Etd

Download as pdf or txt
Download as pdf or txt
You are on page 1of 274

ABSTRACT

QUAYYUM, SHAHRIAR. Advanced Finite Element Analyses of Moment Resisting


Connections for Improving Seismic Performance and Exploring Effects of Residual Stress
and Fire Damage. (Under the direction of Dr. Tasnim Hassan).

This dissertation research deals with rigorous finite element (FE) analysis and
experimental investigation for exploring failure mechanisms of welded steel moment
connections (WSMCs) and developing seismic performance enhancement techniques for
moment resisting steel buildings. Two types of moment resisting connections in steel
buildings were studied, viz. eight bolt stiffened extended end-plate (8ES) connection and
welded unreinforced flange bolted/welded web (WUF-B/WUF-W) connection. The AISC
prequalified eight bolt stiffened extended end plate (8ES) connection showed good
performance in energy dissipation and ductility in many experimental studies under seismic
loading. However, fracture of beam flange or stiffener or weld between beam flange/end
plate and stiffener, leading to early strength degradation or brittle failure of the connection
has also been observed. Hence, the 8ES connection was analyzed using advanced nonlinear
finite element (FE) modeling to explore these failure responses. Based on the simulation
responses, seismic performance enhancement techniques were developed and analytically
evaluated. The modified 8ES connection has been experimentally validated to demonstrate
excellent seismic performance.
The aftermath of the 1994 Northridge Earthquake led to significant modifications in the
design of moment resisting connections. Although the post-Northridge modified connections
showed improved performance in terms of ductility and energy dissipation capacity, low-
cycle fatigue failures have been repeatedly observed at the weld toe or access hole of the
welded unreinforced flange bolted web (WUF-B) connections. Weld induced residual
stresses are considered to be one of the leading contributors to the low-cycle fatigue failures
observed in the post-Northridge connections. An experimental study was performed on
WUF-B connection to study the influence of welding procedure on the failure mechanism of
the connections. Two specimens of WUF-B connections were tested under constant
amplitude cyclic loading history and accumulation of axial strain near the welded joint was
observed for both the specimens. However, the rate and magnitude of strain accumulation
was different for different welding sequence and hence, the fatigue lives of the specimens
were different. Two specimens used in the experimental investigation were simulated, using
advanced thermo-mechanical finite element modeling techniques. Numerical techniques,
incorporating the W-shape manufacturing processes and real time weld sequences, were
developed for simulating initial and welding residual stresses. The analysis model was
validated based on the measured temperature history during welding and strain data from the
experimental study, and initial residual stress data from literature. The simulation responses
of the WUF-W connection demonstrated that under low-cycle fatigue loading despite the
welding residual stress relaxation, the axial strain at the weld toe continued to grow with
cycle (strain ratcheting) to induce brittle fracture similar to what was observed in the
experiments. The critical locations of high stress and strain concentrations are also depicted
by the simulated contour plots. These novel numerical techniques can be implemented to
optimize weld sequence and heat treatment techniques for enhancing seismic or fatigue life
of beam-column connections.
Finally, implications of fire damage to moment resisting steel building under seismic
loading have been explored. It is not known how a rehabilitated fire damaged building would
behave under a seismic event. Properties of structural steel become heterogeneous because of
different peak temperature exposure at different locations of the building. Depending on the
rate of cooling or water quenching steel properties will have a much wider range compared to
those before fire exposure. Finite element analysis was performed to shed light on the
performance of the fire damaged building under simulated seismic loading. The seismic
analysis results demonstrated that the lateral drift demands were significantly influenced by
the post-fire strength degradation and heterogeneity of structural steel. It was observed that
nearly all of the earthquake-induced lateral displacement occurs at the story level where the
fire exposed compartments are located. This may lead to single story mechanism commonly
known as soft story mechanism at the fire exposed story level which may lead to catastrophic
failure of the structure.
Copyright 2014 Shahriar Quayyum

All Rights Reserved


Advanced Finite Element Analyses of Moment Resisting Connections for Improving Seismic
Performance and Exploring Effects of Residual Stress and Fire Damage

by
Shahriar Quayyum

A dissertation submitted to the Graduate Faculty of


North Carolina State University
in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy

Civil Engineering

Raleigh, North Carolina

2014

APPROVED BY:

_______________________________ ______________________________
Dr. Tasnim Hassan Dr. Jim Nau
Committee Chair

_______________________________ _______________________________
Dr. Mervyn Kowalsky Dr. Gracious Ngaile
DEDICATION

To my lovely wife Simia,


Very happy son Aariz
and
Loving Parents

ii
BIOGRAPHY

Shahriar Quayyum was born in Dhaka, Bangladesh on April 26, 1983. He received a BSc
in Civil Engineering from Bangladesh University of Engineering and Technology (BUET),
Dhaka, Bangladesh in 2006. After graduating from BUET, he joined as a lecturer at
Presidency University in Bangladesh, where he taught theoretical courses in Civil
Engineering at the undergraduate level. He worked at Presidency University for two years,
and in 2009 he pursued a Masters degree in Civil Engineering from the University of British
Columbia, BC, Canada. During the Masters degree, his research focus was on bond behavior
of fiber reinforced polymer reinforcing bars in concrete. He graduated from the University of
British Columbia in 2010 and started his PhD in Civil Engineering at North Carolina State
University, NC, USA. His PhD research work focused on the seismic performance
enhancement of steel moment resisting connections, residual stress and fire damage analysis
of steel buildings.

iii
ACKNOWLEDGEMENTS

The author wishes to convey his profound gratitude to the almighty for allowing him to
bring this effort to fruition. The study was undertaken under the supervision of Dr. Tasnim
Hassan, Professor, Department of Civil, Construction and Environmental Engineering, North
Carolina State University, Raleigh, NC. His systematic guidance and constant persuasion has
immensely helped the author throughout this work. The author sincerely acknowledges his
heartiest gratitude and indebtedness to him for his superior technical expertise and solutions
to many projects and computer-related problems. The author is thankful to his doctoral
committee members Dr. Jim Nau, Dr. Mervyn Kowalsky and Dr. Gracious Ngaile for their
encouragement, insightful comments, and time throughout this research.
The author also pays his deepest homage to his parents, whom he believes to be cardinal
source of inspiration for all his achievement.
The author would like to acknowledge his research partner Machel Morrison for his
theoretical and experimental inputs which helped author during his dissertation research. The
contribution provided by Machel Morrison and Doug schweizer in experimental validation of
the concept developed in study is greatly acknowledged. In addition, the author would like to
thank all his colleagues, namely Nazrul Islam, Patrick Graham Pritchard, Raashed Ahmed,
Paul Barret for sharing their knowledge and ideas during the course of the research.
Finally, the author would like to take the opportunity to thank his spouse Simia
Madhuchhanda for being by his side all the time and act as a cradle of motivation for
everything.

iv
TABLE OF CONTENTS

LIST OF SYMBOLS .........................................................................................................ix


LIST OF TABLES ............................................................................................................xii
LIST OF FIGURES......................................................................................................... xiii

CHAPTER 1: INTRODUCTION ............................................................................................. 1


1. Background and Motivation ................................................................................... 1
2. Scope and Organization .......................................................................................... 4
3. References.............................................................................................................. 7

CHAPTER 2: AN IMPROVED UNSTIFFENED 8 BOLT EXTENDED END PLATE MOMENT


CONNECTION FOR SPECIAL AND INTERMEDIATE MOMENT FRAMES, PART I: NUMERICAL
MODELING AND MODIFIED DESIGN DEVELOPMENT ............................................................. 11

Abstract ................................................................................................................... 11
1. Introduction .......................................................................................................... 12
2. Finite Element Modeling of Extended End Plate Connection ................................ 17
2.1. Finite Element Discretization ............................................................................ 18
2.2. Material Models ................................................................................................ 20
2.3. Loads and Boundary Conditions........................................................................ 24
3. Validation of Finite Element Model...................................................................... 25
4. Finite Element Analysis of Extended End Plate Connections ................................ 27
4.1. Eight Bolt Stiffened Extended End Plate Connection (8ES)............................... 28
4.2. Eight Bolt Unstiffened Extended End Plate Connection (8E)............................. 31
4.3. Four Bolt Wide Unstiffened Extended End Plate Connection (4W) ................... 34
4.4. Eight Bolt Unstiffened Modified Extended End Plate Connection (8EM) .......... 37
5. Further Development of the Modified Design of Extended End Plate Connection.....
............................................................................................................................. 43
6. Conclusions .......................................................................................................... 55
7. References............................................................................................................ 56

v
CHAPTER 3: AN IMPROVED UNSTIFFENED 8 BOLT EXTENDED END PLATE MOMENT
CONNECTION FOR SPECIAL AND INTERMEDIATE MOMENT FRAMES, PART II: EXPERIMENTAL
VALIDATION AND ADDRESSING FURTHER DEVELOPMENT.................................................... 63

Abstract ................................................................................................................... 63
1. Introduction .......................................................................................................... 64
2. Review of Proposed Novel Performance Enhancing Techniques .......................... 65
3. Connection Design and Experimental Setup ......................................................... 67
4. Instrumentation .................................................................................................... 70
5. Test Results .......................................................................................................... 71
5.1. Global Response of Modified BEEP Connection ............................................... 71
5.2. Plastic Hinge Formation in Modified BEEP Connection.................................... 72
5.3. Bolt Response in Modified BEEP Connection ................................................... 76
5.4. Failure Mechanism............................................................................................ 76
6. Post-Test FEA and Discussion .............................................................................. 78
7. Further Development of Modified BEEP Connection ........................................... 82
7.1. Future Work on Web Stiffener ........................................................................... 83
7.2. Design Development ......................................................................................... 84
8. Conclusions .......................................................................................................... 84
9. References............................................................................................................ 85

CHAPTER 4: INFLUENCE OF WELD SEQUENCE ON THE SEISMIC RESPONSE OF MOMENT


RESISTING CONNECTIONS .................................................................................................... 88

Abstract ................................................................................................................... 88
1. Introduction .......................................................................................................... 89
2. Cyclic Testing of WSMCs .................................................................................... 91
2.1. Welding Procedure and Welding Details ........................................................... 92
2.2. Test Setup ......................................................................................................... 97
2.3. Loading History ................................................................................................ 98
3. Cyclic Response of WUF-B Connections ........................................................... 100
3.1. Global Response ............................................................................................. 100

vi
3.2. Local Response ............................................................................................... 104
4. Conclusions ........................................................................................................ 110
5. References.......................................................................................................... 111

CHAPTER 5: SIMULATION OF INITIAL RESIDUAL STRESSES IN HOT-ROLLED WIDE FLANGE


SHAPES .............................................................................................................................. 116

Abstract ................................................................................................................. 116


1. Introduction ........................................................................................................ 116
2. Manufacturing Process of Structural Shapes and Formation of Initial Residual
Stresses .............................................................................................................. 119
3. Finite Element Modeling of the Cooling Process ................................................ 121
3.1. Thermal Analysis ............................................................................................ 122
3.2. Residual Stress Analysis ................................................................................ 127
4. Results and Discussion ....................................................................................... 130
5. Influence of IRS on Buckling Response of Wide Flange Steel Columns ............. 133
6. Conclusions ........................................................................................................ 141
7. References.......................................................................................................... 142

CHAPTER 6: INFLUENCE OF WELDING ON THE SEISMIC PERFORMANCE OF MOMENT


RESISTING CONNECTION ................................................................................................... 147

Abstract ................................................................................................................. 147


