids which only
differ by the wall treatment. ECN/CNRS has conducted such computations with
ISIS-CFD on two grids built with HEXPRESST M . Both grids have a viscous layer
close to the hull but the wall resolved grid has a first point with a y+ value below one
180 M. Visonneau
(a) Experiments (b) RSM
Fig. 32 Case 1-3a—comparisons at station S2 between HHI/Star-CCM+ and experiments
in average over the hull while the wall modeled grid has a first point located around
y+ = 80. Figures 44, 45 and 46 show a comparison of the isowake distribution at the
usual experimental stations. Apart from the shapes of isowakes u = 0.2 and 0.3, one
does not notice large differences between the results at the experimental stations.
3.6.1 Codes Comparison
Finally, based on the available contributions, it is possible to perform a code com-

parison in very fair conditions. By fair, one means using exactly the same grid and
(a) Experiments (b) Hybrid RANS-LES - DES
turbulence closure. SOTON delivered computations in such conditions with Open-

FOAM and StarCCM+ and Figs. 47, 48 and 49 show a comparison of the isowakes
at the usual experimental cross-sections. It is remarkable to see that the figures are
almost identical, which underlines the level of maturity reached by computational
codes nowadays.
182 M. Visonneau
(a) Experiments (b) Hybrid RANS-LES - IDDES
3.7 Case 1-3a —Temporary Conclusions
On the basis of this first analysis based on some representative examples taken from
the numerous contributions, one can draw the following conclusions:
• With the typical grids which are used by the contributors (5–10 M points), the
influence of the grid discretisation is moderate for RANSE, which means that the
numerical error appears to be under control.
• As noticed during the Gothenburg2010 workshop for the analysis of the local
flow around the KVLCC2, the major influence comes from the turbulence closure.
Linear isotropic closures significantly under-predict the longitudinal vorticity at
S2 while full RSM closures slightly over-predict it at the same station.
• Non-linear anisotropic closures (EARSM) offer a good compromise from the
standpoint of the local flow although they slightly underpredict the vorticity at
S2.
• New results from hybrid RANS-LES are promising but IDDES seems to over-
predict the vorticity again. The agreement with the isowakes distribution is worse
than what is achieved with the best RANS statistical turbulence closures.
184 M. Visonneau
(a) Fr=0.142 (b) Fr=0.0
Fig. 41 Case 1-3a—ISIS-CFD—influence of the free-surface at station S2
• Computational codes seem to be mature since one does not observe large discrep-
ancies when the same turbulence closures on similar grids are used.
• The influence of the free-surface on the local flow is not negligible but not large.
• The influence of the wall boundary treatment on the development of the stern flow
is weak.
(a) Fr=0.142 (b) Fr=0.0
(a) Fr=0.142 (b) Fr=0.0
3.8 Local Vortex Flow Analysis
3.8.1 Objectives of This Local Analysis
In the previous editions of the Gothenburg/Tokyo workshops on numerical ship

hydrodynamics, the local flow analysis was uniquely based on the inspection of the
flow characteristics at specific cross-sections where experiments were available. Most
of the time, this analysis was based on the visual inspection of iso-lines of character-
istic quantities like the iso-wakes, iso values of turbulent stresses, etc. Although, this
186 M. Visonneau
(a) Wall modeled (b) Wall resolved
Fig. 44 Case 1-3a—ISIS-CFD—influence of the wall treatment at station S2
approach provides a global analysis of the flow for each experimental cross-section,
it can be misleading since it is based on visual inspection of iso-lines for which local
gradients are difficult to appreciate. For the first time, during the Tokyo 2015 edition,
it was decided to enrich this cross-section based evaluation by a more detailed and
local vortex flow analysis in order to draw more elaborate conclusions about the gen-
eration and evolution of the longitudinal vortices. What is aimed by this study is to
provide a detailed inspection and comparison of the experiments and computational
results inside the core of the main bilge vortex.
Fig. 47 Case 1-3a—SOTON—comparison between OpenFOAM and StarCCM+ at station S2
3.8.2 Procedure, Difficulties and Perils of the Exercise
What we wished to compare was the y and z transversal evolutions of characteristic

flow data across the vortex center at stations S2, S4 and S7. The main difficulty is
to locate the center of the main bilge vortex since (i) it is implicitly stated that there
is only one such vortex which can be detected without any ambiguity and (ii) that it
is feasible to locate precisely its center at several cross-sections. First of all, it was
decided to use the same information in computations and experiments to locate the
vortex center. This means that, since we had no three-dimensional information from
the measurements (no tomographic PIV), we were obliged to use the local max(ωx )
188 M. Visonneau
to locate the vortex center in each section instead of max(Q). This procedure is
relatively reliable for stations S4 an S7 where the main vortex is roughly aligned
with the x direction but it is less justified for station S2 where some computations
indicate that the vortex is not aligned with the x axis. This is the reason why some
additional figures were also produced based on max(Q) instead of max(ωx ) to locate
the center of the vortex at station S2.
3.8.3 Genesis from NATO-AVT183—DTMB5415 at Straight Ahead

Condition
The original idea came from the Tomographic PIV experiments performed at IIHR
by Prof. Frederick Stern’s team on the flow around the US Navy frigate DTMB5415
at straight ahead condition. These remarkable experiments were used in the frame-
work of the NATO/AVT-183 collaborative project to assess the ability of various
codes to simulate accurately the flow physics in the core of the longitudinal vortices
shed at the sonar dome (Bhushan et al. 2019). Figure 50b gives a global overview
of the various vortices emanating from the sonar dome and a typical view of the
three-dimensional box where TPIV experiments are performed. Figure 51 shows the
longitudinal evolution in the core of the Sonar Dome Vortex of the invariant Q and
turbulence kinetic energy TKE for experiments and various computations. Figures 52
and 53 show the radial evolution in the core of the SDV vortex of the longitudinal
vorticity, the invariant Q, the turbulence kinetic energy TKE and the longitudinal
component of the velocity U. Here, computations performed by IIHR using two
different turbulence modeling approaches are compared to the experiments.
From these figures, one can notice on one hand that the longitudinal evolution of
Q and TKE were satisfactorily predicted by all the computations. On the other hand,
large differences between the statistical and hybrid RANS/LES turbulence models
run by IIHR were revealed by the comparison on the radial evolution in the core of
the SDV vortex, hybrid RANS/LES DES (resp. statistical) model over (resp. under)
predicting the magnitudes of Q and ωx . It is worthwhile to underline also the large
differences observed in the radial distribution of TKE between statistical and hybrid
RANS/LES turbulence closures.
(a) 3D view (SDV and FBKV vortices) (b) IIHR 3D experiments - ωx behind the sonar dome
Fig. 50 NATO AVT183—longitudinal vortices evolution and 3D tomographic PIV around the
DTMB5415 at straight ahead condition
190 M. Visonneau
(a) Invariant Q (b) TKE
Fig. 51 NATO AVT183—longitudinal evolution of Q and TKE in the core of the SDV vortex
(a) ωx (b) Q
Fig. 52 NATO AVT183—radial evolution of ωx and Q in the core of the SDV vortex
(a) TKE (b) U
Fig. 53 NATO AVT183—radial evolution of TKE and U in the core of the SDV vortex
Based on these comparisons, the organizing committee of T2015 decided to pro-

pose such a detailed analysis to see if the flow physics observed in the core of the SDV
was somewhat generic and to assess the predictive capacities of CFD to reproduce
it in the case of the main bilge vortex developing at the aft part of the JBC.
3.8.4 Local Vortex Flow Analysis—Longitudinal Evolution
As indicated previously, the core of the main bilge vortex was detected by looking
for the point where ωx is maximum since only 2D experiments at sections S2, S4 and
S7 were available. In the computations, a downstream streamline was shed from the
vortex core at S2 and the data was extracted along this streamline. The data obtained in
that way were compared with that extracted along an upstream streamline originating
from the vortex core at S7 and no major differences were observed. Figure 54 showing
the longitudinal evolution of the Y and Z coordinates of the core, establishes that
the trajectory of this vortex is reasonably well predicted by most of the contributors.
A better agreement with the experiments is observed on the vertical location than
on the horizontal one. All the computations predict a wavy longitudinal evolution of
the Y and Z, which means that the core of the vortex follows a kind of helicoidal
trajectory. Globally, all the computations indicate the same trend from upwind to
downwind, although a relatively large dispersion can also be noticed.
Figure 55 shows the longitudinal evolution of the second invariant Q and the
longitudinal component of the vorticity. Apart from Chalmers which predicts an
(a) Y (b) Z
Fig. 54 Case 1-3a—longitudinal evolution of the lateral position of the core of the main vortex
(a) Q (b) ωx
Fig. 55 Case 1-3a—longitudinal evolution of Q and ωx in the core of the main vortex
192 M. Visonneau
(a) U (b) Tke
Fig. 56 Case 1-3a—longitudinal evolution of U and TKE in the core of the main vortex
erroneous longitudinal evolution of Q, all the contributions appear globally similar

in terms of magnitude. The magnitude of the longitudinal vorticity measured by
NMRI is initially under-evaluated in the near-wake by all the computations. After
X/Lpp = 1.0, the agreement between measurement and computations looks improved
but it is difficult to draw any solid conclusions because of the lack of experiments
after X/Lpp = 1.00 (Fig. 56).
3.8.5 Local Vortex Flow Analysis—Transversal Evolution
In order to better understand the local flow physics in the core of this bilge vortex,
it was also decided to report the transversal evolution across the core along y and z
directions. The core was determined by locating the point where ωx was maximum to
stick with the experiments in the absence of any measurement of the second invariant
Q. This is an important limitation which should be underlined since the point where
ωx is maximum coincides with the core of the vortex only if the vortex is aligned with
the x direction, which is not always true for this validation exercise. Figures 57, 58,
59 and 60 show the transversal evolution of the second invariant Q, the longitudinal
vorticity ωx , the longitudinal velocity component U and the turbulence kinetic energy
TKE across the core of the main vortex at station S2. First of all, one notices that the
typical marked Gaussian shape of Q found in the core of the SDV DTMB5415 vortex
core is hardly visible here in most of the computations (except Marin) although more
marked in the Y than in Z directions. This trend is confirmed by Fig. 58 showing the
transversal evolution of ωx . The agreement between measured and computed ωx
and U is reasonable for most of the computations if one takes into account the
underlying uncertainties of such a comparison. The longitudinal component of the
velocity appears to be smooth as it is expected in the core of a vortex and this trend
is reasonably well represented by most of the contributors. This analysis on Q, ωx
and U applies to the other stations S4 and S7 as well (see Figs. 61, 62, 63, 64, 65, 66
and 67).
(a) Along y (b) Along z
Fig. 57 Case 1-3a—transversal evolution along y and z of Q across the core of the main vortex at
station S2
Fig. 58 Case 1-3a—transversal evolution along y and z of ωx across the core of the main vortex
at station S2
Fig. 59 Case 1-3a—transversal evolution along y and z of U across the core of the main vortex at
station S2
194 M. Visonneau
Fig. 60 Case 1-3a—transversal evolution along y and z of TKE across the core of the main vortex
at station S2
station S4
Contrary to the previous comparisons, the study of the turbulence kinetic energy
in the core of the vortex reveals unexpected trends. The experiments show a relatively
high level of TKE which is not predicted by the computations based on statistical
turbulence closures. On the contrary, the contribution from University of Rostock
using a time-resolved hybrid RANS-LES approach is the only one able to capture
the level of turbulence kinetic energy indicated by the NMRI measurements (see
Fig. 60). The level of TKE computed with this hybrid RANS-LES formulation, in
very good agreement with NMRI measurements, is three to ten times higher than
what is simulated by the isotropic or anisotropic RANSE models. This trend is fully
confirmed by the comparisons at stations S4 and S7 (see Figs. 64 and 68) where the
agreement between NMRI measurements and the hybrid RANS-LES computations
of URO appears even better.
Figures 61, 62, 63 and 64 show the transversal evolution of the second invariant
Q, the longitudinal vorticity ωx , the longitudinal velocity component U and the
turbulence kinetic energy TKE across the core of the main vortex at the second
experimental station S4.
at station S4
station S4
at station S4
196 M. Visonneau
station S7
at station S7
station S7
Figures 65, 66, 67 and 68 show the transversal evolution of the second invariant Q,
the longitudinal vorticity ωx , the longitudinal velocity component U and the turbu-
lence kinetic energy TKE across the core of the main vortex at the last experimental
station S7.
at station S7
Fig. 69 Case 1-3a—total TKE distribution at stations S2, S4 and S7—hybrid RANS-LES compu-
tations from URO
Fig. 70 Case 1-3a—ratio between resolved and total TKE distributions at stations S2, S4 and
S7—hybrid RANS-LES computations from URO
To analyze the balance between modeled and resolved TKE in URO’s hybrid
RANS-LES computations, Figs. 69 and 70 show the TKE distributions at S2, S4 and
S7 and the ratio between total and resolved TKE. It is clear that most of the turbulence
kinetic energy computed by URO comes from the unsteady resolved part, while the
modeled part is negligible in the core of the bilge vortex.
It is of course unexpected to find a relatively high level of TKE in the core of
longitudinal vortex since this high level is often associated with a high level of
turbulent dissipation if one traditionally thinks in terms of linear eddy-viscosity
198 M. Visonneau
Fig. 71 Case 1-3a—horizontal and vertical evolutions of TKE around the vortex center at station
S2 (from Visonneau et al. 2016)
turbulence models. Whether the NMRI TKE measurements are reliable or not was
debated during the workshop. NMRI, which realized the experiments, was (and still
is) skeptical about the reliability of their TKE measurements because of a seemingly
too low PIV frequency. Chapter “Experimental Data of Resistance, Sinkage, Trim,
Self-propulsion Factors, Longitudinal Wave Cut and Detailed Flow for JBC With and
Without an Energy Saving Circular Duct” of the present book addresses this topic
by comparing NMRI, OU and TUHH measurements of TKE but, up to now, no solid
conclusion can be drawn by comparing experiments which were not performed at
the same Reynolds number. To shed some light on this enigma, ECN/CNRS decided
to carry out additional unsteady post-workshop computations with a similar hybrid
RANS-LES closure based on a DES-SST turbulence model to check if similar trends
were observed independently of the flow solver. These new DES-SST computations
were performed on a grid around the complete double-body hull comprised of 66
million points locally refined in the core of the bilge vortex and a time step t =
0.006 s. Figures 71, 72 and 73 showing these new results for TKE, fully confirm the
results obtained by University of Rostock during the Tokyo2015 workshop.
(a) Time 205.536s (b) Time 205.596s
Fig. 74 Case 1-3a—instantaneous views of the longitudinal vorticity at section S2 (from Visonneau
et al. 2016)
To support this interpretation, Fig. 74 provides also two instantaneous views of the
longitudinal vorticity at section S2 separated by ten time steps i.e. 0.06s, extracted
from the DES-SST computations. These figures show that the unsteady flow predicted
by a hybrid RANS-LES closure exhibits large scale unsteadiness in the core of the
vortex associated with ring-like vortices which play a major role in the local flow
physics. These intense and strongly unsteady smaller vortical structures contribute
to the high level of TKE through the strong unsteady velocity fluctuations at the
point of measurement located in the core of the time-averaged longitudinal vortex,
making the resolved part of TKE far much important than the modelled one. More
informations could be found in (Visonneau et al., 2016).
It is believed that this is the fundamental reason which can explain concomitant
large levels of averaged turbulence kinetic energy and longitudinal vorticity. The
unsteady motion of these smaller scale vortical structures contributes to a high level
of TKE which is associated with relatively low frequency macroscopic fluctuations.
This level of temporal fluctuations is probably correctly measured by NMRI since
the frequency of this evolution is clearly lower than the experimental measurement
frequency of 6 Hz reported by NMRI. For instance, Fig. 75 showing a FFT decom-
position of TKE at point (X = −3.391 m (i.e. 0.984428 Lpp ), Y = −0.065949 m, Z
200 M. Visonneau
Fig. 75 Case 1-3a—FFT decomposition of TKE (from Visonneau et al. 2016)
Fig. 76 Case 1-3a—instantaneous view of Q invariant colored by the helicity (from Visonneau
et al. 2016)
= 0.102806 m), exhibits two peaks at 0.833 and 1.18 Hz, peaks which could have
been captured by NMRI’s experiments (6 Hz).
To understand the origins of this large scale unsteadiness, one should refer to
Fig. 76 which gives an instantaneous view of the iso-surfaces of the Q invariant
colored by the helicity. The figure clearly shows a succession of ring vortices which
are created after the onset of an open separation linked with the initial thickening of
the boundary layer. This large scale unsteadiness is likely to be due to the peculiar
design of JBC (C B = 0.858). The rapid reduction of the hull sections at the stern,
reflected by the high value of C B , creates the condition of an open separation followed
by a flow reversal and a strong unsteadiness revealed by the shedding of ring vortices.
In chapter “Experimental Data of Resistance, Sinkage, Trim, Self-propulsion Fac-

tors, Longitudinal Wave Cut and Detailed Flow for JBC With and Without an Energy
Saving Circular Duct”, several additional experiments are compared in order to assess
the existence of this high level of turbulence kinetic energy in the core of this lon-
gitudinal vortex. Experiments conducted on the same hull by Osaka University and
TUHH (in a wind tunnel) can be compared to check if similar observations are made
concerning the value reached by TKE in the core of the bilge vortex. Unfortunately,
these measurements are not made at the same Reynolds number. Figure 77 taken from
chapter “Experimental Data of Resistance, Sinkage, Trim, Self-propulsion Factors,
Longitudinal Wave Cut and Detailed Flow for JBC With and Without an Energy
Saving Circular Duct” shows the distribution of the squared root of TKE normalized
by the friction velocity extracted along the same horizontal line at the shaft height at
the same cross-section S4 for the three experiments mentioned above (see Fig. 78 for
the location of the line of extraction). Moreover, data coming from the measurements
on the KVLCC2 (Lee et al. 2003) approximately at the propeller plane are added to
see if one can observe a generic behaviour. Figure 78 shows the location of√the line
of extraction (in red) for each experiment. The distributions are plotted as utkeτ
ver-
+
sus y . Also plotted is the same scaled data from the propeller plane of KVLCC2.
The data is taken from the wind tunnel hot-wire measurements (Lee et al. 2003)
where Reynolds number is Re = 4.6106 . The wall distance y is measured from the
shaft edge and the friction velocity u τ is estimated from the local c f of a turbulent
boundary layer of a flat plate using the formula of Prandtl-Schlichting as below:
c f = (2.0log10 (Rex ) − 0.65)−2.3 (1)

with u τ /U = c f /2.
One clearly observes an increase of the peak value of normalised TKE when
the Reynolds number increases. This trend is correctly captured by a posteriori
computations performed on the JBC by ECN/CNRS with the ISIS-CFD code (see
Fig. 79). This figure shows a comparison between JBC experiments and two compu-
tations based either on the anisotropic EARSM turbulence closure or on the hybrid
RANS/LES DES-SST turbulence model. Although the location of the peak is slightly
shifted outwards compared to the experiments, it is worthwhile to notice that only
the unsteady DES-SST closure is able to capture the right peak level of TKE, thanks
to the resolved large scale unsteadiness present at this location, while the statistical
EARSM model fails to capture this local peak of TKE. Figure 80 shows the isowake
contours at the same S4 section used to generate the previous figures, both for NMRI
experiments and ECN/CNRS computations using EARSM and DES-SST turbulence
models. Black color is used for the NMRI experiments while blue (resp. red) color is
used for the EARSM (resp. DES-SST) simulations. First of all, one can notice that the
line of extraction does not pass through the center of the vortex for the computations,
a location where the peak value of TKE is reached in NMRI experiments. More-
over, the agreement between NMRI measurements and both simulations is good,
202 M. Visonneau
Fig. 77 Case 1-3a—normalized transversal distribution of the turbulent kinetic energy for various
experiments (JBC and KVLCC2)
Fig. 78 Case
1-3a—cross-sections used to
extract the TKE transversal
distribution for NMRI,
TUHH, OU and KVLCC2
experiments
Fig. 79 Case 1-3a—normalized transversal distribution of the turbulent kinetic energy for various
experiments (JBC and KVLCC2)—comparisons with ECN/CNRS-ISIS-CFD computations for the
JBC test case using EARSM and DES-SST
(a) NMRI experiments vs EARSM simu- (b) NMRI experiments vs DES-SST sim- (c) EARSM vs DES-SST simulations
lation ulation
Fig. 80 Case 1-3a—station S4—isowakes from NMRI experiments (black) and ECN/CNRS
computations—EARSM (blue) and DES-SST (red)
which means that the two turbulence closures considered here provide a satisfactory
evaluation of the velocity field. Figure 81 shows the distribution of TKE measured
by NMRI with labels identifying the maximum and average values at section S4.
One can also observe that the maximum value of TKE (0.05) is reached in small
islands in the NMRI experiments while the first well-defined and regular iso-level
surrounding these irregular islands is 0.03. Then, Fig. 82 shows the distribution of
turbulence kinetic energy normalised by the square of the ship velocity at the same S4
section with the same conventions of color, as described previously. The maximum
normalised TKE level is 0.05 while the minimum level is 0.01. Five isolevels are
drawn. As observed previously in the transversal extraction, the TKE distribution
depends strongly on the turbulence closures. The maximum TKE value reached by
EARSM is 0.027 while it is 0.050 for DES-SST like in the NMRI experiments.
204 M. Visonneau
Fig. 81 Case 1-3a—station S4—isowakes from from NMRI experiments
(a) NMRI experiments vs EARSM simu- (b) NMRI experiments vs DES-SST sim- (c) EARSM vs DES-SST simulations
lation ulation
Fig. 82 Case 1-3a—station S4—isowakes from from NMRI experiments (black) and ECN/CNRS
computations—EARSM (blue) and DES-SST (red)
3.8.6 Temporary Conclusions on the Core Vortex Analysis
This first analysis of the flow in the core of the main bilge vortex was surprisingly
interesting, despite the known limitations due to the absence of 3D local PIV mea-
surements. First of all, it appears necessary to use locally refined grid in the center of
the bilge vortex to capture more accurately the flow physics. The grids employed in
the computations were probably locally too coarse to capture accurately the detailed
physics. Despite these limitations, the agreement between most of the computations
and the rather coarse experiments was satisfactory in terms of global trends. The
longitudinal vorticity was somewhat underestimated and the longitudinal velocity
distribution was fairly reproduced. These experiments pointed out very large differ-
ences on the turbulence kinetic energy and this was the most striking (and unexpected)
result of this comparison. The linear or non-linear statistical models were unable to
reproduce this result, providing TKE results which were three to ten times smaller
than what was measured. Only the unsteady hybrid RANS-LES model from Uni-
versity of Rostock produced results in remarkably good agreement with the NMRI
measurements, thanks to the major contribution of the resolved turbulence kinetic
energy. Whether the NMRI TKE measurements are reliable or not is still a matter
of intense debate but the remarkable agreement shown by the blind URO hybrid
RANS-LES computations (confirmed one year later by ECN/CNRS computations)
and the typical frequency of occurence of the large scale structures contained in the
core of the bilge vortex, seem to indicate that NMRI’s measurements correctly repre-
sented the local large scale unsteadiness associated with the onset and progression of
a vortex along a hull with such a large block-coefficient hull. The additional experi-
ments provided by Osaka University and Technical University of Hamburg-Harburg
compared with NMRI results seem to plead in favor of a Reynolds number influence
on the peak value of TKE. However, further much more detailed measurements will
be necessary to draw safer conclusions.
3.9 Case 1-3a —Concluding Remarks
On the basis of this first analysis based on some representative examples taken from
the numerous contributions, one can draw the following conclusions:
• With the typical grids which are used by the contributors (5–10 M points), the
influence of the grid discretisation is moderate for RANSE, which means that the
numerical error appears to be under control.
• As noticed during the Gothenburg2010 workshop for the analysis of the local
flow around the KVLCC2, the major influence comes from the turbulence closure.
Linear isotropic closures significantly under-predict the longitudinal vorticity at
S2 while full RSM closures slightly over-predict it at the same station.
• Non-linear anisotropic closures offer a good compromise from the standpoint of
the local flow although they slightly underpredict the vorticity at S2.
• New results from hybrid RANS-LES are promising but IDDES seems to over-
predict the vorticity again. The agreement with the isowakes distribution is worse
than what is achieved with the best RANS statistical turbulence closures.
• Computational codes seem to be mature since one does not observe large discrep-
ancies when the same turbulence closures on similar grids are used.
• The influence of the free-surface on the local flow is not negligible but not large.
• The influence of the wall boundary treatment on the development of the stern flow
is weak.
This first analysis of the flow in the core of the main bilge vortex was surprisingly
interesting, despite the known limitations due to the absence of 3D local PIV mea-
surements. First of all, it appears necessary to use locally refined grid in the center of
206 M. Visonneau
the bilge vortex to capture more accurately the flow physics. The grids employed in
the computations were probably locally too coarse to capture accurately the detailed
physics. Despite these limitations, the agreement between most of the computations
and the rather coarse experiments was satisfactory in terms of global trends. The
longitudinal vorticity was somewhat underestimated and the longitudinal velocity
distribution was fairly reproduced. These experiments pointed out very large differ-
ences on the turbulence kinetic energy and this was the most striking (and unexpected)
result of this comparison. The linear or non-linear statistical models were unable to
reproduce this result, providing TKE results which were three to ten times smaller
than what was measured. Only the unsteady hybrid RANS-LES model from Uni-
versity of Rostock produced results in remarkably good agreement with the NMRI
measurements, thanks to the major contribution of the resolved turbulence kinetic
energy. Whether the NMRI TKE measurements are reliable or not is still a matter
of intense debate but the remarkable agreement shown by the blind URO hybrid
RANS-LES computations (confirmed one year later by ECN/CNRS computations)
and the typical frequency of occurence of the large scale structures contained in the
core of the bilge vortex, seems to indicate that NMRI’s measurements correctly rep-
resented the local large scale unsteadiness associated with the onset and progression
of a vortex along a hull with such a large block-coefficient. However, further much
more detailed measurements will be necessary to draw safer conclusions. The reader
can refer to chapter “Experimental Data of Resistance, Sinkage, Trim, Self-propul-
sion Factors, Longitudinal Wave Cut and Detailed Flow for JBC With and Without
an Energy Saving Circular Duct” for a more detailed discussion and comparison
between NMRI experiments and two other experiments conducted later on in a tow-
ing tank on a smaller model at Osaka University and in a wind tunnel on a double
body JBC by Technical University of Hamburg-Harburg (TUHH).
4 Test Case 1-3b
4.1 Description of Experimental Results
Table 4 provides a comparison between the two experimental conditions, namely

cases 1-3a and 1-3b. The wind tunnel experiments are conducted on a double body
while the free-surface is taken into account in the towing tank. But the Froude number
is low and one does not think that the free-surface deformation influences the flow at
the experimental locations. More significant is the difference in terms of Reynolds
number, the TUUH experiments having been performed at a Re which is 2.72 times
smaller.
Figure 83 shows the locations of the experimental stations and a global view of
the longitudinal vortex developing in the stern part of the ship, represented by an iso
Q surface colored by the normalized helicity.
Table 4 Case 1-3b—flow conditions for the towing tank experiments (NMRI) and wind-tunnel
tests (TUHH)
NMRI TUHH
Type Towing tank (calm water) Wind tunnel
Model Single body Double body
Scale 1:40 1:80
Lpp (m) 7.00 3.513
V0 (m/s) 1.179 11.8
Re 7.46 106 2.74 106
Fr 0.142 –
ρ (kg/m3 ) 998.2 1.2
ν (m2 /s) 1.107 106 1.5 105
(a) Hull (b) Flow
Fig. 83 Case 1-3b—view of the locations of the three experimental stations S2, S4 and S7 and of
the main longitudinal vortex visualized with the second invariant iso-surface
To understand the difference between NMRI and TUHH experiments, Figs. 84, 85
and 86 show the experimental isowakes measured by NMRI and TUHH at stations
S2, S4 and S7. Globally, the same averaged intense bilge vortex is visible in both
experiments. However, the experiments conducted at TUHH reveal a flow charac-
terised by a slightly less intense longitudinal vortex since the isowakes 0.3 and 0.2
are not visible at station S2 in TUHH measurements. Similarly, at station S7, the
isowake 0.4 is way less developed in TUHH than in NMRI measurements.
Four participants provided results for the test case 1-3b. The main characteristics of
their contributions are listed in Table 5 in terms of turbulence closures, wall treatments
and discretization. ECN/CNRS provided two different contributions with and without
the automatic grid refinement functionality while FOI showed a comparison between
RANS and LES with different wall treatments. Finally, SRC provided a reference
wall-resolved LES computation on a grid composed of 4.9G cells.
208 M. Visonneau
(a) NMRI (b) TUHH
Fig. 84 Case 1-3b—station S2—case without duct—experimental isowakes from NMRI and
TUHH
(a) NMRI (b) TUHH
TUHH
4.3 Grid Sensitivity
As for the case 1-3a, the grids used by the three participants can be classified into
four categories, based on their sizes:
• Not So Fine: Ncell < 2M cells (None)
• Fine: 2M cells < Ncell < 10M cells (ECN without AGR)
• Very Fine: 10M cells < Ncell < 50M cells (ECN with AGR)
• Tremendously Fine: Ncell > 50M cells (FOI (150M cells), SRC (4.9G cells !!!))
(a) NMRI (b) TUHH
TUHH
Table 5 Case 1-3b—main characteristics of the computations

Organization code Turbulence model Wall model Grid characteristics
name
ECN-CNRS SST/EARSM Wall function Unstructured grid
ISIS-CFD aka Wall resolved 5.7/9.2M cells
FINET M /Marine
FOI LES RANSE/LES Wall modeled Unstructured grid
OpenFOAM Wall resolved 143M cells
SRC LES Wall resolved Unstructured grid
FrontFlow/blu 4.9B cells
First of all, one can start to study the grid sensitivity by comparing the results provided
by ECN/CNRS with or without local grid adaptation. Since the same code and same
turbulence model is used, such an internal comparison should provide unbiased
informations on the grid sensitivity which are therefore easier to interprete. The
automatic grid refinement (AGR) used in these computations is based on the Hessian
of the convective fluxes (see Wackers et al. 2014), which increases significantly the
grid density in the regions of high shear stress. Figures 87, 88 and 89 show the
isowake distribution at stations S2 to S7 with and without activating the automatic
grid refinement. With AGR, at station S2, the intensity of the longitudinal is clearly
increased as it is visible through the addition of the closed isowake 0.2 in the core of
the averaged bilge vortex. The same trend is observed at station S4 with the addition
of an isowake 0.1 when AGR is activated. It is also worthwhile to notice that the flow
close to the bottom part of the vertical plane of symmetry is significantly modified
by the addition of grid points.
210 M. Visonneau
(a) Without AGR (b) With AGR
Fig. 87 Case 1-3b—grid influence—station S2

Figures 90, 91 and 92 show the distribution of the secondary velocity components
(v and w) with and without automatic grid refinement. One can indirectly sees the
regions with high shear-stress where points are automatically added.
212 M. Visonneau
4.4 Wall Treatment
Figures 93, 94 and 95 show a comparison between a wall-resolved and a wall-

modelled RANS (k-ω SST) solution performed by FOI. Some minor differences
are visible but one can not conclude that the wall treatment has a crucial influence on
the computed isowakes. One might notice a slight asymmetry in the results indicating
that the computations are not fully converged, probably because of the existence of
large scale unsteadiness near the stern of the ship.
Fig. 93 Case 1-3b—station S2—FOI RANS wall-modeled and wall-resolved solutions
4.5 Turbulence Closures
For Case-1.3b, the turbulence models can be organized into three groups:
• the isotropic linear closures: FOI (k-ω SST),
• the anisotropic non-linear models: ECN-CNRS (EARSM),
• the LES models: FOI (NWM-LES), SRC (LES).
The next subsections will be devoted to a comparison of these various solutions
and turbulence models mainly by assessing two different flow characteristics, the
distributions of isowakes and turbulence kinetic energy at the three experimental
stations S2, S4 and S7.
214 M. Visonneau
4.5.1 Assessment of Turbulence Closures
First of all, results provided by FOI using the code Open FOAM are considered.
The turbulence model is the linear isotropic k-ω SST closure equipped with a low
Re near-wall formulation. Figures 96, 97 and 98 show comparisons on the isowakes
between TUHH experiments and the wall-resolved FOI RANS solution. This solution
appears similar to the one obtained by ECN/CNRS (see Figs. 99, 100 and 101) with
EARSM and local grid adaptation. No closed contours are observed in the core of
the vortex at station S2 and at station S4, U = 0.3 is the lowest iso-contour which is
captured in the core while ECN/CNRS captures U = 0.2. A similar trend is observed
at station S7. Based on the TUHH measurements, it is not possible to draw any firm
conclusion about the best agreement in the core of this vortex. Despite the use of a
locally finer grid generated by AGR for ECN/CNRS, it is believed that the difference
between FOI and ECN/CNRS RANS simulations is to be mainly attributed to the
use of the non-linear anisotropic EARSM turbulence closure by the French team, a
closure which is known to lead to more intense longitudinal vortices.
Interestingly, several teams proposed results based on various LES turbulence
models. FOI provided wall-modeled LES computations on a grid composed of 150M
cells while SRC proposed a remarkable wall-resolved LES simulation computed on a
grid made of 4.9 G cells. Figures 102, 103, 104, 105, 106 and 107 show comparisons
on the iso-wake contours between these two results at stations S2 to S7. First of all,
one may notice that both computations predict a more intense longitudinal vortex, as
indicated by the lower values of the isowake in the core. At S2, the vortex predicted
by FOI (with a closed contour U = 0.1 in the core) appears way too large compared
to the measurements and the SRC LES result. It is impossible to know whether this
might be attributed to a mesh which is too coarse, too large a time step, too small
an averaging period or to the influence of the wall model. Reassuringly, the wall-
Fig. 96 Case 1-3b—station S2—results from FOI-Open Foam

216 M. Visonneau
Fig. 99 Case 1-3b—station S2—results from ECN-ISIS-CFD


218 M. Visonneau
(a) Experiments (b) LES
Fig. 102 Case 1-3b—station S2—results from FOI-OpenFoam
resolved LES SRC simulation appears in good agreement with TUHH measurements
and other RANS computations at station S2. The same remarks hold for the stations
S4 and S7, although LES SRC results provide a vortex which appears more intense
than what is measured at station S7 as indicated by the comparison on the isowake U
= 0.4. However, the solution provided by the wall resolved LES SRC computation
appears to agree globally very well with both TUHH measurements and RANS
computations. A comparison limited to the iso-U contours does not do justice to
the LES simulations since this flow around the JBC is not characterised by a strong
unsteady detached flow. However, the SRC simulation should provide a kind of
reference prediction of the Reynolds stresses and turbulence kinetic energy far much
accurate than the statistical turbulence models and the current experiments.
4.5.2 Focus on the Turbulence Kinetic Energy in the Core of the Main
Longitudinal Vortex
In the case 1-3a, the local core vortex analysis revealed large differences concerning
the prediction of the turbulence kinetic energy between the various teams and consis-
tently, between (U)RANS and hybrid RANS-LES turbulence closures. In particular,
hybrid RANS-LES models computed significantly higher levels of TKE in the core
of the main vortex, due to the contribution of the resolved velocity fluctuations. The
validity of the NMRI measurements was also questioned because of a too low fre-
quency of acquisition, although the resolved velocity fluctuations captured by every
hybrid RANS-LES models led to values of TKE in good agreement with NMRI
measurements.
Fig. 103 Case 1-3b—station S2—results from SRC
Although the Reynolds number is 2.72 times smaller for the case 1-3b, it is
interesting to examine the levels of TKE predicted by the various simulations, and
particularly, by the wall resolved LES simulation provided by SRC.
Figure 108 shows a comparison between the TKE contours computed by a statis-
tical closure (here EARSM) and the wall-resolved LES at station S2. Globally, the
iso-contour maps look similar but the extension of the region where TKE is not neg-
ligible is larger with LES than with EARSM. This is the first noticeable difference. If
one looks at the maximum values of TKE, one sees that they occur for both closures
220 M. Visonneau
(a) Experiments (b) NWM-LES

(a) ECN-CNRS/EARSM (b) SRC/LES
Fig. 108 Case 1-3b—TKE contours—station S2
at the bottom of the region a bit away from the vertical plane of symmetry. Here,
the maximum value predicted by LES is around 0.022 while it is around 0.018 for
EARSM. Moreover, in the region where the center of the vortex should be located,
the values of TKE predicted by LES are significantly higher than what is predicted
by EARSM.
Figure 109 shows a comparison between the TKE contours computed by a statis-
tical closure (here EARSM) and the wall-resolved LES at station S4. The differences
on the TKE contours which were already visible at station S2 are more marked at
station S4. The region with high TKE is much more developed for LES than for
RANS closures with maximum values ranging from 0.03 for LES to 0.02 for RANS.
222 M. Visonneau

Here again, in the core of the vortex, the turbulence kinetic energy computed by LES
is significantly higher than what is modeled by the EARSM RANSE closure.
Finally, Fig. 110 shows a comparison between the TKE contours computed by
a statistical closure (here EARSM) and the wall-resolved LES at station S7. The
analysis made for station S4 can be repeated for station S7 without any alteration. In
the core of the vortex, TKE predicted by LES appears to be about two times higher
than what is modeled by EARSM.
Globally, the comparisons of the TKE maps at stations S2 to S7 computed by
LES and RANSE closures confirm what was observed in the local vortex analysis
conducted for the case 1-3a. The levels of turbulence kinetic energy simulated by
LES are significantly higher than what is modeled by RANS. This suggests that,
even if this flow does not exhibit any strong coherent structures as it is found for
separation around bluff bodies, there is a significant amount of large scale velocity
fluctuation which is resolved by LES in the core of the vortex and is maintained
during its progression in the wake, that is not modelled by the RANSE closures. It is
believed that this is a major flow characteristic which is associated to the large scale
unsteadiness of the flow in the stern region of JBC.
4.6 Case 1-3b—Concluding Remarks
On the basis of the analysis presented above, a few conclusions can be drawn:
• A sensitivity to the local grid density is observed, which contradicts somewhat the
conclusions drawn for the previous case 1-3a,
• However, the major influence comes from the turbulence closure, as expected.
Linear isotropic closures under-predict the longitudinal vorticity at S2 while full
RSM closures tend to over-predict it at the same station.
• For the first time, two LES computations are presented. The (almost) wall-resolved
LES results from SRC predict a level of TKE in the core of the bilge vortex
in agreement with NMRI’s measurements and higher than what is modeled by
EARSM. Results from NWM-LES are promising but this model seems to over-
predict the vorticity. It would have been interesting to check TKE in the bilge
vortex but this information was not available at the time of the workshop.
5 Test Case 1-4
5.1 Description of Experimental Results—Influence of the

Duct
Test case 1-4 concerns the local flow analysis around the JBC with duct but without
propeller towed in the NMRI towing tank, in free trim and sinkage conditions. The
location of the experimental stations with respect to the duct are shown in Figs. 111
and 112 shows the stern equipped with its duct. Figures 113, 114, 115, 116, 117 and
118 show the influence of the duct on the experimental iso-wakes and iso-contours
of the two crossflow velocity components at three sections S2, S4 and S7 already
defined before.
At station S2, no significant influence of the duct can be noticed since the slight
differences observed on the isowakes close to hull might be attributed to measurement
uncertainties. This indicates that the upwind influence of the duct is confined to a
close neighborhood since no modification is visible on the flow velocity distribution
at station S2. Station S4 is located in the close wake of the duct and consequently,
224 M. Visonneau
Fig. 111 Case 1-4—appended hull geometry and experimental stations
Fig. 112 Case 1-4—view of the stern equipped with its duct
(a) Without duct (b) With duct
Fig. 113 Case 1-4—experiments—influence of the duct on the isowakes distribution at station S2
Fig. 114 Case 1-4—experiments—influence of the duct on the secondary velocities distribution at
station S2
one can notice large differences on the flow velocity distribution due to the presence
of the duct. First of all, the duct does not remove the longitudinal vortex as indicated
by the hook shape of the isowakes behind the duct. The location of the core of the
longitudinal vortex is moved down and closer to the vertical plane of symmetry but
the vortical intensity does not seem to have been significantly modified by the duct, if
one looks at the distortion of the isowakes. Correlated with the vortex translation, the
flow appears slightly accelerated in the upper part of duct’s wake where the isowake
U = 0.4 is replaced by U = 0.5. In the bottom part of the duct’s wake, the flow
appears decelerated with an extended region characterized by the new isowakes U
= 0.1 and U = 0.0, which means that one can not exclude the existence of a local
226 M. Visonneau
station S4
flow separation in the core of this vortex. Likewise, one can notice the existence
of closed contour U = 0.4 in the center of the duct’s wake, close to the vertical
plane of symmetry. This peculiar flow region is confirmed by the distribution of the
secondary velocity components. The plots at station S7 confirm the flow topology
detected at station S4. Instead of a unique distorted U = 0.4 isowake, one can notice
one additional iso-contour U = 0.3, which indicates that the flow in the core of the
main vortex is slightly decelerated at station S7 when the duct is mounted on the
hull. Accordingly, the vortex is slightly shifted down in presence of a duct.
station S7
Seventeen participants provided results for the test case 1-4. The main characteris-
tics of their contributions are listed in Table 6 in terms of turbulence closures, wall
treatments and discretizations. Some participants provided several different contri-
butions, which makes possible a comparison of various turbulence closures or wall
treatments for the same grid and same code. Moreover, SOTON performed a compar-
ison of codes (OpenFoam vs. Star-CCM+) with the same grid and same turbulence
closure.
228 M. Visonneau
Table 6 Case 1-4—main characteristics of the computations

Organization code name Turbulence model Wall model Grid characteristics
ABS k-ω SST Wall function Unstructured grid
Chalmers-Flowtech EARSM Wall resolved Structured grid
12M cells
CNR-INSEAN Spalart-Allmaras Wall resolved Structured grid
Xnavis with overlapping
8.6M cells
ECN/CNRS SST/EARSM Wall function Unstructured grid
ISIS-CFD aka FINET M /Marine Wall resolved 8.7/14.1M cells
HHI Realizable Wall function Unstructured grid
HiFOAM/StarCCM+ k-/RSTM 1.4M cells
HSVA 2 equations Wall function Unstructured grid
FreSCo+ from 2.1 to 14.6M
cells
KRISO Realizable Wall function Structured grid
MARIC k-ω SST Wall function Unstructured grid
FINET M /Marine 650 000 cells
MARIN k-ω SST Wall resolved Unstructured grid
ReFRESCO +DM correction 6.67M cells
MHI k- RNG Wall resolved Unstructured grid
Fluentv14 k-ω SST 14.4M cells
MIJAC k-ω SST Wall resolved Unstructured grid
OpenFOAM 8.5M cells
NMRI EARSM Wall resolved Structured grid
Nagisa 10 056 704 cells
PNU 2 equations Wall resolved Unstructured grid
SHIME k-ω SST Wall resolved Unstructured grid
SJTU k-ω SST Wall function Unstructured grid
naoeOpenFOAM 2.25M cells
SOTON k-ω SST Wall resolved Structured grid
OpenFOAM/StarCCM+ 21.3M cells
YNU EARSM Wall resolved Structured grid
SURFv7 3 267 584 cells
Among the seventeen participants, eleven used unstructured grids while six com-
puted on structured grids sometimes locally enriched with overlapping patches. It is
well known that the prediction of the nominal velocity in the stern region depends
strongly on the turbulence closure, as soon as a reasonable dense grid is used. The
turbulence models employed by the participants computing this test case can be
classified into two groups: (i) anisotropic non-linear statistical turbulence modeling:
Reynolds stress transport models (HHI), Explicit Algebraic Reynolds Stress Models
(Chalmers, ECN/CNRS, NMRI, YNU), (ii) isotropic linear eddy-viscosity model
(ECN/CNRS, HSVA, MARIN, URO, …). Concerning the linear isotropic turbu-
lence models, most of the contributors use the k-ω SST model but HHI, MHI and
KRISO use also the realizable k-ε closure while CNR-INSEAN and Marin prefer
to use a one equation model (Spalart-Allmaras or 1 eq. Menter with Dacles-Mariani
modification). Therefore, in five years, we can observe a slight increase of more
sophisticated turbulence closures used for ship flow computations, probably related
with the number of publications illustrating the crucial role of turbulence modeling
for the prediction of local flow in the stern region. All the discretization methods are
formally second-order accurate and based on multi-block structured or unstructured
grids which are all body-fitted, with the exception of Chalmers and CNR-INSEAN
which are using an overset approach combining structured grids to reach an optimal
grid density in the stern region.
5.3 Influence of the Turbulence Closures
5.3.1 Linear Isotropic Turbulence Closures
Starting the analysis of turbulence closures by the one equation models, Figs. 119, 120
and 121 provide the results with the Spalart-Allmaras model. Clearly, the longitudinal
vorticity is strongly under-estimated as illustrated by the open iso-contour u = 0.2
and the wrong shapes of u = 0.3 and u = 0.4 at station S2. These trends are confirmed
at stations S4 and S7. The results provided by MARIN which employs also a k-ω SST
model with Dacles-Mariani correction, are very different as illustrated by Fig. 122 for
station S2. Here, the longitudinal vortex is far more intense, even if the iso-contour
u = 0.2 appears to be too extended, which indicates an over-estimated longitudinal
vorticity. The same trend is recognizable at station S4 (resp. S7) where the iso-contour
u = 0 and the reversed flow region (resp. u = 0.3) appear also significantly over-
estimated (see Figs. 123 and 124). It is interesting to notice that these two results
obtained with a one equation model do not show at all the same trend, which perhaps
indicates that the Dacles-Mariani correction plays a crucial role here.
To illustrate the behaviour of two equation linear isotropic models, the results
from HSVA using a linear EARSM model and SHIME using a more classical k-ω
SST closure have been retained to illustrate the general trend observed with these tur-
bulence closures. Globally, the intensity of the longitudinal vortex is under-estimated
as indicated by the iso-contours in the core of this vortex at station S2 (see Figs. 125
230 M. Visonneau
(a) Experiments (b) INSEAN/Spalart-Allmaras
Fig. 119 Case 1-4—linear isotropic turbulence closures—station S2
and 128). At station S4 just behind the duct, the results from HSVA are in very good
agreement with the experiments while those provided by SHIME show a notice-
able asymmetry and a deteriorated agreement with the measurements as shown by
Figs. 126 and 129. It is interesting to notice that the same asymmetry is present at
the same station in ABS and SJTU results using the same code (OpenFoam) and the
same turbulence model. Most the computations based on k-ω SST are able to predict
reasonably well the flow at station S4 (Figs. 127 and 130).
(a) Experiments (b) MARIN/k-ω SST + DM correction
Fig. 122 Case 1-4—MARIN—k-ω SST + DM correction—station S2
5.3.2 Non-linear Anisotropic Turbulence Closures
Reynolds Stress Transport and non-linear anisotropic EARSM models are used by
HHI, ECN/CNRS, Chalmers, NMRI and YNU. All these models exhibit very similar
trends which are now summarized. At station S2, they are the only models able to
predict, without any ad-hoc corrections, the right shape of the iso-contour u = 0.2 with
a trend towards a slight over-estimation for the RSTM closure shown in Fig. 140.
EARSM models perform reasonably well as illustrated by Chalmers results (see
Fig. 131). However, as confirmed by ECN/CNRS (not shown here), the location
of the core of the vortex is located slightly further from the hull, compared to the
232 M. Visonneau
experiments, which indicates that the computed vortex is likely to be thicker than
the measured one. At station S4, the distorsion of the outer iso-contours (u = 0.4
and 0.5) is better represented by the non-linear anisotropic turbulence closures but
there is a common trend to overestimate the size of the reversed flow region in the
center of the vortex (see for instance EARSM (Fig. 132) or RSTM (Fig. 141) based
computations (Figs. 133, 134, 135, 136, 137, 138, 139 and 142).
(a) Experiments (b) HSVA/Linear EARSM
Fig. 125 Case 1-4—linear EARSM from HSVA—station S2
What is clear one more time is that the turbulence closure plays a central role
in the prediction of the vortex core flow physics as indicated by Figs. 143, 144 and
145 which compare on the same grid the k-ω SST and EARSM turbulence models
implemented in the ECN/CNRS code.
234 M. Visonneau
(a) Experiments (b) SHIME/k-ω SST
Fig. 128 Case 1-4—SHIME—k-ω SST—station S2
5.4 Case 1-4a—Concluding Remarks
Based on the comparisons performed during this workshop, one can say that the
turbulence models have a critical role to play once the grid is fine enough. The
main modifications of the flow created by the presence of the duct are correctly
reproduced by linear isotropic models which provide the correct qualitative trend
with the remarkable exception of the linear model implemented by HSVA in Fresco+.
To get a more quantitative agreement, moving from isotropy to anisotropy appears

necessary although the inner part of the vortex is not perfectly reproduced as shown
by the occurence of a reversed flow region which is over-estimated by both RSTM
and EARSM turbulence closures.
236 M. Visonneau
(a) Experiments (b) CHALMERS/EARSM
Fig. 131 Case 1-4—CHALMERS—EARSM turbulence closure—station S2
6 Test Case 1-7
6.1 Description of the Experiments
Test case 1-7 concerns the local flow analysis around the JBC without duct but with
propeller towed in the NMRI towing tank, in free trim and sinkage conditions. The
location of the experimental stations are indicated in Figs. 146 and 147 shows the
stern with its five blades propeller.
(a) Experiments (b) ECN/CNRS - EARSM
Fig. 134 Case 1-4—ECN/CNRS—EARSM turbulence closure—station S2
Figures 148, 149, 150, 151, 152 and 153 show the influence of the propeller
on the experimental iso-wakes and averaged crossflow velocity components at three
sections S2, S4 and S7 already defined before. At station S2, the propeller has already
an influence on the flow as shown by Fig. 148. With propeller, the isowakes u = 0.3
and u = 0.4 are no more visible, which means that the core of the longitudinal vortex
is accelerated, probably due to the succion effect associated with the presence of
the propeller. Interestingly, the iso-contour u = 0.5 has evolved from a distorted
hook shape (without propeller) to a closed contour in the vicinity of the hull (with
propeller), which indicates also a significant acceleration of the flow close to the hull.
At station S4, the experiments are noisy and hard to interprete (see Fig. 150). One can
238 M. Visonneau
at least observe the strong succion effect illustrated by the respective locations of the
isowakes u = 0.7 and 0.8. At station S7 (see Fig. 152), the influence of the propeller
is way more obvious and easier to analyze. One notices two circular isowakes (u
= 0.9 and 1.0) bounding an inner region in the wake of the propeller. This wake is
asymmetric and characterized by the presence of large zone u = 1.2 with a shape of
Moon’s crescent mainly located on the right part of the wake but also present in the
left part. On the left side, one can see isocontours with high levels (u = 1.3) while on
the right side, one notices the existence of lower level iso-contours (u = 1.1 to 0.7).
(a) Experiments (b) KRISO - EARSM
Fig. 137 Case 1-4—KRISO—EARSM turbulence closure—station S2
Finally, phase-averaged isowakes and secondary velocities from station S4 to

S7 for 0◦ , 24◦ and 48◦ were also measured by NMRI. At station S4, these results
are difficult to exploit because of the absence of measurements in the main part of
the wake. This is the reason why we will restrict the analysis to the isowakes and
secondary velocities measured at station S7. These experimental results are shown in
Figs. 154 and 155. The main interest of these phase-averaged experiments is to show
the structure of the wake flow behind the rotating propeller for several positions of its
five blades. In particular, we expect to be able to detect the phase-averaged position of
the tip vortices which are shed behind each rotating blade. The locations of these five
tip vortices should be indicated by the presence of several closed isowakes (one per
240 M. Visonneau
(a) Experiments (b) HHI - RSM
Fig. 140 Case 1-4—HHI—RSM turbulence closure—station S2
blade) associated to a dimensionless longitudinal velocity higher than one. However,

one must remember that these tip vortices are progressing around an helicoidal path,
which means that they are never strictly orthogonal to the plane of measurements.
This means that the measured Cartesian secondary velocity components (v, w) will
probably not bring any significant information regarding the local rotational motion
inside each tip vortex since the plane of measurement is not orthogonal to the local
axis of this vortex.
Eighteen participants provided results for the test case 1-7. The main characteristics
of their contributions are listed in Table 7 in terms of turbulence closures, wall treat-
ments, discretization and propeller treatment. Some participants provided several
different contributions, which makes possible a comparison of various turbulence
closures or wall treatments for the same grid and same code. Moreover, SOTON
performed a comparison of codes (OpenFoam vs. Star-CCM+) with the same grid
and same turbulence closure, as previously.
242 M. Visonneau

Organization Code name Turbulence Wall model Grid Propeller
model characteristics treatment
ABS k-ω SST Wall function Unstructured grid Actual propeller
Chalmers-Flowtech EARSM Wall resolved Structured grid Lifting line
18.4M cells
CNR-INSEAN Spalart- Wall resolved Structured grid Body-force/actual
Xnavis Allmaras with overlapping propeller
10.6M cells
ECN/CNRS SST/EARSM Wall function Unstructured grid Body-
ISIS-CFD aka Wall resolved 18.7M cells force/Actual
FINET M /Marine propeller
HHI Realizable Wall function Unstructured grid Moving reference
HiFOAM/StarCCM+ k-/RSTM 3.5M cells frame
HSVA 2 equations Wall function Unstructured grid Vortex lattice
FreSCo+ 14.6M cells
IIHR DDES Wall resolved Unstructured grid Actual propeller
REX 67.7M cells
KRISO Realizable Wall function Structured grid Lifting surface
MARIC k-ω SST Wall function Unstructured grid Body force
FINET M /Marine 2M cells
MARIN k-ω SST Wall resolved Unstructured grid Actual propeller
ReFRESCO + DM 17M cells
correction
MIJAC k-ω SST Wall resolved Unstructured grid Simplified
OpenFOAM 8.7M cells propeller theory
NMRI EARSM Wall resolved Structured grid Simplified
Nagisa 7.8M cells propeller theory
PNU 2 equations Wall resolved Unstructured grid Actual propeller
SHIME k-ω SST Wall resolved Unstructured grid N/A
SJTU k-ω SST Wall function Unstructured grid Simplified
naoeOpenFOAM 4.7M cells propeller theory
SOTON k-ω SST Wall resolved Structured grid Body force
OpenFOAM/StarCCM+ 19.4M cells prescribed
UniRostock Hybrid LES Wall resolved Structured grid Actual propeller
OpenFOAM k-ω SST 22.8M cells
YNU EARSM Wall resolved Structured grid Simplified
SURFv7 2.5M cells propeller theory
(a) ECN/CNRS - k-ω SST (b) ECN/CNRS - EARSM
Fig. 143 Case 1-4—ECN/CNRS—influence of turbulence anisotropy—station S2
Among the eighteen participants, twelve used unstructured grids while six com-
puted on structured grids sometimes locally enriched with overlapping patches. It is
well known that the prediction of the nominal velocity in the stern region depends
strongly on the turbulence closure, as soon as a reasonably dense grid is used. The
turbulence models employed by the participants computing this test case can be
classified into three groups: (i) hybrid RANS/LES turbulence closures (University of
Rostock, IIHR), (ii) anisotropic non-linear statistical turbulence modeling: Reynolds
stress transport models (HHI), Explicit Algebraic Reynolds stress models (Chalmers,
ECN/CNRS, NMRI, YNU), (iii) isotropic linear eddy-viscosity model (ECN/CNRS,
HSVA, MARIN, URO, …). Concerning the linear isotropic turbulence models, most
244 M. Visonneau
Fig. 146 Case 1-7—hull geometry and experimental stations
of the contributors use the k-ω SST model but HHI, MHI and KRISO employ also
the realizable k-ε closure while CNR-INSEAN and MARIN prefer to use a one
equation model (Spalart-Allmaras or one eq. Menter with Dacles-Mariani modifica-
tion). Although the influence of turbulence models was assessed by various authors
at model scale without propeller, it will be interesting to check if this is also true in
presence of a rotating propeller. All the discretization methods are formally second-
order accurate and based on multi-block structured or unstructured grids which are
all body-fitted, with the exception of Chalmers and CNR-INSEAN which are using
an overset approach combining structured grids to reach an optimal grid density in
the stern region.
The propeller is treated in various ways. Seven contributors (ABS, ECN/CNRS,
CNR/INSEAN, IIHR, MARIN, PNU, University of Rostock) implemented an actual
propeller while seven other organizations (Chalmers, KRISO, MARIC, MIJAC,
NMRI, SJTU, SOTON and YNU) used various simplified propeller models described
in Table 7. Moreover, HHI used a propeller computational model based on a Mov-
Fig. 147 Case 1-7—view of the stern with its propeller
(a) Without propeller (b) With propeller
Fig. 148 Case 1-7—experiments—influence of the propeller on the isowakes distribution at station
S2 in the absence of duct
246 M. Visonneau
Fig. 149 Case 1-7—experiments—influence of the propeller on the distribution of mean secondary
velocities at station S2
ing Reference Frame model while HSVA used a Vortex Lattice approach to repre-
sent the propeller influence. It is also interesting to notice that some participants
(CNR/INSEAN, ECN/CNRS) provided results with both an actual and modelled
propeller, which can be used to quantify the impact of the propeller model on the
local flow.
6.3 Influence of Propeller Treatment
Figures 156 and 157 show a comparison between a body force representation of the
propeller and the computation of the rotating propeller based on sliding grids. Both
computations were performed by ECN/CNRS. It is interesting to notice that the
propeller treatment has already a significant influence on the isowake distribution at
station S2 due to a different modelling of the succion effect (see Fig. 156). But, the
occurence of the isowake u = 0.5 visible close to the wall in NMRI experiments, is
captured by none of these computations. In the wake of the propeller, at station S7,
248 M. Visonneau
(a) O ◦ (b) 24 ◦ (c) 48 ◦
Fig. 154 Case 1-7—experiments—distribution at station S7 of isowakes at 0◦ , 24◦ and 48◦
(a) O◦ (b) 24◦ (c) 48◦
Fig. 155 Case 1-7—experiments—distribution at station S7 of secondary velocities at 0◦ , 24◦ and

48◦
(a) Body Force (b) Actual Propeller
Fig. 156 Case 1-7—ECN/CNRS—distribution of isowakes at station S2 for body force and actual
propeller
(a) Body Force (b) Actual Propeller
Fig. 157 Case 1-7—ECN/CNRS—distribution of isowakes at station S7 for body force and actual
propeller
the only valid approach is the one based on the computation of the rotating propeller
which provides the correct asymmetry of the mean isowakes (see Fig. 157) since this
very simplified body force approach provides a symmetric wake flow. The isowake
u = 1.2 is correctly captured on the right hand side of the figure but its extent on the
left hand side is much more limited than in NMRI’s experiments.
Other examples of computations based on an actual propeller are provided by IIHR
using Rex and a DDES turbulence model at station S7 (see Fig. 158a), MARIN with
ReFRESCO (see Fig. 158b), PNU with Fluent (see Fig. 158c) or CNR/INSEAN with
250 M. Visonneau
(a) IIHR/Rex (b) MARIN/Refresco
(c) PNU/Fluent (d) ABS/OpenFoam
Fig. 158 Case 1-7—distribution of isowakes at station S7 by various contributors discretizing the
actual propeller
Xnavis (see Fig. 158d). Globally, all these computations are in very good agreement
with NMRI’s measurements. The location of the u = 1.2 isowake is globally very
well captured and the isowake u = 1.3 is even present in MARIN contribution.
Results from IIHR/Rex employing an IDDES hybrid RANS/LES model are also
very promising, although the noisy nature of the isowakes might indicate a too short
averaging time.
More sophisticated body force models can be used since they offer a very inter-
esting alternate choice to the very expensive Actual Propeller simulations. A good
example is provided by Chalmers which uses a lifting line model to represent the flow
around the propeller. Figure 159 shows their computed isowakes at stations S2 and
S7, which are in reasonable agreement with the experiments although the isowake
(a) Station S2 (b) Station S7
Fig. 159 Case 1-7—Chalmers—distribution of isowakes at station S2 and S7 for a lifting line
approach
u = 1.2 is not captured at station S7. HSVA developed a coupling with a vortex
lattice approach to represent the action of the propeller. Their results at stations S2
and S7 are shown in Fig. 160. At station S7, it seems that the isowake u = 1.2 is not
located on the right side of the propeller’s wake, which is difficult to explain. Finally,
Fig. 161 shows the isowake distribution at station S7 computed by NMRI and YNU.
Both institutions used a simplified propeller theory coupled with their RANS solver
to represent the propeller effect. As mentioned previously for Chalmers results, one
can observe that the isowake u = 1.2 is hardly captured in these results, NMRI being
marginally better than YNU from that standpoint.
In order to complete the analysis, it is interesting to compare the phase-averaged
isowake distribution at 0◦ , 24◦ and 48◦ since they were also measured by NMRI (see
Fig. 162). Figures 163, 164, 165, 166 and 167 show the isowakes obtained by different
teams. At first glance, one notices a very large dispersion of the results, mainly due to
the lack of discretisation points in this specific region. It is clear that very fine grids
probably adapted to the instantaneous location of the tip vortices would be necessary
to capture all these flow details. This can be observed, for instance, in Fig. 165 or
Fig. 166 where no isolated vortical structures are visible. The results obtained by
ABS with OpenFoam (Fig. 163), and to a lesser degree, by INSEAN using Xnavis
(Fig. 164) are more convincing. The grids used by these contributors range from 7M
to 17M cells. The most detailed solution is provided by IIHR using Rex (Fig. 167) on
a grid comprised of 68 M cells. One can clearly distinguish the location of tip vortices
inside the propeller disk characterised by a very large longitudinal velocity gradient.
In the center of the wake, a spot of isowakes probably indicates the location of the
propeller hub wake and around this structure, on can distinctly detect five islands of
isowakes which are associated with the traces of tip vortices. It is well known now,
that it is mandatory to use a locally adapted mesh refinement to capture accurately
252 M. Visonneau
(a) Station S2 (b) Station S7
Fig. 160 Case 1-7—HSVA—distribution of isowakes at station S2 and S7 for a vortex lattice
approach
(a) NMRI/Nagisa (b) YNU/Surfv7
Fig. 161 Case 1-7—NMRI and YNU—distribution of isowakes at station S7 using simplified
propeller theories
these structures or, by default, an extremely fine grid in the propeller wake. But a
very fine discretization is not enough since classical RANS closures dissipate very
quickly the vorticity of each of these tip vortex during their progression in the wake,
contrary to the hybrid RANS/LES closure which is able to maintain the right level of
longitudinal vorticity far from the tip vortex onset (see for instance Guilmineau et al.
2018 for more details on the turbulence model’s influence in the wake of a propeller).
Based on an hybrid RANS/LES closure, IIHR/Rex fully confirms this observation.
(a) O◦ (b) 24◦ (c) 48◦
Fig. 162 Case 1-7—experiments—distribution at station S7 of isowakes at 0◦ , 24◦ and 48◦
(a) O◦ (b) 24◦ (c) 48◦
Fig. 163 Case 1-7—ABS/OpenFoam—distribution at station S7 of isowakes at 0◦ , 24◦ and 48◦
(a) O◦ (b) 24◦ (c) 48◦
Fig. 164 Case 1-7—INSEAN/Xnavis—distribution at station S7 of isowakes at 0◦ , 24◦ and 48◦
6.4 Influence of Turbulence Closures
It is more difficult to analyze the role played by the turbulence modelling in case
of a rotating propeller since one does not have at our disposal results combining
the same propeller treatment with different turbulence closures. Moreover, the flow
is strongly accelerated after the propeller disk, which means that the global flow is
254 M. Visonneau
(a) O◦ (b) 24◦ (c) 48◦
Fig. 165 Case 1-7—MARIN/ReFresCo—distribution at station S7 of isowakes at 0◦ , 24◦ and 48◦
(a) O◦ (b) 24◦ (c) 48◦
Fig. 166 Case 1-7—PNU/Fluent—distribution at station S7 of isowakes at 0◦ , 24◦ and 48◦
rather controlled by the pressure than by turbulence. However, as briefly mentioned

in the previous section, the evolution of the individual tip vortices and their local flow
structure are strongly dependent on the turbulence closure as long as a very fine local
grid is generated to capture these flow details. It was noticed in previous systematic
computational studies that RANS models are able to predict the onset of tip vortex
but only hybrid RANS/LES can simulate accurately the detailed flow physics in the
core of these vortices (accurate value of TKE, for instance) and their progression
without dissipation in the wake of the propeller.
In order to get a first idea on the influence of turbulence closures, we are going
to select computations performed with the actual propeller and compare RANS with
hybrid RANS/LES models. Of course, at this stage, it is impossible to perform a true
verification study and evaluate the role played by the discretisation error. However,
we will choose on one hand, PNU/Fluent to represent the RANS category using
a k-ε and on the other hand, IIHR/Rex to represent the hybrid RANS/LES class.
In terms of mesh density, IIHR/Rex has generated a grid comprised of 68 M cells
while PNU/Fluent uses a coarser grid comprised of 17M cells. Figure 168 shows the
averaged iso-wake distribution at station S7 for both contributors. At first glance,
it is obvious that the result from IIHR is richer in terms of vortical structures. The
isowake distribution delimitating the propeller disk is also steeper for IIHR than for
(a) O◦ (b) 24◦ (c) 48◦
Fig. 167 Case 1-7—IIHR/Rex—distribution at station S7 of isowakes at 0◦ , 24◦ and 48◦
(a) IIHR/Rex (b) PNU/Fluent
Fig. 168 Case 1-7—distribution of isowakes at station S7 with various turbulence closures
PNU. However, this could be associated with a different grid distribution. In the
propeller disk, the island u = 1.2 is significantly more extended in IIHR than in PNU
computations. This is a striking difference which might or might not be attributed
to the turbulence closure but IIHR results are there in better agreement with the
experiments. The iso-line u = 1.3 located in the upper part of the propeller disk is
predicted by none of the contributors. PNU predicts a well defined iso-contour u
= 1.2 while IIHR gets a set of lower iso-contours u = 1.2 associated probably to
various vortical structures. This effect is probably due to an averaging time which is
still too short to get a statistically converged solution.
Figures 169 and 170 recall the comparisons of phase-averaged isowakes for PNU
and IIHR. Here, as mentioned previously, the two solutions strongly differ. IIHR is
able to capture the traces and evolution of the five tip vortices while PNU provides
globally identical pictures for the three phase-averaged results which do not exhibit
any temporal evolution. It seems that the natural unsteadiness of the flow in the wake
of the rotating propeller is not captured by this simulation. The same remark applies
256 M. Visonneau
(a) O◦ (b) 24◦ (c) 48◦
Fig. 169 Case 1-7—PNU/Fluent—distribution at station S7 of isowakes at 0◦ , 24◦ and 48◦
(a) O◦ (b) 24◦ (c) 48◦
Fig. 170 Case 1-7—IIHR/Rex—distribution at station S7 of isowakes at 0◦ , 24◦ and 48◦
to MARIN’s results which computed the flow with a rotating propeller (see Fig. 165).
The agreement between IIHR and NMRI’s phase-averaged experiments is globally
good, taking into account the experimental uncertainty.
6.5 Case 1-7—Concluding Remarks
This section provided a short analysis of the local flow around the JBC with a rotating
propeller at one station S2 located before the propeller and two stations S4 and
S7, located after the propeller. Various meshes, turbulence models and propeller
treatments were used by the contributors, which renders a rigourous analysis almost
impossible at this stage. However, a few general conclusions based on the comparison
of the results can be drawn.
• The computations based on the simulation of the actual propeller are the only
ones which are able to represent all the characteristics of the flow in the wake
of propeller, specifically at station S7. Various codes using different meshes and
different turbulence models yield globally comparable results in terms of isowake
distribution when the real rotating propeller is taken into account in the simulation.
In particular, the highest isowake level is systematically captured by the simulations

and its extent is very correctly reproduced.
• However, the price to pay to get this level of agreement is very high, due to the
use of a very small time step (about one time step per degree of rotation), which
makes this approach not realistic for computations in waves for instance. Various
alternate coupling strategies based on simplified propeller models are therefore of
great practical interest although less accurate in terms of local representation of the
isowake distribution. Globally, the local flow produced by these simplified models
is quite satisfactory for some of them, even if they are not able to capture all the
characteristics of the wake flow. Some coupled models are not satisfactory but at
this stage, it is difficult to identify the reasons which explains these disappointing
performances.
• A systematic study of the influence of turbulence models is hardly feasible on the
basis of these results. It seems that, due to the fact that the flow after the propeller is
pressure-controlled, the turbulence model plays a more minor role on the averaged
flow field. Classical linear eddy viscosity based closures are doing reasonably well.
Hybrid RANS/LES model used by IIHR/Rex yields very good results but not
necessarily way superior for the average flow field to what is obtained with usual
k-ω-like models to justify the very high cost associated with this time-accurate
approach. However, this hybrid RANS/LES model is the only one able to capture
most of the details of the flow revealed by the phase-averaged measurements, which
makes it probably useful when propeller/rudder interaction during maneuvers are
considered, in the absence of any new RANS model.
7 Test Case 1-8
7.1 Description of the Experiments
Test case 1-8 concerns the local flow analysis around the JBC with a ducted propeller
towed in the NMRI towing tank, in free trim and sinkage conditions. The location of
the experimental stations are shown in Figs. 171 and 172 shows the ducted propeller.
With these last experiments, we are going to be able to determine the influence of
the duct on the local flow with a rotating propeller behind it while the test case 1-4
provided the same kind of information without the presence of a propeller.
Figures 173, 174, 175, 176, 177 and 178 show the influence of the duct on the
experimental time-averaged iso-wakes and transversal velocity components at three
sections S2, S4 and S7 already defined before. Already at station S2, the influence
of the duct is visible through an intensification of the longitudinal vorticity close to
the wall. This is indirectly indicated by the isowake u = 0.4 which is closed when
the duct is present. The same influence was noticed previously without the presence
of a propeller. It is also worthwhile to notice the presence of the closed isowake u
= 0.5 which is more extended when the duct is there. The experiments at station
258 M. Visonneau
Fig. 171 Case 1-8—hull geometry and experimental stations
Fig. 172 Case 1-8—view of the stern with its ducted propeller
S4 are hardly exploitable but station S7 provides very interesting information on

the influence of the duct in the wake of the propeller. The duct leads to a general
reduction of the longitudinal velocity. For instance, the isowake u = 1.3 has dis-
appeared when the duct is mounted. The region where u is comprised between 1.1
and 1.2 is considerably extended with a duct, which means that the duct creates a
noticeable homogeneization of the longitudinal velocity distribution in the wake of
the propeller, even in the central part of the propeller disk. A by-product of this effect
is the shape of the isowake u = 1.2 which is strongly reduced to the right part of the
propeller disk when the duct is there.
station S2
Figures 179, 180, 181, 182, 183 and 184 provide the same information for the
phase-averaged iso-wakes at stations S4 and S7 for the same angles as previously,
i.e. 0◦ , 24◦ and 48◦ . As mentioned before, experiments at station S4 are hard to
analyze but station S7 provides again very interesting information on the influence
of the duct on the phase-averaged flow. Figure 182 confirms that the wake is more
homogeneous in case of the presence of a duct. The region where u is greater than 1.2
is strongly reduced at 0◦ in case of duct and the central part of the wake appears less
noisy. At 24◦ (see Fig. 183), the region where u is greater than 1.3 is also strongly
reduced when the duct is present. At 48◦ (see Fig. 184), an opposite influence is
260 M. Visonneau
Fig. 175 Case 1-8—experiments—influence of the duct on the mean isowakes distribution at station
S4
station S4
noticed: the region where u is greater than 1.2 is slightly more extended when the
duct is mounted than without it. This compensation for various blade angle might
contribute to the global homogeneization of the averaged flow field observed when
the duct is present.
Fig. 177 Case 1-8—experiments—influence of the duct on the mean isowakes distribution at station
S7
station S7
Seventeen participants provided results for the test case 1-8. The main characteristics
of their contributions are listed in Table 8 in terms of turbulence closures, wall treat-
ments, discretization and propeller treatment. Some participants provided several
different contributions, which makes possible a comparison of various turbulence
closures or wall treatments for the same grid and same code. Moreover, SOTON
262 M. Visonneau
Fig. 179 Case 1-8—experiments—influence of the duct on the isowakes distribution at 0◦ and
station S4
station S4
performed a comparison of codes (OpenFoam vs. Star-CCM+) with the same grid
and same turbulence closure, as previously.
Except University of Rostock which is no more present in this list, the contributors
to case 1-8 were the same as the ones who contributed to case 1-7. For the sake of
conciseness, the characteristics of their contributions are not recalled here and the
reader could refer to the comments accompanying Table 7 in case of irrepressible
need.