1. Introduction ........................................................................................................ 148
2. Numerical Scheme for Weld Residual Stress Simulation .................................... 150
3. Thermal Analysis ............................................................................................... 153
4. Residual Stress Analysis ..................................................................................... 162
5. Influence of Initial Residual Stresses .................................................................. 172
6. Effect of Residual Stresses on the Seismic Performance of WSMCs ................... 174
7. Post Weld Heat Treatment of WSMCs ............................................................... 188
8. Conclusions ........................................................................................................ 196
9. References.......................................................................................................... 198

vii
CHAPTER 7: SEISMIC PERFORMANCE OF A FIRE DAMAGED MOMENT RESISTING FRAME
.......................................................................................................................................... 208
Abstract ................................................................................................................. 208
1. Introduction ........................................................................................................ 208
2. Post-Fire Mechanical Properties of Steel ............................................................ 212
3. Seismic Analysis of Moment Resisting Frames with Fire Damage ...................... 214
3.1. Finite Element Modeling ................................................................................. 215
3.2. Distortion Analysis of a Single Story Moment Resisting Frame during Fire
............................................................................................................................. 216
3.3. Seismic Response of Fire Damaged Moment Resisting Frames ....................... 225
3.4. Seismic Response of an Exterior WUF-W Connection with Fire Damage ....... 232
4. Conclusions ........................................................................................................ 237
5. References.......................................................................................................... 238

CHAPTER 8: CONCLUSIONS AND RECOMMENDATIONS.................................................. 243


1. Conclusions ........................................................................................................ 243
2. Recommendations for Future Works .................................................................. 245
2.1. Modified Extended End Plate Connection with Extended Shear Tab ............... 246
2.2. Further Study on Seismic Analysis of WSMCs with Residual Stresses ............ 246
2.3. Fire Damage Analysis with Water Quenched Structural Steel.......................... 247
2.4. Simulation with Advanced Constitutive Models .............................................. 247

viii
LIST OF SYMBOLS

db Bolt diameter
E Elastic modulus
0 Initial size of the yield surface
Ci Kinematic hardening parameter
i Kinematic hardening parameter
Stress tensor
Current center of the yield surface in the total stress space
s Deviatoric stress tensor
a Current yield surface center in the deviatoric space
R Isotropic hardening variable
p Accumulated plastic strain
Q Size of the plastic strain surface
x Axial stress
x Axial strain
M Moment at the centerline of the column
Interstory drift angle
b Beam rotation
pz Panel zone rotation
c Column rotation
ep End plate rotation
N Number of loading cycles
T Temperature
TP Peak Temperature
th Hold time at peak temperature
Distance along the width of the beam flange from the centerline of the web
(mm)
Distance along the length of the beam flange from the weld toe
k Thermal conductivity
Q Generated heat energy
Density
cp Specific heat
Coefficient of thermal expansion
Poissons ratio

ix
t Time
Mechanical strain
x, y, Cartesian coordinates
zh Surface coefficient of heat transfer
n Coordinate normal to surface
Tatm Atmospheric temperature
{T} Vector of nodal temperatures
[C] Heat-capacity matrix
{K} Heat conductivity matrix
{Q} Vector of heat energy
p
Plastic strain increment
Plastic multiplier
dp Equivalent plastic strain increment
Total strain increment
e
Elastic strain increment
T
Thermal strain increment
{d } Stress increment matrix
[D] Consistent tangent modulus matrix
{d } Strain increment matrix
[Cth] Thermal stiffness matrix
{dT} Temperature increment matrix
T Temperature difference
Rx Longitudinal residual stress
Ux Displacement in the x direction
Uy Displacement in y direction
Uz Displacement in z direction
Rx Rotation about x axis
Ry Rotation about y axis
Rz Rotation about z axis
Arc efficiency
U Arc voltage
I Arc current
Qs Surface heat flux

x
Qb Body heat flux
qs Specific surface heat flux
qb Specific body heat flux
J2 Second invariant
ai Decomposed deviator of back stress
i Initial yield stress as a function of T,Tc
xm Mean axial stress
zm Mean axial strain
aza Axial strain amplitude
0RT Initial yield stress at room temperature
Tc Conditioning temperature

xi
LIST OF TABLES

CHAPTER 2: AN IMPROVED UNSTIFFENED 8 BOLT EXTENDED END PLATE MOMENT


CONNECTION FOR SPECIAL AND INTERMEDIATE MOMENT FRAMES, PART I: NUMERICAL
MODELING AND MODIFIED DESIGN DEVELOPMENT

Table 1 Nonlinear kinematic hardening parameters for different steel materials ... 21

CHAPTER 4: INFLUENCE OF WELD SEQUENCE ON THE SEISMIC RESPONSE OF MOMENT


RESISTING
Table 1 Details of weld sequence for specimen MC1 ........................................... 94
Table 2 Details of weld sequence for specimen MC2 ........................................... 96

CHAPTER 5: SIMULATION OF INITIAL RESIDUAL STRESSES IN HOT-ROLLED WIDE FLANGE


SHAPES
Table 1 Width-thickness ratio for different column sections ............................... 137

xii
LIST OF FIGURES

CHAPTER 1: INTRODUCTION
Figure 1 Beam-column connection fractures during the 1994 Northridge earthquake
[1] ........................................................................................................................ 1

CHAPTER 2: AN IMPROVED UNSTIFFENED 8 BOLT EXTENDED END PLATE MOMENT


CONNECTION FOR SPECIAL AND INTERMEDIATE MOMENT FRAMES, PART I: NUMERICAL
MODELING AND MODIFIED DESIGN DEVELOPMENT

Figure 1 Different configurations of extended end plate moment connections .......... 13


Figure 2 Finite element mesh of an eight bolt stiffened extended end plate connection
.......................................................................................................................... 19
Figure 3 Experimental and simulated response of strain controlled hysteresis loop
with Chaboche model parameters. (a) ASTM A572 Gr.50 steel; (b) ASTM A36
steel; (c) ASTM A992 steel ............................................................................... 22
Figure 4 Monotonic stress-strain curves for bolt and weld materials. (a) ASTM A490
bolt; (b) E70 weld ............................................................................................. 23
Figure 5 (a) A typical experimental setup for seismic testing of moment connections
[31]; (b) SAC loading protocol [14] ................................................................... 24
Figure 6 Comparison between experimental and simulated moment-rotation response
of eight bolt stiffened extended end plate connection (8ES-1.25-1.75-30) tested by
Sumner et al. [10] .............................................................................................. 26
Figure 7 Comparison of experimental and simulated deformed shape for 8ES-1.25-
1.75-30 specimen of Sumner et al. [10] showing formation of plastic hinge at the
base of the stiffener ........................................................................................... 26
Figure 8 Eight bolt stiffened extended end plate connection (8ES). (a) FE mesh; (b)
Equivalent plastic strain contour showing formation of plastic hinge after cyclic
loading upto 6% interstory drift angle ............................................................... 28
Figure 9 Cyclic response of 8ES configuration of connection. (a) Moment-rotation
hysteresis response for cyclic loading upto 6% interstory drift angle; (b)
Contribution of different components of the connection in the total rotation ....... 29
Figure 10 Bolt strains for 8ES configuration during loading cycles at the top flange of
the beam ........................................................................................................... 30

xiii
Figure 11 Eight bolt unstiffened extended end plate connection (8E). (a) FE mesh; (b)
Equivalent plastic strain contour showing formation of plastic hinge after cyclic
loading upto 6% interstory drift angle ............................................................... 32
Figure 12 Axial strain profile along the width of beam flange at the weld toe for 8E
configuration of the connection ......................................................................... 32
Figure 13 Bolt strains for 8E configuration during loading cycles at the top flange of
the beam ........................................................................................................... 33
Figure 14 Four bolt wide unstiffened extended end plate connection (4W). (a) FE
mesh; (b) Equivalent plastic strain contour showing bolt failure and high weld
strain at 3% interstory drift angle ...................................................................... 35
Figure 15 (a) Equivalent plastic strain contour of end plate at the end of 3% interstory
drift angle; (b) Moment-rotation hysteresis response of 4W connection ............. 35
Figure 16 Bolt strains for 4W configuration during loading cycles at the top flange of
the beam ........................................................................................................... 36
Figure 17 Axial strain profile along the width of beam flange at the weld toe for 4W
configuration of the connection ......................................................................... 37
Figure 18 (a) Proposed bolt rearrangement in a hexagonal pattern; (b) FE mesh with
the modified bolt arrangement for 8EM configuration of the connection; (c)
Equivalent plastic strain contour of 8EM configuration of the connection showing
formation of plastic hinge .................................................................................. 38
Figure 19 Cyclic response of 8EM configuration of connection. (a) Moment-rotation
hysteresis response for cyclic loading upto 6% interstory drift angle; (b)
Contribution of different components of the connection in the total rotation ....... 39
Figure 20 Bolt strains for 8EM configuration during loading cycles at the top flange
of the beam ....................................................................................................... 41
Figure 21 Moment-rotation envelopes of the extended end plate connections for
different bolt arrangements ................................................................................ 42
Figure 22 Axial strain profile along the width of the beam bottom flange for different
arrangements of the bolts. (a) At weld toe during loading cycles at 4% interstory
drift angle; (b) At plastic hinge location during loading cycles at 6% interstory
drift angle .......................................................................................................... 43
Figure 23 Participation of end plate and panel zone in the total rotation capacity at 6%
interstory drift angle for different configurations of bolt. (a) End plate; (b) Panel
zone .................................................................................................................. 44

xiv
Figure 24 8EM configuration of the connection with W30148 beam section. (a)
Equivalent plastic strain contour showing formation of plastic hinge; (b) Axial
strain profile along the width of the beam flange for loading cycles upto 6%
interstory drift angle .......................................................................................... 45
Figure 25 Heat treatment of modified extended end plate connection for seismic
performance enhancement ................................................................................. 46
Figure 26 (a) Monotonic stress-strain response for unheat treated and heat treated
ASTM A992 specimens; (b) Experimental and simulated response of strain
controlled hysteresis loop with Chaboche parameters for heat treated ASTM A992
steel .................................................................................................................. 47
Figure 27 Cyclic response of 8EMH configuration of the connection with W30148
beam section. (a) Equivalent plastic strain contour showing formation of plastic
hinge in the heat treated region of the beam flanges; (b) Moment-rotation
response upto 6% interstory drift angle .............................................................. 47
Figure 28 Cyclic response of the connection for 8EMH configuration of the bolts. (a)
Axial strain profile along the beam flange at the weld toe; (b) Participation of
different components of the connection in the total rotation capacity of the
connection.......................................................................................................... 49
Figure 29 Bolt strains for 8EMH configuration of the connection with W30148
beam section ..................................................................................................... 50
Figure 30 Local buckling of beam web and flanges at different loading cycles for
8EMH configuration of the connection with W30148 beam section ................ 51
Figure 31 (a) Modified extended end plate connection (8EMH-W) with heat treatment
and web stiffener; (b) Formation of plastic hinge in the heat treated region of
beam flanges under simulated seismic loading with 8EMH-W configuration of the
connection ......................................................................................................... 52
Figure 32 Cyclic response of 8EMH-W configuration of the connection upto 6%
interstory drift angle. (a) Moment-rotation hysteresis response; (b) Axial strain
profile along beam flange at the weld toe .......................................................... 53
Figure 33 Bolt strains for 8EMH-W configuration of the connection with W30148
beam section ..................................................................................................... 54
Figure 34 Moment-rotation envelopes of different configurations of the extended end
plate connections. (a) W30148 beam section; (b) W3099 beam section ......... 55

xv
CHAPTER 3: AN IMPROVED 8 BOLT EXTENDED END PLATE MOMENT CONNECTION FOR
SPECIAL AND INTERMEDIATE MOMENT FRAMES: PART II. EXPERIMENTAL VALIDATION AND
ADDRESSING FURTHER DEVELOPMENT

Figure 1 Sketch of Modified BEEP Connection ...................................................... 66