Organization Code name Turbulence Wall model Grid Propeller
model characteristics treatment
ABS k-ω SST Wall function Unstructured grid Actual propeller
Chalmers-Flowtech EARSM Wall resolved Structured grid Lifting line
22.4M cells
CNR-INSEAN Spalart- Structured grid Body-force/actual
Xnavis Allmaras with overlapping propeller
11.4M cells
ECN-CNRS SST/EARSM Wall function Unstructured grid Body-
ISIS-CFD aka Wall resolved 25.3M cells force/Actual
FINET M /Marine propeller
HHI Realizable Wall function Unstructured grid Moving
HiFOAM/StarCCM+ k-/RSTM 3.5M cells Reference Frame
HSVA 2 equations Wall function Unstructured grid Vortex lattice
FreSCo+ 14.6M cells
IIHR DDES Wall resolved Unstructured grid Actual propeller
REX 70.9M cells
KRISO Realizable Wall function Structured grid Lifting surface
MARIC k-ω SST Wall function Unstructured grid Body force
FINET M /Marine 2M cells
MARIN k-ω SST Wall resolved Unstructured grid Actual propeller
ReFRESCO + DM 23M cells
correction
MIJAC k-ω SST Wall resolved Unstructured grid Simplified
OpenFOAM 8.7M cells propeller theory
NMRI EARSM Wall resolved Structured grid Simplified
Nagisa 10M cells propeller theory
PNU 2 equations Wall resolved Unstructured grid Actual propeller
SHIME k-ω SST Wall resolved Unstructured grid N/A
SJTU k-ω SST Wall function Unstructured grid Simplified
naoeOpenFOAM 4.9M cells propeller theory
SOTON k-ω SST Wall resolved Structured grid Body force
OpenFOAM/StarCCM+ 21.3M cells prescribed
YNU EARSM Wall resolved Structured grid Simplified
SURFv7 3.3M cells propeller theory
264 M. Visonneau
station S4
station S7
7.3 Influence of the Duct on the Local Flow
In order to see if the computations are able to reproduce the salient features charac-
terising the influence of the duct on the local flow and the unsteady nature of the flow
in the wake of the propeller, one will start with the more complete computational
approach, i.e. the one based on an actual rotating propeller and an hybrid RANS/LES
turbulence model. Then, let us analyse the contribution of IIHR/Rex to check if these
computations are able to catch the local influence of the duct on the flow before and
after the propeller. Figures 185 and 186 show the isowakes with and without duct
station S7
station S7
computed by IIHR/Rex at stations S2 and S7. These computations are based on an

hybrid RANS-LES model (DDES). At station S2 (see Fig. 185), computations do
not show any difference between the two configurations, which means that, either
the experiments are not accurate enough to assess this level of detailed comparisons,
or the computational model is not able to capture the modification of the upstream
influence of the duct at station S2. Once again, it should be noticed that the pecu-
liar shape of the closed isowake u = 0.5 in the vicinity of the hull at station S2 is
not captured by these computations. We will come back to this feature in the next
subsection devoted to the influence of turbulence closures.
266 M. Visonneau
Fig. 185 Case 1-8—IIHR/Rex—influence of the duct on the distribution of isowakes at station S2
with a propeller
Fig. 186 Case 1-8—IIHR/Rex—influence of the duct on the distribution of isowakes at station S7
with a propeller
At station S7 (see Fig. 186), the trend towards an homogeneisation of the wake
in presence of a duct is well captured by IIHR/Rex. Moreover, the isowake u = 1.2
has almost completely disappeared in the simulations with a duct, which is in good
agreement with the experimental observation.
Figures 187, 188 and 189 show a comparison of computations performed by
IIHR/Rex with the same computational approach as before at station S7 with and
without duct for the phase-averaged flow measurements at 0◦ , 24◦ and 48◦ . At 0◦ ,
the number of vortices seems reduced (associated with closed isowakes) when the
Fig. 187 Case 1-8—IIHR/Rex—influence of the duct on the distribution of isowakes at 0◦ and
station S7 with a propeller
duct is present, which is in line with the measurements (see Fig. 187) but one notices
an isowake u = 1.3 which is not visible in the experiments. At 24◦ , (see Fig. 188),
computations with duct look again more homogeneous with less local intense struc-
tures. Moreover, now the isowakes u = 1.3 have disappeared when the duct is there,
which is in agreement with the experimental observation. At 48◦ , (see Fig. 189),
the isowake distribution with duct appears more homogeneous than without duct as
noticed for the other blade positions. It is very difficult to analyze in more details
these computational results, due to the complexity of the flow in the wake of the
propeller.
The previous analysis was performed by using an hybrid RANS/LES turbulence
model. It is interesting to see what can be obtained with a more classical RANS
model activated in URANS mode. As previously, Figs. 190 and 191 show the time-
averaged isowakes at stations S2 and S7, computed by PNU/Fluent using a RNG
k-ε turbulence model and an actual rotating propeller. This specific contribution is
chosen because it is computed on a fine grid comprised of 17.5 M cells, which should
reduce the discretization errors. It is clear from these figures that the presence of the
duct has no influence at station S2, contrary to what is shown by the experiments. At
station S7, the main effect of the duct is to reduce the extent of the iso-wake u = 1.2
region and to increase significantly the area associated with the iso-wake u = 1.1.
This is in very good agreement with the experimental trend. This leads to think that
one can rely on RANS simulations for global predictions of time averaged velocity,
even in the wake of the propeller.
Figures 192, 193 and 194 show the phase-averaged distributions of isowake at
station S7 for 0◦ , 24◦ and 48◦ . Contrary to the previous results of IIHR/Rex, all these
three figures are very similar and do not show any occurence of tip vortices. Probably,
268 M. Visonneau
the tip vortices are already dissipated at this station, due to the combined influence
of a grid which is locally too coarse and the use of RANS turbulence model.
All the previous computations were made on the basis of computations taking into
account the actual rotating propeller. It is also interesting to see what can be obtained
with a modelled propeller without taking into consideration the axisymmetric body-
force model which is obviously not able to represent the flow in the wake of a
propeller. For instance, Chalmers provided us with a simulation using a propeller
modelled on the basis of a lifting line model. Figure 195 shows the time-averaged
Fig. 190 Case 1-8—PNU/Fluent—influence of the duct on the distribution of isowakes at station
S2 with a propeller
Fig. 191 Case 1-8—PNU/Fluent—influence of the duct on the distribution of time-averaged

isowakes at station S7 with a propeller
isowakes at station S7 with and without duct as usual. Although the isowake u = 1.2 is
not captured by the computations without duct, the trend towards an homogeneization
of the wake due to the action of the duct is also well captured by this contribution.
This leads us to think that it is not necessary to use an actual propeller to get the trend
right, which, if it is confirmed, is a valuable result in the perspective of duct shape
optimisation.
270 M. Visonneau
Fig. 192 Case 1-8—PNU/Fluent—influence of the duct on the distribution of isowakes at 0◦ and
7.4 Influence of Turbulence Closures
As previously, the isowakes at station S2 are better predicted with non-linear EARSM
turbulence models as shown by Fig. 196 but this station S2 is not significantly influ-
enced by the presence of the ducted propeller. IIHR/Rex hybrid RANS/LES closure
does not behave well at station S2 probably because the LES model is not yet acti-
vated there and the flow prediction relies on the linear isotropic k-ω SST which is
known to under-predict the longitudinal vorticity.
Fig. 195 Case 1-8—CHALMERS/ShipFlow—influence of the duct on the distribution of time-

averaged isowakes at station S7 with a propeller
Let us now compare a classical RANS model with an hybrid RANS/LES model
implemented in the same code, IIHR/Rex, in the wake of the ducted propeller.
Figure 197 shows the time-averaged isowakes at station S7 for k-ω SST and DDES
turbulence closures with a ducted rotating propeller. At first glance, the k-ω SST
model seems to provide a solution in good agreement with the experiments since it
is able to predict the existence of the crescent-shaped zone u = 1.2 even if its extent
is under evaluated. On the other hand, hybrid RANS/LES results appear much more
272 M. Visonneau
(a) Experiments (b) EARSM Chalmers/ShipFlow (c) Hybrid RANS/LES IIHR/Rex
Fig. 196 Case 1-8—comparison of different turbulence closures at station S2
(a) Experiments (b) SST (c) DDES
Fig. 197 Case 1-8—IIHR/Rex—distribution of the averaged isowakes at station S7 for RANS and
hybrid RANS/LES turbulence closures
noisy as if the statistical convergence was not yet reached (see the irregular boundary
of the isowake u = 1.1 at the top of the propeller disk) and the isowake u = 1.2 is
hardly visible. Globally, based on the comparison of isowakes, it is difficult to justify
having recourse to the hybrid RANS/LES model which is, at least, ten times more
expensive.
Figures 198, 199 and 200 give the same comparison at station S7 for phase-
averaged isowakes. Hybrid RANS/LES based simulation provides richer isowakes
than what is computed by the URANS approach, which is expected but can be related
with a lack of statistical convergence, as well. For each computation, one can follow
the evolution of vortical structures with respect to the blade angles, which means that
even with a RANS approach, it is possible to capture the traces of the tip vortices at
this station, contrary to what was obtained previously by PNU/Fluent using a RNG
k-ε turbulence model. It is hard to assess the results in a more detailed way, due to
the complexity of the phase-averaged flow.
Fig. 198 Case 1-8—IIHR/Rex—distribution of isowakes at 0◦ and station S7 for RANS and hybrid
RANS/LES turbulence closures
Fig. 199 Case 1-8—IIHR/Rex—distribution of isowakes at 24◦ and station S7 for RANS and
Fig. 200 Case 1-8—IIHR/Rex—distribution of isowakes at 48◦ and station S7 for RANS and
274 M. Visonneau
7.5 Case 1-8—Concluding Remarks
This section provided a short analysis of the local flow around the JBC with a ducted
rotating propeller at one station S2 located before the propeller and two stations S4
and S7, located after the propeller. Various meshes, turbulence models and propeller
treatments were used by the contributors, which renders a rigourous analysis almost
impossible at this stage. However, a few general observations that stem from this
study are summarized here:
• When a duct is mounted before the propeller, the flow in the wake of the propeller is
more homegeneous as indicated by the experimental distribution of isowakes. This
influence is correctly captured by most of the contributors, whatever the turbulence
model or propeller representation used.
• The isowake distribution behind the ducted propeller is better represented when
the actual rotating propeller is accounted for. However, some simplified propeller
models work reasonably well and are able to indicate the right trend concerning
the influence of the duct.
• RANS models provide a reasonably accurate prediction of the time-averaged flow
behind the propeller. For phase-averaged quantities, the analysis of experimen-
tal and computational results is hard because of the complexity of the phase-
averaged flow field. Some URANS-based computations are in fair agreement with
the experiments while some others do not show any unsteadiness in the wake of
the propeller, probably due to the combined influence of a grid which is locally
too coarse and a turbulence model which is too diffusive. From that standpoint, the
hybrid RANS/LES model appears more consistent with the physics of the flow.
This is encouraging if one considers the level of details revealed by this compu-
tation although the provided contribution from IIHR/Rex might suffer from an
unsufficient statistical convergence.
8 General Conclusions
In this chapter, we have reviewed computations of the flow around the Japanese
Bulk Carrier (JBC) provided by the contributors to the Tokyo T2015 workshop from
the standpoint of the local flow analysis. Five flow configurations were studied: (i)
the naked hull without propeller nor duct (case 1-3a), (ii) the double body model
(case 1-3b), (iii) the hull with a duct but without propeller (case 1-4), (iv) the hull
with a propeller and without duct (case 1-7) and finally, (v) the hull with a ducted
propeller (case 1-8). This chapter aimed at understanding the physics of the flow from
a local point of view with the help of the flow measurements performed by NMRI at
three stations S2, S4 and S7. But several additional objectives were pursued during
this analysis. The first one, in the continuity of the G2010 Gothenburg workshop
concerned the verification and validation of stern flows over the JBC hull. This hull
is characterised by a relatively high block coefficient, which means that the flow is
more complex in the stern region. An intense bilge vortex was detected and some
questions arose about the steadiness of the flow in the stern region. It was therefore
decided to focus the study on the bilge vortex core and try to compare the various
contributions in this specific region. The second objective was to understand the
mechanisms which might improve the propulsive efficiency when a ship is equipped
with a ducted propeller. To separate physical effects, several configurations with and
without duct combined with the presence or absence of a propeller were studied with
the help of the measurements made by NMRI at stations S2, S4 and S7. This gave us
the opportunity (i) to compare the various turbulence closures in presence of a duct
or a propeller, (ii) to assess the reliability of the various propeller representations
implemented by the contributors, and (iii) to check if the observed influence of the
duct on the local flow is accurately captured by the computations.
The main conclusions that stem from all these studies are summarized as follows.
For the case 1-3a, since the typical grids which are used by the contributors (5–
10 M points) are fine enough, the influence of the grid discretisation is moderate
for RANSE, which means that the numerical error appears to be under control. It
was feasible therefore to focus the study on the influence of the turbulence closures
and see if it was possible to draw general conclusions form all the contributions. As
noticed during the Gothenburg2010 workshop for the analysis of the local flow around
the KVLCC2, the major influence comes from the turbulence closure. One noticed
that the linear isotropic closures significantly under-predict the longitudinal vorticity
while full RSM closures slightly over-predict it. Non-linear anisotropic closures
(EARSM) seemed to offer a good compromise from the standpoint of the local flow
although they slightly underpredict the vorticity at the key station S2. Contributions
employing hybrid RANS-LES were also taken into account. The results appeared
promising although IDDES seemed to over-predict the vorticity again. No spectacular
advantage was noticed compared to the best RANS models at the measurement
stations.
This study was completed by a local core vortex analysis which provided a first
interesting attempt to carry out a more local analysis of the time-averaged bilge
vortex. Globally, the agreement between most of the computations and the rather
coarse experiments was satisfactory in terms of global trends. The longitudinal vor-
ticity was somewhat underestimated and the longitudinal velocity distribution was
fairly reproduced. These experiments pointed out very large differences on the tur-
bulence kinetic energy and this was the most striking (and unexpected) result of
this comparison. Only the hybrid RANS/LES closures were able to reproduce this
characteristic, thanks to the contribution of the resolved turbulence kinetic energy.
Whether the NMRI TKE measurements are reliable or not to provide an accurate
turbulence kinetic energy at this location is still a matter of intense debate which is
partly addressed in chapter “Experimental Data of Resistance, Sinkage, Trim, Self-
-propulsion Factors, Longitudinal Wave Cut and Detailed Flow for JBC With and
Without an Energy Saving Circular Duct” of this book. Further much more detailed
measurements will be necessary to draw safer conclusions.
276 M. Visonneau
In the case 1-3b, the turbulence closure study led to very similar conclusions;
i.e. that the major influence on the computed distribution of isowakes in the stern
region comes from the turbulence closure. Linear isotropic closures under-predict
the longitudinal vorticity at station S2 while full RSM closures tend to over-predict
it at the same station. But what was remarkable and quite unique was that two LES
computations were presented and especially, one from SRC on an extremely fine
grid (4 billion cells) which was in good agreement with measurements and predicted
the level of turbulence kinetic energy measured by NMRI in the core of the bilge
vortex (at least three times higher than what was modeled by EASM). Unfortunately,
it was not possible to exploit further the results of the (almost) wall-resolved LES
simulation, due to the unavailability of this flow database.
For the case 1-4, it was again noticed that the turbulence models have a critical role
to play once the grid is fine enough. The main modifications of the flow created by
the presence of the duct were correctly reproduced by linear isotropic models which
provided the correct qualitative trend. A better quantitative agreement was reached by
anisotropic closures although the inner part of the vortex was not perfectly reproduced
as illustrated by the occurence of a reversed flow region which was over-estimated
by both RSTM and EARSM turbulence closures.
For the test case 1-7, it was observed that the computations based on the simulation
of the actual propeller were the only ones able to represent all the characteristics of the
flow in the wake of propeller, specifically at station S7. Various codes using different
meshes and different turbulence models yielded globally comparable results in terms
of isowake distribution when the real rotating propeller is taken into account in
the simulation. However, alternate coupling strategies based on simplified propeller
models appeared to produce a local flow which was quite satisfactory for some
of them, even if they were not able to capture all the characteristics of the wake
flow. Hybrid RANS/LES model used by IIHR/Rex yielded very good results but
not necessarily way superior for the averaged flow field to what is obtained with
usual k-ω-like models to justify the very high cost associated with this time-accurate
approach. However, this hybrid RANS/LES model was the only one able to capture
most of the details of the flow revealed by the phase-averaged measurements.
For the test case 1-8, the computations confirmed the fact that when a duct is
mounted before the propeller, the flow in the wake of the propeller is more homoge-
neous. This influence was correctly captured by most of the contributors, whatever
the turbulence model or propeller representation used. As expected, the isowake dis-
tribution behind the ducted propeller was better represented when the actual rotating
propeller was accounted for. However, some simplified propeller models worked rea-
sonably well and were able to indicate the right trend concerning the influence of the
duct. RANS models provided a reasonably accurate prediction of the time-averaged
flow behind the propeller. For phase-averaged quantities, some URANS-based com-
putations were in fair agreement with the experiments while some others did not
exhibit any unsteadiness in the wake of the propeller, probably due to the combined
influence of a locally too coarse grid and a too diffusive turbulence model. From that
standpoint, the hybrid RANS/LES model appeared again more consistent with the
physics of the flow.
Finally, some overall considerations can be drawn to conclude this sixth chapter:
(i) Most of the contributors were able to predict the main characteristics of the
flows considered in this chapter, provided that the grid is fine enough. And
computing on a fine enough grid is no more out of reach, nowadays,
(ii) Once discretization error is under control, the main source of discrepancy is the
modelling error, i.e. the turbulence model. From that standpoint, the situation
has not much changed compared to the previous edition of this workshop in
2010. Linear isotropic k-ω SST closure remains the usual workhorse because
of its robustness and global reliability. However, better results in terms of local
flow field like longitudinal vorticity are obtained if one takes into account the
turbulence anisotropy. This improvement has an non-negligible impact (a few
percents) on the prediction of the resistance.
(iii) The influence of the duct on the flow with and without propeller is correctly
captured by a linear isotropic model, at least in terms of trend. This means that it
does not seem mandatory to have recourse to sophisticated turbulence models to
perform a design study aiming at improving the efficiency of a ducted propeller.
However, special care should be used to model the downwind influence of the
propeller which is not represented by a simple body force model. Here again,
the propeller model appears to be the weak point which has to be improved if
one wishes to avoid lengthy computations of the actual rotating propeller.
(iv) Having recourse to more sophisticated and expensive turbulence models like
hybrid RANS-LES closures brings much more physics to the local description
of flow fields. Illustrations were given in this chapter for instance, by the study
of the wake of a rotating propeller or by the fine analysis of the turbulence
characteristics in the core of the vortex. Although RANS statistical turbulence
closures are able to capture the main flow characteristics, the unsteady hybrid
RANS/LES closures provide reliable information on the turbulence charac-
teristics that are out of reach of statistical turbulence models, whatever their
complexity.
(v) In several parts of this analysis, it was difficult to draw any firm conclusion
because of the lack of reliable data to characterize turbulent fluctuations. There
is clearly a need of better experiments in that respect to deepen the analysis
of the physical models. I would therefore recommend for the next workshop
to complete, as much as possible, the mean flow measurements by reliable
experimental information on the velocity fluctuations like the Reynolds Stress
tensor or at least the turbulence kinetic energy.
Acknowledgements The study performed by ECN/CNRS was granted access to the HPC resources
under the allocation 2015-2a1308 made by GENCI (Grand Equipement National de Calcul Intensif).
278 M. Visonneau
References
Bhushan, S., Yoon, H., Stern, F., Guilmineau, E., Visonneau, M., Toxopeus, S., et al. (2019).
Assessment of computational fluid dynamics for surface combatant 5415 at straight ahead and
static drift β=20◦ . Journal of Fluids Engineering, 141(5), 051101.
Guilmineau, E., Deng, G., Leroyer, A., Queutey, P., Visonneau, M., & Wackers, J. (2018). Numerical
simulations for the wake prediction of a marine propeller in straight ahead flow and oblique flow.
Journal of Fluids Engineering, 140, 1–11.
Larsson, L., Stern, F., & Visonneau, M. (Eds.). (2014). Numerical ship hydrodynamics: An assess-
ment of the Gothenburg 2010 workshop. Springer.
Lee, S.-J., Kim, H.-R., Kim, W.-J., & Van, S.-H. (2003). Wind Tunnel tests on flow characteristics
of the KRISO 3,600 TEU containership and 300K VLCC double-deck ship models. Journal of
Ship Research, 47(1), 24–38.
Visonneau, M., Deng, G., Guilmineau, E., Queutey, P., & Wackers, J. (2016, September). Local
and global assessment of the flow around the Japan bulk carrier with and without energy saving
devices at model and full scale. In 31st Symposium on Naval Hydrodynamics. CA: Monterey.
Wackers, J., Deng, G. B., Guilmineau, E., Leroyer, A., Queutey, P., & Visonneau, M. (2014).
Combined refinement criteria for anisotropic grid refinement in free-surface flow simulation.
Computers & Fluids, 92, 209–222.
Evaluation of Self-propulsion and Energy
Saving Device Performance Predictions
for JBC
Takanori Hino
Abstract Self-propulsion cases for the JBC hull with and without an energy saving
duct are analyzed in this chapter. Test cases are set up for self-propulsion condition
at the ship point. No rudder is fitted in either case. About half of the submissions
employ actual propeller models in which a propeller geometry is discretized using a
moving mesh and the remainder uses body force models in which propeller effects are
considered as a body force computed using external potential-flow based programs.
Self-propulsion simulations are carried out in two ways. The first is to follow the
self-propulsion test procedure in a towing tank and a propeller revolution rate is
adjusted in such a way that propeller thrust and towing force or SFC (Skin Friction
Correction) for the ship point condition are balanced with ship’s resistance. The other
way is to fix the propeller revolution rate equal to the experimental value and the
force invariance is computed. Thrust and torque coefficients, propeller revolution and
ship’s resistance components in self-propulsion condition are items to be submitted.
Analysis of grid uncertainty is carried out based on the submission data with multiple
grids. Average of comparison errors and the standard deviations of propeller thrust
and torque together with revolution rates are estimated using the towing tank test data
of 7.0 m model. Self-propulsion factors are estimated using the submitted data and
compared with the measured data. Finally, model scale delivered powers are used
to evaluate overall accuracy of the current CFD analysis in terms of the prediction
accuracy of energy saving duct performance.
1 Introduction
In addition to the resistance test cases of 1.1 for the bare hull and 1.2 for the hull with
a stern duct as an energy saving device (ESD), the self-prolusion cases are set for JBC
without (Case 1.5a) and with (Case 1.6a) ESD. There are 26 submissions for Case
1.5a and 25 submissions for Case 1.6a, respectively. The results are analyzed with
T. Hino (B)
280 T. Hino
respect to thrust, torque and revolution rate of a propeller and self-propulsion factors.
Finally, the performance predictions of the energy saving device are compared and
their accuracy is discussed.
2 Overview of Test Cases
Computational conditions are set in accordance with the self-propulsion test

conducted in NMRI towing tank (see Chap. 2 for detail).
Both Cases 1.5a and 1.6a are for the self-propulsion simulation at the ship-point
condition. No rudder is fitted and flow condition is Fn = 0.142 and Rn = 7.46 ×
106 which corresponds to the experiment with a ship speed of 1.179 [m/s] and ship
length (L PP ) of 7.000 [m]. Ship motion is free in heave and pitch. Only difference
between Cases 1.5a and 1.6a is the presence of ESD.
Geometry of the model propeller and its open water characteristics are provided to
participants. Requested procedure for self-propulsion simulation follows the exper-
imental approach, i.e., the revolution rate of a propeller is adjusted to obtain force
equilibrium in the longitudinal direction considering the applied towing force (Skin
Friction Correction, SFC) for the ship-point condition. Thus, the thrust T is T =
RT (SP) − SFC, where RT (SP) is the total resistance in self propulsion. SFC is pre-
computed as 18.2 [N] for Case 1.5a and 18.1 [N] for Case 1.6a which are the same
amounts as in the experiment. The items to be reported are the propeller revolution
rate n, the thrust coefficient K T , the torque coefficient K Q and the total, pressure and
frictional resistance coefficients in the self-propulsion condition, C T (SP) , C P(SP) and
C F(SP) , respectively.
In case the procedure above cannot be carried out, the propeller revolution rate
n is set to the measured value of 7.8 [rps] for Case 1.5a and 7.5 [rps] for Case 1.6a
and the resulting towing force (RT (SP) − T ) is to be reported together with, K T , K Q ,
C T (SP) , C P(SP) , C F(SP) defined above.
3 Submissions
3.1 Bare Hull (Case 1.5a)
Total of 26 solutions are submitted for Case 1.5a as shown in Tables 1 and 2. Table 1
summarizes 14 submissions using actual propeller models in which the propeller
geometry is discretized and rotated using moving grids. Table 2 is for 12 submissions
with body force propeller models which include lifting line (BL), lifting surface (BS),
prescribed (BP), vortex lattice (BV) or simplified propeller (BX) models. Free surface
effects are not considered in 10 submissions. The other 16 submissions include free
surface effects via VOF model (12), level-set model (3) or potential flow model (1).
Table 1 Submissions with actual propeller model for Case 1.5a
Organization Free surface Prop. M. CT CF C PV KT KQ n(rps) RT -T (N)
(Code) × 1000 × 1000 × 1000
EFD – – D 4.811 – – 0.217 0.0279 7.8 18.2
(NMRI)
ABS V A S 4.768 3.205 1.564 0.213 0.0284 7.842 –
(OpenFOAM) E%D 0.895 – – 2.057 −1.682 −0.538 –
CNR/INSEAN No A S N/A N/A N/A 0.210 0.0304 – N/A
(XNAVIS) E%D N/A – – 3.230 −8.960 – N/A
CSSRC No A S 4.757 3.139 1.619 0.223 0.0291 – 17.272
(FLUENT) E%D −1.094 – – 2.765 4.194 – 5.100
ECN/CNRS V A S 4.661 3.149 1.512 0.214 0.0290 – 17.470
(ISISCFD/WF-EASM-RANSE) E%D 3.110 – – 1.470 −5.550 – 4.010
HHI V A S 4.841 3.128 1.713 0.217 0.0279 – 18.600
(HiFoam) E%D −0.630 – – −0.140 0.100 – −2.400
Evaluation of Self-propulsion and Energy Saving Device …
HHI V A S 4.855 – – 0.224 0.0294 – 18.100

(STRA-CCM+) E%D −0.910 – – −3.320 −5.380 – 0.770
IIHR LS A S 4.812 3.210 1.600 0.219 0.0291 7.818 –
(REX-DDES) (continued)
281
Table 1 (continued)
282

(Code) × 1000 × 1000 × 1000
E%D −0.021 – – −0.922 −4.301 −0.231 –
MARIC V A S 4.986 – – 0.228 0.0278 7.800 18.200
(FINEMarine) E%D 3.637 – – 5.069 −0.358 –
MARIN V A S 4.663 3.065 1.598 0.212 0.0282 – 17.744
(ReFreSCo) E%D 3.085 – – 2.517 −1.143 – 2.504
PNU V A S 5.177 3.166 2.011 0.244 0.0285 7.871 18.515
(FLUENTv15) E%D −7.608 – – −12.442 −2.151 −0.910 −1.732
SHIME No A S 4.852 3.070 1.782 0.216 0.028 – 18.800
(OpenFOAM) E%D −0.855 – – 0.631 −1.853 – −3.355
UDE V A S 4.808 2.946 1.862 0.215 0.0264 – 18.600
(STAR-CCM+) E%D 0.062 – – 0.829 5.376 – −2.198
UNI-ROSTOCK No A S 5.210 3.554 1.656 0.215 0.027 – 21.450
(OpenFOAM-SST) E%D −8.293 – – 0.921 4.659 – −17.850
UNI-ROSTOCK No A S 5.075 2.830 2.245 0.232 0.029 – 19.518
(OpenFOAM-HYBRID) E%D −5.487 – – −6.912 −2.150 – −7.241
T. Hino
Table 2 Submissions with body force propeller model for Case 1.5a
(Code) × 1000 × 1000 × 1000
EFD – – D 4.811 – – 0.217 0.0279 7.8 18.2
(NMRI)
CHALMERS-FLOWTECH Pot.Flow BL S 4.754 3.097 1.657 0.226 0.0279 7.572 –
(SHIPFLOW) E%D 1.180 – – −4.340 −0.080 2.930 –
CNR/INSEAN No BX S N/A N/A N/A 0.187 0.0268 – N/A
(XNAVIS-HO) E%D N/A – – 13.825 3.943 – N/A
ECN/CNRS V BX S 4.625 3.108 1.517 0.214 0.0291 7.600 –
(ISISCFD/LRN-EASM-AD) E%D 3.870 – – 1.240 −4.410 2.560 –
ECN/CNRS V BX S 4.620 3.125 1.495 0.213 0.0291 7.620 –
(ISISCFD/WF-EASM-AD) E%D 3.970 – – 1.840 −4.190 2.310 –
HSVA V BV S 4.652 – – 0.206 0.0255 7.803 –
(FreSCo+) E%D −3.310 3.126 1.526 −5.020 −8.650 0.040 –
KRISO LS BS S 4.813 3.207 1.606 0.216 0.0280 7.863 –

(WAVIS) E%D −0.042 – – 0.599 0.000 −0.812 –
MIJAC No BX S 4.555 2.957 1.598 0.210 0.0270 – 16.784
(OpenFOAM) (continued)
283
Table 2 (continued)
284

(Code) × 1000 × 1000 × 1000
E%D 5.319 – – 3.273 3.292 – 7.783
NMRI LS BX S 4.922 3.211 1.711 0.233 0.0293 7.560 –
(NAGISA) E%D −2.310 – – −7.520 −5.090 3.130 –
SJTU V BX S 4.621 2.839 1.782 0.222 0.0285 7.691 –
(naoeFOAM) E%D 3.949 – – −2.304 −2.151 1.397 –
SOTON No BP S 5.190 3.636 1.554 0.180 0.0248 – 11.339
(OpenFOAM) E%D −7.872 – 17.143 11.111 – 37.696
SOTON No BP S 4.762 3.391 1.371 0.158 0.0226 – 11.601
(STAR-CCM+) E%D 1.021 – – 27.189 18.996 – 36.256
YNU No BX S 4.573 3.084 1.490 0.218 0.0281 7.450 –
(SURFv7) E%D 4.950 – – −0.460 −0.720 4.490 –
T. Hino
Evaluation of Self-propulsion and Energy Saving Device … 285
Of 14 submissions using an actual propeller model, 4 solutions are with the force
balance approach and 10 are with the fixed revolution approach. Of 12 submissions
using body force models, 8 are with the force balance and 4 are with the fixed
revolution.
3.2 With ESD (Case 1.6a)
Table 3 (with an actual propeller model) and Table 4 (with body force models)
show 25 solutions submitted for Case 1.6a. Again, 14 submissions are using actual
propeller models and 11 submissions are with body force propeller models. Free
surface effects are not considered in 7 submissions. The other 17 submissions include
free surface effects via VOF model (12), level-set model (4) or potential flow model
(1). Of 14 submissions using an actual propeller model, 4 solutions are with the force
balance approach and 10 are with the fixed revolution approach. Of 12 submissions
using body force models, 8 are with the force balance and 4 are with the fixed
revolution. Although most of the submissions cover are both Cases 1.5a and 1.6a,
some organizations submitted one case only.
4 Uncertainty Estimation
Participants are requested to perform uncertainty analysis and to report the results.
Iterative uncertainty U I of C T , K T and K Q in the self-propulsion case is larger than that
of C T in the resistance case. The criterion of iterative convergence, |U I /ε12 | < 0.1,
is not satisfied in 38% of the submissions in case of C T and K Q and 50% in case of
KT .
Grid convergence for K T and K Q for Cases 1.5a and 1.6a is visualized in Figs. 1
and 2. Most of submissions seem to show convergent behaviors though divergence
and oscillation are observed in some cases. The averaged values of the estimated U G
except outliers are 4.84%S 1 and 3.59%S 1 for K T of Cases 1.5a and 1.6a, respectively
and 6.42%S 1 and 4.07%S 1 for K Q of Cases 1.5a and 1.6a, respectively as shown in
Tables 5 and 6.
5 Propeller Parameters
Figure 3 shows the comparison errors of all data of the thrust coefficient K T for Case
1.5a. All the submissions except the one with 67.7 M grid points employ the grid
less than 25 M points. The comparison errors are generally in the range between −
10%D and +5%D except for some outliers not depending on the number of grid
points. Larger errors are observed for the cases using body force models with fixed
Table 3 Submissions with actual propeller model for Case 1.6a
286

(Code) × 1000 × 1000 × 1000
EFD – – D 4.762 – – 0.233 0.0295 7.5 18.1
(NMRI)
ABS V A S 4.786 3.159 1.627 0.227 0.0297 7.673 –
(OpenFOAM) E%D −0.497 – – 2.746 −0.540 −2.303 –
CNR/INSEAN No A S N/A N/A N/A 0.242 0.0337 – N/A
(XNAVIS) E%D N/A – – −0.039 −0.142 – N/A
CSSRC No A S 4.682 3.122 1.560 0.236 0.0302 – 17.257
(FLUENT) E%D −0.016 – – 0.014 0.024 – −0.047
ECN/CNRS V A S 4.572 3.133 1.439 0.227 0.0310 – 17.320
(ISISCFD/WF-EASM-RANSE) E%D 3.990 – – 2.780 −3.520 – 4.310
HHI V A S 4.884 3.254 1.629 0.217 0.0279 – 20.300
(HiFoam) E%D −2.550 – – 0.223 5.520 – −12.280
HHI V A S 4.758 – – 0.236 0.0303 – 18.100
(STRA-CCM+) E%D 0.080 – – −1.120 −2.710 – 0.280
IIHR LS A S 4.731 3.190 1.540 0.237 0.0308 7.430 –
(REX-DDES) (continued)
T. Hino
Table 3 (continued)
(Code) × 1000 × 1000 × 1000
E%D 0.650 – – −1.720 −4.410 0.930 –
IIHR LS A S 5.158 3.130 2.030 0.238 0.030 7.995 –
(REX-DES) E%D −8.320 – – −2.150 −2.930 −6.600 –
IIHR LS A S 4.697 3.180 2.330 0.231 0.031 7.481 –
(REX-RANS) E%D 1.360 – – 1.000 −5.080 0.250 –
MARIC V A S 5.025 – – 0.240 0.0286 7.500 18.100
(FINEMarine) E%D 5.523 – – 3.005 −3.051 –
MARIN V A S 4.674 3.074 1.619 0.225 0.0291 – 18.392
(ReFreSCo) E%D 1.857 – – 3.582 1.462 – −1.611
PNU V A S 5.190 3.173 2.017 0.250 0.0300 7.829 18.202
(FLUENTv15) E%D −8.988 – – −7.296 −1.695 −4.387 −0.564
SHIME No A S 4.841 3.069 1.772 0.219 0.029 – 20.300
(OpenFOAM) E%D −1.669 – – 6.087 2.240 – −11.923

UDE V A S 4.858 3.005 1.862 0.235 0.0280 – 18.900
(STAR-CCM+) E%D −2.008 – – −0.944 5.085 – −4.420
287
Table 4 Submissions with body force propeller model for Case 1.6a
288

(Code) × 1000 × 1000 × 1000
EFD – – D 4.762 – – 0.233 0.0295 7.5 18.1
(NMRI)
CHALMERS-FLOWTECH Pot.Flow BL S 4.740 3.109 1.632 0.243 0.0292 7.306 –
(SHIPFLOW) E%D 0.450 – – −4.480 0.980 2.590 –
CNR/INSEAN No BX S N/A N/A N/A 0.222 0.0300 – N/A
(XNAVIS-HO) E%D N/A – – 0.047 −0.017 – N/A
ECN/CNRS V BX S 4.660 3.103 1.563 0.239 0.0306 7.310 –
(ISISCFD/LRN-EASM-AD) E%D 2.140 – – −2.360 −3.660 2.530 –
ECN/CNRS V BX S 4.617 3.116 1.501 0.233 0.0305 7.330 –
(ISISCFD/WF-EASM-AD) E%D 3.040 – – 0.130 −3.250 2.270 –
HSVA V BV S 4.584 3.111 1.473 0.219 0.0272 7.522 –
(FreSCo+) E%D −3.730 – – −6.190 −7.640 0.290 –
KRISO LS BP S 4.760 3.204 1.556 0.231 0.0290 7.580 –
(WAVIS) E%D 0.042 – – 0.858 1.017 −1.067 –
MIJAC No BX S 4.523 2.926 1.597 0.233 0.0287 – 16.146
(OpenFOAM) (continued)
T. Hino
Table 4 (continued)
(Code) × 1000 × 1000 × 1000
E%D 5.013 – – 0.214 2.766 – 10.795
NMRI LS BX S 4.893 3.214 1.680 0.240 0.0300 7.440 –
(NAGISA) E%D −2.760 – – −3.050 −1.760 0.760 –
SJTU V BX S 4.596 2.548 2.048 0.239 0.0299 7.402 –
(naoeFOAM) E%D 3.486 – – −2.575 −1.356 1.307 –
SOTON No BP S 4.468 3.196 1.272 0.180 0.0248 – 9.226
(OpenFOAM) E%D 6.172 – 22.833 15.932 – 49.029
YNU No BX S 4.552 3.082 1.470 0.233 0.0294 7.220 –
(SURFv7) E%D 4.410 – – 0.000 0.340 3.730 –
289
290 T. Hino
Fig. 1 Grid convergence of K T and K Q for Case 1.5a
revolution. Figure 4 is the same plot for the torque coefficient K Q . The error range is
from −10%D to +5%D, again except for outliers. Figure 5 is the errors of the towing
force RT (SP) − T for 13 solutions using the fixed revolution approach and Fig. 6 is
the errors of propeller revolution n for 11 solutions using the force balance approach.
Scatter of the towing force is in the range of ±10%D and it does not depend on the
grid size, whereas, the error of propeller revolution is from −1%D to 5%D.
Averages and standard deviations of the comparison errors of K T and K Q are
listed in Table 5. Average of absolute errors are 3 to 4%D. Differences between
actual propeller models and body force models are not large.
Figures 7 and 8 show the comparison errors of K T and K Q in Case 1.6a, respec-
tively. Numbers of grid points in Case 1.6a are distributed in the range less than
25 M except the large grid cases with 70.9 M points. The grid points are, in general,
slightly larger than those of Case 1.5a since the presence of the duct increases the
complexity of the geometry. The ranges of errors in K T and K Q are from −10%D
and 5%D which is almost the same as Case 1.5a shown in Figs. 5 and 6.
Figures 9 and 10 is the comparison errors of the towing force RT (SP) − T and the
propeller revolution n. The errors of RT (SP) − T are in the range of ±10%D except
Fig. 2 Grid convergence of K T and K Q for Case 1.6a
Table 5 Errors of thrust and torque coefficients in Case 1.5a

Items KT KQ
E%D |E|%D σSD %D UG%D E%D |E|%D σSD %D UG%D
Actual Prop (14/24) −0.30 3.09 4.41 −1.37 3.42 3.94
Modeled Prop (10/24) 0.11 4.04 5.60 −1.81 3.25 3.70
Force balance (12/24) −1.85 3.65 4.62 −2.82 2.82 2.48
Fixed n (12/24) 1.59 3.32 4.64 −0.29 3.88 4.50
All (24) −0.13 3.49 4.94 4.84 −1.55 3.35 3.85 6.42
the outliers and this is in the same level as Case 1.5a. The scatter of the revolution
rate n is also in the same range as Case 1.5a.
Averages and standard deviations of the comparison errors of K T and K Q are
listed in Table 6. Average of the absolute errors |E| of K T and K Q are about 2%D in
Case 1.6a with EFD and lightly smaller than those in Case 1.5a without ESD which
is 3% D. Average error of actual propeller models is slightly better than body force
propeller models in this case.
292 T. Hino
Table 6 Errors of thrust and torque coefficients in Case 1.6a

Items KT KQ
E%D |E|%D σSD %D UG%D E%D |E|%D σSD %D UG%D
Actual Prop (14/26) 0.44 2.34 3.11 −0.70 2.74 3.19
Modeled Prop (12/26) 2.36 5.46 9.39 1.61 4.55 6.91
Force balance (15/26) −1.58 2.68 2.98 −2.36 2.69 2.36
Fixed n (11/26) 4.71 5.06 8.35 3.55 4.61 6.11
All (26) −0.43 2.02 2.87 3.59 −0.86 2.35 2.96 4.07
Fig. 3 Comparison errors of K T in Case 1.5a
6 Self-prolusion Factors
One of the goals of self-propulsion simulations is to obtain self-propulsion factors,

i.e., thrust deduction and the wake coefficients, 1 − t and 1 − w, respectively and the
relative rotative efficiency η R for power estimation. To estimate the self-propulsion
factors, the resistance data and the propeller open water characteristics are needed in
addition to the self propulsion data. Although the resistance data is available from the
corresponding submissions to Cases 1.1a and 1.2a, the open water characteristics data
of individual propeller models are not included in the requested items. Therefore,
the result of the experimental propeller open test is used for the present analysis.
Figure 11 shows the propeller open water characteristics of the experiment for the
model propeller MP.687 which was provided to participants.
The procedure to estimate self-propulsion factors is essentially same as that in the
experiment as summarized below:
Fig. 4 Comparison errors of K Q in Case 1.5a
Fig. 5 Comparison errors of RT (SP) − T in Case 1.5a
First, from the computed K T , K Q and n, the wake coefficient 1 − w is obtained by

using the thrust identity method with the measured propeller open water character-
istics and the propeller efficiency behind hull η B is computed using thrust T, torque
Q and the ship speed U as,
T (1 − w)U
ηB =
2π n Q
294 T. Hino
Fig. 6 Comparison errors of n in Case 1.5a
Fig. 7 Comparison errors of K T in Case 1.6a
The relative rotative efficiency is η R = η B /ηo , where ηo is the propeller open

water efficiency. From thrust T and total resistance in towing condition R(tow) from
Ct (tow) in Case 1.1a or 1.2a with the towing force SFC for the ship point condition,
thrust deduction coefficient 1 − t is obtained by
R(tow) − S FC
1−t =
T
Fig. 8 Comparison errors of K Q in Case 1.6a
Fig. 9 Comparison errors of RT (SP) − T in Case 1.6a
Finally, the model scale Delivered Power (DP) is estimated by
T (1 − w)U
DP =
ηB
Estimated self-propulsion factors and DPs in Case 1.5a are listed in Tables 7 for
actual propeller models and Table 8 for body force propeller models. Comparison
296 T. Hino
Fig. 10 Comparison errors of n in Case 1.6a
Fig. 11 Measured propeller open characteristics of model propeller MP. 687
errors of 1 − t and 1 − w are shown in Figs. 12 and 13, respectively. Error ranges
are from −20%D to +10%D for 1 − t and from −10%D to 5%D for 1 − w without
outliers and the scatters are not dependent on the number of grids. Average of errors
are listed in Table 9. Absolute errors for 1 − t and 1 − w are 11%D and 7%D,
which are considerably larger than errors of K T and K Q which are around 3%D as
shown in Table 5. The standard deviations values are also quite large (24%D and
11%D). From the definition, self-propulsion factors are the results of complicated
hydrodynamic interaction between a ship hull and a propeller. Not only K T , K Q
Table 7 Self-propulsion factors with actual propeller models for Case 1.5a
Organization Free surface Prop. M. 1−t 1−w ηR ηO DP(Kw)
(Code)
EFD – – D 0.804 0.554 1.015 0.501 0.0290
(NMRI)
ABS V A S 0.808 0.568 0.984 0.511 0.0296
(OpenFOAM) E%D −0.440 −2.481 3.834 −1.840 −2.026
CSSRC No A S – 0.532 0.992 0.487 0.0299
(FLUENT) E%D – 4.082 2.859 2.857 −2.870
ECN/CNRS V A S 0.792 0.562 0.967 0.508 0.0298
(ISISCFD/WF-EASM-RANSE) E%D 1.482 −1.336 5.969 −1.385 −2.517
HHI V A S 0.828 0.552 1.015 0.501 0.0287
(HiFoam) E%D −2.919 0.421 0.000 0.000 1.258
HHI V A S 0.804 0.530 0.985 0.485 0.0304
(STRA-CCM+) E%D −0.002 4.471 3.733 3.345 −4.652
IIHR LS A S – 0.546 0.979 0.497 0.0301

(REX-DDES) E%D – 1.435 4.419 0.939 −3.584
MARIC V A S 0.790 0.515 1.054 0.475 0.0285
(FINEMarine) E%D 1.747 7.136 −4.935 5.329 1.725
(continued)
297
Table 7 (continued)
298

(Code)
MARIN V A S 0.803 0.568 0.987 0.513 0.0289
(ReFreSCo) E%D 0.177 −2.527 3.374 −2.291 0.311
PNU V A S 0.728 0.464 1.078 0.432 0.0300
(FLUENTv15) E%D 9.516 16.365 −7.930 13.768 −3.525
SHIME No A S 0.773 0.555 1.008 0.504 0.029
(OpenFOAM) E%D 3.897 −0.140 0.859 −0.465 1.018
UDE V A S – 0.559 1.065 0.506 0.0271
(STAR-CCM+) E%D – −0.739 −6.303 −0.926 6.675
UNI-ROSTOCK No A S 0.712 0.559 1.042 0.506 0.028
(OpenFOAM-SST) E%D 11.513 −0.739 −3.359 −0.926 4.553
UNI-ROSTOCK No A S 0.637 0.501 1.023 0.464 0.030
(OpenFOAM-HYBRID) E%D 20.734 9.602 −1.055 7.364 −2.517
T. Hino
Table 8 Self-propulsion factors with body force propeller models for Case 1.5a
(Code)
EFD – – D 0.804 0.554 1.015 0.501 0.0290
(NMRI)
CHALMERS-FLOWTECH Pot.Flow BL S 0.813 0.506 1.044 0.480 0.0262
(SHIPFLOW) E%D −1.150 8.661 −3.659 4.331 9.771
ECN/CNRS V BX S 0.871 0.547 0.963 0.508 0.0276
(ISISCFD/LRN-EASM-AD) E%D −8.306 1.263 6.382 −1.385 4.842
ECN/CNRS V BX S 0.766 0.552 0.960 0.511 0.0278
(ISISCFD/WF-EASM-AD) E%D 4.795 0.420 6.777 −1.840 4.089
HSVA V BV S 0.770 0.588 1.070 0.526 0.0262
(FreSCo+) E%D 4.230 −6.114 −6.909 −4.936 9.752
KRISO LS BS S 0.805 0.560 1.008 0.504 0.0294
(WAVIS) E%D −0.147 −0.949 0.859 −0.465 −1.399
MIJAC No BX S 0.816 0.575 1.024 0.517 0.0277

(OpenFOAM) E%D −1.418 −3.713 −1.230 −3.186 4.553
NMRI LS BX S 0.829 0.482 1.016 0.462 0.0274
(NAGISA) E%D −3.046 12.984 −0.132 7.881 5.693
(continued)
299
Table 8 (continued)
300

(Code)
SJTU V BX S 0.807 0.528 1.009 0.489 0.0280
(naoeFOAM) E%D −0.331 4.824 0.661 2.373 3.416
SOTON No BP S 1.393 0.671 1.000 0.577 0.0254
(OpenFOAM) E%D −73.253 −20.998 1.768 −15.047 12.331
SOTON No BP S 1.471 0.738 1.002 0.610 0.0232
(STAR-CCM+) E%D −82.955 −33.059 1.595 −21.692 20.108
YNU No BX S 0.795 0.524 1.011 0.499 0.0251
(SURFv7) E%D 1.083 5.500 0.492 0.468 13.446
T. Hino
Fig. 12 Comparison errors of 1 − t in Case 1.5a
Fig. 13 Comparison errors of 1 − w in Case 1.5a
and n from self-propulsion simulations but also C T from resistance simulations are
involved in the analysis. These points are attributed to the larger errors and scatters
of self-propulsion factors.
For Case 1.6a, the estimated values of self-propulsion factors and DPs are listed in
Tables 10 and 11 for actual propeller models and for body force models, respectively.
Comparison errors of 1 − t and 1 − w are shown in Figs. 14 and 15, respectively.
Errors of 1 − t are from −10%D to 5%D and errors of 1 − w are in the range of
302 T. Hino
Table 9 Errors of self-propulsion factors and delivered power in Case 1.5a

Items 1−t 1−w DP
E%D |E|%D σSD %D E%D |E|%D σSD %D E%D |E|%D σSD %D
Actual Prop −4.57 5.24 6.88 −2.27 3.94 5.39 0.97 3.19 3.64
(14/26)
Modeled Prop 14.59 16.43 30.20 4.02 9.63 12.85 −7.66 7.89 5.47
(12/26)
Force −0.72 3.16 6.22 −4.09 5.68 6.22 −3.52 5.27 5.34
balance(12/26)
Fixed n (14/26) 12.28 19.84 33.66 4.67 7.33 10.97 −2.58 5.44 6.96
All (26) 5.47 11.10 24.33 0.63 6.57 10.09 −3.01 5.36 6.28
± 10%D except outliers. Error ranges are slightly smaller than Case 1.5a. Average
errors and the standard deviations are shown in Table 12. Absolute errors for 1 − t
and 1 − w are 3%D and 5%D and smaller than Case 1.5a. Due to presence of ESD,
stern flow fields are smoothened to some extent and this may improve the accuracy
of self-propulsion factors.
Absolute errors of model scale DPs are 5%D for Case 1.5a and 4%D for Case1.6a,
which are smaller than errors of self-propulsion factors in Case 1.5a. Since DPs are
integrated quantity depending on all factors such as resistance, thrust and torque,
some errors may cancel out during estimations.
7 ESD Performance
Based on the model scale Delivered Powers (DPs) derived above, ESD performance
which is the reduction DP due to the ESD is estimated. DP reduction rate which
is defined as DP(ESD) /DP(Bare) is listed in Tables 10 and 11. From the experimental
result, DP reduction is 0.94, that is, 6% power reduction in model scale. The estimated
DP reductions are plotted Fig. 16. Most of simulations show DP reductions to some
extent. However, the scatters are considerably large and reduction rates are from 0.88
to 1.04.
In view that a few percent power reduction is the typical ESD performance, the
current status of ESD performance estimation is not sufficient. In order to achieve
reliable prediction of ESD effects, the accuracy improvement in all the aspects,
resistance, wake, propeller effect is required, particularly in the estimations of self-
propulsion factors.
Table 10 Self-propulsion factors with actual propeller models for Case 1.6a
Organization Free surface Prop. M. 1−t 1−w ηR ηO DP(Kw) DP(Kw) ESD/
(Code) bare Bare
EFD – – D 0.81107 0.48087 1.00878 0.462 0.0273 0.0290 0.940
(NMRI)
ABS V A S 0.807 0.510 0.984 0.477 0.0290 0.0296 0.980
(OpenFOAM) E%D 0.460 −6.026 3.080 −3.314 −6.312 −2.026 −4.200
CSSRC No A S – 0.469 0.994 0.454 0.0275 0.0299 0.923
(FLUENT) E%D – 2.534 1.789 1.703 −0.953 −2.870 1.864
ECN/CNRS V A S 0.852 0.498 0.943 0.477 0.0283 0.0298 0.950
(ISISCFD/WF-EASM-RANSE) E%D −4.992 −3.636 8.167 −3.314 −3.627 −2.517 −1.083
HHI V A S 0.793 0.531 1.015 0.501 0.0254 0.0287 0.888
(HiFoam) E%D 2.246 −10.358 −0.718 −8.555 6.736 1.258 5.547
HHI V A S 0.829 0.469 0.991 0.454 0.0276 0.0304 0.910
(STRA-CCM+) E%D −2.218 2.534 2.194 1.703 −1.287 −4.652 3.215
IIHR LS A S – 0.461 0.978 0.451 0.0273 0.0301 0.909

(REX-DDES) E%D – 4.129 3.821 2.277 −0.103 −3.584 3.361
IIHR LS A S – 0.493 1.007 0.449 0.033 0.0285 1.162
(REX-DES) E%D – −2.421 0.241 2.855 −21.480 1.725 −23.613
(continued)
303
304
Organization Free surface Prop. M. 1−t 1−w ηR ηO DP(Kw) DP(Kw) ESD/

(Code) bare Bare
IIHR LS A S – 0.484 0.954 0.467 0.028 0.0289 0.970
(REX-RANS) E%D – −0.652 6.732 −1.118 −2.842 0.311 −3.163
MARIC V A S 0.815 0.455 1.062 0.443 0.0261 0.0300 0.868
(FINEMarine) E%D −0.514 5.312 −6.599 4.020 4.396 −3.525 7.651
MARIN V A S 0.783 0.505 0.998 0.482 0.0265 0.029 0.924
(ReFreSCo) E%D 3.464 −4.991 1.347 −4.390 2.724 1.018 1.723
PNU V A S 0.718 0.440 1.042 0.415 0.0311 0.0271 1.149
(FLUENTv15) E%D 11.486 8.515 −4.081 10.052 −14.069 6.675 −22.227
SHIME No A S 0.762 0.524 0.982 0.497 0.026 0.028 0.955
(OpenFOAM) E%D 6.004 −9.025 3.248 −7.535 3.058 4.553 −1.566
UDE V A S – 0.472 1.069 0.457 0.0255 0.030 0.858
(STAR-CCM+) E%D – 1.842 −7.450 1.132 6.401 −2.517 8.699
T. Hino
Table 11 Self-propulsion factors with body force propeller models for Case 1.6a
Organization Free surface Prop. M. 1−t 1−w ηR ηO DP (Kw) DP(Kw) ESD/
(Code) bare bare
EFD – – D 0.81107 0.48087 1.00878 0.462 0.0273 0.0290 0.940
(NMRI)
CHALMERS-FLOWTECH Pot.Flow BL S 0.822 0.434 1.050 0.435 0.0246 0.0262 0.940
(SHIPFLOW) E%D −1.369 9.805 −5.023 5.793 9.770 9.771 0.000
ECN/CNRS V BX S 0.823 0.447 0.990 0.446 0.0258 0.0276 0.936
(ISISCFD/LRN-EASM-AD) E%D −1.450 7.031 2.318 3.436 5.289 4.842 0.470
ECN/CNRS V BX S 0.789 0.468 0.976 0.462 0.0260 0.0278 0.933
(ISISCFD/WF-EASM-AD) E%D 2.736 2.721 4.078 0.000 4.821 4.089 0.764
HSVA V BV S 0.774 0.526 1.047 0.497 0.0250 0.0262 0.956
(FreSCo+) E%D 4.532 −9.345 −4.768 −7.535 8.273 9.752 −1.639
KRISO LS BP S 0.810 0.490 1.020 0.467 0.0273 0.0294 0.928
(WAVIS) E%D 0.142 −1.984 −1.382 −1.118 −0.077 −1.399 1.304
MIJAC No BX S 0.817 0.479 1.037 0.462 0.0262 0.0277 0.945

(OpenFOAM) E%D −0.678 0.464 −3.467 0.000 4.061 4.553 −0.516
NMRI LS BX S 0.827 0.452 1.013 0.443 0.0267 0.0274 0.976
(NAGISA) E%D −1.909 6.069 −0.486 4.020 2.103 5.693 −3.806
(continued)
305
306
Organization Free surface Prop. M. 1−t 1−w ηR ηO DP (Kw) DP(Kw) ESD/

(Code) bare bare
SJTU V BX S 0.811 0.453 1.013 0.446 0.0262 0.0280 0.935
(naoeFOAM) E%D −0.032 5.861 −0.539 3.436 3.917 3.416 0.519
SOTON No BP S 1.575 0.645 1.000 0.577 0.0226 0.0254 0.889
(OpenFOAM) E%D −94.159 −34.148 1.036 −24.889 17.098 12.331 5.438
YNU No BX S 0.802 0.461 1.012 0.462 0.0239 0.0251 0.952
(SURFv7) E%D 1.177 4.180 −0.423 0.000 12.323 13.446 −1.298
T. Hino
Fig. 14 Comparison errors of 1 − t in Case 1.6a
Fig. 15 Comparison errors of 1 − w in Case 1.6a
8 Conclusions
Conclusions from the analysis of submitted data for JBC self-propulsion cases are
summarized as follows:
• Grid uncertainty U G in general is about 4–5%S for K T and K Q .
308 T. Hino
Table 12 Errors of self-propulsion factors and delivered power in Case 1.6a

Items 1−t 1−w DP
E%D |E|%D σSD %D E%D |E|%D σSD %D E%D |E|%D σSD %D
Actual Prop 1.99 3.92 4.79 −0.94 4.77 5.49 −0.34 4.00 4.87
(8, 13, 8)
Modeled Prop 0.35 1.56 2.01 2.76 5.27 5.42 0.89 2.82 3.71
(9, 9, 9)
Force balance 1.39 2.35 3.67 2.37 5.29 5.44 1.63 3.50 4.11
(11, 14, 14)
Fixed n (6, 8, 0.64 3.27 3.67 −2.58 4.42 4.87 −2.41 3.54 3.89
8)
All (17, 22, 1.12 2.67 3.69 0.57 4.97 5.76 0.16 3.51 4.48
17)
Fig. 16 Comparison of DP Reductions due to the Duct
• Averages of absolute comparison errors of K T and K Q are 3–4%D for a bare hull
case and 2–3%D with ESD case. ESD case is slightly smaller than a bare hull
case. Actual propeller models give slightly better results than body force models
though the difference is smaller in case of a bare hull.
• Averages of absolute comparison errors of self-propulsion factors, 1 − t and 1 −
w are 7–10%D in a bare hull case, which are considerably larger than the errors
of thrust and torque. This is due to the increased number of quantities involved to
estimate the factors. Errors become smaller with the presence of a duct, maybe
due to the smoothening of stern flow fields by a duct.
• Average absolute errors of model scale delivered power (DP) is in the range of 5–
6%, about the same level of other quantities. However, the estimated DP reduction
rates due to a duct scatter from 0.88 to 1.0, whereas the experimental value is
0.94. The current accuracy of CFD estimation of ESD performance seems to be
reasonably well in terms of the magnitude, although this may not be sufficient for
design/performance prediction of an energy saving duct which aims the reduction
rate of a few percent.
Evaluation of Resistance, Sinkage, Trim
and Self-propulsion Predictions for KCS
Jin Kim
Abstract This chapter discusses the results of resistance predictions including trim
and sinkage (the workshop case 2.1) and the results of self-propulsion predictions
(the workshop case 2.5) for KCS. The resistance predictions for 6 different Froude
numbers in trim and sinkage free condition were requested. The comparison error
at the design speed (Fr = 0.26) is −0.2% and the standard deviation is 1.5% of
the data value. The mean comparison error for all 6 speeds is 0.43% and the mean
standard deviation is 2.48%. The increased error and standard deviation is caused by
the results for low Fr simulation submissions. The submitted trim and sinkage also
shows larger comparison error in low Fr region (Fr < 0.2). Self-propulsion results are
reported with both towing force (FD) fixed and propeller revolution (rps) fixed. The
participants using a body force propeller model based on potential theory selected FD
fixed condition for self-propulsion and the participants who adopted actual propeller
rotating simulation preferred rps fixed condition. The mean comparison errors of KT
and KQ are 0.5 and −3.5% respectively. The standard deviations are 2.7 and 2.4%
respectively. Self-propulsion parameters are slightly better predicted by body force
models (FD fixed). However, local flow characteristics are better predicted by actual
propeller rotating simulation (rps fixed).
1 Resistance, Sinkage & Trim
1.1 Participants
The KCS resistance cases of the Workshop were the case 2.1 for the bare hull with
a rudder, which was the same case in Gothenburg 2010 case 2.2b (Larsson et al.
2010, 2013). The KCS model ship is towed in a condition of free to heave and pitch
motion with 6 different speeds in a calm water. This case received 13 submissions
J. Kim (B)
Korea Research Institute of Ships & Ocean Engineering (KRISO), Daejeon, Korea
312 J. Kim
of computational results from 11 different organizations. The 13 cases with reported

computational scheme and grid information are listed in the Table 1.
The number of participants is increased compared to the previous Gothenburg
2010 workshop. Most participants used commercial and open source codes except
of WAVIS, an in-house code of KRISO. The finite volume method with unstructured
grid is used by most participants. The grid size varies from 1.1 to 6.6 M. All submis-
sions except one use two equation turbulence models. For the ship motion of pitch
and heave, all moving, deforming grid and overset schemes are used.
The Fig. 1 shows the total resistance (C T ) results of all participants for 6 different
Froude numbers (Fr). The submitted results are still showing large deviation from
the experimental values. The results are more scattered as Fr decreases. The Figs. 2
and 3 show sinkage and trim results of all participants for 6 different Fr’s. The error
bars show the standard deviation of CFD submissions except the outliers.
Table 1 Computational schemes and grid information of all participants for the case 2.1
Participants Grid Free Turbulence Grid Discretization Grid Code
type surface motion size
CTO Unstr. VOF Two-eqn. All FV − Star-CCM
moving
MARIC − VOF Two-eqn. Deforming FV − FINEMarine
grid
UNIZAG Unstr. VOF Two-eqn. All FV 4.7 M NavalFoam
moving
HHI Unstr. VOF RSM All FV 1.1 M Star-CCM
moving
UDE(C) − − − − − − Comet
UDE(F) Unstr. VOF Two-eqn. Deforming FV 1.1 M FINEMarine
grid
UDE(O) Unstr. VOF Two-eqn. Deforming FV 1.1 M OpenFoam
grid
SJTU Unstr. VOF Two-eqn. Overset FV 3.4 M NaoeFoam
KRISO Str. Level-Set Two-eqn. All FV 4.3 M WAVIS
moving
DAMEN Unstr. VOF Two-eqn. Deforming FV 2.6 M FINEMarine
grid
NUMECA Unstr. VOF Two-eqn. Deforming FV 6.2 M ISIS
grid
UM Unstr. VOF Two-eqn. All FV 2.6 M OpenFoam
moving
PNU Unstr. VOF Two-eqn. All FV 6.6 M Fluent
moving
Evaluation of Resistance, Sinkage, Trim and Self-Propulsion … 313
Fig. 1 Total resistance results for all participants
Fig. 2 Sinkage results for all participants

314 J. Kim
Fig. 3 Trim results for all participants
1.2 Statistics and Error Estimation
The case is exactly same as the case 2.1b of the previous Gothenburg 2010 work-
shop. We are interested in both how well resistance, sinkage and trim can be predicted
and how much the results are improved compared to the previous workshop. Espe-
cially, resistance is the primary parameter for the evaluation of ship hydrodynamic
performances.
All computational results are shown in the previous Sect. 1.1 and we can see that
a couple of them are very different from other results. The results are considered
as outliers when the absolute error, |E| is larger than 2σ, where E = D-S and σ is
the standard deviation. The Tables 2, 3 and 4 show all submitted results for total
resistance, sinkage and trim respectively and outlier (|E| > 2σ) is marked as X.
The analyzed statistics and error estimation with and without outliers are shown
for each Fr in Tables 5, 6, 7, 8, 9 and 10, where T2015 means the statistics with outlier
and T2015*, without outliers. The errors and standard deviations are compared with
the previous Gothenburg 2010 workshop and the absolute error |E| and standard devi-
ation based on the absolute error is analyzed in the present analysis. The submitted
frictional resistance (C F ) and pressure resistance (C P ) are separately shown in Figs. 4,
5, 6, 7, 8 and 9 compared with the experimental results (EFD) and averaged computa-
tional results (CFDave ). In the group discussion for this case, Dr. Starke from MARIN
pointed out that a couple of entries unrealistically had friction resistances well below
a plate friction line (ATTC line). We saw the misuse of wall functions as a likely
Table 2 All submitted results for total resistance

Fn 0.108 0.152 0.195 0.227 0.260 0.282
CTO O O O O O O
MARIC O O O O O O
UNIZAG-FSB O O O O O O
HHI O O O O O O
UDE-Comet O N/A N/A O N/A O
UDE-FINEMarine O O O O O N/A
UDE-OpenFOAM N/A N/A N/A O N/A O
SJTU O O O O O O
KRISO O O O O O O
DAMEN O N/A N/A N/A N/A N/A
NUMECA O O O O O O
UM O O O O O O
PNU O O O O O X
Table 3 All submitted results for sinkage

Fn 0.108 0.152 0.195 0.227 0.260 0.282
CTO O O O O O O
MARIC O O O O O O
HHI O O O O O O
SJTU X O O O O O
KRISO O O O O O O
NUMECA O O O O O O
UM O O O O O O
PNU O O X X X X
candidate to blame: two of the participants with low Cf-values, also reported mean
y + values around 15, i.e. in the buffer layer. The ITTC and ATTC lines based on a
flat plate friction are also given in the Figs. 4−9.
Mean values of the total resistance, sinkage and trim for corresponding Fr are
plotted in the Figs. 10, 11 and 12 respectively. The Table 11 shows statistics and
error estimation based on the mean values of computational results. The error, E
and absolute error, |E| are obtained as 0.43 and 2.0%D respectively. Both errors are
316 J. Kim
Table 4 All submitted results for trim

Fn 0.108 0.152 0.195 0.227 0.260 0.282
CTO O O X X O O
MARIC O O O O O O
HHI O O O O O O
SJTU O X O O O O
KRISO O O O O O O
NUMECA O O O O O O
UM O O O O O O
PNU O O O O X X
Table 5 Statistics and error estimation for Fn = 0.108

G2010 T2015 T2015*
Participants 4 12
CT Emean %D +1.6 −0.4 0.3
CT σSD %D 1.4 4.2 3.5
Sinkage Emean %D −67.1 −89.0 −78.7
Sinkage σSD %D 8.2 50.3 36.8
Trim Emean %D −60.3 −54.8 −48.0
Trim σSD %D 38.6 45.5 40.8
CT |E|mean %D N/A 3.4 3.0
CT σSD |E|%D N/A 2.2 1.6
Sinkage |E|mean %D N/A 89.0 78.7
Sinkage σSD |E|%D N/A 50.3 36.8
Trim |E|mean %D N/A 59.7 53.4
Trim σSD |E|%D N/A 38.1 32.7
slightly increased since E = −0.3%D and |E| = 1.64%D in G2010. The standard
deviation is also increased from 1.3 to 2.48%D.
Note that the number of participants is three times larger than at G2010.
Presumably more unexperienced users participated in 2015.
For the predictions of resistance, sinkage and trim, the number of grid points are
varied from 1.1 to 6.6 M. The submitted results do not show large dependency of grid
points as shown in Fig. 17. It means that CFD users who have efficient grid making

G2010 T2015 T2015*
Participants 3 10
CT Emean %D −0.1 0.3 0.7
CT σSD %D 1.1 3.1 3.0
Sinkage Emean %D −44.1 −41.9 −37.2
Sinkage σSD %D 15.1 23.5 19.3
Trim Emean %D 0.7 0.4 5.1
Trim σSD %D 4.9 19.0 12.7
CT |E|mean %D N/A 2.3 2.4
CT σSD |E|%D N/A 1.7 1.8
Sinkage |E|mean %D N/A 43.9 38.2
Sinkage σSD |E|%D N/A 21.7 17.0
Trim |E|mean %D N/A 14.7 10.0
Trim σSD |E|%D N/A 13.0 8.9

G2010 T2015 T2015*
Participants 4 10
CT Emean %D −0.9 1.3 1.6
CT σSD %D 1.7 2.4 2.5
Sinkage Emean %D −12.9 −20.2 −19.0
Sinkage σSD %D 6.6 9.5 5.5
Trim Emean %D 1.6 15.0 2.0
Trim σSD %D 3.8 30.8 8.0
CT |E|mean %D NA 2.0 2.2
CT σSD |E|%D NA 1.7 1.9
Sinkage |E|mean %D NA 20.2 19.0
Sinkage σSD |E|%D NA 9.5 5.5
Trim |E|mean %D NA 18.8 6.7
Trim σSD |E|%D NA 28.4 4.2
techniques can have an acceptable simulation results for low block coefficient hull
form such as KCS (Fig. 13).
318 J. Kim

G2010 T2015 T2015*
Participants 6 12
CT Emean %D −1.0 0.5 0.5
CT σSD %D 1.4 2.7 2.7
Sinkage Emean %D −4.2 −7.5 −7.0
Trim Emean %D −3.2 7.9 −3.5
Trim σSD %D 4.6 29.8 7.7
CT σSD |E|%D NA 1.5 1.7

G2010 T2015 T2015*
Participants 4 10
CT Emean %D −0.3 −0.2 −0.5
CT σSD %D 1.2 1.5 1.3
Sinkage Emean %D 1.5 −1.5 0.2
Trim Emean %D −3.1 2.6 −0.2
Trim σSD %D 3.0 10.3 5.6
CT σSD |E|%D NA 1.0 1.0
2 Self-Propulsion
2.1 Participants
The KCS self-propulsion case of the Workshop was the case 2.5 for the bare hull with
a rotating propeller and without a rudder. Both heave and pitch motions are fixed.

G2010 T2015 T2015*
Participants 6 11
CT Emean %D −0.4 0.7 −0.0
CT σSD %D 0.8 2.9 1.9
Sinkage Emean %D −2.1 −2.9 −1.6
Trim Emean %D 3.6 3.6 5.9
Trim σSD %D 4.2 9.0 5.2
CT σSD |E|%D NA 2.1 1.1
Fig. 4 Comparison of the resistance components and the experimental data for Fn = 0.108
The computations were requested for the model at the ship point that the model ship
was towed to account for the larger skin friction at model scale compared to full
scale. The towing force, the skin friction correction, SFC, was pre-computed and
was the same as in the experiments. In the experiments the thrust T, was adjusted by
varying the rps, n, such that T = RT(SP) − SFC, where RT(SP) is the resistance with
a rotating propeller. 3 participants followed this procedure, but the others computed
320 J. Kim
the flow with the measured rps from the experiment. In the first case, the achieved n
from numerical force balancing was requested, while in the second case the resulting
towing force, RT(SP) − SFC, was to be reported. The participants for the first case
used a body force propeller with a Lifting Surface method, the others used an actual
rotating propeller. The computational schemes, grid information and the ship point
are shown in the Table 12.
322 J. Kim
Fig. 10 Comparison with mean total resistance and experiment
2.2 Statistics and Error Estimation
All computational results of the resistance at self-propelled condition, RT(SP) , are

shown in the Fig. 14. For better statics and error estimation, two submissions are
considered as outliers because the absolute error, |E| is larger than 2σ, where E =
Fig. 11 Comparison with mean sinkage and experiment
Fig. 12 Comparison with mean computed trim and experiment
D-S and σ is the standard deviation. The statics and errors are compared with the
previous workshops in the Table 13. The errors of RT(SP) are 1%D which is slightly
bigger than −0.3%D at G2010. Tables 14 and 15 show the error estimation for the
FD fixed cases and rps fixed cases, respectively.
324 J. Kim
Table 11 Statistics and error estimation based on mean values for each Fn
G2010 T2015*
Participants 4 13
CT Emean %D −0.3 0.43
CT σSD %D 1.3 2.48
Sinkage Emean %D −21.9 −23.87
Sinkage σSD %D 9.9 12.09
Trim Emean %D −9.6 −6.45
Trim σSD %D 11.2 13.34
CT |E|mean %D 1.64 2.00
Sinkage |E|mean %D (Fr < 0.2) 55.6 45.26
Sinkage |E|mean %D (Fr ≥ 0.2) 7.5 4.25
Trim |E|mean %D (Fr < 0.2) 30.5 23.34
Trim |E|mean %D (Fr ≥ 0.2) 3.62 5.29
Fig. 13 CT comparison at Fn = 0.26 w.r.t. no. of grids
2.3 Local Flow Characteristics
Figure 15 shows the computed axial velocity contours and cross flow vectors at
propeller plane and the velocity profiles downstream of propeller plane at z/Lpp =
−0.03 are depicted in Fig. 16. It is confirmed that the methods using actual propeller
Table 12 Computational schemes and grid information of all participants for the case 2.5
Participants Grid Type Turbulence Grid Size Propeller Ship point Code
model
CTO Unstructured Two-eqn. − − FD fixed Star-CCM
HHI Unstructured RSM 4.5 M Actual RPS fixed Star-CCM
rotating
MARIN Structured One-eqn. 12 M Lifting FD fixed PARNASSOS
Surface
MARIC Unstructured Two-eqn. − − RPS fixed FINEMarine
KRISO Structured Two-eqn. 8.6 M Lifting FD fixed WAVIS
Surface
UM Unstructured Two-eqn. 2.6 M Actual RPS fixed OpenFoam
rotating
PNU Unstructured Two-eqn. 15 M Actual − Fluent
rotating
Fig. 14 The resistance at self-propelled condition
model better predict the crescent shape of high-speed region of the axial velocity
contour. Figure 17 shows the comparison of hull surface pressure distributions.
326 J. Kim
Table 13 Statistics and error estimation for the case 2.5

T2005 G2010 T2015
Participants 4 13 5/7
CT Emean %D −0.9 −0.3 1.0
σSD %D 1.0 3.1 1.0
KT Emean %D −0.6 0.5
σSD %D 7.2 2.7
KQ Emean %D −4.6 −3.5
σSD %D 6.1 2.4
Rps Emean %D 0.6 −0.3
σSD %D 2.8 1.8
RT -T Emean %D −7.8 3.9
σSD %D 4.4 1.1
Table 14 Error estimation for the case 2.5 (fixed FD case)

2015 2015
Participants 3
CT Emean %D 0.6 |E|mean %D 0.6
KT Emean %D 0.6 |E|mean %D 2.9
KQ Emean %D −3.0 |E|mean %D 3.4
Rps Emean %D −0.3 |E|mean %D 1.5
RT -T Emean %D − |E|mean %D −
Table 15 Error estimation for the case 2.5 (fixed rps case)
2015 2015
Participants 2
CT Emean %D 1.5 |E|mean %D 1.5
KT Emean %D 0.3 |E|mean %D 0.9
KQ Emean %D −4.2 |E|mean %D 4.2
Rps Emean %D − |E|mean %D −
RT -T Emean %D 3.9 |E|mean %D 3.9
3 Conclusions
• For the predictions of resistance, sinkage and trim, the number of grid points are
varied from 1.1 M to 6.6 M. The submitted results do not show large dependency
of grid points as shown in Fig. 17. It means that CFD users who have efficient
<EFD>
<KRISO/WAVIS>
<MARIC/FINEMarine>
Fig. 15 Axial velocity contours and flow vectors
grid making techniques can have an acceptable simulation results for low block
coefficient hull form such as KCS.
• Most participants use two equation turbulence model and only few participants
use an advanced turbulence model. Based on these submissions, there is no
visible improvement in accuracy of the resistance prediction for more advanced
turbulence model than the two equation models.
• The comparison error for resistance prediction at the design speed (Fr = 0.26) is
−0.2% of the data and the standard deviation is 1.5% of the data value. The mean
328 J. Kim
<MARIN/PARNASSOS>
<PNU/Fluent>
<UM/OpenFOAM>
Fig. 15 (continued)
comparison error for all 6 speeds is 0.43% and the mean standard deviation is
2.48%. The increased error and standard deviation is caused by the results for low
Fr simulation submissions. The mean absolute error |E| is 2%, which is slightly
larger than 1.64% the results of Gothenburg 2010.
• The comparison errors and standard deviation of the sinkage and trim are much
larger than for the resistance. The errors come from the results in a speed range
below Fr = 0.2. Two possible reasons can be considered. This range generates very
short waves and so it requires better grid distribution to resolve the waves. Another
<EFD> <KRISO/WAVIS>
<MARIC/FINEMarine> <MARIN/PARNASSOS>
<PNU/Fluent> <UM/OpenFOAM>
Fig. 16 Velocity downstream of propeller plane at z/L pp = −0.03
problems are likely to be due to the difficulties of measuring two quantities at low
speed in the experiment. The data might have larger uncertainties than for high
speed.
• Self-propulsion submissions are slightly improved compared to the results of
Gothenburg 2010 Workshop. The mean comparison errors of K T and K Q are 0.5
and −3.5% respectively (−0.6 and −4.6% for Gothenburg 2010 workshop). The
standard deviations for of K T and K Q are 2.7 and 2.4% respectively, which are
330 J. Kim
<EFD>
<KRISO/WAVIS>
<MARIC/FINEMarine>
<MARIN/PARNASSOS>
Fig. 17 Hull surface pressure contours (pressure coefficient)
smaller than the results of 7.2 and 6.1% respectively for Gothenburg 2010 work-
shop. However, prediction for K Q is still less accurate than K T . Self-propulsion
parameters are slightly better predicted by body force models (FD fixed) but local
flow characteristics are comparatively well predicted by actual propeller rotating
simulation (rps fixed).
<PNU/Fluent>
<UM/OpenFOAM>
Fig. 17 (continued)
• For the future workshops, new experimental data for KCS or other container hull
would be recommended to simulate not only captive but free running 6DoF calm
water self-propulsion condition.
References
Larsson, L., Stern, F., & Visonneau, M. (2010). Proceedings of Gothenburg 2010 A Workshop on
Numerical Ship Hydrodynamics. Gothenburg, Sweden.
Larsson, L., Stern, F., Visonneau, M. (Eds.). (2013). Numerical Ship Hydrodynamics—An
assessment of the Gothenburg 2010 Workshop, Springer.
Assessment of CFD for KCS Added
Resistance and for ONRT Course
Keeping/Speed Loss in Regular Head
and Oblique Waves
Frederick Stern, Hamid Sadat-Hosseini, Timur Dogan, Matteo Diez,

Dong Hwan Kim, Sungtek Park, and Yugo Sanada
Abstract CFD is assessed for added resistance for KCS (captive test cases 2.10 and
2.11) and course keeping/speed loss for ONRT (free running test cases 3.9/3.12/3.13)
in head and oblique waves. The number of submissions were 10, 2, and 8 for test
cases 2.10, 2.11, and 3.9/3.12/3.13, respectively. The assessment approach uses
both solution and N-version validation. The former considers whether the abso-
lute error |E i | = |D − Si | is less, equal or greater than the validation uncer-
tainty, which is the root sum square of the numerical and experimental uncer-
tainties, i.e., |E i | ≤ UVi = U S2Ni + U D2 . The latter considers whether the abso-
lute error is less, equal or greater than the state-of-the-art SoAi uncertainty, i.e.,
|E i | ≤ U So Ai = UV2i + P|E 2
i|
where P|Ei | = kσ|E| is the uncertainty due to the
scatter in the solution absolute error. Errors and uncertainties are normalized using
both the data value D and its dynamic range DR. The captive resistance CT and
free running self-propulsion propeller revolutions RPS |E| < 2%D with UD less
than but comparable P|Ei | < 3%D such that 3/1 solutions were validated but 8/4
codes/solutions were N-version validated for CT /RPS. The head waves captive and
free running heave and pitch |E| is less than 8%DR with UD less than 5%DR and
P|Ei | less than 13%DR such that about 5 for captive and 2 for free running solutions
were validated and about 7 for captive and 5 for free running codes/solutions were
N-version validated. The errors for added resistance and speed loss were less than
13%DR with UD less than 7%DR and P|Ei | = 25 and 13%DR such that about 4 for
captive and 1 for free running solutions were validated and 7 for captive 4 for free
F. Stern (B) · T. Dogan · S. Park · Y. Sanada

University of Iowa and IIHR-Hydroscience & Engineering, Iowa City, IA, USA
H. Sadat-Hosseini
University of North Texas, Denton, TX, USA
M. Diez
CNR-INM, National Research Council-Institute of Marine Engineering, Rome, Italy
D. H. Kim
Korea Research Institute of Ocean Science & Technology, Daejeon, South Korea
334 F. Stern et al.
running codes/solutions were N-version validated. For captive head waves, the errors
and scatter are smaller than those for potential flow. The oblique waves captive and
free running motion errors |E| are less than 10%DR except for roll with UD large
23%DR for captive pitch and other wise small <2%DR. The errors for added resis-
tance and speed loss were <8%DR with UD < 9%DR. The largest errors were for roll,
which had errors for captive of 12%DR and for free running of 20%DR. None of the
KCS and ONRT test cases were able to achieve a programmatic requirement of 5%D
and 5%DR for calm water and waves, respectively. Note that not all the experimental
uncertainties are able to meet this requirement either such that both experiments and
CFD need to reduce their uncertainties and in addition CFD its errors. Nonetheless,
in view the comparable CFD capability for ONRT free running vs. KCS captive
conditions, the prognosis for CFD capability is excellent. Future CFD assessment
should extend the current test cases for consideration of added power in regular and
irregular waves. Verification should be required, and more limited/most important
validation variables should be confirmed and used for the V&V and SoA assessment.
Nomenclature
A Desired wave amplitude (m)

AR Added resistance (N)
B Breadth (m)
BWL Maximum beam at the waterline (m)
CB Block coefficient
CM Midship Section Coefficient
CG Center of gravity
CT Total resistance coefficient
CTi I-th harmonic of total resistance coefficient
D Measurement value
DDES Delayed detached eddy simulation
E Error
Ei Error of i-th simulation
DR Dynamic range
fe Wave encounter frequency (Hz)
fhz Hydrostatic estimates of natural heave frequency (Hz)
fhθ Hydrostatic estimate of natural pitch frequency (Hz)
fhφ Hydrostatic estimate of natural roll frequency (Hz)
Fr Froude number
g Gravitational acceleration (m/s2 )
H Wave height (m)
IMO International Maritime Organization
k Wave number/Spring constant (N/m)
KCS KRISO/MOERI Container Ship
KG Vertical center of gravity (m)
Assessment of CFD for KCS Added Resistance and for ONRT… 335
KQ Torque coefficient
KT Thrust coefficient
kxx Radius of gyration in roll (m)
kyy Radius of gyration in pitch (m)
kzz Radius of gyration in yaw (m)
KVLCC2 KRISO Very Large Crude Carrier
L Ship length (m)
LCB Longitudinal center of buoyancy
LPP Length between perpendiculars (m)
LS Level set method
LWL Length at the waterline (m)
n Propeller revolutions (rps)
ONRT ONR Tumblehome
P The random/precision uncertainty
Re Reynolds number
RSM Reynolds stress turbulence model
S Wetted surface area (m2 )/simulation value
Si I-th simulation value
SoA State of art
SR Wetted surface area of rudder (m2 )
SW Wetted surface area of hull (m2 )
t Time (s)
T Draft (m)
Te Wave encounter period (s)
u Surge velocity (m/s)
ui I-th harmonic surge velocity amplitude (m/s)
U Ship speed (m/s)
UD Experimental uncertainty
USoA Mean code uncertainty
USoAi Uncertainty of i-th code
v Sway velocity (m/s)
vi I-th harmonic sway velocity amplitude (m/s)
V Ship speed (m/s)
VA Desired approach speed (m/s)
VoF Volume of fluid method
X X force (N)/Coordinate (m)
XC X axis carriage coordinate (m)
Xi Measurement value
XH Hydrodynamic force (N)
XS X axis ship coordinate (m)
YC Y axis carriage coordinate (m)
YS Y axis ship coordinate (m)
z Heave (m)
336 F. Stern et al.
δ Rudder angle (deg)

Mass of ship (kg)
ζi Wave elevation (m)
ζi I-th harmonic wave amplitude (m)
θ Pitch angle (rad)
θi I-th harmonic pitch amplitude (rad)
λ Wave length (m)
μ Wave encounter angle (deg)
v Kinematic viscosity of water (m2 /s)
ρ Density of water (kg/m3 )
σ Standard deviation
σ/L Sinkage
τ Trim (deg)
φ Roll angle (rad)
φi I-th harmonic roll amplitude (rad)
ψ Yaw angle (rad)
ψi I-th harmonic yaw amplitude (rad)
ψc Target yaw angle (deg)
∇ Displacement (m3 )
1 Introduction
T2015 included test cases for KCS Fr = 0.26 and Re = 1.074 × 107 and Re = 3.613
× 106 , respectively, for added resistance in regular head (2.10) and oblique (2.11)
waves; and for course keeping/speed loss for ONRT Fr = 0.2 and Re = 3.497 ×
106 for free running in regular head (3.12) and oblique (3.13) waves. Test cases 2.10
and 2.11 included calm water resistance, sinkage and trim. ONRT calm water self-
propulsion propeller revolutions, sinkage and trim was included as test case (3.9).
Section 5 covers the test case 2.10; Sect. 6 covers test case 2.11; and Sect. 7 covers
test cases 3.9, 3.12, and 3.13.
Section 2 summarizes the quantitative approach for the CFD assessment, which
uses both solution and N-version validation. The former considers whether the abso-
lute error |E i | = |D − Si | is less, equal or greater than the validation uncertainty,
which
is the RSS of the numerical and experimental uncertainties, i.e., |E i | ≤ UVi =
U S2Ni + U D2 . The latter considers whether the absolute error is less, equal or greater

than the state-of-the-art SoAi uncertainty, i.e., |E i | ≤ U So Ai = UV2i + P|E 2
i|
where
P|Ei | = kσ|E| is the uncertainty due to the scatter in the solution absolute error. The
tables, figures and discussions use these variables; thus, the reader is encouraged to
familiarize themselves with this terminology.
Section 3 summarizes the experimental N-order and M×N-order level testing and
uncertainty analysis. The individual facility data value is Xi and when available its
individual facility N-order level testing uncertainty estimate

M is UXi . The multiple
facility M×N order level testing average value X̄ = M1 i=1 X i and precision limit
1 M
2
are PXi = 2σXi , where σ X i = M−1 i=1 X i − X̄ is the standard deviation SD such
2
the individual facility MxN-order level testing uncertainty estimate is UDi =

that
2
U Xi + PXi
2
. The discussions use both Xi and X̄ , which are often simply referred
to as D and similarly for UXi and UDi as UD for convenience. Chapter 4 summarizes
the available experimental data used for the KCS and ONRT test cases. Multiple
facility data is available for KCS calm water resistance and head waves such that σXi
is evaluated, which was used for the KCS calm water resistance, sinkage and trim
validation.
Section 4 summarizes the validation variables and experimental uncertainties used
for the KCS and ONRT test cases. The validation variables for calm water KCS are
resistance, sinkage and trim and for ONRT self-propulsion are propeller revolutions,
sinkage and trim. Resistance and propeller revolutions are considered primary and
motions secondary variables. For KCS head waves, added/0th harmonic resistance,
and 1st harmonic heave and pitch are considered primary and 1st harmonic resis-
tance as secondary variables; with additionally 1st harmonic roll as primary variable
for oblique waves. Similarly, for ONRT free running in head and oblique waves
primary variables are speed loss (0th harmonic axial velocity) and first harmonic
motions which additionally include yaw are the primary variables and other 0th and
2nd harmonic variables as secondary. Most of the discussions focus on averages
over wavelength or heading, including averages for roughly speaking first order (1st
harmonic) and second order (0th and 2nd harmonic) terms. For KCS head waves test
case 2.10, comparisons are made with the previous G2010 workshop, which had the
same test case; and with results from 8 different potential flow codes provided by
Bunnik et al. (2010).
2 Verification, Validation and SoA Approaches
The individual code solution verification and validation (V&V) follows Stern et al.
(2001) overall procedures. For ship hydrodynamics, verification usually uses either
the factor of safety (Xing and Stern 2010) or least square (Eça and Hoekstra 2014)
methods. The numerical uncertainties U S Ni use the root-sum-square of the iterative
U Ii , grid UG i and time step UTi uncertainties

U S Ni = UG2 i + UT2i + U I2i %Si (1)
i indicates an individual code and S i is the solution on the finest grid. ASME
(2009) advocates adding U Ii with UG2 i and UT2 . The comparison error E i is
i
338 F. Stern et al.
(D − Si )
E i %D = × 100. (2)
D
D is the experimental benchmark data. Validation compares the errors with the
validation interval.