Figure 2 Engineering stress-strain response of A992 and heat-treated A992 steel .... 66
Figure 3 HBS 4 Test setup. (a) Sketch of test setup, (b) photograph showing test setup
.......................................................................................................................... 67
Figure 4 HBS 4 connection details. (a) Elevation of connection, (b) photograph of
HBS 4 connection prior to testing, (c) section A-A showing connection of beam
web and webstiffeners to endplate, (d)elevation of end plate showing modified
bolt arrangement ............................................................................................... 68
Figure 5 Heat-treatment setup. (a) Photograph showing electric surface heating pads
during installation on a beam flange, (b) photograph showing insulation of beam
flange for well controlled heating and cooling ................................................... 69
Figure 6 Instrumentation of HBS 4. (a) LED markers placed on beam top flange,
beam web and webstiffeners, (b) LED markers placed on the sides of beam top
and bottom flange, beam web and webstiffners ................................................. 71
Figure 7 Experimental response of HBS 4. (a) Moment rotation hysteresis response,
(b) total rotation contributions ........................................................................... 72
Figure 8 Plastic rotation contributions by components of HBS 4. (a) Total connection
plastic rotation, (b) panel zone shear plastic rotation, (c) column flexural plastic
rotation, and (d) beam plastic rotation ............................................................... 73
Figure 9 Recorded longitudinal strains along the center of top flange of the beam
from HBS 4 ....................................................................................................... 74
Figure 10 Progression of yielding and plastic hinge formation in HBS 4. (a) 1% drift,
(b) 2% drift, (c) 3% drift, and (d) 4% drift ......................................................... 74
Figure 11 Measured flexural strains from strain gages placed on the bottom flange of
HBS 4. Strains are shown from the beginning of test to the second cycle of 3%
drift ................................................................................................................... 75
Figure 12 Recorded axial bolt strain vs column centerline moment ......................... 77
Figure 13 Local buckling failure of HBS 4. (a) Post test photograph showing flange
and web local buckling, (b) photograph showing fracture of beam flange in the
crest of flange buckle ........................................................................................ 78

xvi
Figure 14 Validation of post-test FEA of HBS 4. (a) FEA discretization including
initial distortion of web stiffener plates, (b) comparison between FEA prediction
and the recorded moment-rotation response, (c) photograph of HBS 4 during
loading cycles at 6% drift, (d) FEA prediction of deformation at 6% drift, (e)
comparison between FEA predictions and recorded top flange vertical
displacements during various stages of loading history, (f) photograph of HBS 4
top flange during loading at 4% drift ................................................................. 80
Figure 15 Post-test FEA simulation results showing the effect of initial web stiffener
plate distortion. (a) Moment-rotation envelop, (b) plastic equivalent strain
prediction at 5% drift for HBS 4 (post-test simulation with initial web stiffener
distortion), (c) plastic equivalent strain prediction at 5% drift for HBS 4 (pre-test
simulation without initial web stiffener distortion) ............................................ 81

CHAPTER 4: INFLUENCE OF WELD SEQUENCE ON THE SEISMIC RESPONSE OF MOMENT


RESISTING CONNECTIONS
Figure 1 Sketch of the full-scale exterior WUF-B connection tested in this study .... 92
Figure 2 Sketch of weld sequences at the beam top and bottom flange for specimen
MC1 showing number of welding passes in order of completion ...................... 93
Figure 3 Sketch of weld sequences at the beam top and bottom flanges for specimen
MC2 showing number of welding passes in order of completion ...................... 95
Figure 4 Schematic of the test setup used for cyclic testing of WUF-B Connections
................................................................................................................................ 98
Figure 5 (a) SAC loading protocol [35], (b) non-standard loading history prescribed
in the experiments ............................................................................................. 99
Figure 6 Response of specimen MC1. (a) moment-rotation hysteresis response, (b)
rupture in the beam top flange ......................................................................... 101
Figure 7 Response of specimen MC2. (a) moment-rotation hysteresis response, (b)
cracks in the reinforcing fillet weld of bottom beam flange ............................. 102
Figure 8 Response of specimen MC2. (a) rupture of the top flange initiated at the
flattened portion of the weld access hole, (b) cracks in the weld-base metal
interface of top beam flange ............................................................................ 103
Figure 9 Mean axial strain and amplitude plotted against number of loading cycles
recorded at the top flange (inside surface) of the connections .......................... 105

xvii
Figure 10 Mean axial strain and amplitude plotted against number of loading cycles
recorded at the beam bottom flange (inside surface) of the connections ........... 106
Figure 11 Mean axial strain and amplitude plotted against number of loading cycles
recorded at the beam bottom flange (outside surface) of the connections ......... 108
Figure 12 Mean axial strain and amplitude plotted against number of loading cycles
recorded at the column flange of the connections ............................................ 109

CHAPTER 5: SIMULATION OF INITIAL RESIDUAL STRESSES IN HOT-ROLLED WIDE FLANGE


SHAPES

Figure 1 Schematic diagram showing the manufacturing process of structural steel


shapes [26] ....................................................................................................... 120
Figure 2 (a) Beam blank continuous caster mold, (b) cross-sectional view of beam
blank, (c) cross-sectional view of finished product after hot-rolling ................. 121
Figure 3 Finite element mesh for thermal finite element analysis of W36150 shape
........................................................................................................................ 122
Figure 4 Material coefficients of structural steel as a function of temperature used in
thermal finite element analysis. (a) Specific heat [32], (b) thermal conductivity
[32], (c) heat transfer coefficient [33], (d) enthalpy [34] .................................. 124
Figure 5 Contour showing temperature distribution (in K) during the cooling process
of hot rolled W36150 section from initial 1300C reheat temperature. (a) after
15 minutes, (b) 30 minutes, (c) 45 minutes ...................................................... 125
Figure 6 (a) Cooling curves at three different locations over the cross section of a
W36150 section, (b) temperature difference as a function of cooling time over
the cross section of a W36150 section ........................................................... 126
Figure 7 (a) Poissons ratio [37] and (b) thermal expansion coefficient [38] of
structural steel plotted as a function of temperature ......................................... 128
Figure 8 (a) Monotonic stress-strain response of ASTM A36 steel at different
temperatures, (b) monotonic stress-strain response of ASTM A992 steel at
different temperatures [36] ............................................................................... 129
Figure 9 Contour showing longitudinal stress distribution during the cooling process
of hot rolled W36150 section. (a) after 15 minutes, (b) 1 hour, (c) 2 hours ... 130
Figure 10 Residual stress analysis results for a W36150 section. (a), (b) Measured
[3] and simulated longitudinal residual stresses at the centerline along the beam

xviii
length, (c) simulated longitudinal residual stress distribution along the length of
beam, (d) simulated longitudinal residual stresses plotted against cooling time at
different locations of the beam section ............................................................. 132
Figure 11 Measured [1] and simulated longitudinal residual stresses at the centerline
along the beam length for a W1443 section ................................................... 133
Figure 12 Measured [20] and simulated longitudinal residual stresses at the centerline
along the beam length for a W1650 section ................................................... 134
Figure 13 Measured [20] and simulated longitudinal residual stresses at the centerline
along the beam length for a W27102 section ................................................. 134
Figure 14 (a) Equivalent plastic strain contour of W1287 column under axial
compression, (b) simulated and experimental [12] axial load-displacement
response, (c) simulated and experimental [12] out-of-plane deflection response
........................................................................................................................ 135
Figure 15 (a) Finite element mesh used in the cyclic response evaluation of wide
flange column sections, (b) displacement-based simulated earthquake loading
history [42] ..................................................................................................... 136
Figure 16 Cyclic response of W24131 column without considering IRS. (a)
Equivalent plastic strain contour, and (b) moment-rotation hysteresis response at
the end of 6% interstory drift angle ................................................................. 138
Figure 17 Cyclic response of W24131 column considering IRS. (a) Equivalent
plastic strain contour, and (b) moment-rotation hysteresis response at the end of
6% interstory drift angle .................................................................................. 139
Figure 18 Moment-rotation envelope showing peak moment at each loading cycle for
different column sections with and without IRS. (a) W14132, (b) W24131, (c)
W24250, (d) W36230 .................................................................................. 140

CHAPTER 6: INFLUENCE OF WELDING ON THE SEISMIC PERFORMANCE OF MOMENT


RESISTING CONNECTION

Figure 1 Steps performed in the thermo-mechanical analysis of WUF-W connection


........................................................................................................................ 151
Figure 2 Finite element mesh of WUF-W connection used in the study ................. 152
Figure 3 Material coefficients of structural steel as a function of temperature used in
thermal analysis. (a) Specific heat [55], (b) thermal conductivity [55], (c) heat
transfer coefficient [20], (d) enthalpy [56] ....................................................... 154

xix
Figure 4 Element birth and death technique in the FE simulation of welding process
........................................................................................................................ 155
Figure 5 Illustration of moving temperature (in Kelvin) field during the final welding
pass in the top flange of a welded steel moment connection (a-e) and cooling (f)
........................................................................................................................ 156
Figure 6 Sketch of weld sequences in beam top and bottom flanges for specimen
MC1 showing number of welding passes in order of completion ..................... 157
Figure 7 Sketch of weld sequences in beam top and bottom flanges for specimen
MC2 showing number of welding passes in order of completion ..................... 158
Figure 8 Measured and simulated temperature profiles at the beam top flange for the
two specimens. (a) Specimen MC1, (b) specimen MC2 ................................... 159
Figure 9 Simulated temperature profiles at the top and bottom flange of the beam
showing variation for two different welding sequences. (a) Top flange, (b) bottom
flange .............................................................................................................. 160
Figure 10 Temperature distribution (in Kelvin) at the end of weld passes 1 and 2 of
welding sequence A at the beam bottom flange of specimen MC1 showing
nonuniform temperature on both sides of the beam web .................................. 161
Figure 11 Temperature distribution (in Kelvin) at the end of weld passes 1 and 2 of
welding sequence B at the beam bottom flange of specimen MC2 showing
uniform temperature on both sides of the beam web ........................................ 161
Figure 12 Monotonic tension test response of ASTM A992 steel at different
temperatures [57]. (a) Monotonic stress-strain response, (b) initial yield stress
plotted as a function of temperature, (c) elastic modulus plotted against
temperature ..................................................................................................... 163
Figure 13 (a) Poissons ratio [59] and (b) thermal expansion coefficient of structural
steel plotted as a function of temperature [60] ................................................. 164
Figure 14 (a) Stress-strain response of ASTM A992 coupons conditioned at different
peak temperatures [62], (b) temperature dependent yield stress of ASTM A992 for
five temperature levels to account for the material heterogeneity ..................... 165
Figure 15 Residual stress distribution (in Pa) for specimen MC1. (a) Longitudinal
direction, (b) transverse direction .................................................................... 167
Figure 16 Residual stress distribution (in Pa) for specimen MC2. (a) Longitudinal
direction, (b) transverse direction .................................................................... 167

xx
Figure 17 Welding induced residual stress distribution in the beam top flange of
specimen MC1. (a) Across width, (b) across length ......................................... 168
Figure 18 Welding induced residual stress distribution in the beam bottom flange of
specimen MC1 for weld sequence A. (a) Across width, (b) across length ........ 169
Figure 19 Welding induced residual stress distribution in the beam top flange of
specimen MC2. (a) Across width, (b) across length ......................................... 170
Figure 20 Welding induced residual stress distribution in the beam bottom flange of
specimen MC2 for weld sequence B. (a) Across width, (b) across length ........ 171
Figure 21 Welding induced residual stress distribution across the width of beam
flange near weld toe for the two specimens. (a) Top flange, (b) bottom flange
............................................................................................................................. 171
Figure 22 Residual stress distribution along the length of the beam for specimen MC1
showing the influence of initial residual stresses. (a) Beam top flange, (b) beam
bottom flange .................................................................................................. 173
Figure 23 (a) SAC loading protocol [70], (b) non-standard constant amplitude loading
history prescribed in this study ........................................................................ 175
Figure 24 Experimental and simulated response of strain controlled hysteresis loop
for ASTM A992 steel with Chaboche parameters ............................................ 177
Figure 25 Plastic strain contour in axial direction showing accumulation of strains
near weld toe and weld access hole in presence of residual stresses ................. 178
Figure 26 Low-cycle fatigue response of the connection for specimen MC1. (a) Axial
mean strain and amplitude plotted against number of loading cycles, (b) axial
stress-strain hysteresis cycles showing ratcheting ............................................ 179
Figure 27 Low-cycle fatigue response of the connection for specimen MC2. (a) Axial
mean strain and amplitude plotted against number of loading cycles, (b) axial
stress-strain hysteresis cycles showing ratcheting ............................................ 179
Figure 28 Low-cycle fatigue response of the connection at the beam bottom flange for
specimen MC1. (a) Axial mean strain, and (b) strain amplitude plotted against
number of loading cycles at different locations across the width near weld toe; (c)
axial mean strain and (d) strain amplitude across the length ............................. 181
Figure 29 (a) Axial mean stress plotted against number of loading cycles for
specimen MC1 at different locations across the width of near weld toe showing
stress relaxation, (d) axial mean strain and amplitude near the top and bottom
beam flange weld access hole for specimen MC1 ............................................ 183

xxi
Figure 30 Axial mean stress and strain, and von-Mises mean stress plotted against
number of loading cycles at the top flange for specimen MC1. (a) 25 mm from
weld toe, (b) 190 mm from weld toe ................................................................ 184
Figure 31 Comparison of stress-strain response at the beam top flange for the two
specimens. (a) Stress-strain hysteresis response, (b) axial mean strain and
amplitude plotted against number of loading cycles ......................................... 185
Figure 32 Comparison of stress-strain response at the beam bottom flange for the two
specimens with different weld sequences. (a) Stress-strain hysteresis response, (b)
axial mean strain and amplitude plotted against number of loading cycles ....... 186
Figure 33 Axial mean strain and amplitude plotted against number of loading cycles
at different locations of the beam flanges showing influence of weld sequence
............................................................................................................................. 187
Figure 34 A schematic of the time-temperature curve for the post weld heat treatment
specified by AWS D1.1 [79] ........................................................................... 189
Figure 35 Post weld heat trreatment cases investigated in the study ....................... 190
Figure 36 Residual stress contours showing distribution of residual stresses after
PWHT for the six cases under investigation for specimen MC1 ....................... 192
Figure 37 Residual stress distribution in the beam top flange for specimen MC1
showing effect of post weld heat treatment (Case-VI). (a) Across the width of the
flange near the weld toe, (b) along the length of the beam ............................... 193
Figure 38 Residual stress distribution in the beam bottom flange for specimen MC1
with weld sequence A showing effect of post weld heat treatment (Case-VI). (a)
Across the width of the flange near the weld toe, (b) along the length of the beam
........................................................................................................................ 194
Figure 39 Axial strain mean and amplitude plotted against number of loading cycles
at (a) beam top flange near weld toe, (b) bottom flange near weld toe, (c) around
weld access hole near beam top flange of the connection for specimen MC1
showing influence of PWHT (Case-VI) ........................................................... 195

CHAPTER 7: SEISMIC PERFORMANCE OF A FIRE DAMAGED MOMENT RESISTING FRAME

Figure 1 Monotonic stress-strain response of ASTM A992 steel heated at different


peak temperatures followed by air cooling [29] ................................................ 213

xxii
Figure 2 (a) A schematic of the special moment frame used in this study (adapted
from Adan [31]), (b) finite element mesh developed for the frame analysis ..... 214
Figure 3 Displacement based simulated earthquake loading history [30] ............... 216
Figure 4 (a) Time-temperature curves to represent a fire scenario, (b) monotonic
stress-strain response of ASTM A992 steel at elevated temperatures [37]......... 218
Figure 5 Finite element mesh developed for thermo-mechanical analysis of a single
story two bay moment frame ........................................................................... 219
Figure 6 Response of a single story frame with fire in a single compartment. (a)
Temperature distribution (in Kelvin) at the end of peak temperature exposure (50
minutes), (b) vertical deflection (in meter), and (c) out-of-plane deflection (in
meter) at the end of the cooling phase .............................................................. 220
Figure 7 Response of a single story frame with fire in both compartments. (a)
Temperature distribution (in Kelvin) at the end of peak temperature exposure (50
minutes), (b) vertical deflection (in meter), and (c) out-of-plane deflection (in
meter) at the end of the cooling phase .............................................................. 221
Figure 8 Stress (in Pa) in longitudinal direction of a single story frame with fire in a
single compartment. (a) At the end of peak temperature exposure (50 minutes),
(b) at the end of the cooling phase. ................................................................... 223
Figure 9 (a) Stress in longitudinal direction, and (b) out-of-plane deformation at three
different locations across the beam cross section along the centerline of the beam
length. .............................................................................................................. 224
Figure 10 Cyclic response of the frame at room temperature. (a) Equivalent plastic
strain contour of the frame at 6% interstory drift angle, (b) base shear plotted
against interstory drift angle showing hysteresis response of the frame, (c)
equivalent plastic strain contour of the exterior (left), and (d) interior connection
at the 3rd level at 6% interstory drift angle ........................................................ 226
Figure 11 Displacement contours (in meter) of the frame at room temperature at 6%
interstory drift angle. (a) Lateral displacement, (b) out-of-plane displacement
............................................................................................................................. 227
Figure 12 Hysteresis response of the frame for different fire cases ........................ 228
Figure 13 Comparison of frame responses at room temperature and different fire
cases. (a) Peak positive and negative lateral forces at each loading cycle plotted
against interstory drift angle, (b) interstory drift angles at each story level........ 229

xxiii
Figure 14 Formation of plastic hinge in the interior column at the 2nd level. (a) Fire
Case-III, (b) Fire Case-IV ............................................................................... 229
Figure 15 Out-of-plane displacement contours (in meter) of the frame for different
fire cases ......................................................................................................... 231
Figure 16 (a) Reference special moment frame showing an exterior WUF-W
connection taken for the study, (b) finite element mesh of the external WUF-W
connection ....................................................................................................... 233
Figure 17 Response of WUF-W connection at room temperature under simulated
seismic loading history. (a) Equivalent plastic strain contour showing formation
of plastic hinges, (b) moment-rotation hysteresis response ............................... 234
Figure 18 (a) Temperature distribution (in Kelvin) at the peak temperature level of
fire exposure in the WUF-W connection, (b) equivalent plastic strain contour of
fire damaged WUF-W connection showing formation of plastic hinges ......... 235
Figure 19 (a) Moment-rotation hysteresis response of WUF-W connection with fire
damage under simulated seismic loading history, (b) comparison of moment-
rotation envelopes showing peak positive and negative moments at each loading
cycle ................................................................................................................ 236

xxiv
CHAPTER 1: INTRODUCTION

1. Background and Motivation


Before the 1994 Northridge Earthquake, welded unreinforced flange bolted web (WUF-
B) moment connections were used extensively in the construction of multistory buildings in
North America and it was anticipated that under seismic loading, damage would be limited to
ductile yielding of members or connection. The Northridge Earthquake in 1994 challenged
the connections in meeting seismic demands. During the Northridge Earthquake many steel
buildings were extensively damaged at the WUF-B connections and several exhibited brittle
fractures [1-4]. These brittle fractures were observed at the weld between the bottom beam
flange and the column flange as shown in Figure 1(a). In some cases, the cracks continued
through the column flange right above the weld as shown in Figure 1(b). Interestingly, the
same types of brittle connection failures were observed during the Kobe Earthquake exactly
one year after the Northridge Earthquake.

(a) Fracture at fused zone (b) Fracture through column flange

Figure 1 Beam-column connection fractures during the 1994 Northridge earthquake [1].

Many failure mechanisms were observed, and many were addressed through research
over the last two decades. The lack of ductility displayed in the pre-Northridge welded steel
moment connections (WSMCs) prompted research by the Federal Emergency Management

1
Agency (FEMA) as well as the SAC Joint Venture, a partnership between the Structural
Engineers Association of California (SEAOC), the Applied Technology Council (ATC), and
the California Universities for Research in Earthquake Engineering (CUREE). Weld-base
metal mismatch, base metal and weld metal toughnesses, backing bar, fillet weld
reinforcement, residual stress, continuity plates, panel zone, reduced beam section, weld
access hole, and column flange through-thickness fractures have all been investigated and
modifications were made in the exsisting design of moment connections [1-2, 4-11].
Based on a decade of research activities, deformation based performance requirements
for WSMCs in steel moment frame was developed [1, 12]. These requirements are specified
in the 2010 AISC Seismic Provisions for structural steel buildings [12]. According to these
provisions, beam-to-column moment connections in special moment frames (SMFs) are
required to sustain a total interstorey drift angle of 4% without significant loss of strength.
Significant loss of strength is defined to occur when the flexural resistance of the connection
drops below 80% of the nominal plastic moment capacity of the beam. These interstorey drift
angle requirements include both elastic and inelastic components of frame deformation. In
order to fulfill this performance criteria, many modifications were made in the pre-
Northridge connection to enhance the connections ductility and energy dissipation capacity,
and the seismic performance of the modified connections were tested under standard seismic
loading protocol [13-19]. Based on the results of these experimental studies, AISC 358-10
Standards approved five moment connections for special and intermediate moment frames
(SMF andf IMFs) [20] application. These are reduced beam section (RBS) moment
connection, bollted stiffened and unstiffned extended end-plate moment connection, bolted
flange-plate (BFP) moment connection, welded unreinforced flange welded web (WUF-W)
moment connection, and Kaiser bolted bracket (KBB) moment connection. Recently, Side
Plate and ConX connections are prequalified for SMF and IMFs.
The design modifications to the pre-Northridge connections showed significant
improvement in the ductility and energy dissipation capacity of welded steel moment frames;
however, the fundamental reasons for the brittle failures at the local level were not
uncovered. As a result, some of the prequalified moment connections often suffered localized

2
fractures. Testing of post-Northridge welded steel moment connections have shown that
localized low-cycle fatigue cracks may initiate near the welds [6-7, 21-23]. Research by the
European community has also demonstrated localized low-cycle fatigue fractures and
concluded that an exhaustive understanding of the low-cycle fatigue strength of steel
members and joints is required [23]. Further, several researchers found that under constant,
small amplitude tests where the beam doesnt develop its full plastic moment capacity, the
connection is susceptible to brittle fractures near the weld metal and heat affected zone [24].
Brittle fractures were also observed in connections with strong panel zones prior to the
development of the plastic moment in the beam [25].
Some of the European tests have demonstrated problems with the RBS beam column
connection which is the most widely used post-Northridge connection in USA. Failures have
been observed in the welds due to high stress concentrations, thus RBS is not considered as
an acceptable solution in Europe [23]. The RBS moment connection also results in lower
stiffness and lower capacity of the beam. This could lead to selecting larger members, thus
higher material costs. In addition, the resistance to lateral and torsional stability of RBS are
reduced compared to the WSMC beam. As a result, special attention must be paid to lateral
bracing of the flanges at the reduced beam section. The extended end-plate connection for
smaller size beam sections showed significant seismic performance improvement, but for
larger size beam sections the connection requires an end-plate stiffener. The stiffener induces
high stress concentration at the base of the stiffener which may lead to premature failure of
the beam flanges and stiffener welds [27-29].
Thus based on the observations made from the literature it is apparent that some of the
post-Northridge connections may still be exposed to brittle failure because of low-cycle
fatigue mechanism, and lateral or torsional buckling before reaching the interstory drift 4%
without significant loss of strength [6, 18]. While many design procedures can predict the
performance of a global structural parameter under prescribed load conditions, those same
procedures do not attempt to correlate these quantities with local responses. This is not a
shortcoming of the current guidelines and design procedures; however it is a limitation on the
knowledge on local failure mechanisms. Hence it is essential that the cause and mechanism

3
of local failure of welded steel moment frames be understood to effectively and reliably meet
the seismic demands. In order to fully understand the connection failure mechanisms,
localized failure mechanisms should be studied along with global failure mechanisms, which
is the subject of this study.
This study implemented advanced finite element analysis techniques and experimental
study to analyze and understand the local and global failure mechanisms of welded steel
moment connections. The finite element models were validated against experimental results
prior to analyzing the failure mechanisms. The failure mechanism of two prequalified
moment connections such as eight bolt stiffened extended end plate (8ES) and welded
unreinforced flange welded web (WUF-W) moment connection were studied through finite
element analysis. Seismic performance of welded unreinforced flange bolted web (WUF-B)
connection was investigated through experimental study. The outcome of the study
demonstrated the influence of various parameters in influencing the failure mode of the
connections. Some of these parameters are not accounted for in the design of moment
resisting buildings. Based on analysis results of the force transfer and failure mechanism of
the 8ES connection, techniques to enhance its seismic performance are developedand the
modified connection is validated through experimental investigation. The influence of
residual stress are investigated through complex thermo-mechanical finite element analysis
and validated against experimental responses of WUF-B connections. Analysis responses
demonstrated that the fire damages in steel buildings may result in soft story mechanism
towards collapse of the building. These results are presented in three parts: i) seismic
performance enhancement of extended end-plate moment connection via detailed
investigation of local and global response of the connection, ii) effect of residual stresses on
the localized failure of WUF-B and WUF-W connections, iii) effect of fire induced material
heterogeneity on the seismic performance of welded steel moment frame.