UVi = U S2Ni + U D2 %D (3)
UD is the experimental benchmark data uncertainty. The simulations are validated

at an interval UVi if
|E i |%D ≤ UVi %D. (4)
A shortcoming of this and other V&V approaches is that the justification for
uncertainty estimates at the 95% confidence level is based on reasoning like that
used for experimental bias uncertainties at the 0-order-replication level without addi-
tional statistical 1-order replication level precision uncertainties, which in combina-
tion provides N-order replication level and increased confidence for experimental
uncertainty analysis.
Available solutions from multiple codes for the same validation variables enable
evaluation of the mean and standard deviation, which is referred to as N-version
testing (Hemsch 2004). Stern et al. (2006, 2007) extended N-version testing by also
considering the individual code V&V, which was referred to as CFD certification: a
process for assessing probabilistic confidence intervals for CFD codes for specific
benchmark applications and certification variables. Certification of CFD codes is
at the multiple codes or users, models, grid types, etc. level (code level). Multiple
users are appropriate for many user codes, whereas multiple codes are appropriate
for various few user codes (which is often the case for industrial applications).
Differences between versions and implementations are due to a myriad of possi-
bilities for modeling, numerical methods, and their implementation as CFD codes
and simulations.
Stern et al. (2016, 2017) modified this approach to use the mean and standard
deviation of the absolute value of the error, which provides a more conservative error
estimate and smaller standard deviation. Here we follow this approach by consid-
ering the mean error Ē, mean absolute error |E| and its standard deviation σ|E| and
uncertainty U So A in assessing the results. The use of the terminology “certification”
was criticized as implies/requires an authorizing body; therefore, the new approach
is called a statistical approach for CFD SoA assessment: N-version verification and
validation.
At the multiple code level, the individual code/solution uncertainty includes both
deterministic/bias and random/precision components
U S2i = BS2i + PS2i (5)

BSi = U S Ni (6)
The mean code solution, bias and standard deviation are
1
N
S̄ = Si (7)
N i=1
1 2
N
BS̄2 = U (8)
N i=1 S Ni

1 N
2 1 N
2
σS = Si − S̄ = E i − Ē = σ E (9)
N − 1 i=1 N − 1 i=1
The individual code/solution and mean code random/precision numerical uncer-

tainties are
√
PSi = 2σ S and PS̄ = 2σ S / N . (10)
Outliers can be identified and rejected similarly as with experimental data using,
e.g., Chauvenet’s criterion. Herein, for simplicity, a solution is rejected if its deviation
from the mean is larger than 2σ S , i.e. N ≈ 10. The mean code solution uncertainty is
U S̄2 = BS̄2 + PS̄2 . (11)
The average error and average absolute error are
1
N
Ē = Ei (12)
N i=1
1
N
|E| = |E i | ≥ Ē (13)
N i=1
The average absolute error is always greater than or equal to the absolute value
of the signed average error. The average absolute error standard deviation is

1
N
2
σ|E| = |E i | − |E| (14)
N − 1 i=1
with absolute value statistical property

2
σ|E|
2
= σ E2 + Ē 2 − |E| ≤ σ E2 = σ S2 (15)
340 F. Stern et al.
The random/precision uncertainty component for the individual code/solution and

mean code are
P|Ei | = 2σ|E| (16)
2σ|E|
P|E| = √ (17)
N
The uncertainty of the individual code/solution error E i is
Ai = U D + U S Ni + P|E i |
2 2 2 2
U So (18)
The uncertainty of the mean code error |E| is
A = U D + Ū S N + P|E|
2 2 2
U So (19)
The SoA of the individual code/solution and mean code are defined by
|E i | ≤ U So Ai (20)
|E| ≤ U So A (21)
3 Experimental Uncertainties
The experimental uncertainty analysis considers both individual facility/model N-

order testing U X i and M × N-order level testing U Di and U X (Stern et al. 2005).
Head waves U X i also includes post-workshop analysis of the statistical convergence
uncertainty U SCi .
N-order level testing:
1 j
N
Xi = X (22)
N j=1 i
1 j
N
σ X2 j = (X − X i )2 (23)
i N − 1 j=1 i

UX j = B X2 i + (2σ X j )2 + U 2 j (24)
i i SCi
√
UXi = B X2 i + (2σ X j / N )2 + U SC
2
i
(25)
i
M × N-order level testing:
1
M
X= Xi (26)
M i=1
1
M
BX = UXi (27)
M i=1
1
M
σ X2 i = (X i − X̄ )2 (28)
M − 1 i=1
PX i = 2σ X i (29)
2σ X
PX = √ i (30)
M

U X = B X2 + PX2 (31)

U Di = U X2 i + PX2i (32)
The facility bias uncertainty is included as follows.
i = X̄ − X i (33)

Ui = U X2 i + U X2 (34)
If the absolute value of i is less than Ui
|i | ≤ Ui (35)
the individual facility is certified at interval Ui . Whereas if the absolute value of i
is greater than Ui
|i | > Ui (36)
the facility bias and total uncertainty are defined as

U F Bi = i2 − U2 i (37)

UTi = U X2 i + U F2 Bi . (38)
342 F. Stern et al.
Lastly, U Di including facility bias is

U Di = U X2 i + U F2 Bi + PX2i = UT2i + PX2i (39)
M × N-order level testing U Di is only available for test case 2.10 and thus far
only used for calm water resistance.
4 Validation Variables and Experimental Uncertainties
4.1 KCS and ONRT Particulars and Coordinate Systems
Tables 1 and 2 list the model particulars for the KCS and ONRT full and model
scales. Figures 1 and 2 shows the carriage- and ship-fixed coordinate system used
to report the validation variables for KCS and ONRT, respectively. For KCS, the
Table 1 Main particulars for

Full scale IIHR/FORCE FORCE
KCS 2.10 and 2.11 (Fr =
L = 2.7 m L = 6.0702 m
0.26)
Main particulars
L( = LPP ) (m) 230 2.7 6.0702
LWL (m) 232.5 2.729 6.1357
BWL (m) 32.2 0.378 0.8498
T (m) 10.8 0.1268 0.285
∇ (m3 ) 52030 0.0842 0.9571
SW (m2 ) 9424 1.2987 6.6177
SR (m2 ) 115 0.0158 0.0801
CB 0.6505 0.651 0.651
CM 0.9849
LCB (%LPP ), −1.48 – −1.48
fwd+
KG/L 0.0622 0.0622 0.0622
kxx /B 0.4 0.4 0.4
kyy /L, kzz /L 0.25 0.25 0.252
fhφ (Hz) 0.025 0.23 0.15
fhz (Hz) 0.11 1.09 0.73
fhθ (Hz) 0.11 1.09 0.73
Ship speed
U (m/s) 12.35 1.24 2.00
Re 2.84 × 109 3.613 × 106 1.074 × 107
Table 2 Main particulars for ONRT 3.9, 3.11, 3.12 (Fr = 0.2)
Full scale IIHR L = 3.147 m
Main particulars
L( = LWL ) (m) 154.0 3.147
BWL (m) 18.78 0.384
T (m) 5.494 0.112
(kg) 8507000 72.6
SW (m2 ) NA 1.5
SR (m2 ) NA 0.012 × 2
CB 0.535 0.535
LCB (%LPP ), fwd+ NA 1.625
KG/L NA 0.156
kxx /B 0.444 0.444
kyy /L, kzz /L 0.25 0.246
Propeller diameter NA 0.1066
Propeller center, lateral location (from CL) NA 0.9267
Propeller center, vert. location (below WL) NA 0.02661
Propeller shaft angle (downward positive) NA 0.03565
Propeller rotation direction (view from stern) Inward Inward
Shaft angle (deg) 5 5
Maximum rudder rate NA 35
fhφ (Hz) NA 0.65
fhz (Hz) NA 1.6
fhθ (Hz) NA 1.6
Ship speed
U (m/s) 7.77 1.11
Re 1.197 × 109 3.497 × 106
carriage-fixed coordinate system is located at the center of gravity, with x pointing

toward the stern, y to starboard and z upward. Figure 1 shows the IIHR and Osaka
University (OU) surge free mounts. FORCE used a fixed surge added resistance
mount designed for large motions (same as used in waves) and not a conventional
calm water resistance test setup. For ONRT, the ship-fixed coordinate system is also
located at the center of gravity but with x pointing toward the bow, y to starboard
and z downward. The evaluation of the motions and forces is at the center of gravity.
344 F. Stern et al.
Fig. 1 Coordinate system for KCS 2.10 and 2.11 experiments and simulations
Fig. 2 Coordinate system for ONRT 3.9, 3.11, 3.12 experiments and simulations
4.2 Wave Elevation
For test cases 2.10, 2.11, 3.12 and 3.13 wave elevations are measured at FP, but the
y/L values are different: for 2.10, y/L = 0.81; for 2.11, y/L = 0.037; and for 3.12
and 3.13, y/L = 0.032. The test case 2.11, 3.12 and 3.13 y/L locations are less than
desirable to insure that they are not affected by the model diffraction/radiation waves
but were restricted due to the experimental setup. For 2.10, the CFD did not report
the location for the wave elevation submissions, whereas for 2.11, 3.12 and 3.13, the
location was the same as the experiments.
For 2.10, the experimental 1st harmonic wave amplitude ζ1 and wavelength λ were
used as the CFD input, whereas for 2.11, 3.12 and 3.13, the ideal wave amplitude A
and length λ values were used. For 2.11 and 3.12/3.13 the experimental differences
with the ideal values was ≤3% based on stationary basin fixed wave gauge data,
which meets IMO criteria.
For 2.10, the wavelength conditions are λ/L = 0.65, 0.85, 1.15, 1.37 and 1.95
with the wave steepness of H/λ = 1/60. For 2.11, λ/L = 1, H/λ = 1/60 and wave
headings of μ = 0, 45, 90, 135 and 180°. For 3.12 and 3.13, the conditions are the
same as 2.11, except H/λ = 1/50.
The wave conditions are close to the usual assumptions for small amplitude ship
motions response. The experimental waves were largely 1st harmonic dominant. For
2.10, the second harmonic of wave amplitude ζ2 is ≤ 1% of the first harmonic wave
amplitude ζ1 for short waves and ζ2 ≤ 10%ζ1 for long waves. The experimental wave
uncertainty was not reported for 2.10. For 2.11 and 3.12/3.13, ζ2 = 5% and 3%ζ1 ,
respectively. The experimental wave uncertainty was ≤1.5%D for 2.11 and ≤3%D
for 3.12 and 3.13. Herein only qualitative assessment of the experimental and CFD
wave elevations is conducted via comparisons of the time histories.
4.3 Motions and Forces
4.3.1 Test Case 2.10
FORCE conducted experiments for Test Case 2.10 using L = 6.1 m model. IIHR
used the 2.7 m model for Test Case 2.11. Both models are appended with a horn
type rudder. The hydrostatic estimates of heave, pitch and roll natural frequencies
for IIHR/FORCE models are (fhz , fhθ , fhφ ) = (1.09/0.73, 1.09/0.73, 0.23/0.15 Hz).
The validation variables for calm water are resistance C T = 1/2ρXSV 2 , sinkage
σ
L
and trim τ [deg]. FORCE provided the experimental values Xi of the validation
variables, as shown in Table 3. The data uncertainty U X i is only provided for CT . PX i
and U Di values are evaluated based on data from 5 facilities using 7.3 m (NMRI and
Table 3 Uncertainties for KCS resistance tests in calm water at Fr = 0.26 (Case 2.10)
Variable Xi U X i %D PX i %D U Di %D
C T × 103 3.835 0.95 1.47 1.75
σ/L × 102 −0.2074 NA 7.38 7.38
τ [deg] −0.165 NA 14.25 14.25
346 F. Stern et al.
KRISO), 4.4, 6.1, 2.7 m (FORCE) and 2.7 m (IIHR) and 3.2 m (OU) models. The
PX i and U Di values in Table 3 exclude the data for the 2.7 and 3.2 m models. U Di is
used for the CFD assessment, as provides a more robust UD estimate than U X i .
The validation variables for head waves are wave elevation ζA , resistance C T =
T 0 −X T Calm
2
X
1/2ρ SV 2
, added resistance A R = XρgA B /L
2 2 = 2gASVB /L
2 2 (C T 0 − C T Calm ), heave Az
θ
and pitch Ak (where A = ζ1 for 2.10 and A for 2.11 and ONRT test cases). Resistance
coefficient is based on wetted surface area S/L2 = 0.1818 with rudder in calm water.
The computational and experimental values were compared for time histories of
wave elevation, resistance, heave and pitch. Fourier series 0th, 1st and 2nd harmonic
amplitudes and phases for resistance, heave and pitch were also evaluated for both
experiments and computations, as follows:
P0
N
P(t) = + Pn cos(2π f e t + Pεn )
2 n=1
2 T
an = ∫ P(t)cos(2π f e t)dt
T 0
2 T
bn = ∫ P(t)sin(2π f e t)dt (40)
T 0

Pn = an2 + bn2
bn
Pεn = tan −1 − − γI
an
where P is time series, Pn is the n-th harmonic amplitude and Pεn is the corresponding
phase, f e = ωe /2π, ωe = ω − ωg U cosμ is the wave encounter frequency, ω is the
2
wave frequency, μ is the wave encounter angle (μ = 0 for head waves) and γ I is the
incident wave phase at the bow at t = 0.
The experimental values X i for the 0th, 1st and 2nd harmonic amplitudes and
phases of the validation variables are provided by FORCE, as shown in Tables 4, 5,
6, 7 and 8. U X i is also provided for AR and 1st harmonic amplitude of heave and
pitch. IIHR evaluated the statistical convergence uncertainty U SCi for the FORCE Xi
using a running Fourier series analysis. For each wave length, two different window
sizes were selected for the running Fourier series analysis;
√ one was based on two
encounter wave periods, and the other was based on the N encounter wave periods
where N was√ the number of encounter wave periods for which the FORCE data was
available. N was 8, 7, 6, 6, 5 for λ/L = 0.65, 0.85, 1.15, 1.37, and 1.95, respectively.
U SCi values based on two encounter wave periods are reported in Tables 4, 5, 6, 7
√
and 8, which are slightly larger than those based on N encounter wave periods.
Overall, the U SCi values are often large and are included in UD estimate.
Table 4 0th harmonic amplitude uncertainties for KCS seakeeping in waves at Fr = 0.26 (Case
2.10)
Variable/Dynamic range λ/L fe / fhz/θ Xi U X i %D U SCi %D U X i %D U X i %D R
CT × 103 DR = 5.927 0.65 1.50 8.237 5.2 5.2 7.2
0.85 1.24 9.269 5.6 5.6 8.7
1.15 1.00 14.164 2.8 2.8 6.7
1.37 0.89 13.948 2.9 2.9 6.7
1.95 0.70 10.798 9.1 9.1 16.5
Ave. 5.1 5.1 9.2
A R DR = 8.002 0.65 1.50 3.386 8.2 6.3 10.3 4.4
0.85 1.24 6.102 7.9 4.2 8.9 6.8
1.15 1.00 9.911 7.9 2.0 8.2 10.1
1.37 0.89 6.512 7.9 3.1 8.5 6.9
1.95 0.70 1.909 8.0 25.7 26.9 6.4
Ave. 8.0 8.3 12.6 6.9
z/A DR = 0.610 0.65 1.50 −0.810 6.2 6.2 8.3
0.85 1.24 −0.628 5.5 5.5 5.7
1.15 1.00 −0.278 9.4 9.4 4.3
1.37 0.89 −0.249 8.1 8.1 3.3
1.95 0.70 −0.200 18.6 18.6 6.1
Ave. 9.6 9.6 5.5
θ/Ak DR = 0.128 0.65 1.50 −0.108 9.1 9.1 7.7
0.85 1.24 −0.130 4.6 4.6 4.6
1.15 1.00 −0.002 701.3 701.3 11.8
1.37 0.89 −0.007 212.3 212.3 12.4
1.95 0.70 −0.057 57.9 57.9 25.8
Ave. 197.0 197.0 12.5
4.3.2 Test Case 2.11
The test case 2.11 validation variables for calm water and waves are the same as
φ
2.10, but also include roll Ak . IIHR provided the Xi and U X i for resistance, sinkage
and trim in calm water, as shown in Table 9. IIHR provided the Xi and U X i for added
resistance and 0th, 1st and 2nd harmonic amplitudes of wave elevation and motions
in waves, as shown in Tables 10, 11 and 12. Only one facility/model is available and
M × N analysis is not possible.
348 F. Stern et al.
Table 5 1st harmonic amplitude uncertainties for KCS seakeeping in waves at Fr = 0.26 (Case
2.10)
ζ /ADR = 0.011 0.65 1.50 0.005 1.3 1.3 0.6
0.85 1.24 0.006 0.8 0.8 0.5
1.15 1.00 0.010 0.2 0.2 0.2
1.37 0.89 0.012 0.3 0.3 0.3
1.95 0.70 0.016 0.8 0.8 1.2
Ave. 0.7 0.7 0.6
CT × 103 DR = 21.749 0.65 1.50 3.348 20.4 20.4 3.1
0.85 1.24 4.206 6.6 6.6 1.3
1.15 1.00 3.929 5.5 5.5 1.0
1.37 0.89 12.768 2.7 2.7 1.6
1.95 0.70 25.097 0.7 0.7 0.8
Ave. 7.2 7.2 1.6
z/A DR = 0.803 0.65 1.50 0.130 4.0 18.0 18.4 3.0
0.85 1.24 0.244 4.0 13.2 13.8 4.2
1.15 1.00 0.899 4.0 1.8 4.4 4.9
1.37 0.89 0.874 4.0 1.9 4.4 4.8
1.95 0.70 0.933 4.0 1.1 4.2 4.8
Ave. 4.0 7.2 9.0 4.3
θ/Ak DR = 1.102 0.65 1.50 0.017 3.8 45.4 45.5 0.7
0.85 1.24 0.146 4.0 9.6 10.4 1.4
1.15 1.00 0.748 4.0 1.7 4.3 3.0
1.37 0.89 0.966 4.0 1.5 4.2 3.7
1.95 0.70 1.119 4.0 0.5 4.0 4.1
Ave. 4.0 11.7 13.7 2.6
4.3.3 Test Case 3.9, 3.12 and 3.13
The validation variables for calm water self-propulsion are propeller RPS n, sinkage
σ
L
and trim τ [deg]. The validation variables for waves are wave elevation ζA , heave
φ θ ψ
z
A
, roll Ak , pitch Ak , yaw Ak , surge velocity Uu and sway velocity Uv . The computa-
tional and experimental values were compared for time histories of wave elevation,
resistance, heave and pitch. Fourier series 0th, 1st and 2nd harmonic amplitudes and
phases for resistance, heave and pitch were also evaluated for both experiments and
computations, as per Eqs. (40).
IIHR provided Xi and U X i for propeller RPS, sinkage and trim for self-propulsion
in calm water, as shown in Table 13. Xi and U X i were also provided by IIHR for
0th, 1st and 2nd harmonic amplitudes and phases of wave elevation and motions
Table 6 1st harmonic phase uncertainties for KCS seakeeping in waves at Fr = 0.26 (Case 2.10)
Variable λ/L Xi U X i %2π U SCi %2π U X i %2π
ζ /A 0.65 0.525
0.85 −2.641
1.15 0.483
1.37 −1.123
1.95 −1.040
Ave.
C T × 103 0.65 −0.712 3.0 3.0
0.85 −1.286 3.6 3.6
1.15 −1.546 0.7 0.7
1.37 −1.333 0.4 0.4
1.95 −0.610 0.2 0.2
Ave. 1.6 1.6
z/A 0.65 1.907 2.4 2.4
0.85 0.970 3.4 3.4
1.15 −3.101 0.3 0.3
1.37 −2.397 0.5 0.5
1.95 −1.971 0.2 0.2
Ave. 1.4 1.4
θ/Ak 0.65 1.098 7.4 7.4
0.85 1.337 3.3 3.3
1.15 −2.495 0.3 0.3
1.37 −1.702 0.4 0.4
1.95 −0.832 0.2 0.2
Ave. 2.3 2.3
for course keeping/speed loss, as shown in Tables 14, 15, 16, 17 and 18. Only one
facility/model is available and M × N analysis is not possible.
After T2015 in July 2016, Dr. Michel Visonneau pointed out inconsistent differ-
ences between the CFD and the measured surge velocity for ONRT. A sign error
was discovered in the equation to get the ship coordinates (XS , YS ) in the earth
fixed coordinate system from the carriage coordinates (Xc, Yc) in the earth fixed
coordinate system and ship deviations (X, Y) from the carriage, which led to
incorrect values for Xs, Ys, u and v. The final proceedings use the corrected Xs, Ys,
u and v values. The maximum differences between original and corrected values are
1.6%DR for u0 at μ = 135° and 87%DR for v0 at μ = 90°. The differences for the
first harmonic u and v amplitudes are 102%DR at μ = 45° and 193%DR at μ = 90°,
respectively. The differences for the second harmonics and phases were larger.
After the T2015 in September 2018, Dr. Michel Visonneau pointed out a time
delay between the yaw and rudder of about 0.233 s, which is attributed to a latency
350 F. Stern et al.
Table 7 2nd harmonic amplitude uncertainties for KCS seakeeping in waves at Fr = 0.26 (Case
2.10)
CT × 103 DR = 2.045 0.65 1.50 0.652 88.6 88.6 28.3
0.85 1.24 0.755 53.3 53.3 19.7
1.15 1.00 2.697 11.0 11.0 14.5
1.37 0.89 2.201 14.4 14.4 15.5
1.95 0.70 0.675 55.5 55.5 18.3
Ave. 44.6 44.6 19.3
z/A DR = 0.015 0.65 1.50 0.005 115.6 115.6 39.3
0.85 1.24 0.008 100.2 100.2 54.2
1.15 1.00 0.013 17.2 17.2 14.4
1.37 0.89 0.007 52.2 52.2 23.4
1.95 0.70 0.020 46.9 46.9 62.9
Ave. 66.4 66.4 38.8
θ/Ak DR = 0.036 0.65 1.50 0.001 96.3 96.3 3.2
0.85 1.24 0.002 94.0 94.0 5.8
1.15 1.00 0.014 12.3 12.3 4.7
1.37 0.89 0.012 26.4 26.4 8.7
1.95 0.70 0.037 18.4 18.4 19.0
Ave. 49.5 49.5 8.3
of the yaw angle data acquisition in the experiments. The yaw angle used for the
rudder control in the course keeping/speed loss experiments at IIHR is obtained
by the carriage tracking system. There is a latency due to internal processing and
communication between the systems. This latency cannot be eliminated based on
the current technology even if we use an onboard gyroscope. For future validation
studies, we propose the following:
1. For rudder control based on yaw angle in the CFD simulations, the latency for
the yaw angle data acquisition should be considered as the simulation condition
in addition to the rudder speed. In the CFD codes, the yaw angle obtained from
the previous time step which is corresponding to the latency needs to be used for
the rudder control to mimic the latency of the yaw angle data acquisition, as per
the following new rudder controller equation:
ψ(t) = ψC − ψ(t − t L )
t (41)
δ(t) = K δ P ψ(t) + K δ I 0 ψ(τ )dτ + K δ D dψ(t)
dt
Here, tL is latency of yaw angle acquisition.

Table 8 2nd harmonic phase uncertainties for KCS seakeeping in waves at Fr = 0.26 (Case 2.10)
Variable λ/L Xi U X i %2π U SCi %2π U X i %2π
C T × 103 0.65 0.068 24.8 24.8
0.85 −1.122 11.4 11.4
1.15 1.829 2.2 2.2
1.37 2.903 16.6 16.6
1.95 −0.446 11.3 11.3
Ave. 13.3 13.3
z/A 0.65 −0.046 61.4 61.4
0.85 0.810 64.8 64.8
1.15 0.214 2.7 2.7
1.37 0.580 9.3 9.3
1.95 0.794 8.7 8.7
Ave. 29.4 29.4
θ/Ak 0.65 0.605 78.8 78.8
0.85 1.215 40.9 40.9
1.15 0.201 2.1 2.1
1.37 0.708 4.2 4.2
1.95 1.646 3.0 3.0
Ave. 25.8 25.8
Table 9 Uncertainties for

Variable Xi U X i %D
KCS resistance test in calm
water at Fr = 0.26 (Case 2.11) CT × 103 4.888 2.5
σ/L × 102 −0.1704 0.6
τ [deg] −0.022 65.5
For new experiments, it is desirable to develop a new method to evaluate the

latency for the yaw angle data acquisition in course keeping/speed loss experiments.
It is necessary to evaluate the worst execution time for each process inside the system
and to obtain the actual latency due to the communication and controlling using a
global master clock to issue time stamps for each measurement variable. In addition,
when sensor fusion is done to get the maneuvering data, it is also necessary to
evaluate time jittering of the internal clock inside each device based on the global
master clock.
352 F. Stern et al.
Table 10 0th harmonic amplitude uncertainties for KCS seakeeping in waves (Case 2.11)
Variable/Dynamic range μ [deg] fe / fhz/θ Xi U X i %D U X i %D R
C T × 103 DR = 3.696 0 0.57 13.135 0.8 5.8
45 0.50 12.572 0.7 4.7
90 0.35 9.994 1.2 6.3
135 0.19 9.683 0.8 4.1
180 0.12 9.438 0.8 4.1
Ave. 0.9 5.0
A R DR = 8.343 0 0.57 9.296 6.0 6.7
45 0.50 8.027 6.0 5.8
90 0.35 2.208 26.3 7.0
135 0.19 1.504 27.4 4.9
180 0.12 0.953 43.4 5.0
Ave. 21.8 5.9
z/A DR = 0.214 0 0.57 −0.322 7.7 23.1
45 0.50 −0.260 5.3 12.9
90 0.35 −0.298 4.9 13.6
135 0.19 −0.269 5.3 13.3
180 0.12 −0.474 3.4 14.9
Ave. 5.3 15.6
θ/Ak DR = 0.135 0 0.57 0.088 21.8 28.5
45 0.50 −0.027 105.0 42.1
90 0.35 −0.047 61.2 42.2
135 0.19 −0.021 164.4 50.7
180 0.12 0.031 524.0 240.6
Ave. 175.3 80.8
φ/Ak DR = 1.69 0 0.57 0.647 5.8 4.5
45 0.50 −0.264 4.3 1.4
90 0.35 1.177 4.2 5.8
135 0.19 1.026 9.6 11.6
180 0.12 −0.514 5.8 3.5
Ave. 5.9 5.4
4.4 Conclusions
More attention is needed to quantify the CFD input wave condition and uncertainty;
and to specify the CFD wave elevation evaluation locations.
KCS calm water and head waves experimental data was available from several
facilities such that resistance, sinkage and trim data provided relatively robust vali-
dation/SoA assessment. Nonetheless, more attention is needed for individual facility
Table 11 1st harmonic amplitude uncertainties for KCS seakeeping in waves (Case 2.11)
ζ /LDR = 0 0.57 0.0083 1.4
45 0.50 0.0083 1.5
90 0.35 0.0083 1.4
135 0.19 0.0083 1.1
180 0.12 0.0083 1.2
Ave. 1.3
C T × 103 DR = 5.551 0 0.57 2.647 2.5 1.2
45 0.50 6.193 1.0 1.2
90 0.35 1.424 5.7 1.5
135 0.19 3.775 3.4 2.3
180 0.12 0.642 10.2 1.2
Ave. 4.6 1.5
z/A DR = 1.101 0 0.57 0.608 3.5 1.9
45 0.50 1.161 1.2 1.3
90 0.35 1.132 1.8 1.9
135 0.19 0.404 4.4 1.6
180 0.12 0.059 24.6 1.3
Ave. 7.1 1.6
θ/Ak DR = 0.637 0 0.57 0.492 6.1 4.7
45 0.50 0.685 4.4 4.8
90 0.35 0.048 60.0 4.5
135 0.19 0.365 8.4 4.8
180 0.12 0.138 451.2 97.4
Ave. 106.0 23.2
φ/Ak DR = 2.777 0 0.57 0.012 236.6 1.0
45 0.50 0.558 5.3 1.1
90 0.35 0.378 8.3 1.1
135 0.19 2.789 1.0 1.0
180 0.12 0.247 14.3 1.3
Ave. 53.1 1.1
uncertainty analysis and for comparing data that is measured in conventional or dedi-
cated calm water resistance tests setups in all facilities, since calm water resistance
normally is measured that way.
KCS oblique waves and ONRT head and oblique wave data and uncertainties need
more data from more facilities including uncertainties for more robust validation/SoA
assessment.
354 F. Stern et al.
Table 12 2nd harmonic amplitude uncertainties for KCS seakeeping in waves (Case 2.11)
Variable/Dynamic range μ [deg] fe /fhz/θ Xi U X i %D U X i %D R
C T × 103 DR = 2.932 0 0.57 0.402 16.3 1.0
45 0.50 3.219 2.0 1.0
90 0.35 0.908 7.2 1.0
135 0.19 0.715 9.9 1.1
180 0.12 0.287 23.7 1.1
Ave. 11.8 1.1
z/A DR = 0.033 0 0.57 0.020 70.7 0.2
45 0.50 0.008 161.1 0.2
90 0.35 0.034 40.9 0.2
135 0.19 0.001 1374.9 0.2
180 0.12 0.024 67.3 0.3
Ave. 343.0 0.2
θ/Ak DR = 0.018 0 0.57 0.013 213.2 0.4
45 0.50 0.010 275.9 0.4
90 0.35 0.020 146.3 0.5
135 0.19 0.028 113.7 0.5
180 0.12 0.019 172.1 0.5
Ave. 184.2 0.5
φ/Ak DR = 0.142 0 0.57 0.006 5076.3 4.7
45 0.50 0.148 19.4 0.5
90 0.35 0.067 45.1 0.5
135 0.19 0.094 32.9 0.5
180 0.12 0.046 63.7 0.5
Ave. 1047.5 1.3
Table 13 Uncertainties for