2. Scope and Organization


This dissertation deals with rigorous finite element (FE) analysis and experimental
investigation for exploring failure mechanisms of welded steel moment connections

4
(WSMCs) and developing seismic performance enhancement techniques for moment
resisting steel buildings. The findings of this study are presented through the six chapters
written in journal paper format. Chapter 1 is the current chapter introducing the dissertation.
Chapters 2 to 7 are the six journal papers describing the various findings in detail. Chapter 8
is the concluding chapter discussing recommendations for future research.
Chapter 2 is a journal paper to be submitted to the Engineering Structures on the
development of a modified design of extended end plate moment connection to enhance the
seismic performance. The AISC prequalified eight bolt stiffened extended end plate
connection has been investigated in detail by using finite element modeling techniques.
Firstly, the finite element models were validated against experimental results from the
literature, and afterwards, the seismic performance of the connection with design
modifications was studied. The goal of this study was to eliminate the necessity of end plate
stiffener with an optimized bolt arrangement that ensures even distribution of bolt forces.
Based on global and local responses of all the connections studied, a modified design of
extended end plate connection was proposed which consists of a hexagonal pattern of bolt
arrangement, heat treatment of the beam flanges near the welded joint and inclusion of a
stiffener plate at the web of the beam near connection region. The study discussed
sequentially the development of the proposed modified design and shows comparison of the
performance of the different connections.
The modified design of extended end plate connection developed in Chapter 2 has been
experimentally validated through full scale cyclic testing of the connection under simulated
seismic loading history which is presented in Chapter 3. The experiment was conducted at
the Constructed Facilities Laboaratory (CFL) by the graduate students, Machel Morrison and
Doug Schweizer. My contributions in the experimental study were through the pre and post-
test analyses. This is a journal paper to be submitted to the Engineering Structures. This
chapter gives detail of the experimental setup of the connection and the results obtained from
the experiments, which demonstrated excellent performance under seismic loading. The post-
test analysis simulated the seismic performance of the connection considering the initial
imperfections and actual loading.

5
Chapter 4 is a journal paper to be submitted to the Journal of Structural Engineering on
the experimental investigation of the influence of welding sequence on the seismic response
of the WUF-B connection. This study was motivated by the low-cycle fatigue cracks at the
weld toe that was observed in the cyclic testing of post-Northridge connections which was
attributed to the residual stresses induced by welding near the welded region. This study
takes into account two exterior sub-assemblages of WUF-B moment connections having
different sequence of welding for connecting the beam flange to the column flange. The
objective was to investigate the effect of welding procedure on the localized low-cycle
fatigue failure mechanism of the connections especially the local strain responses. The results
demonstrates the influence of weld sequence on the strain responses near the welded joint
and weld access hole and thereby influence the crack initiation life.
Chapters 5 and 6 are the continuation of the Chapter 4 study and demonstrate the
influence of initial and welding residual stresses on the seismic response of moment resisting
connections. These studies dealt with complex finite element modeling of moment resisting
connections. Chapter 5 deals with the development of advanced numerical technique to
simulate the residual stresses in the wide flange shapes and will be submitted to the Journal
of Structural Engineering. This chapter describes the detail of the modeling feature for the
heat transfer analysis followed by a thermo-mechanical analysis to simulate the initial
residual stresses. The residual stress measurements obtained from the finite element
simulation were validated against experimental data from the literature. An example was
presented on the application of the numerical simulation technique on the influence of initial
residual stresses on the seismic performance of wide flange steel columns. The results of the
analysis demonstrated the vulnerability of the steel columns with initial residual stresses.
Chapter 6 deals with the understanding of the residual stresses and their influence on the
localized failure mechanism of welded steel moment resisting connections. This study
developed a numerical technique based on sequentially coupled transient nonlinear thermo-
mechanical analysis to simulate the welding residual stresses and their influence on localized
fatigue responses. The finite element simulation technique was validated against the
experimental temperature and strain measurements. The simulated response of residual

6
stresses showed development of very high residual stresses close to the weld region. The
welding induced residual stresses together with the initial residual stresses in wide flange
shapes were included for the seismic analysis of moment resisting connections, which
demonstrated the influence of the residual stresses on the low-cycle fatigue failure
mechanism of the connections. The results of the finite element investigation also explain the
variability of the low-cycle fatigue failure response for different weld sequences. Finally, a
method to reduce the magnitude of the residual stresses has been proposed based on the
concept of post weld heat treatment which showed improved performance under seismic type
of loading. Chapter 6 on the welding residual stress will be submitted to the Journal of
Constructional Steel Research.
Chapter 7 is a journal paper to be submitted to the Journal of Constructional Steel
Research on the seismic performance assessment of fire damaged steel moment resisting
frames. The objective of the study was to investigate how a fire damaged steel building will
respond under a seismic event if it is rehabilitated for continued service. The post-fire
material strength degradation and material heterogeneity was considered to account for the
damage caused by fire. A three story reference steel moment resisting frame was modeled by
using finite element software ANSYS and its performance was investigated under different
fire scenarios where different compartments at different stories were exposed to fire. The
results of the fire exposed structures are compared to that of the structure at ambient
temperature. Moreover, to characterize the influence of the fire damage on the seismic
response of an isolated connection, an exterior beam-column connection was analyzed under
different fire scenarios. The results of the study demonstrate the vulnerability of fire damaged
steel buildings during seismic events.

3. References
[1] FEMA-350. Recommendation and seismic design criteria for new steel moment-frame
buildings. Prepared by the SAC Joint Venture for the Federal Emergency Management
Agency, Washington, DC, 2000.

7
[2] FEMA-353. Recommended specifications and quality assurance guidelines for steel
moment- frame construction for seismic applications. Prepared by the SAC Joint Venture
for the Federal Emergency Management Agency, Washington, DC, 2000.

[3] FEMA-355B. State of the art report on welding and inspection. Prepared by the SAC
Joint Venture for the Federal Emergency Management Agency, Washington, DC, 2000.

[4] FEMA-355D. State of the art report on connection performance. Prepared by the SAC
Joint Venture for the Federal Emergency Management Agency, Washington, DC, 2000.

[5] Nakashima M, Suita K, Marisako K, Maruoka Y. Tests of welded beam-column


subassemblies. I. Global behavior. Journal of Structural Engineering 1998;124:1236-
1244.

[6] SAC/BD-99/23. Failure analysis of welded beam to column connections. by Barsom, JM,
Pellegrino JV.

[7] Uang CM, Yu QS, Noel S, Gross J. Cyclic testing of steel moment connections
rehabilitated with RBS or welded haunch. Journal of Structural Engineering
2000;126:57-68.

[8] Stojadinovic B, Goel SC, Lee KH, Margarian AG, Choi JH. Parametric tests on
unreinforced steel moment connections. Journal of Structural Engineering 2000;126:40-
49.

[9] SAC/BD-00/01. Parametric tests on unreinforced connections. by Lee KH, Stojandinovic


B, Goel SC, Margarian AG, Choi J, Wongkaew A, Reyher BP, Lee DY.

[10] SAC/BD-00/24. Development and evaluation of improved details for ductile welded
unreinforced flange connections. by Ricles JM, Mao C, Lu LW, Fisher JW.

8
[11] Lee D, Cotton SC, Hajjar JF, Dexter RJ. Cyclic behavior of steel moment-resisting
connections reinforced by alternative column stiffener details, I. Connection performance
and continuity plate detailing. Engineering Journal 2005;4th Quarter:189-213.

[12] AISC (2010a). Seismic provisions for structural steel buildings. American Institute of
Steel Construction 341, AISC 341-10.

[13] Tsai KC, Popov EP. Cyclic behavior of end-plate moment connections. Journal of
Structural Engineering 1990;116(11):2917-2930.

[14] Roeder CW, Douglus AF. Experimental results for seismic resistant steel moment frame
connections. Journal of Structural Engineering 1996;122(6):581-588.

[15] Engelhardt MD, Winneberger T, Zekany AJ, Polyraj TJ. Experimental investigation of
dogbone moment connections. Engineering Journal 1998;35:128-139.

[16] Adey BT, Grondin GY, Cheng JJR. Cyclic loading of end plate moment connections.
Canadian Journal of Civil Engineering 2000;27:683-701.

[17] Popov EP, Takhirov SM. Bolted large seismic steel beam-to-column connections part 1:
experimental study. Engineering Structures 2002;24:1523-1534.

[18] Ricles JM, Fisher JW, Lu LW, Kaufmann EJ. Development of improved welded
moment connections for earthquake-resistant design. Journal of Constructional Steel
Research 2002;58:565-604.

[19] Sumner EA, Murray TM. Behavior of extended end-plate moment connections subject
to cyclic loading. Journal of Structural Engineering 2002;128(4):501-508.

[20] AISC (2010b). Prequalified connections for special and intermediate steel moment
frames for seismic applications. American Institute of Steel Construction 358, AISC 358-
10.

9
[21] Stojadinovic B, Goel SC, Lee KH, Margarian AG, Choi JH. Parametric tests on
unreinforced steel moment connections. Journal of Structural Engineering 2000;126:40-
49.

[22] Suita K, Nakashima M, Morisako K. Tests of welded beam-column subassemblies II:


Detailed behavior. Journal of Structural Engineering 1998;124(11):1245-1252.

[23] Mazzolani FM. (Editor). Moment resistant connections of steel frames in seismic areas:
design and reliability. E&FN Spoon, 2000.

[24] Castiglioni CA. Effects of the loading history on the local buckling behavior and failure
mode of welded beam-to-column joints in moment-resisting steel frames. Journal of
Engineering Mechanics 2005;131:568-585.

[25] Han SW, Kwon GU, Moon KH. Cyclic behavior of post-Northridge WUF-B
connections. Journal of Constructional Steel Research 2009;63(3):365-374.

[26] Rodgers JE, Stephen AM. Effects of connection fractures on global behavior of steel
moment frames subjected to earthquakes. Journal of Structural Engineering
2006;132(1):78-88.

[27] Shi G, Shi Y, Wang Y. Behavior of end-plate moment connections under earthquake
loading. Engineering structures 2007;29(5):703-716.

[28] Guo B, Gu Q, Liu F. Experimental behavior of stiffened and unstiffened end-plate


connections under cyclic loading. Journal of Structural Engineering 2006;132(9):1352-
1357.

[29] Ghobarah A, Osman A, Korol RM. Behavior of extended end-plate connections under
cyclic loading. Engineering Structures 1990;12(1):15-27.

10
CHAPTER 2: AN IMPROVED UNSTIFFENED 8 BOLT EXTENDED
END PLATE MOMENT CONNECTION FOR SPECIAL AND
INTERMEDIATE MOMENT FRAMES
PART I: NUMERICAL MODELING AND MODIFIED DESIGN
DEVELOPMENT

Abstract
Eight bolt stiffened extended (8ES) end plate connection is one of the prequalified
connections proposed by AISC 358 Standards for special moment frames (SMFs) in seismic
regions. In 8ES connection, a gusset plate is welded between the end plate and the beam
flanges to strengthen the extended portion of the end plate. In addition to increasing the
flexural strength, it helps in forming plastic hinge away from the connection region as well as
distributing the flange forces evenly to the connecting bolts. In experimental studies, this
connection showed good performance in energy dissipation and ductility in many cases,
whereas premature fracture of beam flange or stiffener or weld between beam flange/end
plate and stiffener, leading to a brittle mode of failure of the connection have also been
observed due to high stress concentration. The study reported herein performed detailed
finite element analysis to understand the failure mechanism of the eight bolt
unstiffened/stiffened extended end plate moment connections, and thereby develop a
modified design of eight bolt extended end plate connection. In the modified design, the end
plate stiffeners were excluded from the connection to eliminate the stress concentration
associated with the stiffener, and the bolts were rearranged into a hexagonal pattern to ensure
uniform distribution of bolt forces. It was observed that the hexagonal rearrangement of the
bolts improves connection performance, but does not necessarily ensure formation of plastic
hinge away from the connection region for different beam sections. However, with a recent
idea of heat treating the beam flanges integrated into the connection design ensures formation
of the plastic hinge at the desired location regardless of the size of the beam. The modified
design with the new bolt pattern and heat treatment showed improvement in the seismic

11
response of the connection in terms of ductility and energy dissipation capacity. However, in
terms of strength degradation, no noticeable improvement was observed because of the
lateral instability of the connection. To delay the onset of strength degradation for the
modified connection, the web of the beam was strengthened by a stiffener to reinforce the
connection region. It was observed that the modified connection was able to sustain 6%
interstory drift angle without losing significant strength and also, exhibited very good
ductility and energy dissipation capacity under simulated seismic loading. Overall, the
seismic resilience of the extended end plate connection has been improved with the proposed
design. The novel connection design has been validated through full scale experimental
testing of extended end plate moment connections and has demonstrated excellent
performance under simulated seismic loading, which will be presented in a separate study.

Keywords: Extended end plate connection, Seismic performance, Hexagonal bolt


arrangement, Heat treated beam section, Beam web stiffener

1. Introduction
Before the 1994 Northridge Earthquake, welded unreinforced flange bolted web (WUF-
B) moment connections were used extensively in the construction of steel buildings in North
America. However, during the 1994 Northridge Earthquake many steel buildings were
seriously damaged at the WUF-B connections and several exhibited brittle failures [1]. Since
then a considerable research activities have been performed to find alternatives to the WUF-
B moment connections [1-8]. Based on a decade of research activities after the 1994
Northridge earthquake, five moment resisting connections were prequalified by AISC 358
Standards for special (SMFs) and intermediate (IMFs) moment frames [9]. One of these five
prequalified connections is extended end plate connection which is the subject of this study.
In extended end plate connections, a steel plate is shop welded to the end of a beam which is
then field bolted to the connectiing member. These connections are primarily used to connect
a beam to a column or to splice two beams together. The end plate is extended beyond the
beam flanges such that the bolts can be positioned outside the flanges on the extended part of

12
the end plate. Extended end plate connection can be stiffened or unstiffened. The stiffened
configuration has a gusset plate (stiffener) welded to the outside of the beam flange and to
the end plate. The stiffener is aligned with the web of the connecting beam to strengthen the
extended portion of the end plate. Extended end plate moment connections are further
classified based on the number and arranegement of the bolts. The two most common types
of the extended end plate connections prequalified by the AISC 358 Standards are four bolt
extended unstiffened and eight bolt extended stiffened end plate connections (Figure 1a, b).
The prequalification of these connections emanated from the study of Sumner and co-
workers [10-13], where they performed experimental investigation of the connections under
simulated seismic loading recommended by the Protocol for Fabrication, Inspection, Testing,
and Documentation of Beam-column Connection Tests and Other Experimental Specimens
[14]. The study reported herein dealt with the eight bolt configuration of the extended end
plate connection.

(a) Four bolt unstiffened (b) Eight bolt stiffened (c) Four bolt wideunstiffened

Figure 1 Different configurations of extended end plate moment connections.

The eight bolt configuration of the extended end plate connection is prequalified with a
stiffener welded between the end plate and the beam flanges commonly referred to as eight
bolt stiffened extended end plate connection or 8ES connection. The stiffener is usually
provided to strengthen the extended portion of the end plate. In addition to enhancing the

13
flexural strength, it helps in forming plastic hinge away from the connection region as well as
ensure uniform distribution of bolt forces. In many experimental studies, stiffened
connections showed good performance in energy dissipation and ductility [10-13, 15-23]
which demonstrated the applicability of the connection for seismic application. However,
there have been some experimental studies where undesirable mode of failure was observed
which was associated with the presence of the stiffener in the extended portion of the end
plate. In the studies by Ghobarah et al. [17], Korol et al. [16] and Guo et al. [24] crack was
observed at the toe of the weld between the beam and the end plate stiffener at large inelastic
loading cycles which was attributed to the stress concentration induced by the stiffener. In
the experiments performed by Adey et al. [13], the end plate ruptured around the beam
flange and stiffener. They also observed pullout of the stiffener in some of their experiments.
Shi et al. [25] observed fracture and buckling of the end plate stiffener during cyclic testing
of extended end plate connections. The observations made in these studies led to the notion
that the end plate stiffener imposes a high stress concentration at the interface between the
stiffener and the beam flange or end plate. As a result, the inelastic actions are localized at
these locations which may lead to fracture of beam flange, end plate or stiffener, fracture of
the welds between the stiffener and beam flange/end plate and local buckling of the stiffener.
In addition to the unfavorable failure modes caused by the stiffener, a portion of the stiffened
end plate may extend above the finished floor requiring a larger column closer and reduced
useable floor area [26]. Moreover, the cost of construction increases due to the weld works
required for attaching the stiffener to beam flange and end plate. These shortcomings of the
8ES connection induced by the presence of the stiffener motivate further investigation for
improving the resilience of the connection by removing the end plate stiffener.
In the works of Sumner and co-workers [10], a different configuration of the eight bolt
extended end plate connection was investigated, where four bolts were arraneged in a single
row at each of the beam flanges eliminating the necessity of end plate stiffener as shown in
Figure 1c. In their experiments, this type of connection was referred to as four bolt wide
unstiffened extended end plate connection or 4W connection. In the experiments, the 4W
connection exhibited non-ductile behavior due to an uneven distribution of flange forces to

14
the connecting bolts, which disqualified this connection for seismic applications. The uneven
distribution of the flange force between the bolts generated an unfavorable load path where
the flange forces attempted to reach only the interior bolts, hence the interior bolts ruptured
leading to tearing of end plate. The observations made from this study led to the conclusion
that bolt arrangement can be a critical parameter in the flange force distribution as well as
overall performance of the connection. Not much literature is available on the study of bolt
arrangement on the seismic performance of extended end plate moment connection.
Borgsmiller et al. [27] and Meng [28] showed that the connection geometry greatly
influences the load path to the bolts and for some configurations, half of the bolts are
ineffective in carrying load at all. Mays [29] performed parametric studies on the effect of the
pitch distances between the bolt and the beam flange and also, between bolt colums for four
bolt wide unstiffened extended end plate moment connection. It was observed that when the
outer bolts remained at least 6.35 mm inside the beam flange tip, all the bolts performed
effectively in distributing the flange forces. When the distance between the beam flange tip
and bolt increased, an unfavorable load path was generated with very high strains in the inner
bolts leading to bolt rupture type of failure. Adey et al. [21-23] observed greater flexibility
and energy dissipation capacity of both four bolt and eight bolt extended end plate
connections with and without stiffeners, when the distance between the bolt and beam flange
was increased from 2db to 3db (db-bolt diameter). They also noticed that with increase in the
pitch distance, the web to flange junction remained remote from the bolts which reduced the
streses at the web to flange junction. This could be beneficial because of the lack of fusion
that may result when weld access holes are not used may not be detrimental to the behavior
of the end plate connection. Gerami et al. [30] noticed an increased participation in the
connections components when the pitch distance was increased between the bolt and the
beam flange. They also observed change in the failure mode based on the arrangement of the
bolts for T-stub connection. These observations from the literature demonstrated that to get a
better ductility and energy dissipation from the connection, it is important to generate a
favorable load path so that the flange forces can be distributed evenly to the connecting bolts.

15
This study is motivated by these two findings from the literature-one is the problems
associated with the presence of the stiffener and the other is the influence of bolt arrangement
on the performance of the connection. In this study an attempt has been made to design an
eight bolt extended end plate connection without end plate stiffener to eliminate the stress
concentration, while the bolts have been rearranged to generate a favorable loading path
through even distribution of bolt forces. A detailed finite element analysis was performed on
extended end plate moment connections to understand the failure mechanisms for different
types of bolt arrangement and thereby, a hexagonal pattern of bolt arrangement has been
proposed such that all the bolts remain effective in distributing the flange forces equally.
Althoug the hexagonal bolt arranagement alone showed improved performance of the
connection for smaller size beam sections, it was obsereved that for larger size beam sections
plastic hinge forms very close to the connection region leading to very high strain demands
on the welds. Hence, a recent idea of heat treating the beam flanges proposed by Morrison et
al. [31] was implemented to ensure formation of plastic hinge at a desired location by
weakening the material strength. The modified bolt arrangement with heat treatment
eliminates the necessity of using end plate stiffener, and is applicable to all the beam sections
available. This connection showed improved performance in ductility and energy dissipation
capacity compared to the 8ES connection. Nonetheless, the strength degradation started
almost at the same interstory drift level for both the 8ES and modified connection due to the
lateral instability of the connection as a result of reduced strength of the material. This
indicates that the performance can be further improved if the onset of strength degradation
can be delayed which was necessitated by the local buckling of the beam flanges and web.
Therefore, the web of the beam was strengthened by welding a stiffener extended from the
end plate to the end of the heat treated region to reinforce the connection region. The
modified connection with heat treatment and web stiffener showed significant improvement
in the seismic response of the connection, especially the strength degrdation was significantly
delayed because of the presence of web stiffener.
It must be mentioned here that during the course of developing the modified design of
extended end plate connection in this study, Kiamanesh et al. [32] investigated a circular type

16
of bolt arrangement for extended end plate connection which is very similar to the type of
bolt arrangement proposed in this study. The circular bolt arrangement was proposed to
reduce the hysteresis pinching observed in extended end plate connection with rectangular
bolt pattern without end plate stiffener. It was shown that the circular bolt pattern reduces
hysteresis pinching noticeably for end plate connections with large bolts and thick end plates,
whereas for connections with small bolts and thin end plates, the influence of circular bolt
arrangement was insignificant. Although this study showed the promise of circular bolt
pattern on the cyclic response of extended end plate connections, the necessity of modified
bolt arranegement has not been addressed with proper comparison with the prequalified
connection, where an end plate stiffener is used to strengthen the extended portion of the end
plate. Moreover, the influence of circular bolt arrangment on the plastic hinge formation as
well as the applicability of the circular bolt pattern based on the size of the beam section has
not been addressed which should be a particular concern while designing a special moment
frame for seismic application. The study reported herein addresses the necessity for
developing the modified bolt arrangement with particular attention to the bolt force
distribution, plastic higne formation, energy dissipation, and ductility of the connection.
Further, the applicabiltiy of the modified design for different beam sections available has
been explained.

2. Finite Element Modeling of Extended End plate Connection


It has been shown that finite element methods can be used to predict the behavior of
extended end plate connections reasonably [28-30, 33-39]. However, to capture the behavior
of the extended end plate connection, it is important to model each individual component and
their interaction precisely. In this study advanced finite element modeling technique has been
employed to model each component of the connection and their interaction accurately. A
brief description of the finite element modeling scheme adopted in this study is presented
next.