Variable Xi U X i %D
ONRT self-propulsion tests in
calm water at Fr = 0.2 (Case n [RPS] 8.97 0.0834
3.9) σ/L × 102 0.0718 20.0
τ [deg] −0.0386 134.2
5 CFD Assessment Test Case 2.10
As shown in Table 19, 10 institutes from 8 countries using 9 different CFD codes
submitted results for both calm water and waves. Two equation turbulence models
were used with 8 using wall functions and 2 using near wall models. The solvers
were mostly Finite Volume (FV) except one is Finite Difference (FD). All but 2 used
unstructured grids. 4 used level set and 6 used VoF free surface models. Grid motion
Table 14 0th harmonic amplitude uncertainties for ONRT course keeping in waves (Case 3.12 &
3.13)
z/A DR = 0.134 0 0.62 0.057 27.0 11.5
45 0.58 0.013 108.3 10.4
90 0.44 −0.049 27.5 10.1
135 0.29 0.008 159.3 9.7
180 0.22 0.085 16.1 10.2
Ave. 67.7 10.4
θ/Ak DR = 0.025 0 0.62 −0.026 30.1 31.9
45 0.58 −0.030 15.4 18.8
90 0.44 −0.022 30.2 27.5
135 0.29 −0.023 12.7 12.0
180 0.22 −0.005 200.6 45.1
Ave. 57.8 27.1
φ/Ak DR = 0.279 0 0.62 0.019 68.7 4.7
45 0.58 −0.139 10.9 5.5
90 0.44 0.098 18.7 6.6
135 0.29 −0.181 12.0 7.8
180 0.22 0.003 494.1 5.1
Ave. 120.9 5.9
ψ/Ak DR = 0.612 0 0.62 −0.078 82.4 10.5
45 0.58 −0.690 4.3 4.8
90 0.44 −0.139 6.4 1.5
135 0.29 −0.245 17.4 7.0
180 0.22 −0.091 14.9 2.2
Ave. 25.1 5.2
u/U DR = 0.180 0 0.62 0.818 0.993 4.516
45 0.58 0.912 2.885 14.639
90 0.44 0.981 2.277 12.431
135 0.29 0.981 1.389 7.579
180 0.22 0.998 1.671 9.272
Ave. 1.84 9.69
v/U DR = 0.037 0 0.62 0.004 0.000 0.000
45 0.58 −0.030 44.079 35.896
90 0.44 −0.001 5023.561 85.149
135 0.29 −0.016 140.405 59.808
180 0.22 0.007 510.349 94.739
Ave. 1143.68 55.12
356 F. Stern et al.
Table 15 1st harmonic amplitude uncertainties for ONRT course keeping in waves (Case 3.12 and
3.13)
ζ /L DR = 0.001 0 0.62 0.010 1.0 13.7
45 0.58 0.010 1.2 15.5
90 0.44 0.010 1.0 13.4
135 0.29 0.010 3.0 39.3
180 0.22 0.009 1.6 20.1
Ave. 1.6 20.4
z/A DR = 0.89 0 0.62 0.597 2.2 1.4
45 0.58 0.821 1.7 1.6
90 0.44 1.044 1.3 1.5
135 0.29 0.517 2.7 1.6
180 0.22 0.154 8.4 1.5
Ave. 3.3 1.5
θ/Ak DR = 0.672 0 0.62 0.690 0.6 0.6
45 0.58 0.684 1.1 1.1
90 0.44 0.018 10.5 0.3
135 0.29 0.436 0.9 0.6
180 0.22 0.248 1.6 0.6
Ave. 2.9 0.6
φ/Ak DR = 1.788 0 0.62 0.001 1421.0 0.6
45 0.58 0.386 3.0 0.7
90 0.44 1.180 1.6 1.0
135 0.29 1.789 1.0 1.0
180 0.22 0.017 82.8 0.8
Ave. 301.9 0.8
ψ/Ak DR = 0.568 0 0.62 0.003 7.1 0.0
45 0.58 0.158 0.4 0.1
90 0.44 0.033 2.8 0.2
135 0.29 0.571 0.8 0.8
180 0.22 0.011 22.4 0.4
Ave. 6.7 0.3
u/U DR = 0.071 0 0.62 0.005 15.495 1.053
45 0.58 0.018 109.131 27.647
90 0.44 0.003 900.697 36.884
135 0.29 0.074 27.517 28.644
180 0.22 0.062 35.378 31.291
(continued)
Ave. 217.64 25.10
v/U DR = 0.074 0 0.62 0.000
45 0.58 0.003
90 0.44 0.054 2.168 1.581
135 0.29 0.074 27.776 27.946
180 0.22 0.001
Ave. 14.97 14.76
used dynamic patched, dynamic overset and deforming grids. Grid sizes varied from
0.6–29 M with average 9.8 M. Average number CPU and time is 60 processors and
3484 CPU hours, respectively.
It is very unfortunate that participants did not submit individual code grid and
time step verification and an important oversight of the workshop organizers. One
submission did verification for head waves but did not submit results and others have
done for different hulls and similar conditions but not for present test case. Larsson
et al. (2014) Table 11 summarizes results from several studies. For current grid
sizes/time steps, individual solution verification U S Ni was estimated at about 4 and
8%S1 for 1st and 2nd order variables, respectively, which would increase intervals of
validation. Therefore, neglecting U S Ni provides estimates of validation intervals with
warning that intervals are optimistic. Therefore, future workshops need to require
that all participants submit verification.
Table 20 provides the calm water V&V and SoA variables. Figure 3 shows the
CT , sinkage and trim absolute error running record and SoA metrics. In this case the
experimental and computational uncertainties include multiple facilities/models and
codes/solutions, respectively.
At the solution level, E i shows 60% over, 100% under and 56% over predictions
for CT , sinkage and trim, respectively. 3, 5 and 9 solutions are validated at UVi =
U D = 1.75, 7.38 and 14.25%D for CT , sinkage and trim, respectively.
At the code level, the mean code shows over, under and over predictions with
|E| = 2, 6 and 5%D for CT , σ/L and τ, respectively. |E| values are larger than
signed E. The code level solution absolute error scatter P|Ei | = 3, 4 and 7%D such
that U So Ai = 3, 8 and 16%D for CT , σ/L and τ, respectively. There is one outlier for
sinkage. Thus, 8, 7 and 9 codes/solutions1 are SoA at U So Ai intervals |E i | ≤ U So Ai .
The mean code is not N-version validated for resistance but is for sinkage and trim
at U So A = 2, 8 and 14%D, respectively.
For calm water resistance, the experimental 2%D and absolute error scatter 3%D
uncertainties are comparable to the mean absolute error 2%D; thus, all three must be
1 Code solutions are N-version validated; however, to distinguish between solution and N-version
levels when discussing the latter for simplicity the terminology code is used when referring to each
code solution.
358 F. Stern et al.
Table 16 1st harmonic phase uncertainties for ONRT course keeping in waves (Case 3.12 and
3.13)
Variable/Dynamic range μ [deg] fe /fhz/θ Xi U X i %2π
z/A 0 0.62 0.225 4.1
45 0.58 0.415 4.3
90 0.44 2.279 0.6
135 0.29 −0.951 4.0
180 0.22 0.314 2.5
Ave. 3.1
θ/Ak 0 0.62 −2.458 5.1
45 0.58 −1.791 4.4
90 0.44 −2.333 3.5
135 0.29 0.695 62.8
180 0.22 2.240 2.4
Ave. 15.6
φ/Ak 0 0.62 −0.534 36.3
45 0.58 −1.673 11.1
90 0.44 1.563 10.2
135 0.29 −0.238 11.3
180 0.22 1.706 11.0
Ave. 16.0
ψ/Ak 0 0.62 −1.334 3.6
45 0.58 0.296 5.1
90 0.44 1.734 2.0
135 0.29 2.072 2.0
180 0.22 −1.899 24.3
Ave. 7.4
u/U 0 0.62 −0.700 4.080
45 0.58 0.037 2.740
90 0.44 3.139 7.050
135 0.29 2.326 8.260
180 0.22 −2.409 2.970
Ave. 5.02
v/U 0 0.62 −1.083 4.820
45 0.58 1.015 3.190
90 0.44 2.143 0.600
135 0.29 −0.987 4.010
180 0.22 −0.949 3.030
Ave. 3.13
Table 17 2nd harmonic amplitude uncertainties for ONRT course keeping in waves (Case 3.12
and 3.13)
Variable/Dynamic range μ [deg] fe /fhz/θ Xi U X i %D U X i %D R
z/A DR = 0.076 0 0.62 0.019 66.2 16.8
45 0.58 0.012 102.9 16.9
90 0.44 0.067 19.2 17.0
135 0.29 0.088 14.6 17.0
180 0.22 0.019 69.7 17.1
Ave. 54.5 17.0
θ/Ak DR = 0.047 0 0.62 0.020 5.6 2.4
45 0.58 0.005 25.8 2.7
90 0.44 0.016 10.9 3.6
135 0.29 0.052 11.9 13.1
180 0.22 0.015 7.8 2.4
Ave. 12.4 4.8
φ/Ak DR = 0.046 0 0.62 0.004 296.7 23.8
45 0.58 0.016 67.7 24.3
90 0.44 0.040 28.6 24.9
135 0.29 0.050 27.3 29.5
180 0.22 0.013 103.8 29.7
Ave. 104.8 26.4
ψ/Ak DR = 0.022 0 0.62 0.001 45.7 1.8
45 0.58 0.008 3.6 1.4
90 0.44 0.004 28.2 5.3
135 0.29 0.023 31.4 32.7
180 0.22 0.003 44.2 5.5
Ave. 30.6 9.3
u/U DR = 0.004 0 0.62 0.004 13.834 13.320
45 0.58 0.005 0.000 0.000
90 0.44 0.001 2292.002 448.644
135 0.29 0.004 160.672 159.293
180 0.22 0.001 2519.754 494.136
Ave. 997.25 223.08
v/U DR = 0.002 0 0.62 0.000
45 0.58 0.000
90 0.44 0.001
135 0.29 0.002
180 0.22 0.001
360 F. Stern et al.
Table 18 2nd harmonic phase uncertainties for ONRT course keeping in waves (Case 3.12 and
3.13)
Variable/Dynamic range μ [deg] fe /fhz/θ Xi U X i %2π
z/A 0 0.62 −2.796 8.8
45 0.58 0.092 6.7
90 0.44 −0.082 0.8
135 0.29 2.077 11.1
180 0.22 0.089 6.6
Ave. 6.8
θ/Ak 0 0.62 1.327 7.8
45 0.58 −2.080 58.2
90 0.44 2.228 2.2
135 0.29 0.242 10.3
180 0.22 3.094 54.1
Ave. 26.5
φ/Ak 0 0.62 1.342 12.3
45 0.58 −3.007 58.0
90 0.44 −1.004 10.4
135 0.29 −1.207 13.5
180 0.22 −1.799 10.8
Ave. 21.0
ψ/Ak 0 0.62 −2.468 35.2
45 0.58 −2.632 8.7
90 0.44 1.868 8.0
135 0.29 −1.318 12.6
180 0.22 −1.765 11.1
Ave. 15.1
u/U 0 0.62 1.672 6.350
45 0.58 −2.466 5.690
90 0.44 1.464 37.600
135 0.29 0.884 8.430
180 0.22 −2.467 5.820
Ave. 12.78
v/U 0 0.62 1.754 21.940
45 0.58 −0.574 54.480
90 0.44 1.850 4.790
135 0.29 −0.193 9.960
180 0.22 −0.848 8.560
Ave. 19.95
Table 19 Submissions for KCS resistance test and seakeeping in calm water and head waves at Fr = 0.26
No. Institute Country Code Turbulence Free Discretization Grid Grid type Grid size No CPU
surface method motion .CPU time
[CPUh]
1 FORCE Denmark StarCCM+ k-ε/k-ω VoF FV Dynamic Unstructured 18– 28 M NA NA
SST Overset
2 HSVA Germany FreSCo+ k-ε/k-ω VoF FV Dynamic Unstructured NA NA NA
3 IIHR United CFDShip-Iowa k-ε/k-ω LS FD Dynamic Structured 29.2 M 224 16128
States V4.5 SST Single Overset
Ph.
4 KRISO South WAVIS k-ε/k-ω LS FV Dynamic Structured 10 M Half B. 38 170.3
Korea EARSM
5 MARIC China FINEMarine k-ε/k-ω VoF FV Deforming Unstructured 1.18-6.9 M NA NA
SST
6 NMRI Japan NAGISA k-ε/k-ω LS FV Deforming Unstructured 4.4 M 2 250
EARSM
7 NUMECA Belgium FINEMarine k-ε/k-ω VoF FV Deforming Unstructured 2–6 M NA 3116
SST
Assessment of CFD for KCS Added Resistance and for ONRT…
8 UDE Germany OpenFOAM k-ε/k-ω VoF FV Deforming Unstructured 1.06 M 8 104

SST
9 UM United OF23x k-ε/k-ω VoF FV Dynamic Unstructured 2.6 M 48 3800
States SST
10 UNIZAG-FSB Croatia swenseFoam k-ε/k-ω LS FV Dynamic Unstructured 0.61–1.6 M 40 817
SST
Ave. 9.80 M 60 3484
G2010 2.62 M
(6 sub.) (0.55–7.6 M)
361
362 F. Stern et al.
Table 20 Comparison errors and uncertainties for KCS calm water resistance test submissions at
Fr = 0.26
Variable C T 0 × 103 σ/L τ (deg) Motion
abs.
average
D 3.835 −0.00207 −0.165
U D %|D| 1.75 7.38 14.25 7.79
# 10 9 9 9
submissions
# outliers 0 1 0 0
CFD Si E i %|D| Si E i %|D| Si E i %|D| |E i |%|D|
submissions i
1 3.741 2.45 −0.00180 13.04 −0.1574 4.37 8.71
2 3.955 −3.13 −0.00198 4.35 −0.1605 2.49 3.42
3 3.845 −0.26 −0.00188 9.18 −0.1797 −9.17 9.18
4 3.738 2.53 −0.00198 4.35 −0.1725 −4.80 4.57
5 3.836 −0.03 −0.00199 3.86 −0.1560 5.22 4.54
6 3.983 −3.86 −0.00195 5.80 −0.1644 0.12 2.96
7 3.898 −1.64 −0.00195 5.80 −0.1698 −3.16 4.48
8 3.730 2.74
9 3.758 2.01 −0.00190 8.21 −0.1820 −10.57 6.93
10 4.015 −4.69 −0.00191 7.73 −0.1754 −6.56 6.33
S 3.850 −0.00194 −0.169
E%|D| −0.39 (−2.55) 6.16 (3.67) −2.45 (−0.24) 4.30 (2.15)
σ E %|D| 2.83 2.00 5.82 3.91
|E|%|D| 2.33 (3.03) 6.16 (4.33) 5.16 (3.46) 5.66 (3.61)
σ|E| %|D| 1.45 (2.83) 2.00 (1.46) 3.25 (2.34) 2.62 (2.21)
P|Ei | %|D| 2.90 3.99 6.50 5.25
P|E| %|D| 0.92 1.33 2.17 1.75
U So Ai %|D| 3.39 8.39 15.66 12.03
U So A %|D| 1.98 7.50 14.41 10.96
# validated 3 5 9 7
# N-version 8 7 9 8
validated
Mean Code No Yes Yes Yes
N-version
validated
Note The gray-shaded numbers are outliers and values in parenthesis are obtained from G2010
Fig. 3 EFD data and submission values for resistance and motions in calm water for KCS at Fr =
0.26
364 F. Stern et al.
reduced to achieve mean code N-version validation. Nonetheless, 8 codes/solutions

were N-version validated at USoAi = 3%D. For sinkage/trim, the experimental uncer-
tainties 7/14%D are larger than the absolute error scatter uncertainties 4/7%D and
the mean absolute error 6/5%D; thus, reducing the experimental uncertainties is the
most critical issue. Note that the present experimental uncertainties are based on the
scatter in the KCS sinkage and trim Fr = 0.26 measurements from several facili-
ties. Current sinkage and trim mean code errors are reasonable. Acceptance of the
accuracy of E i without consideration of the experimental and computational uncer-
tainties excludes the possibility of establishing confidence levels/intervals; therefore,
no assurance of the actual CFD capability. Whereas including uncertainties enables
evaluation of both E i and its confidence level/interval. Decomposition of the uncer-
tainties into systematic and random components for both the experiments and simu-
lations provides robust uncertainty estimates and identification of where to focus
resources for the reduction in the intervals of validation/SoA.
Table 20 includes results from G2010 for CT , σ/L and τ based on 6 CFD submis-
sions for the same test case but using FORCE 4.4 m model validation data (Simonsen
et al. 2013). T2015 results are improved for resistance and comparable for motions.
Since G2010 is only based on 6 submissions vs. 10 for T2015 and the present exper-
imental uncertainty estimate is based on multiple facilities/models, the present data
is considered a more robust estimate of current CFD code capabilities.
Figures 4, 5,6, 7 and 8 show experimental time histories for a ten-period segment
for wave elevation, resistance, heave and pitch for each wavelength condition: λ/L
= 0.65, 0.85, 1.15, 1.37 and 1.95, respectively. Figures display mostly dominant
first harmonic response except for CT , especially for λ/L = 1.15, which is near
heave/pitch resonance i.e. the frequency of encounter is close to 0.73 Hz based on
the hydrostatic estimates as reported in Table 1. FORCE reported that the mount
natural frequency was close to the heave and pitch resonance condition. Most clearly
seen in CT for λ/L equal to 0.85 and 1.15 in Figs. 5b and 6b. CFD shows similar
trends as the experiments but several solutions do not display statistical convergence.
CFD shows seemingly larger scatter for shorter vs. longer waves.
Tables 21 and 22 provide average results for each wave length and over all wave
lengths of the CFD solutions for the 0th/AR and 1st and 2nd harmonic amplitude
validation and SoA variables using %D and %DR, respectively. Table 23 is similar
but provides the 1st and 2nd harmonic phase validation and SoA variables using
%2π. The participants submitted their own harmonic analysis values for amplitudes
and phases, which were used for reconstruction of the time histories. For validation
at a single condition %D is usually used unless D nearly zero, whereas for validation
for a range of conditions both %D and %DR are of interest for evaluation of trends
with recognition that the former and latter weight the errors for small vs. large D
values, respectively. Since large responses are mostly of interest %DR is often more
meaningful. 0th and 2nd harmonics are considered 2nd order and 1st harmonic 1st
order terms from a ship design point of view. The discussions mostly use values for
λ/L averages and %DR.
Figures 9, 10, 11, 12 and 13 show the results for the 0th harmonic, 1st and 2nd
harmonic amplitudes and 1st and 2nd harmonic phases, respectively, versus λ/L. The
Fig. 4 Time series for head waves with λ/L = 0.65, H/λ = 0.0157 (A/L = 0.00511) and fe /fn =
1.5: a Wave; b Resistance coefficient; c Heave; and d Pitch
366 F. Stern et al.
368 F. Stern et al.
Fig. 7 Time series for head waves with λ/L = 1.37, H/λ = 0.0179 (A/L = 0. 0123) and fe /fn =
Table 21 (%D). Evaluations of individual and mean code values on harmonic amplitudes of resistance and motions for KCS seakeeping
370
λ/L Variable 0th harmonic amplitude 1st harmonic amplitude 2nd harmonic amplitude Ave
CT0 × 103 AR z0 /L θ0 /Ak ζ1 /L CT1 × 103 z1 /L θ1 /Ak CT2 × 103 z2 /L θ2 /Ak 1st 2nd
0.65 UD %|D| 5.20 10.34 6.20 9.10 1.30 20.40 18.44 45.56 88.60 115.60 96.30 21.42 47.33
# 10 10 10 9 7 10 10 9 10 9 9 9.00 9.57
submissions
# outliers 0 0 1 1 0 1 0 0 1 0 1 0.25 0.57
P|Ei| %|D| 11.33 122.23 7.11 9.49 4.28 13.98 19.69 38.27 27.22 61.32 112.45 19.06 50.16
P|E| %|D| 3.58 38.65 2.37 3.35 1.62 4.66 6.23 12.76 9.07 20.44 39.76 6.32 16.75
USoAi %|D| 12.47 122.67 9.43 13.14 4.48 24.73 26.98 59.50 92.69 130.86 148.05 28.92 75.62
USoA %|D| 6.31 40.01 6.64 9.70 2.08 20.93 19.46 47.31 89.06 117.39 104.18 22.44 53.33
|E|%|D| 4.57 44.60 8.08 6.93 1.68 21.16 11.45 21.77 60.39 66.21 43.49 14.01 33.47
# validated 8 3 2 6 5 4 7 8 9 9 7 6.00 6.29
# N validated 8 8 7 7 6 7 10 8 9 9 7 7.75 7.86
Mean N Yes No No Yes Yes No Yes Yes Yes Yes Yes – –
validated
0.85 UD %|D| 5.60 8.95 5.50 4.60 0.80 6.60 13.79 10.40 53.30 100.20 94.00 7.90 38.88
# 9 9 9 9 7 9 9 9 9 8 8 8.50 8.71
submissions
# outliers 0 0 1 1 0 0 0 1 0 1 1 0.25 0.57
P|Ei| %|D| 7.01 27.05 14.29 14.35 3.07 9.59 32.68 38.07 27.06 33.48 92.94 20.85 30.88
P|E| %|D| 2.34 9.02 5.05 5.07 1.16 3.20 10.89 13.46 9.02 12.65 35.13 7.18 11.18
USoAi %|D| 8.97 28.49 15.31 15.07 3.18 11.64 35.47 39.46 59.77 105.64 132.19 22.44 52.21
USoA %|D| 6.07 12.70 7.47 6.85 1.41 7.33 17.58 17.01 54.06 101.00 100.35 10.83 41.21
(continued)
F. Stern et al.
|E|%|D| 4.28 21.78 11.38 29.74 1.26 9.14 17.94 56.53 28.64 27.64 42.06 21.22 23.65
# validated 7 2 1 0 4 2 5 0 9 7 6 2.75 4.57
# N validated 7 5 7 0 5 6 7 1 9 7 7 4.75 6.00
Mean N Yes No No No Yes No No No Yes Yes Yes – –
validated
1.15 UD %|D| 2.80 8.15 9.40 701.30 0.20 5.50 4.39 4.35 11.00 17.20 12.30 3.61 108.88
# 10 10 10 9 8 3 10 9 3 9 9 7.50 8.57
submissions
# outliers 1 1 1 1 0 0 1 1 0 1 0 0.50 0.71
P|Ei| %|D| 10.09 20.24 11.13 359.34 4.58 8.51 7.24 6.98 11.40 43.52 16.93 6.83 67.52
P|E| %|D| 3.36 6.75 3.71 127.05 1.62 4.91 2.41 2.47 6.58 15.38 5.64 2.85 24.07
USoAi %|D| 10.48 21.82 14.57 788.00 4.59 10.13 8.47 8.22 15.84 46.79 20.93 7.85 131.20
USoA %|D| 4.38 10.58 10.11 712.71 1.63 7.37 5.01 5.00 12.82 23.08 13.53 4.75 112.46
|E|%|D| 4.73 8.31 14.07 238.14 1.46 19.81 6.40 4.44 23.85 21.35 13.32 8.03 46.25
# validated 5 5 1 8 4 0 4 4 0 4 5 3.00 4.00

# N validated 8 8 4 8 7 0 6 7 0 7 6 5.00 5.86
Mean N No Yes No Yes Yes No No Yes No Yes Yes – –
validated
1.37 UD %|D| 2.90 8.49 8.10 212.30 0.30 2.70 4.43 4.27 14.40 52.20 26.40 2.93 46.40
(continued)
371
372
# 9 9 9 9 8 9 9 9 9 9 9 8.75 9.00
submissions
# outliers 0 0 1 1 0 1 0 0 0 1 1 0.25 0.57
P|Ei| %|D| 12.12 25.87 42.60 238.41 5.19 20.16 10.11 15.69 40.30 27.52 47.25 12.79 62.01
P|E| %|D| 4.04 8.62 15.06 84.29 1.83 7.13 3.37 5.23 13.43 9.73 16.71 4.39 21.70
USoAi %|D| 12.46 27.23 43.37 319.24 5.19 20.34 11.04 16.26 42.79 59.01 54.13 13.21 79.75
USoA %|D| 4.97 12.10 17.10 228.42 1.86 7.62 5.56 6.75 19.69 53.10 31.24 5.45 52.38
|E|%|D| 8.56 18.53 26.07 153.10 1.99 10.31 7.28 8.61 21.81 19.02 21.26 7.05 38.33
# validated 1 4 1 5 3 1 2 4 4 8 6 2.50 4.14
# N validated 6 6 7 7 7 7 7 7 8 8 7 7.00 7.00
Mean N No No No Yes No No No No No Yes Yes – –
validated
1.95 UD %|D| 9.10 26.92 18.60 57.90 0.80 0.70 4.15 4.03 55.50 46.90 18.40 2.42 33.33
# 10 10 10 9 8 10 10 9 10 9 9 9.25 9.57
submissions
# outliers 1 1 1 1 1 1 1 1 1 1 1 1.00 1.00
P|Ei| %|D| 5.06 23.11 13.55 23.24 0.77 6.94 4.28 3.42 46.93 9.88 19.68 3.85 20.21
P|E| %|D| 1.69 7.70 4.52 8.22 0.29 2.31 1.43 1.21 15.64 3.49 6.96 1.31 6.89
USoAi %|D| 10.41 35.48 23.01 62.39 1.11 6.98 5.96 5.29 72.68 47.93 26.94 4.83 39.83
USoA %|D| 9.26 28.00 19.14 58.48 0.85 2.42 4.39 4.21 57.66 47.03 19.67 2.97 34.18
(continued)
F. Stern et al.
|E|%|D| 3.44 10.44 11.53 12.53 0.49 9.77 3.10 4.17 20.35 7.41 8.26 4.38 10.56
# validated 9 8 7 8 6 0 7 6 8 8 7 4.75 7.86
# N validated 9 8 8 8 7 2 7 6 9 8 7 5.50 8.14
Mean N Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes – –
Validated
Ave UD %|D| 5.12 12.57 9.56 197.04 0.68 7.18 9.04 13.72 44.56 66.42 49.48 7.66 54.96
# 9.60 9.60 9.60 9.00 7.60 8.20 9.60 9.00 8.20 8.80 8.80 8.60 9.09
submissions
# outliers 0.40 0.40 1.00 1.00 0.20 0.60 0.40 0.60 0.40 0.80 0.80 0.45 0.69
P|Ei| %|D| 9.12 43.70 17.73 128.97 3.58 11.84 14.80 20.48 30.58 35.14 57.85 12.67 46.16
P|E| %|D| 3.00 14.15 6.14 45.60 1.30 4.44 4.87 7.02 10.75 12.34 20.84 4.41 16.12
USoAi %|D| 10.96 47.14 21.14 239.57 3.71 14.76 17.58 25.75 56.76 78.05 76.45 15.45 75.72
USoA %|D| 6.20 20.68 12.09 203.23 1.57 9.13 10.40 16.06 46.66 68.32 53.80 9.29 58.71
|E|%|D| 5.12 20.73 14.22 88.09 1.38 14.04 9.23 19.11 31.01 28.33 25.68 10.94 30.45
# validated 6.00 4.40 2.40 5.40 4.40 1.40 5.00 4.40 6.00 7.20 6.20 3.80 5.37
# N validated 7.60 7.00 6.60 6.00 6.40 4.40 7.40 5.80 7.00 7.80 6.80 6.00 6.97
373
Table 22 (%DR). Evaluations of individual and mean code values on harmonic amplitudes of resistance and motions for KCS seakeeping
374
0.65 UD %|DR| 7.23 4.38 8.24 7.67 0.61 3.14 2.99 0.71 28.24 39.27 3.21 1.86 14.03
# submissions 10 10 10 9 7 10 10 9 10 9 9 9.00 9.57
# outliers 0 0 1 1 0 1 0 0 1 0 1 0.25 0.57
P|Ei| %|DR| 15.75 51.72 9.45 7.99 2.01 2.15 3.19 0.60 8.68 20.83 3.74 1.99 16.88
P|E| %|DR| 4.98 16.36 3.15 2.83 0.76 0.72 1.01 0.20 2.89 6.94 1.32 0.67 5.50
USoAi %|DR| 17.32 51.91 12.54 11.08 2.10 3.81 4.37 0.93 29.55 44.46 4.93 2.80 24.54
USoA %|DR| 8.78 16.93 8.82 8.17 0.98 3.22 3.15 0.74 28.39 39.88 3.47 2.02 16.35
|E|%|DR| 6.35 18.87 10.74 5.84 0.79 3.26 1.86 0.34 19.25 22.50 1.45 1.56 12.14
# validated 8 3 2 6 5 4 7 8 9 9 7 6.00 6.29
# N validated 8 8 7 7 6 7 10 8 9 9 7 7.75 7.86
Mean N validated Yes No No Yes Yes No Yes Yes Yes Yes Yes – –
0.85 UD %|DR| 8.76 6.82 5.67 4.68 0.47 1.28 4.19 1.38 19.69 54.16 5.76 1.83 15.08
# submissions 9 9 9 9 7 9 9 9 9 8 8 8.50 8.71
# outliers 0 0 1 1 0 0 0 1 0 1 1 0.25 0.57
P|Ei| %|DR| 10.96 20.62 14.73 14.59 1.81 1.85 9.92 5.05 9.99 18.09 5.69 4.66 13.53
P|E| %|DR| 3.65 6.87 5.21 5.16 0.68 0.62 3.31 1.79 3.33 6.84 2.15 1.60 4.75
USoAi %|DR| 14.03 21.72 15.78 15.32 1.87 2.25 10.77 5.23 22.08 57.10 8.10 5.03 22.02
USoA %|DR| 9.49 9.69 7.70 6.96 0.83 1.42 5.34 2.26 19.97 54.59 6.15 2.46 16.36
|E|%|DR| 6.70 16.61 11.73 30.24 0.74 1.77 5.45 7.50 10.58 14.94 2.58 3.86 13.34
# validated 7 2 1 0 4 2 5 0 9 7 6 2.75 4.57
(continued)
F. Stern et al.
# N validated 7 5 7 0 5 6 7 1 9 7 7 4.75 6.00
Mean N validated Yes No No No Yes No No No Yes Yes Yes – –
1.15 UD %|DR| 6.69 10.09 4.29 11.80 0.19 0.99 4.91 2.95 14.51 14.41 4.74 2.26 9.50
# submissions 10 10 10 9 8 3 10 9 3 9 9 7.50 8.57
# outliers 1 1 1 1 0 0 1 1 0 1 0 0.50 0.71
P|Ei| %|DR| 24.12 25.07 5.08 6.04 4.26 1.54 8.11 4.74 15.03 36.45 6.52 4.66 16.90
P|E| %|DR| 8.04 8.36 1.69 2.14 1.51 0.89 2.70 1.67 8.68 12.89 2.17 1.69 6.28
USoAi %|DR| 25.03 27.02 6.65 13.26 4.26 1.83 9.48 5.58 20.89 39.19 8.06 5.29 20.01
USoA %|DR| 10.46 13.10 4.61 11.99 1.52 1.33 5.60 3.39 16.90 19.33 5.21 2.96 11.66
|E|%|DR| 11.32 10.29 6.42 4.01 1.36 3.58 7.17 3.02 31.45 17.88 5.13 3.78 12.36
# validated 5 5 1 8 4 0 4 4 0 4 5 3.00 4.00
# N validated 8 8 4 8 7 0 6 7 0 7 6 5.00 5.86
Mean N validated No Yes No Yes Yes No No Yes No Yes Yes – –
1.37 UD %|DR| 6.82 6.91 3.31 12.39 0.34 1.59 4.82 3.74 15.50 23.37 8.69 2.62 11.00
# submissions 9 9 9 9 8 9 9 9 9 9 9 8.75 9.00
# outliers 0 0 1 1 0 1 0 0 0 1 1 0.25 0.57
P|Ei| %|DR| 28.52 21.05 17.42 13.92 5.85 11.84 11.01 13.74 43.38 12.32 15.56 10.61 21.74
P|E| %|DR| 9.51 7.02 6.16 4.92 2.07 4.19 3.67 4.58 14.46 4.36 5.50 3.63 7.42
USoAi %|DR| 29.32 22.16 17.73 18.64 5.86 11.94 12.02 14.24 46.07 26.42 17.82 11.01 25.45
(continued)
375
376
USoA %|DR| 11.70 9.85 6.99 13.34 2.09 4.48 6.06 5.91 21.20 23.77 10.29 4.64 13.88
|E|%|DR| 20.14 15.08 10.66 8.94 2.24 6.05 7.93 7.55 23.48 8.51 7.00 5.94 13.40
# validated 1 4 1 5 3 1 2 4 4 8 6 2.50 4.14
# N validated 6 6 7 7 7 7 7 7 8 8 7 7.00 7.00
Mean N validated No No No Yes No No No No No Yes Yes – –
1.95 UD %|DR| 16.58 6.42 6.12 25.79 1.18 0.81 4.82 4.09 18.33 62.83 19.01 2.72 22.16
# submissions 10 10 10 9 8 10 10 9 10 9 9 9.25 9.57
# outliers 1 1 1 1 1 1 1 1 1 1 1 1.00 1.00
P|Ei| %|DR| 9.23 5.51 4.46 10.35 1.13 8.01 4.97 3.47 15.50 13.24 20.33 4.40 11.23
P|E| %|DR| 3.08 1.84 1.49 3.66 0.43 2.67 1.66 1.23 5.17 4.68 7.19 1.50 3.87
USoAi %|DR| 18.97 8.46 7.57 27.79 1.63 8.05 6.93 5.37 24.01 64.21 27.84 5.49 25.55
USoA %|DR| 16.86 6.68 6.30 26.05 1.25 2.79 5.10 4.27 19.05 63.01 20.33 3.35 22.61
|E|%|DR| 6.26 2.49 3.79 5.58 0.72 11.28 3.60 4.24 6.72 9.92 8.53 4.96 6.19
# validated 9 8 7 8 6 0 7 6 8 8 7 4.75 7.86
# N validated 9 8 8 8 7 2 7 6 9 8 7 5.50 8.14
Mean N validated Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes – –
Ave UD %|DR| 9.22 6.92 5.53 12.47 0.56 1.56 4.35 2.58 19.25 38.81 8.28 2.26 14.35
# submissions 9.60 9.60 9.60 9.00 7.60 8.20 9.60 9.00 8.20 8.80 8.80 8.60 9.09
# outliers 0.40 0.40 1.00 1.00 0.20 0.60 0.40 0.60 0.40 0.80 0.80 0.45 0.69
(continued)
F. Stern et al.
P|Ei| %|DR| 17.72 24.80 10.22 10.58 3.01 5.08 7.44 5.52 18.52 20.19 10.37 5.26 16.06
P|E| %|DR| 5.85 8.09 3.54 3.74 1.09 1.82 2.47 1.89 6.91 7.14 3.67 1.82 5.56
USoAi %|DR| 20.94 26.25 12.05 17.22 3.14 5.58 8.71 6.27 28.52 46.28 13.35 5.93 23.52
USoA %|DR| 11.46 11.25 6.88 13.30 1.33 2.65 5.05 3.32 21.10 40.12 9.09 3.09 16.17
|E|%|DR| 10.15 12.67 8.67 10.92 1.17 5.19 5.20 4.53 18.30 14.75 4.94 4.02 11.49
# validated 6.00 4.40 2.40 5.40 4.40 1.40 5.00 4.40 6.00 7.20 6.20 3.80 5.37
# N validated 7.60 7.00 6.60 6.00 6.40 4.40 7.40 5.80 7.00 7.80 6.80 6.00 6.97
377
378 F. Stern et al.
Table 23 (%2π). Evaluations of individual and mean code values on harmonic phase of resistance
and motions for KCS seakeeping
λ/L Variable 1st harmonic phase 2nd harmonic phase Ave
CT1 × 103 z1 /L θ1 /Ak CT2 × 103 z2 /L θ2 /Ak 1st 2nd
0.65 UD %2p 3.00 2.40 7.40 24.80 61.40 78.80 4.27 55.00
# 9 9 8 9 8 8 8.67 8.33
submissions
# outliers 0 1 1 1 0 1 0.67 0.67
P|Ei| %2p 1.90 5.76 4.80 5.35 12.16 4.84 4.15 7.45
P|E| %2p 0.63 2.04 1.81 1.89 4.30 1.83 1.49 2.67
USoAi %2p 3.55 6.24 8.82 25.37 62.59 78.95 6.20 55.64
USoA %2p 3.07 3.15 7.62 24.87 61.55 78.82 4.61 55.08
|E|%2p 4.46 5.59 2.46 6.71 39.17 27.12 4.17 24.34
# validated 1 0 7 8 8 7 2.67 7.67
#N 1 6 7 8 8 7 4.67 7.67
validated
Mean N No No Yes Yes Yes Yes – –
validated
0.85 UD %2p 3.60 3.40 3.30 11.40 64.80 40.90 3.43 39.03
# 8 8 8 8 7 7 8.00 7.33
submissions
# outliers 0 1 0 0 1 1 0.33 0.67
P|Ei| %2p 9.14 10.79 5.57 16.03 10.61 10.58 8.50 12.41
P|E| %2p 3.23 4.08 1.97 5.67 4.33 4.32 3.09 4.77
USoAi %2p 9.82 11.32 6.48 19.67 65.66 42.25 9.20 42.53
USoA %2p 4.84 5.31 3.84 12.73 64.94 41.13 4.66 39.60
|E|%2p 8.39 10.92 10.99 19.80 42.13 37.68 10.10 33.20
# validated 2 0 0 1 6 4 0.67 3.67
#N 4 5 0 3 6 4 3.00 4.33
validated
Mean N No No No No Yes Yes – –
validated
1.15 UD %2p 0.70 0.30 0.30 2.20 2.70 2.10 0.43 2.33
# 3 9 8 3 8 8 6.67 6.33
submissions
# outliers 0 1 0 0 1 1 0.33 0.67
P|Ei| %2p 6.59 5.16 3.21 69.59 5.65 12.91 4.99 29.38
P|E| %2p 3.81 1.83 1.14 40.18 2.13 4.88 2.26 15.73
USoAi %2p 6.63 5.17 3.23 69.62 6.26 13.08 5.01 29.65
USoA %2p 3.87 1.85 1.17 40.24 3.44 5.31 2.30 16.33
(continued)
|E|%2p 3.99 2.93 2.32 20.91 3.96 8.08 3.08 10.98
# validated 1 1 0 2 4 1 0.67 2.33
#N 3 7 7 3 5 5 5.67 4.33
validated
Mean N No No No Yes No No – –
validated
1.37 UD %2p 0.40 0.50 0.40 16.60 9.30 4.20 0.43 10.03
# 8 8 8 8 8 8 8.00 8.00
submissions
# outliers 0 0 0 1 0 0 0.00 0.33
P|Ei| %2p 4.42 1.66 2.75 5.46 16.78 7.34 2.94 9.86
P|E| %2p 1.56 0.59 0.97 2.06 5.93 2.60 1.04 3.53
USoAi %2p 4.44 1.73 2.78 17.47 19.19 8.46 2.98 15.04
USoA %2p 1.61 0.77 1.05 16.73 11.03 4.94 1.15 10.90
|E|%2p 2.13 1.74 2.25 3.20 8.23 4.64 2.04 5.36
# validated 0 1 0 7 5 4 0.33 5.33
#N 7 5 6 7 7 7 6.00 7.00
validated
Mean N No No No Yes Yes Yes – –
validated
1.95 UD %2p 0.20 0.20 0.20 11.30 8.70 3.00 0.20 7.67
# 9 9 8 9 8 8 8.67 8.33
submissions
# outliers 1 1 1 1 1 0 1.00 0.67
P|Ei| %2p 1.92 1.47 1.43 15.64 8.49 19.58 1.61 14.57
P|E| %2p 0.68 0.52 0.54 5.53 3.21 6.92 0.58 5.22
USoAi %2p 1.93 1.48 1.44 19.30 12.16 19.81 1.62 17.09
USoA %2p 0.71 0.56 0.57 12.58 9.27 7.54 0.61 9.80
|E|%2p 5.48 5.12 4.92 8.76 6.81 13.36 5.17 9.64
# validated 0 0 0 6 4 0 0.00 3.33
#N 0 0 0 7 7 7 0.00 7.00
validated
Mean N No No No Yes Yes No – –
validated
Ave UD %2p 1.58 1.36 2.32 13.26 29.38 25.80 1.75 22.81
(continued)
380 F. Stern et al.
# 7.40 8.60 8.00 7.40 7.80 7.80 8.00 7.67
submissions
# outliers 0.20 0.80 0.40 0.60 0.60 0.60 0.47 0.60
P|Ei| %2p 4.80 4.97 3.55 22.41 10.74 11.05 4.44 14.73
P|E| %2p 1.98 1.81 1.29 11.07 3.98 4.11 1.69 6.39
USoAi %2p 5.28 5.19 4.55 30.29 33.17 32.51 5.00 31.99
USoA %2p 2.82 2.33 2.85 21.43 30.05 27.55 2.67 26.34
|E|%2p 4.89 5.26 4.59 11.88 20.06 18.18 4.91 16.70
# validated 0.80 0.40 1.40 4.80 5.40 3.20 0.87 4.47
#N 3.00 4.60 4.00 5.60 6.60 6.00 3.87 6.07
validated
left column shows D, Si (including outliers) and S̄ without outliers. The middle and
right columns show UD , U So Ai , USoA , |E i | and |E| using %D and %DR, respectively.
The solutions show large scatter especially for resistance but also motions.
The 0th harmonic experiments show that CT0 /AR peak at the resonance condition
with small/medium values for short/long waves; CT0 is considerably larger than CT
in calm water except for short waves. Heave and pitch magnitudes are mostly larger
than their calm water values, except for outliers. The mean solution S̄ closely follows
the experimental trends; however, the scatter in the solutions is very large. The errors
are largest for CT0 /AR near resonance and for motions for wave lengths just before
and after resonance. The solution scatter mostly correlates with large errors.
The 1st harmonic amplitude experiments show that heave and pitch follow the
expected trends: for short/long waves heave/pitch are small/unity. Heave shows a
maximum at the resonance condition. CT1 increases with λ/L except for the resonance
condition where it shows a dramatic reduction. S̄ closely follows the experimental
trends with reduced scatter compared to the 0th harmonic. The error and scatter for
CT1 is larger for long waves, whereas for the motions its mostly like that for the 0th
harmonic.
The 1st harmonic phase experiments shows expected trends that pitch leads heave
by nearly π/2 for large λ/L and for small λ/L the phase difference for heave and pitch
reduces to nearly zero. CT1 phase is mostly small except near resonance where it is
near π/2. S̄ follows the experimental trends. The errors and scatter for both CT1 and
motions are largest for the wavelength just before resonance and for the longest wave.
The 2nd harmonic amplitude experiments show that CT2 trends are similar CT0 ,
heave is nearly zero and pitch increases with λ/L with similar trends as for CT1 .
For AR, UD = 7%DR and |E| = 13%DR such that only 4 solutions are validated.
The code level scatter P|Ei | = 25%DR is much larger than UD . U So Ai = 26%DR
such that 7 codes/solutions are N-version validated. USoA = 11%DR and the mean
code is not N-version validated.
Fig. 9 EFD data and submission values for 0th harmonic amplitudes of resistance and motions for
KCS seakeeping in regular waves at Fr = 0.26
382 F. Stern et al.
Fig. 10 EFD data and submission values for 1st harmonic amplitudes of resistance and motions
for KCS seakeeping in regular waves at Fr = 0.26
Fig. 11 EFD data and submission values for 1st harmonic phases of resistance and motions for
384 F. Stern et al.
Fig. 12 EFD data and submission values for 2nd harmonic amplitudes of resistance and motions
for KCS seakeeping in regular waves at Fr = 0.26
Fig. 13 EFD data and submission values for 2nd harmonic phases of resistance and motions for
386 F. Stern et al.
For first harmonic heave and pitch amplitudes, UD is about 4%DR and |E| =
5%DR, respectively, such that only about half the solutions validated. The code level
scatter P|Ei | is about 7%DR and about twice as large as UD . U So Ai = 8%DR such
that only about half the solutions are N-version validated. USoA is about 4%DR and
the mean code is N-version validated for heave but not pitch. The trends for the first
harmonic phases are like those for the first harmonic amplitudes with |E| = 5%DR.
The trends for the 2nd harmonic amplitudes and phases for errors and uncertainties
are similar but with much larger intervals.
The solution scatter uncertainty is larger than UD especially for AR. The errors
for AR are six times larger than for calm water resistance and for heave and pitch are
comparable to those for sinkage and trim. The largest errors are for near resonance
for AR, near resonance for heave, and just before/after resonance for pitch.
5.1 Conclusions
The calm water experimental data included assessment of facility biases/scale effects
with reasonable uncertainties for resistance, sinkage and trim. However, more atten-
tion is needed for assessment of individual facility uncertainties and for comparing
data that is measured in conventional or dedicated calm water resistance tests setups
in all facilities, since calm water resistance normally is measured that way. Reducing
experimental uncertainties is a deciding factor in reducing intervals of validation
and SoA for sinkage and trim. The head wave experimental data showed significant
effects of statistical convergence, mount resonance, and scatter due to model size and
facility/mount. More attention is needed to document the CFD input wave condition
and location for evaluation. Nonetheless, the data was useful for CFD validation/SoA
with reasonable uncertainties. Reducing the experimental uncertainties is important
for reducing the validation interval, but not a deciding factor in reducing the SoA
interval since the order of magnitude of the current scatter in simulations is larger
than UD for 1st order variables.
It is very unfortunate that solution verification was not provided by any of the
submitters such that the current intervals of validation are optimistic. A reasonable
estimate for the numerical uncertainty is that it is comparable to that for the exper-
imental uncertainty; thus, a better estimate for the intervals of validation is likely
30%. larger. Hopefully, the current SoA estimates will motivate the 8th CFD Ship
Hydrodynamics Workshop to be held in 2021 to attract more submissions including
verification, which will likely reduce the CFD scatter; and benchmark experiments
from more facilities with reduced uncertainties. Ship hydrodynamics CFD work-
shops should follow the example of the AIAA CFD Drag Prediction Workshops by
requiring the participants to submit verification and possibly use provided or spec-
ified grids. This will provide more robust uncertainty estimates and validation/SoA
confidence levels/intervals.
The current sample size of 10 institutes from 8 countries using 9 different CFD
codes submitting results for both calm water and waves is enough, especially as the
codes differ in modeling and numerical methods. The grid size variation was large
from 0.6–29 M cells.
For calm water, 3, 5 and 9 solutions are validated at UVi = 1.75, 7.38 and 14.25%D
for CT , sinkage and trim, respectively. At the code level, the mean code shows
|E| = 2, 6 and 5%D for CT , σ/L and trim, respectively. The code level U So Ai = 3, 8
and 16%D for CT , sinkage/L and trim, respectively. Thus, 8, 7 and 9 codes are SoA
at the U So Ai intervals. The mean code is not N-version validated for resistance but is
for sinkage and trim at U So A = 2, 8 and 14%DR, respectively. Since G2010 is only
based on 6 submissions vs. 10 for T2015 and the present experimental uncertainty
estimate considers multiple facilities/models, the present results are considered a
more robust estimate of the current CFD code capabilities.
For head waves (averaged over all λ/L), the scatter in the solutions is very large,
i.e., 5 and 16%DR for 1st order and 2nd order variables, respectively. Interestingly the
mean code solution shows relatively close agreement with the experiments both for
1st and 2nd order variables. This suggests that the present estimates using standard
deviations well represent the multiple code level random/precision simulation and
error uncertainty components. 4/5 solutions are validated at 2/14%DR for 1st and 2nd
order variables. 6/7 codes are N version validated SoAi at 6/24%DR for 1st and 2nd
order variables. The mean code is not SoA for 1st order variables with |E| = 4%DR
and U So A = 3%DR, whereas it is SoA for 2nd order variables at interval 16%DR
with |E| = 12%DR
Table 24 summarizes the present results for short, medium and long waves,
including comparison with G2010 and potential flow and CFD results from Bunnik
et al. (2010). Error/scatter from Bunnik et al. (2010) were extracted from the provided
figures, as quantitative values were not provided; also, H/λ values are not known but
presumably in the linear range.
The present error/scatter with 10 submissions are larger for short than long
waves. The average errors/scatter are 8/7, 16/10 and 24/28 for heave, pitch and
AR, respectively. The errors/scatter increase from heave to pitch to AR.
The G2010 error/scatter with 6 submissions for KCS do not include short waves
and AR. The average errors/scatter are 10/9 and 4/4 for heave and pitch, respec-
tively, which are comparable T2015. The G2010 error/scatter with 5 submissions for
KVLCC2 are mostly larger for short waves for heave and pitch, but not AR. The
average errors/scatter are 11/7, 10/4 and 38/1 for heave, pitch and AR, respectively.
The errors are similar for heave and pitch but larger for AR and the scatter is largest
for heave and smaller for pitch and AR. The errors/scatter are larger/similar compared
to those for KCS. Thus, T2015 and G2010 CFD errors/scatter are comparable but
T2015 errors/scatter are considered most reliable since there are more submissions
in T2015 with more rigorous analysis.
The Bunnik et al. (2010) error/scatter with 8 PF submissions for a container ship
are mostly larger for short than long waves for heave, pitch and AR. The average
errors/scatter are 25/13, 31/11 and 39/14 for heave, pitch and AR, respectively. The
errors/scatter mostly increase from heave to pitch to AR. The error/scatter with 3
CFD submissions for a container ship are mostly larger for short than long waves
for heave, pitch and AR. The average errors/scatter are 35/15, 22/12 and 36/13 for
388 F. Stern et al.
Table 24 CFD and PF comparison for motions and added resistance computations in head waves
λ/L |E|%|D| σ|E| %|D|
z1 /A θ1 /Ak AR z1 /A θ1 /Ak AR
Container ship: KCS (T2015; CFD; N = 10)
Short (λ/L = 0.65) 11.5 21.8 44.6 9.8 19.1 61.1
Medium (λ/L = 0.85, 1.15, 1.37) 10.5 23.2 16.2 8.3 10.1 12.2
Long (λ/L = 1.95) 3.1 4.2 10.4 2.1 1.7 11.6
Average 8.4 16.4 23.7 6.8 10.3 28.3
Container ship: KCS (G2010; CFD; N = 6)
Medium (λ/L = 1.15, 1.33) 8.9 3.9 5.2 3.1
Long (λ/L = 2.0) 10.8 4.1 12.7 4.6
Average 9.9 4 9 3.9
Tanker: KVLCC2 (G2010; CFD; N = 5)
Short (λ/L = 0.6, 0.64) 16.6 15 34.5 8.5 6.6
Medium (λ/L = 0.92, 1.1) 10.3 5.9 2.9 9.1 2.8 1.9
Long (λ/L = 1.6) 6.7 9.6 76 3.6 3.9 0.6
Average 11.2 10 37.8 7.1 4.4 1.3
G2010–Average 10.6 7.0 37.8 8.1 4.2 1.3
Container ship (Bunnik et al. 2010; PF; N = 8)
Short (λ/L = 0.3, 0.5, 0.6, 0.7) 36.5 58.8 53.8 9.5 17.4 23.8
Medium (λ/L = 0.9, 1.1) 29.4 23 23.3 23.8 7.4 11.2
Long (λ/L = 1.4) 10 12 39.6 5.7 9 7.8
Average 25.3 31 38.9 13 11 14.3
Container ship (Bunnik et al. 2010; CFD; N = 3)
Short (λ/L = 0.3, 0.5, 0.6, 0.7) 48.1 50.2 48.1 12.6 27.2 4.3
Medium (λ/L = 0.9,1.1) 45.1 13.6 33.7 28.1 8.6 26.3
Long (λ/L = 1.4) 10.7 2.1 26.4 4.5 1 9.7
Average 34.6 22 36.1 15 12 13.4
Ferry (Bunnik et al. 2010; PF; N = 7)
Medium (λ/L = 0.85, 1.0, 1.17) 29.2 27.3 18.1 14
Long (λ/L = 1.4, 1.75, 2.23, 2.90) 23.6 9.4 15.1 6.2
Average 26.4 18 17 10
Bunnik et al. 2010–Average 28.8 23.7 37.5 15.0 11.0 13.9
heave, pitch and AR, respectively. The error/scatter are larger for heave and AR than
pitch. The error/scatter with 7 PF submissions for ferry ship are larger for short than
long waves for heave and pitch. The average errors/scatter are 26/17 and 18/10 for
heave and pitch. Bunnik et al. (2010) concluded that the scatter is large, there is a
large difference between methods due to differences in the restoring force, added
mass and damping, and diffraction models but the best methods used more accurate
forward speed models, added resistance and internal loads were under predicted, and
CFD did not show benefit in consideration larger CPU requirement.
Current CFD results from T2015 shows that CFD can predict head waves heave,
pitch and AR with better accuracy than potential flow for short, medium and long
waves. CFD results also show smaller scatter than potential flow solutions for heave
and pitch. Use of finer grids and improved user best practices will likely reduce CFD
errors and scatter. In addition, improved benchmark experimental data from multiple
facilities and solution verification will provide estimates that are more robust for
intervals of validation and SoA.
6 CFD Assessment Test Case 2.11
As shown in Table 25, only 2 institutes from 2 countries using two different CFD
codes submitted results for both calm water and all the wave heading conditions.
Both institutes also submitted results for test case 2.10 using the same codes and
one used among the coarsest and the other among the finest grids. Although just 2
submissions are available, the results provide an opportunity for the assessment of
CFD for AR variable heading.
Table 26 provides calm water results. Both codes/solutions have larger errors than
shown for test case 2.10 calm water resistance, sinkage and trim.
Figure 14 shows the non-dimensional frequency of encounter fe L/U for each
heading vs. λ/L, including horizontal lines marking the heave/pitch and roll natural
frequencies and parametric roll frequency. Test case 2.11 is for λ/L = 1 and μ = 0,
45, 90, 135 and 180°. Head and bow quartering conditions are close to the heave/pitch
natural frequency. Beam condition is far from all resonance conditions. Note that B/λ
= 0.14. Stern quartering and following wave conditions are close to the roll natural
frequency and parametric roll frequency.
The experimental and CFD time histories are shown in Figs. 15, 16, 17, 18 and
19 for μ = 0, 45, 90, 135 and 180°., respectively. The experiments only cover 5Te
for μ = 0 and 45°, 3Te for μ = 90, 135 and 180° and show more scatter compared
to the 10 Te available for test case 2.10. There is mostly good agreement between
the 2 codes and the data, but clearly larger errors than for test case 2.10. The largest
differences are for roll, phase shifts and poorer agreement for stern waves.
Tables 27 and 28 provide average results for each heading and over all headings
of the CFD solutions for the 0th/AR and 1st and 2nd harmonic amplitude validation
and SoA variables using %D and %DR, respectively. Table 29 is similar but provides
the 1st and 2nd harmonic phase validation and SoA variables using %2π. Similarly,
as for test case 2.10, the participants submitted their own harmonic analysis values
for amplitudes and phases, which were used for reconstruction of the time histories.
The discussions mostly use values for μ averages and %DR.
harmonic amplitudes and 1st and 2nd harmonic phases, respectively, vs. μ. The left
390
Table 25 Submissions for KCS resistance test and seakeeping in calm water and head waves at Fr = 0.26 (Case 2.11)
No Institute Code Turbulence Free Discretization Grid Grid type Grid size No.CPU CPU
surface motion time
1 IIHR CFDShip-Iowa k-ε/k-ω SST LS FD Dynamic Structured 29.2 M 224 16128
V4.5 Single Overset CPUh
Ph.
2 UNIZAG-FSB swenseFoam k-ε/k-ω SST LS FV Dynamic Unstructured 0.61–1.6 M 40 817
CPUh
Ave. 15.15 M 132 8472.5
F. Stern et al.
Table 26 Comparison errors and uncertainties for KCS calm water resistance test submissions at
Fr = 0.26 (Case 2.11)
Variable C T 0 × 103 σ/L τ (deg)
D 4.888 −0.001704 −0.022
U D %|D| 2.5 0.6 65.5
CFD submissions i Si E i %|D| Si E i %|D| Si E i %|D|
1 4.512 7.69 −0.00187 −9.74 −0.1816 −725.45
2 4.528 7.36 −0.00176 −3.29 −0.1717 −680.45
E%D 7.53 −6.51 −702.95
E %D 7.53 6.51 702.95
Fig. 14 Wave encounter