17
2.1. Finite Element Discretization
Three dimensional nonlinear finite element models were developed for extended end
plate connections by using the finite element analysis software ANSYS. Both geometric and
material nonlinearities were incorporated in the finite element models. The beams and
columns were modeled and discretized by eight noded solid brick elements (SOLID 185).
This type of solid element is well suited for modeling 3D solid structures which can be
defined by eight nodes having three degrees of freedom at each node. The modeling of bolt
and end plate is very critical for accurate prediction of extended end plate connections
behavior because the rigid body motion in extended end plate connection is solely
constrained by the contact between the bolt and the end plate. In order to precisely model the
bolts and their interaction with the end plate, the solid bolt model was adopted in the finite
element simulation. Although there are other bolt models available in the literature such as
coupled bolt model, spider bolt model and no-bolt model, the solid bolt model is the most
realistic finite element model since it considers the influence of the bolt stiffness on
simulation, and has the provision of applying bolt pretension and generating contact between
the bolt and the flanges [40]. The end plate and the bolt were discretized by twenty noded
solid brick elements (SOLID 186) which exhibits quadratic displacement behavior. The
interaction between the end plate and column flange, bolt head and end plate, bolt nut and
column flange, and bolt shank and circular holes in the end plate and column flange are
intricate and the modeling of these interactions is one of the important features in the finite
element modeling. These interactions in the finite element modeling were defined by
incorporating small sliding contact surfaces. Surface-to-surface contact elements (CONTA
174) and target elements (TARGE 170) were used to generate the contact regions. The
contact elements themselves overlaid the solid elements describing the boundary of a
deformable body and were potentially in contact with the target surface defined by
TARGE170. This target surface is discretized by a set of target segment elements
(TARGE170) and is paired with its associated contact surface via a shared real constant set.
A penetration tolerance factor was applied in the direction of the surface normal to determine
if penetration compatibility is satisfied. The amount of penetration between contact and

18
target surfaces depends on the normal stiffness. Higher stiffness values decrease the amount
of penetration/slip, but can lead to ill-conditioning of the global stiffness matrix and to
convergence difficulties. Lower stiffness values can lead to a certain amount of
penetration/slip and produce an inaccurate solution. In the finite element modeling of the
extended end plate connection, frictional contact was generated by using penalty stiffness
with the penalty value of 0.1 so that the penetration/slip is acceptably small, but the problem
is well-behaved in terms of convergence. In static analysis, rigid body motion will be an
issue if the body is not sufficiently constrained. Since in extended end plate connection, the
rigid body motion will be constrained by only the presence of contact, it was ensured that the
contact pairs are in contact in the initial geometry i.e. the model was built in such a way that
initially the contact pairs are just touching. This was accomplished through the use of the
coefficients of initial contact closure and contact surface offset.

Figure 2 Finite element mesh of an eight bolt stiffened extended end plate connection.

Pretension in bolts and other structural components often have significant effect on
deflections and stresses and hence, in the finite element modeling, it is essential to include

19
the effect of bolt pretensioning. For the finite element modeling of extended end plate
connection, bolt pretension was modeled by using two ANSYS features, the PRETS179
pretension element and the PSMESH pretension meshing command. The pretension load was
applied as specified by the AISC LRFD Specification [41]. A finer mesh was considered
along the beam and column length in the vicinity of the beam-to-column connection, and also
in the column panel zone where majority of the inelastic action occurs. The bolts and end
plate was also discretized with finer mesh for accurate prediction of the connection behavior.
The other regions of the connection will remain elastic and hence, coarser mesh was used in
those places to reduce the computational time and cost. The mesh density was finalized based
on a convergence study which gave a finite element model of extended end plate connection
with 50569 elements and 71913 nodes. Figure 2 shows an example of the finite element mesh
generated for an eight bolt stiffened extended end plate connection.

2.2. Material Models


The prediction of experimental responses of moment connections using finite element
simulation is highly sensitive to the material models adopted and the parameters of the
adopted material models, especially under cyclic loading. In the literature, most of the works
related to the finite element modeling and analysis of extended end plate connection used
either bilinear or multilinear kinematic hardening material model in the finite element
simulation of the cyclic behavior of moment connections [29-30, 38-39]. Although both
bilinear and multilinear material models obey von-Mises yield criterion and include
Bauschinger effect, both the models are incapable of simulating ratcheting or shakedown,
which is essential for cyclic response simulation of moment connections. Both multilinear
and bilinear material models have a linear kinematic hardening rule and hence, are not
sufficient for simulating ratcheting or shakedown. To simulate ratcheting and shakedown in a
FE simulation, a metal plasticity model is required with a nonlinear kinematic hardening rule.
Nonlinear kinematic hardening implies a shift (or movement) of the yield surface along a
nonlinear path. The Chaboche material model [42], available in ANSYS Mechanical
software, offers such a nonlinear kinematic hardening rule. Since this study deals with cyclic

20
response simulation of extended end plate connections, the advanced nonlinear kinematic
hardening model of Chaboche [42] has been adopted in this study to define the material
characteristics. A brief description of the Chaboche [42] model is given below to demonstrate
different features of the model.

Table 1 Nonlinear kinematic hardening parameters for different steel materials.

Parameters ASTM ASTM ASTM ASTM Heat treated


A36 A572 Gr.50 A992 A490 ASTM A992
E (MPa) 186861 191505 199948 199948 199948
0 (MPa) 261.5 251.7 238.6 777.7 193.74
C1 (MPa) 119996.4 125415.7 383336.8 204615.8 1015116
C2 (MPa) 10873.1 28868.4 280782.1 152250.1 145075.9
C3 (MPa) 537.8 2675.2 50780 101318.5 59398.9
C4 (MPa) 68.9 144.8 1958.3 32219.2 372.2
1 1036 4585 21081.2 4143 60464.5
2 129 324 6256.2 285 5253.1
3 5 42 515 107 619
4 0 0 13.2 0 9.8

The yield criterion:


1/ 2
3
J2 ( ) s a s a 0 R ( p, q) , (1)
2
where is the stress tensor, is the current center of the yield surface in the total stress
space, sis the deviatoric stress tensor, a is the current yield surface center in the deviatoric
space, 0
is the initial size of the yield surface, and R represents the isotropic hardening
variable as a function of the accumulated plastic strain p and the size of the plastic strain
surface q. The rate-independent plastic strain increment will be calculated using the
associated flow rule:

p J2 3 s a
d d dp (2)
2 0 R
The superimposed kinematic hardening rule is given by:

21
4
da dai (3)
i 1

d ai 2 Cd p
a i dp (4)
3 i i

Each of the superposed kinematic hardening rules has a strain hardening term (1st term in Eq.
4) and a dynamic recovery term (2nd term in Eq. 4). These are important modeling features
for simulating cyclic responses of materials.

Figure 3 Experimental and simulated response of strain controlled hysteresis loop with
Chaboche model parameters. (a) ASTM A572 Gr.50 steel; (b) ASTM A36 steel; (c) ASTM
A992 steel.

22
Figure 4 Monotonic stress-strain curves for bolt and weld materials. (a) ASTM A490 bolt;
(b) E70 weld [2].

In the finite element modeling of the extended end plate connection, ASTM A992 and
A570 Gr.50 steels were used as the beam and column materials, whereas ASTM A36 was
used as the end plate material. Bolts were modeled as ASTM A490/A325 bolts. Use of
Chaboche model to simulate the material responses requires that the stable hysteresis loop of
the cyclic stressstrain response be known. In this case, stable hysteresis loops were obtained
from single amplitude strain controlled cyclic test of ASTM A992, ASTM A572 Gr.50 and
ASTM A36 steel coupons [43-44]. Chaboche parameters were determined from the up going
strain controlled experiment of ASTM A992, ASTM A572 Gr.50 and ASTM A36 steels. The
nonlinear kinematic hardening parameters (shown in Table 1) were determined by using a
genetic algorithm based optimization method. Simulations were made at the material level
with obtained parameters for validation. As shown in Figure 3, the simulated responses
harmonized very well with the experimental responses. The connection between the beam
flanges and end plate was modeled with groove welds using the flux core arc welding
procedure with E70TG-K2 electrodes. The material properties of E70 weld metal was
obtained from FEMA-355B [2] and it was implemented in the FE simulation through bilinear

23
material model as shown Figure 4b. For determining the material properties of ASTM A490
bolts, uniaxial test was performed on ASTM A490 bolts, and Chaboche model parameters
(shown in Table 1) were obtained through fitting the monotonic stress-strain curve (Figure
4a) by using genetic algorithm based optimization method.

2.3. Loads and Boundary Conditions


In the finite element modeling of the extended end plate moment connection, the boundary
conditions were applied such that it mimics the boundary conditions applied in the
experimental setup for conventional seismic testing of moment connections. A typical
experimental setup for full scale seismic testing of an exterior beam-column assembly used
by Morrison et al. [31] is shown in Figure 5a and this configuration of the experimental
setup was adopted for the finite element analysis of extended end plate connections. In the
finite element modeling, the nodes at the two ends of the column were hinged i.e. x, y and z-
direction of the nodal displacements were zeroed.

(a)

Figure 5 (a) A typical experimental setup for seismic testing of moment connections [31];
(b) SAC loading protocol [14].

24
Lateral supports were simulated as lateral restraint with z-direction nodal translation
zeroed at their point of contact with the beam. Vertical displacement was applied at all the
central nodes located on the stiffener plate at the free end of the beam so as to simulate the
displacement controlled actuator loading during the actual test. All beam-column assembly
was analyzed with a displacement-based artificial loading protocol offered by SAC/BD-
97/02 [14]. The displacement based time history was based on the SAC Test Protocol and
consisted of imposing a series of prescribed quasi-static cyclic displacements to the end of
the beam [14]. The prescribed displacements included six cycles of 0.375, 0.5 and 0.75%
interstory drift, followed by four cycles of 1% interstory drift and two cycles each of 1.5, 2,
3, 4, 5, 6% and so forth interstory drift [14]. A schematic of the loading protocol is shown in
Figure 5b.

3. Validation of Finite Element Model


The first step of the finite element analysis was to validate the precision of the finite
element model in predicting the cyclic response of extended end plate moment connections.
This was achieved by comparing finite element analysis (FEA) results against experimental
responses. In this study, the finite element model was validated against the experimental
response of the eight bolt stiffened extended end plate connection (8ES-1.25-1.75-30) of
Sumner et al. [10]. Figure 6 shows comparison of simulated and experimental moment-
rotation response of 8ES-1.25-1.75-30 specimen tested by Sumner et al. [10]. It was
observed that the simulated hysteresis loop shapes accorded very well with the experimental
hysteresis loop shapes. The peak moments in each cycle and the initiation of strength
degradation in the experimental responses was captured precisely in the finite element
simulation. The global and local buckling phenomena obserevd in the experiments were also
simulated very well by the finite element model as shown in Figure 7, where the simulated
deformed shape of the specimen is compared with the experimental deformed shape showing
plastic hinge foramtion at the base of the stiffener at the end of loading cycles corresponding
to 6% interstory drift angle. The resemblance of the simulated responses with those of the

25
experimetal responses validates the adequacy of the developed finite element model for
predicting the seismic response of extended end plate moment connections.

Figure 6 Comparison between experimental and simulated moment-rotation response of


eight bolt stiffened extended end plate connection (8ES-1.25-1.75-30) tested by Sumner et al.
[10].

(a) Experiment (b) FE Simulation

Figure 7 Comparison of experimental and simulated deformed shape for 8ES-1.25-1.75-30


specimen of Sumner et al. [10] showing formation of plastic hinge at the base of the stiffener.

26
4. Finite Element Analysis of Extended End plate Connections
In this part of the study, detailed finite element analysis was performed on eight bolt
extended end plate connections with different bolt arrangements in order to investigate the
failure mechanisms involved and come up with an optimized bolt arrangement which does
not require end plate stiffener, yet improves the seismic performance of the connection. An
exterior sub-assemblage was chosen consisting of a W14193 (ASTM A992) column with a
single W3099 (ASTM A992) beam attached to the flange. This is the connection setup that
was used by Sumner et al. [10] as discussed in the validation part of the study, except the fact
that in the experiments, the beam and column material was ASTM A572 Gr.50. The beam-
to-column connection was designed to develop 110% of the nominal plastic moment capacity
of the beam as recommended by AISC 358 Standard [9]. The following four types of
connection configurations were considered consisting of eight bolts at each flange with
different arrangements of bolts.
i. Eight bolt stiffened connection (8ES): four rows of two bolts at each flange, two rows
above and two rows below the flange with end plate stiffener (Sumner et al. [10]).
ii. Eight bolt unstiffened connection (8E): four rows of two bolts at each flange, two
rows above and two rows below the flange without end plate stiffener.
iii. Four bolt wide unstiffened connection (4W): two rows of four bolts at each flange,
one row above and one row below the flange without end plate stiffener (Sumner et
al. [10]).
iv. Modified eight bolt unstiffened connection (8EM): four rows of two bolts at each
flange, two rows above and two rows below the flange without end plate stiffener
with hexagonal bolt pattern.
The end plate material used was ASTM A36 which was bolted to the column flange with
ASTM A490/A325 bolts. The bolts were given a pretension as specified in the AISC LRFD
specification [41]. The column was reinforced with 19.05 mm continuity plates and a 9.525
mm doubler plate. Lateral supports were provided for the beam at a distance of 1.25 m, 2.46
m and 5.72 m from the centerline of the column [10]. No lateral supports were provided for
the column. The loading was applied to the beam at a distance of 6.13 m from the centerline

27
of the column. All the connections were analyzed under displacement controlled cyclic
loading in accordance with the SAC loading protocol [14] as shown in Figure 5b. The results
of the finite element analysis on the above mentioned four types of connection configurations
are presented next.