frequencies compared to
natural frequencies (fz/θ =
2.20; fφ = 0.46; fPR = 0.93)
for KCS test conditions in
waves (WMTC:
Sadat-Hosseini et al. 2015)
column shows D, Si and S̄. The middle and right columns show UD ,|E i | and |E|
using %D and %DR, respectively.
The 0th harmonic experiments show large CT0 , AR and θ0 /Ak for head waves
(also following waves for θ0 /Ak). z0 /A is nearly constant except for following waves.
φ0 /Ak is largest for μ = 90 and 135°. Both CFD submissions mostly over predict
CT0 and AR; show similar trends with each other and the data for z0 /A; show similar
trends with each other but different from the data for θ0 /Ak; and mostly different
with each other and the data for φ0 /Ak. The largest errors are for the motions.
CT1 and θ1 /Ak show similar oscillations vs. μ with peaks at 45 and 135 and trough
at 90°; however, amplitudes are small/large for head/following waves, respectively.
z1 /A is largest for bow and beam waves and smallest for stern waves. Since B/λ =
0.14 for beam waves, pitch is nearly zero and heave and wave amplitudes coincide,
as the model simply moves up and down parallel with the wave. φ1 /Ak is largest
for stern quartering waves. Both CFD solutions show overall good agreement with
each other and the data for both amplitude and phase, except larger amplitude CT1
and some differences in phases for stern waves. The largest errors are for the CT1
amplitude for following waves.
392 F. Stern et al.
2
(a)
1
ζ/A
-1
-2
0 1 2 3 4 5 6 7 8 9 10
t/Te
(b)
15
3
CT×10
10
0
0 1 2 3 4 5 6 7 8 9 10
t/Te
2
(c)
1
z/A
-1
-2
0 1 2 3 4 5 6 7 8 9 10
t/Te
1
(d)
0.5
θ/Ak
-0.5
-1
0 1 2 3 4 5 6 7 8 9 10
t/Te
Fig. 15 Submissions with EFD for KCS course keeping in regular waves with μ = 0°: a Wave;
b Resistance coefficient; c Heave; d Pitch; e Roll; f x-velocity
0.6
(e)
0.4
φ/Ak
0.2
-0.2
0 1 2 3 4 5 6 7 8 9 10
t/Te
1.1
(f)
1.05
u/U
0.95
0.9
0 1 2 3 4 5 6 7 8 9 10
t/Te
Fig. 15 (continued)
The 2nd harmonic experiments and CFD often show somewhat similar trends,
but with larger CFD scatter than shown for the zeroth and first harmonic response.
The trends for |E| versus μ for %D vs. %DR show that.%DR provides a better
assessment of the CFD capability since small D values bias the %D analysis. The
primary variables (AR, z1 /A, θ1 /Ak and φ1 /Ak) |E| averaged over all headings are
all less than about 12%DR, i.e., 7, 8, 6 and 12%DR, respectively. The overall average
1st/2nd order |E| and UD are 18/40 and 7/43%DR for amplitudes and the overall
1st/2nd order |E| are 14/24%DR for phases. The largest errors are bow quartering for
AR, beam waves for z1 /A, following waves for θ1 /Ak and bow and stern quartering
waves for φ1 /Ak.
6.1 Conclusions
Improved benchmark experimental data from multiple facilities and more simulations
from multiple institutes will provide estimates that are more robust for intervals of
validation and SoA assessment. CFD can predict the effects of variable heading albeit
with larger comparison errors than those for head waves. The agreement between
the two codes solutions and the experimental data is remarkable.
394 F. Stern et al.
2
(a)
1
ζ/A
-1
-2
0 1 2 3 4 5 6 7 8 9 10
t/Te
20
(b)
10
3
CT×10
-10
0 1 2 3 4 5 6 7 8 9 10
t/Te
2
(c)
1
z/A
-1
-2
0 1 2 3 4 5 6 7 8 9 10
t/Te
2
(d)
1
θ/Ak
-1
-2
0 1 2 3 4 5 6 7 8 9 10
t/Te
2
(e)
1
φ/Ak
-1
-2
0 1 2 3 4 5 6 7 8 9 10
t/Te
1.1
(f)
1.05
u/U
0.95
0.9
0 1 2 3 4 5 6 7 8 9 10
t/Te
Fig. 16 (continued)
7 CFD Assessment Test Cases 3.9, 3.12, and 3.13
As shown in Table 30, 6 institutes from 4 countries using 5 different CFD codes
submitted results for both calm water and waves. There were 8 submissions for calm
water self-propulsion and (6,3,1,3,1) for μ = (0, 45, 90, 135, 180), respectively.
Two equations, RSM and DDES turbulence models were used with 6 using wall
functions and 2 using near wall models. Mostly discretized but also body force
propeller modeling was used. It is not known which ones have used prescribed and
which ones have applied interactive body forces. The solvers were mostly FV except
one is FD. All but 1 used unstructured grids. 1 used single phase level set and 6 used
VoF free-surface models. Grid motion used dynamic patched, dynamic overset and
deforming grids. Most codes used reduced DOF and fixed rudder for test case 3.9
and in some cases for test cases 3.12 and 3.13. Grid sizes varied from 1–35 M with
average 9.12 M. Average number CPU and CPU time is 87 processors and 100,000 h,
respectively. Only IIHR also submitted to test cases 2.10 and 2.11.
7.1 Test Case 3.9
Table 31 provides the calm water self-propulsion V&V and SoA variables. Figure 25
graphically displays the sinkage, trim and propeller revolutions RPS validation. The
396 F. Stern et al.
2
(a)
1
ζ/A
-1
-2
0 1 2 3 4 5 6 7 8 9 10
t/Te
15
(b)
10
CT×103
-5
0 1 2 3 4 5 6 7 8 9 10
t/Te
2
(c)
1
z/A
-1
-2
0 1 2 3 4 5 6 7 8 9 10
t/Te
0.2
(d)
0.1
θ/Ak
-0.1
-0.2
0 1 2 3 4 5 6 7 8 9 10
t/Te
3
(e)
2
1
φ/Ak
0
-1
-2
0 1 2 3 4 5 6 7 8 9 10
t/Te
1.1
(f)
1.05
u/U
0.95
0.9
0 1 2 3 4 5 6 7 8 9 10
t/Te
Fig. 17 (continued)
average errors are |E| = 1.5%D for RPS and 6 and 15%D for sinkage and trim,
respectively. Most solutions under predict RPS with one outlier, whereas most solu-
tions overpredict sinkage and trim. UD is small for RPS such that only one solution
is validated, whereas UD is large for sinkage and trim such that all solutions are
validated. The scatter in the solutions for RPS is small < 1%D, whereas it is large
for sinkage and trim, i.e., 6 and 16%D, respectively. Since U So Ai is small for RPS
and large for sinkage and trim 4 and 8 codes/solutions are N-version validated for
RPS and sinkage/trim, respectively, and the mean code is not N-version validated
for RPS but it is for sinkage and trim at U So A = <1, 20 and 134%D, respectively.
For calm water RPS, the mean absolute error 2%D and individual code uncertainty
U So Ai = 2%D are acceptably small, which indicates free running self-propulsion
can be predicted with simialr error and uncertainty as captive towed resistance. The
sinkage and trim errors and uncertainties are mostly larger than those for captive
towed resistance.
7.2 Test Cases 3.12 and 3.13
Test cases 3.12 and 3.13 are for λ/L = 1 and μ = 0 and 45, 90, 135 and 180°., respec-
tively. Figure 26 shows head and bow quartering conditions are far from heave/pitch
natural frequency. Beam condition is close to roll natural frequency. Note that B/λ
398 F. Stern et al.
2
(a)
1
ζ/A
-1
-2
0 1 2 3 4 5
t/Te
30
(b)
20
CT×103
10
0
-10
-20
0 1 2 3 4 5
t/Te
1
(c)
0.5
z/A
-0.5
-1
0 1 2 3 4 5
t/Te
0.8
(d)
0.4
θ/Ak
-0.4
-0.8
0 1 2 3 4 5
t/Te
6
(e)
4
2
φ/Ak
0
-2
-4
0 1 2 3 4 5
t/Te
(f)
1.1
u/U
0.9
0 1 2 3 4 5
t/Te
Fig. 18 (continued)
= 0.12. Stern quartering and following wave conditions are far from roll natural
frequency. Parametric roll frequency is close to heave/pitch natural frequency and
far from all heading conditions.
The experimental and CFD time histories are shown in Fig. 27, 28, 29, 30 and 31
for μ = 0, 45, 90, 135 and 180°., respectively. The experiments and CFD cover 7, 6,
4, 6, 7 and 7, 7, 7, 6, 9 Te , respectively, for μ = 0, 45, 90, 135 and 180°. Different
trends are observed for head/following, quartering and beam waves. For μ = 0 and
180°. the errors are largest for φ, ψ, v and δ; and the experiments require much larger
rudder control than the CFD. For μ = 45 and 135, the errors are largest for φ, ψ,
u, v and δ; and the experiments and CFD have similar heading control. For μ = 90,
the errors are largest for φ, ψ, v and δ; the experiments and CFD seem to approach
the same steady state heading control, but with different initial transients. Thus, the
largest differences are for horizontal motions and roll angle.
Tables 32 and 33 provide average results for each heading and over all headings
of the CFD solutions for the 0th/AR and 1st and 2nd harmonic amplitude validation
and SoA variables using %D and %DR, respectively. Table 24 is similar but provides
the 1st and 2nd harmonic phase validation and SoA variables using %2π. The partic-
ipants submitted their own trajectories and time histories, which were harmonically
analyzed using the last two cycles for all headings since the mean heading and
encounter frequency were nearly constant. The discussions mostly use values for μ
averages and %DR.
400 F. Stern et al.
2
(a)
1
ζ/A
-1
-2
0 1 2 3 4 5 6 7 8
t/Te
30
(b)
20
CT×103
10
0
-10
-20
0 1 2 3 4 5 6 7 8
t/Te
0.2
(c)
0
z/A
-0.2
-0.4
-0.6
0 1 2 3 4 5 6 7 8
t/Te
0.8
(d)
0.4
θ/Ak
-0.4
-0.8
0 1 2 3 4 5 6 7 8
t/Te
(e)
0
φ/Ak
-0.5
-1
0 1 2 3 4 5 6 7 8
t/Te
1.1
(f)
1.05
u/U
0.95
0.9
0 1 2 3 4 5 6 7 8
t/Te
Fig. 19 (continued)
harmonic amplitudes and 1st and 2nd harmonic phases, respectively, versus μ. The
left column shows D, Si and S̄. The middle and right columns show UD ,|E i | and |E|
using %D and %DR, respectively.
The 0th harmonic experiments show z0 /A peaks at μ = 0 and 180, trough at 90,
and near zero for 45 and 135°. θ0 /Ak is nearly constant. φ0 /Ak, ψ0 , and v0 have
peaks at μ = 0, 90 and 180 and troughs at 45 and 135°. u0 shows large speed loss for
bow waves, which is attributed to large bow wave pitch motions. CFD shows large
scatter for μ = 0 and differences for other μ but fewer codes. The mean and code
with results for all μ shows similar trends as the experiments, except for z0 /A. The
errors for speed loss are relatively small, whereas the other variables show larger
error with peaks for different headings.
z1 /A is maximum for beam waves and larger for bow vs. stern waves. Since B/λ
= 0.12 for beam waves pitch is nearly zero and heave and wave amplitudes coincide,
as the model simply moves up and down parallel with the wave. θ1 /Ak is maximum
for bow, nearly zero for 90, and about half bow wave values for stern waves. φ1 /Ak
increases from head to stern quartering and reduces for following waves. ψ1/ Ak is
nearly zero for μ = 0, 90 and 180 and large for μ = 45 and especially 135°. u1 /U
is small for bow and beam and larger for bow quartering and very largest for stern
waves. v1 /U is small for bow and following and large for beam and bow and stern
quartering waves. All variables show large response for stern quartering waves. The
CFD trends for 1st harmonic amplitude and phase agree with the experiments but
Table 27 (%D). Evaluations of individual and mean code values on harmonic amplitudes of resistance and motions for KCS in regular wave (Case 2.11)
402
μ[deg] Variable 0th harmonic amplitude 1st harmonic amplitude 2nd harmonic amplitude Ave
CT0 × AR z0 /L θ0 /Ak ϕ0 /Ak ζ1 /L CT1 × z1 /L θ1 /Ak ϕ1 /Ak CT2 × z2 /L θ2 /Ak ϕ2 /Ak 1st 2nd
103 103 103
0 UD %|D| 1.63 6.02 15.38 43.54 – 1.37 2.47 3.52 6.14 – 16.26 70.70 213.17 – 3.37 52.39
# 2 2 2 2 – 1 2 2 2 – 2 2 2 – 1.40 1.56
submissions
|E|%|D| 3.14 9.44 25.10 133.87 – 0.36 19.36 15.44 7.43 – 266.14 35.92 18.11 – 10.65 70.25
# validated 0 0 1 0 – 1 0 0 1 – 0 2 2 – 0.40 0.56
45 UD %|D| 1.39 5.98 10.68 209.94 8.67 1.51 1.04 1.24 4.45 5.29 2.02 161.11 275.85 19.36 2.71 77.22
# 2 2 2 2 2 1 2 2 2 2 2 2 2 2 1.80 2.00
submissions
|E|%|D| 4.77 16.21 33.63 31.70 271.90 0.36 14.39 6.04 6.13 76.67 68.76 22.97 9.99 92.61 20.72 61.39
# validated 0 0 0 2 0 1 0 0 1 0 0 2 2 0 0.40 0.67
90 UD %|D| 2.31 26.29 9.80 122.44 8.40 1.44 5.73 1.81 60.00 8.29 7.17 40.93 146.33 45.11 15.46 45.42
# 2 2 2 2 2 1 2 2 2 2 2 2 2 2 1.80 2.00
submissions
|E|%|D| 2.80 27.81 47.81 114.68 109.48 0.36 69.85 12.79 14.23 40.14 32.73 51.25 79.63 93.30 27.47 62.16
# validated 1 1 0 1 0 1 0 0 2 0 1 0 2 0 0.60 0.67
135 UD %|D| 1.58 27.39 10.63 328.77 19.18 0.94 3.41 4.41 8.41 0.98 9.94 1374.87 113.73 32.89 3.63 213.22
# 2 2 2 2 2 1 2 2 2 2 2 2 2 2 1.80 2.00
submissions
|E|%|D| 0.63 5.51 39.78 394.51 143.91 0.36 214.08 6.08 11.30 11.33 62.16 1472.34 23.99 56.29 48.63 244.35
# validated 2 2 0 1 0 1 0 1 1 0 0 1 2 1 0.60 1.00
180 UD %|D| 1.59 43.44 6.75 1048.03 – 1.22 10.19 24.61 451.21 – 23.74 67.34 172.14 – 121.81 194.72
(continued)
F. Stern et al.
103 103 103
# 2 2 2 2 – 1 2 2 2 – 2 2 2 – 1.40 1.56
submissions
|E|%|D| 0.69 21.02 18.98 451.26 – 0.36 1014.07 137.43 48.40 – 39.12 57.25 66.89 – 300.06 93.60
# validated 2 2 0 2 – 1 0 0 2 – 0 1 2 – 0.60 1.00
Ave UD %|D| 1.70 21.82 10.65 350.55 12.08 1.30 4.57 7.12 106.04 4.85 11.82 342.99 184.24 32.45 29.39 116.59
# 2 2 2 2 1.2 1 2 2 2 1.2 2 2 2 1.2 1.64 1.82
submissions
|E|%|D| 2.41 16.00 33.06 225.21 175.09 0.36 266.35 35.56 17.50 42.71 93.78 327.95 39.72 80.73 81.51 106.35
# validated 1 1 0.2 1.2 0 1 0 0.2 1.4 0 0.2 1.2 2 0.2 0.52 0.78
403
Table 28 (%DR). Evaluations of individual and mean code values on harmonic amplitudes of resistance and motions for KCS in regular wave (Case 2.11)
404
103 103 103
0 UD %|DR| 5.80 6.71 23.08 28.52 – 1.90 1.18 1.94 4.74 – 2.23 42.18 151.51 – 2.44 37.15
# 2 2 2 2 – 1 2 2 2 – 2 2 2 – 1.40 1.56
submissions
|E|%|DR| 11.17 10.52 37.66 87.70 – 0.50 9.23 8.53 5.73 – 36.45 21.43 12.87 – 6.00 31.11
# validated 0 0 1 0 – 1 0 0 1 – 0 2 2 – 0.40 0.56
45 UD %|DR| 4.72 5.75 12.93 42.12 1.59 2.10 1.17 1.30 4.78 1.22 2.21 38.07 146.33 35.48 2.11 32.13
# 2 2 2 2 2 1 2 2 2 2 2 2 2 2 1.80 2.00
submissions
|E|%|DR| 16.24 15.59 40.73 6.36 49.82 0.50 16.05 6.37 6.59 17.74 75.49 5.43 5.30 169.71 9.45 42.74
# validated 0 0 0 2 0 1 0 0 1 0 0 2 2 0 0.40 0.67
90 UD %|DR| 6.26 6.96 13.60 42.23 6.86 2.00 1.47 1.86 4.48 1.30 2.22 42.19 156.43 37.55 2.22 34.92
# 2 2 2 2 2 1 2 2 2 2 2 2 2 2 1.80 2.00
submissions
|E|%|DR| 7.57 7.36 66.37 39.56 89.42 0.50 17.92 13.15 1.06 6.30 10.14 52.84 85.12 77.68 7.79 48.45
# validated 1 1 0 1 0 1 0 0 2 0 1 0 2 0 0.60 0.67
135 UD %|DR| 4.13 4.94 13.32 50.68 13.66 1.30 2.32 1.62 4.81 1.13 2.42 42.44 174.06 38.21 2.24 38.21
# 2 2 2 2 2 1 2 2 2 2 2 2 2 2 1.80 2.00
submissions
|E|%|DR| 1.64 0.99 49.83 60.81 102.48 0.50 145.58 2.23 6.46 13.11 15.16 45.45 36.72 65.39 33.58 42.05
# validated 2 2 0 1 0 1 0 1 1 0 0 1 2 1 0.60 1.00
180 UD %|DR| 4.07 4.96 14.92 240.55 – 1.70 1.18 1.33 97.44 – 2.32 47.81 179.41 – 25.41 70.58
(continued)
F. Stern et al.
103 103 103
# 2 2 2 2 – 1 2 2 2 – 2 2 2 – 1.40 1.56
submissions
|E|%|DR| 1.76 2.40 41.97 103.58 – 0.50 117.21 7.42 10.45 – 3.83 40.64 69.72 – 33.89 37.70
# validated 2 2 0 2 – 1 0 0 2 – 0 1 2 – 0.60 1.00
Ave UD %|DR| 5.00 5.86 15.57 80.82 7.37 1.80 1.46 1.61 23.25 1.22 2.28 42.54 161.55 37.08 6.88 42.60
# 2 2 2 2 1.2 1 2 2 2 1.2 2 2 2 1.2 1.64 1.82
submissions
|E|%|DR| 7.67 7.37 47.31 59.60 80.57 0.50 61.20 7.54 6.06 12.38 28.21 33.16 41.94 104.26 18.14 40.41
# validated 1 1 0.2 1.2 0 1 0 0.2 1.4 0 0.2 1.2 2 0.2 0.52 0.78
*Dynamic range was obtained from Mark Stocker’s experimental data for 1st Amp ζ1 /L* (ζ1 /LDR = 0.006)
405
406 F. Stern et al.
Table 29 (%2π ). Evaluations of individual and mean code values on harmonic phase of resistance
and motions for KCS in regular wave (Case 2.11)
μ[deg] Variable 1st harmonic phase 2nd harmonic phase Ave
CT1 × z1 /L θ1 /Ak ϕ1 /Ak CT2 × z2 /L θ2 /Ak ϕ2 /Ak 1st 2nd
103 103
0 # 2 2 2 – 2 2 2 – 1.5 1.5
submissions
|E|%2p 7.65 8.94 6.75 – 21.27 17.82 17.29 – 7.78 18.79
45 # 2 2 2 2 2 2 2 2 2 2
submissions
|E|%2p 4.54 4.54 4.43 4.08 30.40 25.82 9.43 47.88 4.40 28.38
90 # 2 2 2 2 2 2 2 2 2 2
submissions
|E|%2p 16.16 1.47 2.69 22.23 19.98 6.73 12.45 57.48 10.64 24.16
135 # 2 2 2 2 2 2 2 2 2 2
submissions
|E|%2p 16.52 15.28 14.48 25.25 30.79 9.56 32.02 30.60 17.89 25.74
180 # 2 2 2 – 2 2 2 – 1.5 1.5
submissions
|E|%2p 38.24 17.82 26.20 – 18.34 22.16 21.53 – 27.42 20.68
Ave # 2.00 2.00 2.00 1.20 2.00 2.00 2.00 1.20 1.80 1.80
submissions
|E|%2p 16.62 9.61 10.91 17.19 24.15 16.42 18.55 45.32 13.62 23.55
with large differences between the solutions. The errors for z1 /A and θ1 /Ak are small
compared to those for φ1 /Ak, ψ1/ Ak, u1 /U and v1 /U. Roll and heading have largest
errors for stern quartering waves, u1 /U has largest errors for bow and stern waves,
and v1 /U has largest errors for beam waves.
The 2nd harmonic experiments and CFD often show somewhat similar trends,
but with larger CFD scatter than shown for the zeroth and first harmonic response.
The trends for |E| versus μ for %D versus %DR show that.%DR provides a better
assessment of the CFD capability since small D values bias the %D analysis. For
head waves primary variables (u0 , z1 /A, and θ1 /Ak), the code level scatter P|Ei | is
about 12%DR and much larger than UD ≈ 5%DR for u0 and 1%DR for motions.
|E| = 11, 5 and 8%DR, respectively, for amplitudes and <3%DR for phases such
that only 1, 2 or 0 solutions are validated for amplitudes and 6 and 0 for heave and
pitch phases. U So Ai is about 12%DR such that about 5 codes/solutions are N-version
validated. USoA is about 5%DR and the mean code is only N-version validated for
heave.
For oblique waves primary variables (u0 , z1 /A, θ1 /Ak, φ1 /Ak and ψ1 /Ak), the UD
values are small <2%DR except u0 for which UD = 10%DR. |E| = 9, 2, 3, 20, and
6%DR, respectively, for amplitudes and 12%DR for phases such that only about 1or
0 solutions are validated. The largest errors are bow and stern quartering waves for
Fig. 20 Validations for 0th harmonic amplitudes of resistance and motions for KCS seakeeping in
regular waves at various heading with Fn = 0.26
408 F. Stern et al.
Fig. 21 Validations for 1st harmonic amplitudes of resistance and motions for KCS seakeeping in
regular waves at various heading with Fr = 0.26
Fig. 22 Validations for 1st harmonic phases of resistance and motions for KCS seakeeping in
410 F. Stern et al.
Fig. 23 Validations for 2nd harmonic amplitudes of resistance and motions for KCS seakeeping
in regular waves at various heading with Fr = 0.26
Fig. 24 Validations for 2nd harmonic phases of resistance and motions for KCS seakeeping in
Table 30 Submissions for ONRT course keeping in head and variable heading waves at Fr = 0.2
412
No. Institute Code Cases participated Turbulence Free Propeller Motions Rudders Discretization Grid motion Grid type Grid size No.CPU CPU
surface Method Time
(CPUh)
1 ABS OpenFOAM 3.9 3.12 3.13 Two-eq. VoF Discretized 6DOF Active FV Overset Unstructured 6.3–6.5 M 16 38301
(135) 36002
2 IIHR CFDShip-Iowa V4.5 3.9 3.12 3.13 Two-eq. LS Body Force 4DOF Fixed1 FD Overset Structured 23.2 M 224 215001
(45, (x, z, φ,
90, θ)1
135, Single 6DOF2 Active2 325002
180) Ph.
3 CSSRC FLUENT12 3.9 3.12 Two-eq. VoF Body Force 2DOF Fixed FV Dynamic Unstructured 1.15–2.7 M 16 NA
and (z, θ) patched
Discretized1
Body
Force2
4 ECN ISISCFD-AD 3.9 3.12 3.13 Two-eq. VoF Body Force 3DOF Fixed1 FV Deforming Unstructured 6.5 M 24–64 3201
CNRS (45) (x, z, Grid/Overset
θ)1
6DOF2 Active2 145002
5 IIHR REX 3.9 3.12 3.13 Two-eq. LS Discretized 4DOF Fixed1 FD Overset Structured 35 M 270 70000
(45, DDES (x, z, φ,
135) θ)1
Single 6DOF2 Active2
Ph.
6 HHI STAR-CCM+ 3.9 3.12 RSM VoF Discretized 6DOF Fixed FV Dynamic Unstructured 10.7 M 32 NA
patched
7 ECN ISISCFD-FullRANSE 3.9 Two-eq. VoF Discretized 3DOF Fixed FV Deforming Unstructured 9.3 M 72 20000
CNRS (x, z, θ) Grid/Overset (Half
Domain)
(continued)
F. Stern et al.
No. Institute Code Cases participated Turbulence Free Propeller Motions Rudders Discretization Grid motion Grid type Grid size No.CPU CPU
surface Method Time
(CPUh)
8 SJTU OpenFOAM 3.9 Two-eq. VoF Discretized 6DOF Active FV Overset Unstructured 6.46 M 20 4100
Number of submissions for ONRT course keeping of Tokyo 2015 Workshop
μ[deg] Heave Pitch Roll Yaw u2/U v2/U
0 6 6 5 5 6 5
45 3 3 3 3 3 3
90 1 1 1 1 1 1
135 3 3 3 3 3 3
180 1 1 1 1 1 1
1 Applied for Case 3.9
2 Applied for Case 3.12 & 3.13
Case 3.9: Calm water condition

Case 3.12: Head wave condition (0°)
Case 3.13: Oblique wave condition (45–180°)
413
414 F. Stern et al.
Table 31 Comparison errors and uncertainties for ONRT self-propulsion in calm water submissions
at Fr = 0.2
Variable σ/L τ N Motion abs.
average
D 0.072 −0.0386 8.970
U D %|D| 20 134.2 0.083 77.1
# submissions 8 8 8 8
# outliers 0 0 1 0
CFD Si E i % |D| Si E i % |D| Si E i % |D| |E i |%|D|
submissions i
1 0.074 −3.10 −0.0411 −6.48 8.800 1.90 4.79
2 0.070 2.21 −0.0387 −0.26 9.449 -5.34 1.24
3 0.084 −16.37 −0.0482 −24.87 8.970 0.00 20.62
4 0.078 −8.41 −0.0343 11.14 8.834 1.52 9.77
5 0.073 −1.77 −0.0521 −34.97 8.850 1.34 18.37
6 0.070 2.65 −0.0361 6.48 8.850 1.34 4.57
7 0.074 −3.36 −0.0437 −13.21 8.734 2.63 8.29
8 0.077 −6.64 −0.0464 −20.21 8.819 1.68 13.42
S 0.075 −0.0426 8.837
E%|D| −4.35 −10.30 1.49 7.33
σ E %|D| 6.18 15.97 0.79 11.08
|E|%|D| 5.56 14.70 1.49 10.13
σ|E| %|D| 4.94 11.36 0.79 8.15
P|Ei | %|D| 9.89 22.72 1.59 16.31
P|E| %|D| 3.50 8.03 0.56 5.77
U So Ai %|D| 22.31 136.11 1.59 79.21
U So A %|D|nn 20.30 134.44 0.57 77.37
# validated 8 8 1 8
# N-version 8 8 4 8
validated
Mean code Yes Yes No Yes
N-version
validated
Note The gray-shaded numbers are outliers
**Propeller RPS is adjusted to the EFD value
speed loss, head and stern quartering waves for heave and pitch and stern quartering
waves for roll and heading.
Fig. 25 EFD data and submission values for motions and propeller revolution for ONRT self-
propulsion in calm water at Fr = 0.2 (**Propeller RPS is adjusted to the EFD value)
416 F. Stern et al.
Fig. 26 Wave encounter

frequencies compared to
natural frequencies (fz/θ =
4.53; fϕ = 1.84; fPR = 3.68)
for ONR Tumblehome test
conditions in waves
7.3 Conclusions
ONRT tests cases are the most challenging of all T2015 test cases. Nonetheless,
6 institutes using 5 codes participated. The errors and scatter were larger than test
cases 2.10 and 2.11, however, clearly many institutes and codes displayed promising
capability for free running course keeping simulations.
8 Overall Conclusions
T2015 test cases 2.10 and 2.11 for KCS and 3.9/3.12/3.13 for ONRT provide assess-
ment of CFD capability for captive and free running model conditions in head and
oblique waves. The number of submissions were relatively large for test cases 2.10
and 3.9/3.12/3.13, i.e., 10 and 8, respectively, but small for test case 2.11, i.e., 2.
Nonetheless as a whole the test cases enable assessment of CFD for AR and course
keeping/speed loss including head and oblique waves and captive and free running
conditions, respectively.
Assessment approach includes not only evaluation of errors E using experi-
mental data D, but also including consideration of both deterministic and stochastic
numerical (USN and P|Ei | ) and experimental UD uncertainties, which provides
confidence levels/intervals for validation UV = U S2N + U D2 and SoA (U So Ai )
assessment. However, more attention is needed by individual facilities/codes on
experimental/numerical uncertainty analysis to provide more robust confidence
levels/intervals; and more attention is needed to quantify the CFD input wave condi-
tion and uncertainty and specify the CFD conditions and evaluation locations. KCS
calm water and head waves experimental data was available from several facilities
such that resistance, sinkage and trim data provided relatively robust validation/SoA
assessment. Nonetheless, more attention is needed for individual facility uncertainty
Fig. 27 Submissions with EFD for ONRT course keeping in waves with μ = 0°: a Trajectory;
b Wave; c Heave; d Roll; e Pitch; f Yaw; g x-vel.; h y-vel.; i Rudder angle; j Thrust coeff. (portside);
k Thrust coeff. (starboard); l Torque coeff. (portside); m Torque coeff. (starboard) * j ~ m don’t
include EFD
analysis. KCS oblique waves and ONRT head and oblique wave data and uncer-
tainties need more data from more facilities including uncertainties for more robust
validation/SoA assessment.
Tables 35, 36 and 37 provide an overall summary of the CFD SoA assessment for
calm water 2.10 and 3.9, head waves 2.10 and 3.12 and oblique waves 2.11 and 3.13
test cases, respectively. (Table 34)
418 F. Stern et al.
include EFD
Note that the Table 35 UD %D values include the scatter in the experimental data
due to scale effects/facility biases, i.e., PX i = 2σ X i , which is comparable to UXi for
calm water resistance and used for UD for sinkage and trim since UXi is not available,
as per Table 3. For Table 36, UD only includes UXi and USCi as per Tables 4, 5, 6,
7 and 8, but not PX i since σ X i was not available at the time of the SoA assessment.
Recently, as included in Chapter 4, Tables 4, 5, 6, 7 and 8 PX i = 28, 6 and 24%DR for
AR, heave and pitch, respectively, based on experiments at FORCE (test case 2.10),
include EFD
IIHR (test case 2.11 head waves) and OU (head waves). Including these values would
increase UD considerably, i.e., = 29, 7 and 24%DR, respectively, such that the scatter
in the experimental data is comparable to the scatter in the CFD for AR and heave
and dominant for pitch.
For KCS calm water resistance (10 codes/solutions), the solution scatter P|Ei |
is larger than the experimental uncertainty UD for resistance but not for sinkage
and trim. 3, 5 and 9 solutions are validated at UD = 2, 7 and 14%D; 8, 7 and 9
420 F. Stern et al.
include EFD
codes/solutions are N-version validated at U So Ai = 3, 8 and 16%D; and |E| = 2, 6

and 5%D, respectively, for resistance, sinkage and trim. The mean code is N-version
validated for sinkage and trim, but not for resistance. Priorities for improvements are:
for resistance, reduction in code/solution scatter; and for sinkage and trim, reduction
in code/solution scatter and errors and experimental uncertainties.
Interestingly, scatter in the CFD codes/simulations for resistance is similar as
the most recent AIAA CFD Drag Prediction Workshop (Vassberg, 2016) wherein
include EFD
the standard deviation is 2.1% of the mean value based on 44 CFD submissions,
whereas the comparable value for ship hydrodynamics is 2.8% based on ten CFD
submissions. Note that the AIAA test case is relatively less complex than that for
ship hydrodynamics, i.e., low-Mach number compressible flow, fixed slender body.
For ONRT calm water free running self-propulsion (8 codes/solutions), the CFD
SoA for propeller RPS is comparable to that for resistance. For sinkage and trim both
UD and P|Ei | are much larger than for captive tests especially for trim (partially due
Table 32 (%D). Evaluations of individual and mean code values on harmonic amplitudes of wave and motions for ONRT course keeping
422
z0 /L θ0 /Ak ϕ0 /Ak 0 /Ak u0 /U v0 /U z1 /L θ1 /Ak ϕ1 /Ak 1 /Ak u1 /U v1 /U z2 /L θ2 /Ak ϕ2 /Ak 2 /Ak u2 /U v2 /U 1st 2nd
0 UD %|D| 27.05 30.13 68.74 82.36 0.99 0.00 2.15 0.61 1421.00 7.07 15.50 0.00 66.22 5.64 296.71 45.68 13.83 0.00 241.05 53.11
# 6 6 5 5 6 5 6 6 5 5 6 5 6 6 5 5 6 5 5.50 5.50
submissions
# outliers 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0.17 0.08
P|Ei| %|D| 89.19 289.02 45.78 64.74 2.76 35.74 19.06 8.89 709.83 36.39 52.61 60.70 11.72 63.67 60.78 54.99 8.06 78.12 147.91 67.05
P|E| %|D| 36.41 117.99 20.47 28.95 1.13 15.98 7.78 3.63 317.45 16.28 23.53 27.14 4.78 25.99 27.18 24.59 3.61 34.94 65.97 28.50
USoAi %|D| 93.20 290.58 82.59 104.76 2.93 35.74 19.18 8.91 1588.43 37.07 54.84 60.70 67.25 63.92 302.87 71.48 16.01 78.12 294.85 100.79
USoA %|D| 45.36 121.78 71.73 87.30 1.50 15.98 8.07 3.68 1456.03 17.74 28.17 27.14 66.40 26.60 297.95 51.88 14.30 34.94 256.81 69.64
|E|%|D| 41.83 103.78 106.02 85.41 2.37 113.90 7.42 7.77 294.32 70.56 271.53 75.67 11.07 28.28 80.26 68.03 7.62 68.28 121.21 59.74
# validated 3 3 0 2 1 0 2 0 5 0 0 0 6 3 5 1 5 0 1.17 2.42
#N 5 5 1 4 4 0 5 4 5 0 0 1 6 5 5 2 5 2 2.50 3.67
validated
Mean N Yes yes No yes No No Yes No Yes No No No Yes No Yes No Yes No – –
validated
45 UD %|D| 108.30 15.39 10.94 4.28 2.89 44.08 1.71 1.07 3.02 0.43 109.13 0.00 102.93 25.77 67.66 3.63 0.00 0.00 19.23 32.16
# 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3.00 3.00
submissions
|E|%|D| 231.37 50.89 72.95 28.06 2.80 71.80 1.66 3.24 29.56 4.18 102.69 1392.70 16.29 73.50 11.60 23.62 60.47 1368.20 255.67 167.63
# validated 0 0 2 0 2 2 2 0 0 0 2 0 3 0 3 0 0 0 0.67 1.00
90 UD %|D| 27.55 30.15 18.71 6.40 2.28 5023.56 1.30 10.45 1.56 2.82 900.70 2.17 19.19 10.87 28.59 28.17 2292.00 0.00 153.17 623.96
# 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1.00 1.00
submissions
|E|%|D| 192.47 56.34 73.16 62.42 1.04 764.45 0.43 49.71 40.40 18.94 29.06 99.79 97.99 23.46 69.76 65.70 67.67 55.31 39.72 127.48
# validated 0 0 0 0 1 1 1 0 0 0 1 0 0 0 0 0 1 0 0.33 0.25
135 UD %|D| 159.27 12.74 11.99 17.35 1.39 140.40 2.67 0.90 1.01 0.75 27.52 27.78 14.64 11.88 27.34 31.42 160.67 0.00 10.10 49.09
(continued)
F. Stern et al.
# 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3.00 3.00
submissions
|E|%|D| 514.22 8.32 101.81 155.28 1.33 44.73 4.86 6.24 65.99 26.05 24.38 29.64 49.81 43.55 65.46 68.16 44.61 194.01 26.19 107.61
# validated 1 2 0 0 2 3 2 0 0 0 2 2 0 0 1 0 3 0 1.00 1.00
180 UD %|D| 16.11 200.57 494.08 14.94 1.67 510.35 8.44 1.64 82.75 22.43 35.38 0.00 69.70 7.81 103.77 44.24 2519.75 0.00 25.11 331.92
# 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1.00 1.00
submissions
|E|%|D| 23.41 161.85 365.28 123.71 0.86 127.09 4.75 0.91 84.30 89.61 13.09 87.41 3.81 80.28 89.43 91.37 38.96 97.93 46.68 100.33
# validated 0 1 1 0 1 1 1 1 0 0 1 0 1 0 1 0 1 0 0.50 0.58
Ave UD %|D| 67.66 57.80 120.89 25.07 1.84 1143.68 3.25 2.94 301.87 6.70 217.64 5.99 54.54 12.39 104.81 30.63 997.25 0.00 89.73 218.05
# 2.80 2.80 2.60 2.60 2.80 2.60 2.80 2.80 2.60 2.60 2.80 2.60 2.80 2.80 2.60 2.60 2.80 2.60 2.70 2.70
submissions
|E|%|D| 200.66 76.23 143.84 90.98 1.68 224.40 3.82 13.57 102.92 41.87 88.15 337.04 35.79 49.81 63.30 63.38 43.86 356.75 97.89 112.56
# validated 0.80 1.20 0.60 0.40 1.40 1.40 1.60 0.20 1.00 0.00 1.20 0.40 2.00 0.60 2.00 0.20 2.00 0.00 0.73 1.05
423
Table 33 (%DR). Evaluations of individual and mean code values on harmonic amplitudes of wave and motions for ONRT course keeping
424
0 UD %|DR| 11.46 31.89 4.72 10.46 4.52 0.00 1.44 0.63 0.64 0.04 1.05 0.00 16.77 2.42 23.83 1.80 13.32 0.00 0.63 10.10
# 6 6 5 5 6 5 6 6 5 5 6 5 6 6 5 5 6 5 5.50 5.50
submissions
# outliers 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0.17 0.08
P|Ei| %|DR| 37.79 305.81 3.14 8.23 12.56 3.83 12.79 9.12 0.32 0.20 3.58 0.37 2.97 27.28 4.88 2.16 7.76 3.43 4.40 34.99
P|E| %|DR| 15.43 124.85 1.40 3.68 5.13 1.71 5.22 3.72 0.14 0.09 1.60 0.17 1.21 11.14 2.18 0.97 3.47 1.53 1.82 14.39
USoAi %|DR| 39.49 307.47 5.67 13.31 13.35 3.83 12.87 9.14 0.72 0.20 3.73 0.37 17.03 27.39 24.33 2.81 15.42 3.43 4.50 39.46
USoA %|DR| 19.22 128.85 4.92 11.09 6.83 1.71 5.42 3.78 0.66 0.10 1.92 0.17 16.82 11.40 23.93 2.04 13.77 1.53 2.00 20.18
|E|%|DR| 17.72 109.81 7.27 10.85 10.80 12.22 4.98 7.97 0.13 0.39 18.46 0.46 2.80 12.12 6.45 2.68 7.34 3.00 5.40 16.92
# validated 3 3 0 2 1 0 2 0 5 0 0 0 6 3 5 1 5 0 1.17 2.42
#N 5 5 1 4 4 0 5 4 5 0 0 1 6 5 5 2 5 2 2.50 3.67
validated
Mean N Yes Yes No yes No No Yes No Yes No No No Yes No Yes No Yes No – –
Validated
45 UD %|DR| 10.36 18.85 5.45 4.83 14.64 35.90 1.58 1.08 0.65 0.12 27.65 0.00 16.88 2.66 24.26 1.39 0.00 0.00 5.18 11.27
# 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3.00 3.00
submissions
|E|%|DR| 22.14 62.32 36.36 31.63 14.19 58.47 1.53 3.29 6.39 1.16 26.01 51.71 2.67 7.60 4.16 9.02 72.30 163.16 15.01 40.33
# validated 0 0 2 0 2 2 2 0 0 0 2 0 3 0 3 0 0 0 0.67 1.00
90 UD %|DR| 10.08 27.51 6.58 1.45 12.43 85.15 1.53 0.27 1.03 0.16 36.88 1.58 17.02 3.61 24.93 5.31 448.64 0.00 6.91 53.56
# 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1.00 1.00
submissions
(continued)
F. Stern et al.
|E|%|DR| 70.41 51.41 25.73 14.17 5.67 12.96 0.50 1.30 26.66 1.09 1.19 72.76 86.88 7.80 60.83 12.39 13.25 38.58 17.25 33.34
# validated 0 0 0 0 1 1 1 0 0 0 1 0 0 0 0 0 1 0 0.33 0.25
135 UD %|DR| 9.73 11.97 7.78 6.95 7.58 59.81 1.55 0.59 1.01 0.76 28.64 27.95 17.04 13.11 29.53 32.66 159.29 0.00 10.08 29.62
# 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3.00 3.00
submissions
|E|%|DR| 31.43 7.82 66.01 62.21 7.24 19.05 2.82 4.05 66.02 26.20 25.38 29.82 57.98 48.05 70.72 70.84 44.23 202.54 25.71 57.34
# validated 1 2 0 0 2 3 2 0 0 0 2 2 0 0 1 0 3 0 1.00 1.00
180 UD %|DR| 10.22 45.06 5.15 2.22 9.27 94.74 1.46 0.61 0.80 0.43 31.29 0.00 17.15 2.41 29.67 5.46 494.14 0.00 5.76 59.62
# 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1.00 1.00
submissions
|E|%|DR| 14.85 36.36 3.81 18.35 4.76 23.59 0.82 0.34 0.82 1.70 11.58 0.95 0.94 24.78 25.57 11.28 7.64 36.67 2.70 17.38
# validated 0 1 1 0 1 1 1 1 0 0 1 0 1 0 1 0 1 0 0.50 0.58
Ave UD %|DR| 10.37 27.06 5.93 5.18 9.69 55.12 1.51 0.64 0.83 0.30 25.10 5.91 16.97 4.84 26.44 9.32 223.08 0.00 5.71 32.83
# 2.80 2.80 2.60 2.60 2.80 2.60 2.80 2.80 2.60 2.60 2.80 2.60 2.80 2.80 2.60 2.60 2.80 2.60 2.70 2.70
submissions
|E|%|DR| 31.31 53.54 27.83 27.44 8.53 25.26 2.13 3.39 20.00 6.11 16.53 31.14 30.25 20.07 33.54 21.24 28.95 88.79 13.22 33.06
# validated 0.80 1.20 0.60 0.40 1.40 1.40 1.60 0.20 1.00 0.00 1.20 0.40 2.00 0.60 2.00 0.20 2.00 0.00 0.73 1.05
425
Table 34 (%2π. Evaluations of individual and mean code values on harmonic phase of wave and motions for ONRT course keeping
426

z1 /L θ1 /Ak ϕ1 /Ak 1 /Ak u1 /U v1 /U z2 /L θ2 /Ak ϕ2 /Ak 2 /Ak u2 /U v2 /U 1st 2nd
0 UD %2p 4.09 0.09 0.63 0.06 4.08 4.82 8.80 0.14 0.22 0.61 6.35 21.94 2.30 6.34
# submissions 6 6 5 5 6 5 6 6 5 5 6 5 5.50 5.50
# outliers 0 0 0 0 0 0 0 0 0 0 0 0 0.00 0.00
P|Ei| %2p 2.22 5.51 10.92 42.29 13.91 15.05 19.18 28.51 44.72 36.51 14.90 24.61 14.98 28.07
P|E| %2p 0.91 2.25 4.88 18.91 5.68 6.73 7.83 11.64 20.00 16.33 6.08 11.01 6.56 12.15
USoAi %2p 4.65 5.51 10.94 42.29 14.50 15.81 21.10 28.51 44.72 36.52 16.20 32.97 15.62 30.00
USoA %2p 4.19 2.25 4.92 18.91 6.99 8.28 11.78 11.64 20.00 16.34 8.79 24.55 7.59 15.52
|E|%2p 1.31 2.47 30.48 26.21 6.26 20.22 17.21 11.99 37.67 43.35 14.76 23.70 14.49 24.78
# validated 6 0 0 0 3 0 2 0 0 0 0 3 1.50 0.83
# N validated 6 5 0 3 5 1 4 5 3 2 4 4 3.33 3.67
Mean N Validated Yes No No No Yes No No No No No No Yes – –
45 UD %2p 4.32 0.08 0.19 0.09 2.74 3.19 6.74 1.02 1.01 0.15 5.69 54.48 1.77 11.51
# submissions 3 3 3 3 3 3 3 3 3 3 3 3 3.00 3.00
|E|%2p 8.47 8.57 17.70 13.32 11.41 8.71 13.88 22.66 26.06 26.41 7.59 65.47 11.36 27.01
# validated 0 0 0 0 0 2 0 0 0 0 2 1 0.33 0.50
90 UD %2p 0.62 0.06 0.18 0.04 7.05 0.60 0.77 0.04 0.18 0.14 37.60 4.79 1.42 7.25
# submissions 1 1 1 1 1 1 1 1 1 1 1 1 1.00 1.00
|E|%2p 9.57 9.97 4.12 12.05 27.62 11.76 1.93 4.28 23.34 40.69 55.63 38.25 12.52 27.35
# validated 0 0 0 0 0 0 0 0 0 0 0 0 0.00 0.00
135 UD %2p 3.95 1.10 0.20 0.04 8.26 4.01 11.13 0.18 0.24 0.22 8.43 9.96 2.92 5.03
(continued)
F. Stern et al.
z1 /L θ1 /Ak ϕ1 /Ak 1 /Ak u1 /U v1 /U z2 /L θ2 /Ak ϕ2 /Ak 2 /Ak u2 /U v2 /U 1st 2nd
# submissions 3 3 3 3 3 3 3 3 3 3 3 3 3.00 3.00
|E|%2p 11.11 10.01 15.76 8.92 10.70 8.45 23.38 19.98 14.55 37.14 17.02 19.90 10.83 22.00
# validated 0 0 0 0 2 1 1 0 0 0 0 1 0.50 0.33
180 UD %2p 2.51 0.04 0.19 0.42 2.97 3.03 6.60 0.94 0.19 0.19 5.82 8.56 1.53 3.72
# submissions 1 1 1 1 1 1 1 1 1 1 1 1 1.00 1.00
|E|%2p 1.31 0.98 36.94 3.91 3.28 2.14 8.37 11.65 19.22 37.96 63.38 26.80 8.09 27.90
# validated 1 0 0 0 0 1 0 0 0 0 0 0 0.33 0.00
Ave UD %2p 3.10 0.27 0.28 0.13 5.02 3.13 6.81 0.46 0.37 0.26 12.78 19.95 1.99 6.77
# submissions 2.8 2.8 2.6 2.6 2.8 2.6 2.8 2.8 2.6 2.6 2.8 2.6 2.70 2.70
|E|%2p 6.35 6.40 21.00 12.88 11.86 10.26 12.95 14.11 24.17 37.11 31.68 34.82 11.46 25.81
# validated 1.4 0 0 0 1 0.8 0.6 0 0 0 0.4 1 0.53 0.33
427
428 F. Stern et al.
Fig. 32 Validations for 0th harmonic amplitudes of motions for ONRT seakeeping in regular waves
at Fn = 0.2
Fig. 33 Validations for 1st harmonic amplitudes of wave and motions for ONRT seakeeping in
regular waves at Fr = 0.2
430 F. Stern et al.
Fig. 34 Validations for 1st harmonic phases of motions for ONRT seakeeping in regular waves at
Fr = 0.2
Fig. 35 Validations for 2nd harmonic amplitudes of motions for ONRT seakeeping in regular waves
at Fr = 0.2
432 F. Stern et al.
Fig. 36 Validations for 2nd harmonic phases of motions for ONRT seakeeping in regular waves at
Fr = 0.2
Table 35 Summary of calm water cases

Variable KCS 2.10 calm water ONRT 3.9 calm water
CT × 103 σ/L ς N σ/L ς
UD %|D| 1.75 7.38 14.25 0.083 20 134.2
P|Ei| %|D| 2.90 3.99 6.50 1.59 9.89 22.72
USoAi %|D| 3.39 8.39 15.66 1.59 22.31 136.11
USoA %|D| 1.98 7.50 14.41 0.57 20.3 134.44
|E|%|D| 2.33 6.16 5.16 1.49 5.56 14.7
to the mount, as designed primarily for ship motions testing). For RPS, sinkage and
trim, respectively: 1, 8 and 8 solutions are validated at UD = < 1, 20 and 134%D; 4,
8 and 8 codes/solutions are N-version validated at U So Ai = 2, 22 and 136%D; and
|E| = 2, 6 and 15%D. The mean code is N-version validated for sinkage and trim,
but not for RPS.
For KCS captive head waves averaged over λ/L (10 codes/solutions), the 1st (first
harmonic) and 2nd (AR, 0th and 2nd harmonic) order variables solution scatter P|Ei |
is larger than the experimental uncertainty UD . For AR, heave and pitch, respectively:
4, 5 and 4 solutions are validated at UD = 7, 4 and 3%DR; 7, 7 and 6 codes/solutions
are N-version validated at U So Ai = 26, 9 and 6%DR; and |E| = 13, 5 and 5%DR.
The mean code is not N-version validated. The solution scatter uncertainty is larger
than UD especially for AR. The errors for AR are six times larger than for calm
water resistance and for heave and pitch are comparable to those for sinkage and
trim. The largest errors are near resonance for AR, near resonance for heave, and just
before/after resonance for pitch.
The current CFD results from T2015 shows that CFD can predict head waves
heave, pitch and AR with better accuracy than potential flow for short, medium and
long waves. CFD results also show smaller scatter than potential flow for heave and
pitch. G2010 CFD errors/scatter are comparable to T2015, but T2015 errors/scatter
are considered most reliable since there are more submissions in T2015 with more
rigorous analysis.
Use of finer grids and improved user best practices will likely reduce CFD errors
and scatter. In addition, improved benchmark experimental data from multiple facili-
ties and solution verification will provide estimates that are more robust for intervals
of validation and certification.
For ONRT free running head waves with λ/L = 1 (6 codes/solutions), P|Ei | is
much larger than UD . For u0 , heave and pitch, respectively: 1, 2 and 0 solutions are
validated at UD = 5, 1 and 1%DR; 4, 5 and 4 codes/solutions are N-version validated
at U So Ai = 13, 13 and 9%DR; and |E| = 11, 5 and 8%DR.
For KCS captive oblique waves with λ/L = 1 (2 codes/solutions), UD = 7 and
43%DR and |E| = 18 and 40%DR, respectively, for 1st and 2nd order terms. The
largest errors are bow quartering for AR, beam waves for Z1 /A, following waves for
θ1 /Ak and bow and stern quartering waves for φ1 /Ak.
434
Table 36 Summary of head wave cases

Variable KCS 2.10 head waves λ/L average ONRT 3.12 head waves λ/L = 1
Amplitudes Phases Amplitudes Phases
AR z1 /L θ1 /Ak 1st Ave 2nd Ave 1st Ave 2nd Ave u0 /U z1 /L θ1 /Ak 1st Ave 2nd Ave 1st Ave 2nd Ave
UD %|DR|/%2p 6.92 4.35 2.58 2.26 14.35 1.75 22.81 4.52 1.44 0.63 0.63 10.10 2.30 6.34
P|Ei| %|DR|/%2p 24.80 7.44 5.52 5.26 16.06 4.44 14.73 12.56 12.79 9.12 4.40 34.99 14.98 28.07
USoAi %|DR|/%2p 26.25 8.71 6.27 5.93 23.52 5.00 31.99 13.35 12.87 9.14 4.50 39.46 15.62 30.00
USoA %|DR|/%2p 11.25 5.05 3.32 3.09 16.17 2.67 26.34 6.83 5.42 3.78 2.00 20.18 7.59 15.52
|E|%|DR|/%2p 12.67 5.20 4.53 4.02 11.49 4.91 16.70 10.80 4.98 7.97 5.40 16.92 14.49 24.78
F. Stern et al.
Table 37 Summary of oblique wave cases
Variable KCS 2.11 oblique waves μ average ONRT 3.13 oblique waves μ average
Amplitudes Phases Amplitudes Phases
AR z1 /L θ1 /Ak ϕ1 /Ak 1st 2nd 1st 2nd u0 /U z1 /L θ1 /Ak ϕ1 /Ak 1 /Ak 1st 2nd 1st 2nd
Ave Ave Ave Ave Ave Ave Ave Ave
UD %|DR|/%2p 5.86 1.61 23.25 1.22 6.88 42.60 – – 9.69 1.51 0.64 0.83 0.30 5.71 32.83 1.99 6.77
|E|%|DR|/%2p 7.37 7.54 6.06 12.38 18.14 40.41 13.62 23.55 8.53 2.13 3.39 20.00 6.11 13.22 33.06 11.46 25.81
435
436 F. Stern et al.
For ONRT free running oblique waves (3 or 1 code/solution depending heading),

UD = 6 and 32%DR and |E| = 13 and 33%DR, respectively, for 1st and 2nd order
terms. The largest errors are bow and stern quartering waves for speed loss, head and
stern quartering waves for heave and pitch and stern quartering waves for roll and
heading.
Figure 37 compares the experimental and mean code 0th and 1st harmonic ampli-
tudes for the three test cases (λ/L = 1.0 for ONRT and KCS L = 2.7 m model and
λ/L = 1.15 for KCS 6.1 m model). The 0th harmonic values are small and trends
not discernable. However, 1st harmonic values show clear and consistent trends,
which correlate with wave heading. Heave/pitch values are large for bow waves,
maximum/minimum for beam waves and small for stern waves. B/λ is small for
both models (0.14 and 0.12 for KCS and ONRT, respectively) and they simply oscil-
late with the wave with minimum relative motion for beam waves. Roll is small for
bow waves, increases for beam waves, maximum for stern quartering and reduced
for following waves. Larger heave/pitch for bow quartering waves and roll for stern
quartering waves is due to resonance conditions.
Figure 38 is like 37 but only includes experimental data; since, this condition was
not used as a test case at T2015. Heave and pitch values for bow waves and roll for
beam and stern quartering waves are smaller since the encounter frequency is not
close to the resonance condition at those wave headings.
These trends seem reasonable for course keeping and likely typical for displace-
ment ships with similar Fr, λ/L and H/λ. The agreement between the experiments
and CFD is outstanding and surely provides motivation for increased use of CFD for
industrial ship hydrodynamics applications.
As discussed, Stern et al. (2017), programmatic requirements/tolerances Ureq can
be considered similarly as for solution validation, but with UV replaced by U So Ai .
Since U So Ai is ≥ UV it will always be a more conservative assessment. There are six
possible combinations of |E i |, U So Ai and Ureq assuming none are equal:
(1) |E i | < U So Ai < Ur eq

(2) |E i | < Ur eq < U So Ai
(3) Ur eq < |E i | < U So Ai
(41)
(4) U So Ai < |E i | < Ur eq
(5) U So Ai < Ur eq < |E i |
(6) Ur eq < U So Ai < |E i |
In cases 1, 2 and 3 N-version validation is achieved at the U So Ai interval, i.e., the

comparison error is below the noise level. From an uncertainty perspective, modeling
errors cannot be isolated. In cases 4, 5 and 6, the comparison error is larger than the
noise level, i.e.,U So Ai < |E i | such that from an uncertainty perspective the sign and
magnitude of E can be used to estimate the modeling errors δSM . If U So Ai |E i |, E
= δSM . Only cases 1 and 4 meet the programmatic requirements.
Assuming Ureq = 5%D for calm water resistance and self-propulsion, resistance
and RPS are condition 6, whereas sinkage and trim are condition 3. Assuming Ureq
Fig. 37 0th harmonic (left) and1st harmonic amplitude (right) of motions for KCS L = 6.1 m for
λ/L = 1.15 and KCS L = 2.7 m and ONRT for λ/L = 1.0
438 F. Stern et al.
Fig. 38 0th harmonic (left) and1st harmonic amplitude (right) of motions for KCS L = 6.1 m,
KCS L = 2.7 m, and ONRT for λ/L = 0.5
= 5%DR for head waves AR and speed loss, 1st order terms are conditions 6 and 5,
respectively, and 2nd order terms are condition 3; and for oblique waves all errors
and uncertainties are greater than Ureq . Thus, none of the T2015 test cases were
able to achieve a programmatic requirement of 5%D and 5%DR for calm water and
waves, respectively. Note that not all the experimental uncertainties are able to meet
this requirement either such that both experiments and CFD need to reduce their
uncertainties and in addition CFD its errors. Nonetheless, in view the comparable
CFD capability for ONRT free running vs. KCS captive conditions, the prognosis
for CFD capability is excellent.
Future CFD assessment should extend the current test cases for consideration of
added power in regular and irregular waves. Verification should be required, and
more limited/most important validation variables should be confirmed and used for
the V&V and SoA assessment.
Acknowledgments The Office of Naval Research supports the IIHR research under grants
administered by Drs. Thomas Fu, Woei-Min Lin, Ki-Han Kim, and Robert Brizzolara.
References
ASME. (2009). Standard for verification and validation in computational fluid dynamics and heat
transfer. American Society of Mechanical Engineers (New York).
Bunnik, T., van Daalen, E., Kapsenberg, G., Shin, Y., Huijsmans, R., Deng, G., Delhommeau, G.,
Kashiwagi, M., & Beck, B. (2010). A comparative study on state-of-the-art prediction tools for
Seakeeping. In: 28th Symposium on Naval Hydrodynamics. Pasadena, California, pp. 12–17.
Eça, L., & Hoekstra, M. (2014). A procedure for the estimation of the numerical uncertainty of CFD
calculations based on grid refinement studies. Journal of Computational Physics, 262, 104–130.
Hemsch, M. (2004). Statistical analysis of computational fluid dynamics solutions from the drag
prediction workshop. Journal of Aircraft, 41(1), 95–103.
Larsson, L., Stern, F., & Visonneau, M. (Eds.). (2014). Numerical ship hydrodynamics: An
assessment of the gothenburg 2010 workshop. Springer.
Sadat-Hosseini, H., Toxopeus, S., Kim, D., Castiglione, T., Sanada, Y., Stocker, M., Simonsen, C.
D., Otzen, J. F., Toda, Y., & Stern, F. (2015). Experiments and computations for KCS added
resistance for variable heading, 5th world maritime technology conference november 3–7, Rhode
Island Convention & Omni Hotel Providence, Rhode Island, USA.
Simonsen, C. D., Otzen, J. F., Joncquez, S., & Stern, F. (2013). EFD and CFD for KCS heaving and
pitching in regular head waves. Journal of Marine Science and Technology, 18, 435–459.
Stern, F., Wilson, R. V., Coleman, H. W., & Paterson, E. G. (2001). Comprehensive approach to
verification and validation of CFD simulations—Part 1: Methodology and procedures. Journal
of Fluids Engineering—Transactions of the ASME, 123(4), 793–802.
Stern, F., Olivieri, A., Shao, J., Longo, J., & Ratcliffe, T. (2005). Statistical approach for estimating
intervals of certification or biases of facilities or measurement systems including uncertainties.
Journal of Fluids Engineering—Transactions of the ASME, 127(3), 604–610.
Stern, F., Wilson, R., & Shao, J. (2006). Quantitative approach to V&V of CFD simulations and
certification of CFD codes with examples. International Journal for Numerical Methods in Fluids,
50(11), 1335–1355.
Stern, F. (2007) Quantitative V&V of CFD solutions and certification of CFD codes with examples
for ship hydrodynamics, symposium on computational uncertainty, AVT-147. Athens, Greece.
Stern, F., Diez, M., & Sadat-Hosseini, H. (2016). Improved statistical approach for certification
of CFD codes with examples, verification and validation symposium (pp. 16–20). Las Vegas,
Nevada: The Westin Las Vegas Hotel.
Stern, F., Diez, M., Sadat-Hosseini, H., Yoon, H., & Quadvlieg, F. (2017). Statistical approach
for CFD State-of-the-Art assessment: N-Version verification and validation. ASME Journal of
Verification, Validation and Uncertainty Quantification, 2(3).
Vassberg, J. (2016). Challenges and accomplishments of the AIAA CFD drag prediction workshop
series. In: AVT-246 specialists’ meeting on progress and challenges in validation testing for
computational fluid dynamics. Avila, Spain.
Xing, T., & Stern, F. (2010). Factors of safety for Richardson extrapolation. Journal of Fluids
Engineering-Transactions of the ASME, 132(6), 061403.

Numerical Ship Hydrodynamics - An Assessment of The

Uploaded by

Copyright:

Available Formats

Numerical Ship Hydrodynamics - An Assessment of The

Uploaded by

Document Information

Original Title

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

Numerical Ship Hydrodynamics - An Assessment of The

Uploaded by

Copyright:

Available Formats

Lecture Notes in Applied and Computational Mechanics 94

Takanori Hino · Frederick Stern ·

More information about this series at http://www.springer.com/series/4623

Lars Larsson Michel Visonneau

Nobuyuki Hirata Jin Kim

Lars Larsson Michel Visonneau

Nobuyuki Hirata Jin Kim

ISSN 1613-7736 ISSN 1860-0816 (electronic)

Yokohama, Japan Takanori Hino

Introduction, Conclusions and Recommendations . . . . . . . . . . . . . . . . . 1

Evaluation of Resistance, Sinkage, Trim and Self-propulsion

Moustafa Abdel-Maksoud Hamburg University of Technology, Hamburg,

Frederick Stern University of Iowa and Iowa Institute of Hydraulic Research

Takanori Hino, Nobuyuki Hirata, Frederick Stern, Lars Larsson,

the JBC. Chapter 8 is the evaluations of resistance and self-propulsion simulations

The following three hulls were used in the Workshop:

Several types of computations were requested, namely:

Table 1 Main particulars of the three ships

Fig. 2 Cartesian coordinate system

Table 2 Test cases

4 Participants and Methods

Table 3 Workshop participants and methods

Chapter 2 Experimental Data for JBC Resistance, Sinkage, Trim,

Chapter 3 Experimental Data for KCS Resistance, Sinkage, Trim,

Chapter 5 Evaluation of Resistance, Sinkage, Trim and Wave Pattern

Chapter 6 Analysis of the Local Flow around JBC

perfectly reproduced as illustrated by the occurrence of a reversed flow region

Chapter 7 Evaluation of Self-propulsion and Energy Saving Device

Chapter 8 Evaluation of Resistance, Sinkage, Trim and Self-propulsion

Nobuyuki Hirata, Hiroshi Kobayashi, Takanori Hino, Yasuyuki Toda,

Abstract In this Chapter, measured data of resistance, sinkage, trim, self-propulsion

2 JBC (Japan Bulk Carrier)

2.1 Ship Hull

A propeller used is a conventional five-bladed MAU type propeller. The particulars

Table 1 Principal particulars of JBC in actual ship scale

Fig. 1 Body plan of JBC

Table 2 Principal particulars of a model propeller

2.3 Energy Saving Device

A circular duct installed ahead of a propeller is selected as an energy saving device,

Table 3 Principal particulars of a duct

Fig. 2 Energy saving duct layout

Table 4 Principal particulars of NMRI, OU and TUHH models

Fig. 3 NMRI model

Fig. 5 TUHH model

Model propellers used in self-propulsion tests and SPIV measurements at NMRI

Fig. 6 Propeller open water characteristics of the NMRI model propeller

3 Resistance and Self-propulsion Tests

Table 5 Results of resistance and self-propulsion tests at design speed

Fig. 7 Total resistance coefficient (C T ) of NMRI model

Fig. 8 Wave making resistance coefficient (C W ) of NMRI model

Fig. 9 Trim of NMRI model

Fig. 10 Sinkage of NMRI model

Fig. 11 Thrust deduction coefficient (1 −t) of NMRI model

Fig. 12 Thrust wake coefficient (1 − wT ) of NMRI model

Fig. 13 Relative rotative efficiency (η R ) of NMRI model

Fig. 14 Propeller thrust coefficient (K T ) of NMRI model

4 Detailed Flow Measurements

Fig. 15 Propeller torque coefficient (K Q ) of NMRI model

4.2 Accuracy Estimation of SPIV Measurement at NMRI