4.1. Eight Bolt Stiffened Extended End plate Connection (8ES)


The conventional eight bolt stiffened extended end plate connection (8ES) was designed
according to the AISC [9] design requirements for SMFs in the regions of moderate to high
seismicity. The connection had an end plate thickness of 44.45 mm (ASTM A36) which was
bolted to the column flange with 31.75 mm diameter ASTM A490 bolts. The bolts were
arranged in four rows of two bolts at each flange, two rows above and two rows below the
flange. There was a 12.7 mm thick end plate stiffener placed at the extended portion of the
end plate which was welded to the end plate and the beam flange. The material properties
used for end plate stiffener was ASTM A36. The finite element model was developed
mimicking the test configuration of Sumner et al. [10] which was then subjected to simulated
earthquake loading history upto 6% interstory drift angle to evaluate its seismic performance.

(a) (b)

Figure 8 Eight bolt stiffened extended end plate connection (8ES). (a) FE mesh; (b)
Equivalent plastic strain contour showing formation of plastic hinge after cyclic loading upto
6% interstory drift angle.

28
Figure 9 Cyclic response of 8ES configuration of connection. (a) Moment-rotation hysteresis
response for cyclic loading upto 6% interstory drift angle; (b) Contribution of different
components of the connection in the total rotation.

The finite element mesh of the connection is shown in Figure 8a. In the discussion, this
configuration of the connection has been denoted as 8ES followed by the beam size
analyzed. The connection showed ductile behavior by forming plastic hinge at a distance of
492 mm away from the face of the column which was at the base of the end plate stiffener as
can be seen in Figure 8b. The moment-rotation plot of the connection (shown in Figure 9a)
showed initiation of strength degradation in the 2nd cycle of 3% interstory drift angle which
became significant (30%) during loading cycles for 4% interstory drift angle. As the loading
cycles progressed, the strength of the connection dropped by 40% at 6% interstory drift
angle, because of the instability created due to the lateral buckling of the beam. The
contributions of the beam, panel zone, column and end plate to the maximum interstory drift
angle of 6% were: b=5.51%, pz=0.07% c=0.37% ep=0.05% (shown in Figure 9b). It was
obvious that beam played the most dominant role in carrying most of the rotation capacity of
the connection by forming plastic hinge away from the face of the column. The participation
of the other components of the connection such as end plate, column and panel zone was

29
negligible in comparison to beam, especially the end plates contribution in the rotation
capacity of the connection was minimal. This was expected as the connection was designed
for 110% of the nominal plastic moment capacity of the beam and the presence of end plate
stiffener made the connection more rigid which attenuated the participation of the end plate.

Figure 10 Bolt strains for 8ES configuration during loading cycles at the top flange of the
beam.

It is to be noted here that the contributions from panel zone, column and end plate in the
rotation capacity of the connection started to decrease when the beam started to buckle

30
locally, indicating an increase in the beams participation in the rotation capacity of the
connection. The bolt strains were in the elastic range and a gradual decrease in the bolt
strains was observed during cycling (Figure 10). The bolt strains cycled within the pretension
strain range and it was noticeable that the bolt strains in the outer and inner bolts for both top
and bottom rows of bolts at each flange was uniform. This was achieved through the
presence of the end plate stiffener which generated a favorable load path for the flanges to
distribute the forces to the bolts evenly. Throughout the loading cycles, the axial strains in
both the top and bottom flange groove weld never crossed 1% which shows lower strain
demands on the welds of the beam. The axial strain profile along the width of the beam
bottom flange at the weld toe is shown in Figure 22a for loading cycle at the end of 4%
interstory drift angle.

4.2. Eight Bolt Unstiffened Extended End plate Connection (8E)


For the sake of comparison, the end plate stiffener in 8ES connection was removed and
the connection was analyzed under simulated earthquake loading history to understand the
influence of the stiffener on the performance of the connection. In the discussion, this
configuration of the connection has been denoted as 8E followed by the beam size analyzed.
The finite element mesh and the equivalent plastic strain contour for 8E configuration of the
connection are shown in Figure 11. The 8E configuration also showed global and local
buckling of the beam flanges leading to formation of plastic hinge (Figure 11b), but unlike
8ES connection, the plastic hinge formed close to the column (100 mm) as a result of the
removal of end plate stiffener. Consequently, the axial strains at the toe of the weld was
much higher compared to the 8ES configuration. The weld strain profile shown in Figure 12
showed that the axial strain reached 5% during loding cycles at 5% interstory drift angle
leading to around 8% strain in the bottom beam flange weld during 6% interstory drift angle.
This imposes a large strain demand on the weld, which might lead to brittle weld failure of
the connection. Moreover, two bolt inner rows are relatively closer to the beam flanges
compared to the outer rows of bolts, and since the end plate stiffener has been removed, the
strain distribution in the bolts was not uniform unlike 8ES configuration of the connection.

31
As shown in Figure 13, the inner bolt strains were much higher than the strains in the outer
bolts indicating an uneven distribution of the flange forces to the bolts.

(a) (b)

Figure 11 Eight bolt unstiffened extended end plate connection (8E). (a) FE mesh; (b)
Equivalent plastic strain contour showing formation of plastic hinge after cyclic loading upto
6% interstory drift angle.

Figure 12 Axial strain profile along the width of beam flange at the weld toe for 8E
configuration of the connection.

32
Figure 13 Bolt strains for 8E configuration during loading cycles at the top flange of the
beam.

In other words, the outer bolts were not as effective as the inner bolts in carrying the
flange forces. However, the strains in both the inner and outer bolts were in the elastic range
during the loading cycles. For outer bolts, the strain in the bolts cycled within the pretension
strain range with a gradual decrease in the bolt strain during cycling. On the other hand, for
inner bolts, the bolt strains were observed to cycle from the initial pretension strain upto
strains of approximately 40% above the pretension strain. The finite element analysis results

33
of the 8E configuration of the connection revealed the consequence of removal of end plate
stiffener which showed uneven bolt force distribution and high strain demands on the weld,
which may lead to brittle failure of the connection.

4.3. Four Bolt Wide Unstiffened Extended End plate Connection (4W)
Similar to 8ES and 8E configuration of the connection, the four bolt wide unstiffened
extended end plate connection (4W) utilizes eight bolts at each flange for connecting the
beam to the column, but the bolts are arranged in two rows instead of four rows unlike 8ES
and 8E connections. Each row has four bolts, one row above and one row below each flange
of the beam. This configuration of the connection is particularly suitable for wide flange W
shapes since it requires a larger width of the end plate compared to 8E and 8ES connection.
On the contrary, arranging bolts in single row reduces the depth of the end plate. As already
mentioned, this type of connection configuration was tested by Sumner et al. [10], and the
connection developed unfavorable loading path which led to non-ductile failure of the
connection through interior bolt rupture and end plate tearing. Figure 14 shows the FE mesh
developed in this study for 4W connection and the equivalent plastic strain contour of 4W
connection when subjected to the simulated earthquake loading history. The connection was
not able to sustain more than 3% interstory drift angle and the bolt strains became so high
that the finite element model suffered numerical instability. Initial yielding of the specimen
was observed during loading cycles at 1% interstory drift angle in the beam flanges close to
the weld. During load cycles for 2% interstory drift angle, additional yielding of the beam
flanges was observed with small amount of yielding of beam web, panel zone and end plate.
During cycling in 3% interstory drift angle, severe yielding of the beam flanges was observed
close to the welds. At this point, the panel zone and end plate have yielded significantly. At
the last cycle of 3% interstory drift angle, the two inner bolts in the inside face of the beam
bottom flange showed very high strains which might lead to rupture of the bolts (Figure 14b).
The same bolts ruptured in the tests performed by Sumner et al. [10]. In the test, the end plate
tearing was observed very close to the weld region of the beam flanges. In the FE simulation,
very high strain was observed at the same location which shows sign of possible failure of

34
end plate at that location. Figure 16 shows the equivalent plastic strain contour plots of end
plate at 3% interstory drift angle.

(a) (b)

Figure 14 Four bolt wide unstiffened extended end plate connection (4W). (a) FE mesh; (b)
Equivalent plastic strain contour showing bolt failure and high weld strain at 3% interstory
drift angle.

(a)

Figure 15 (a) Equivalent plastic strain contour of end plate at the end of 3% interstory drift
angle; (b) Moment-rotation hysteresis response of 4W connection.

35
Figure 16 Bolt strains for 4W configuration during loading cycles at the top flange of the
beam.

It is quite apparent that end plate had yielded significantly close to the beam flanges
which might lead to tearing of the end plate. There was no reduction in the strength of the
connection uptp 3% interstory drift angle (Figure 16b) since it did not show any sign of local
flange or web buckling. The bolts strain plots in Figure 16 show that the outside bolts in a
row was not as effective as the inside bolts i.e. the flange forces were not distributed to all the
bolts uniformly. The inside bolts on both the outer and inner rows showed very high strains

36
compared to the outside bolt strains. This uneven distribution of the flange forces to the
connecting bolts generated an unfavorable loading path which may lead to premature bolt
rupture and end plate tearing although a considerable amount of energy can be dissipated. In
addition to high strains in the bolts and the end plate, the beam flanges yielded significantly
very close to the groove weld, and once the bolt in the bottom flange experienced very high
strains during the 2nd cycle of 3% interstory drift angle, there is a sudden jump in the axial
strains (around 10% at the bottom flange as shown in Figure 17) at the weld toe which may
lead to brittle weld failure of the connection. It is apparent that the 4W configuration of the
connection is not suitable for seismic application because of the unfavorable load path
generated during cyclic loading leading to premature bolt rupture, end plate tearing and very
high weld strain demand.

Figure 17 Axial strain profile along the width of beam flange at the weld toe for 4W
configuration of the connection.

4.4. Eight Bolt Unstiffened Modified Extended End plate Connection (8EM)
From the finite element response and discussion in the preceding sections, it is quite
obvious that 8E and 4W configurations of extended end plate connections showed
unfavorable seismic behavior which is not suitable for SMFs. The distribution of flange

37
forces was uneven in the connecting bolts, the weld showed higher strain demands, and
considerable amount of yielding of the end plate was observed. These shortcomings of the
connections necessitate the importance of end plate stiffener in the conventional design of
eight bolt stiffened extended end plate connections. However, as observed in some previous
studies, the presence of the stiffener can degrade performance of the connection and result in
a non-desirable failure mode such as buckling and fracture of the triangular stiffener, fracture
of the beam flanges, and fracture of the weld between the stiffener and beam flange/end plate
[16, 18, 21, 24-25]. In addition to producing undesirable failure modes, the inclusion of the
end plate stiffener reduces the usable floor area as well as increases the construction cost
because of the CJP groove weld that is required to attach the stiffener to the beam flanges
and end plate.

(a) (b) (c)


Figure 18 (a) Proposed bolt rearrangement in a hexagonal pattern; (b) FE mesh with the
modified bolt arrangement for 8EM configuration of the connection; (c) Equivalent plastic
strain contour of 8EM configuration of the connection showing formation of plastic hinge.

To eliminate the uncertainty associated with the triangular stiffeners and to allow for an
equal distribution of demands in the bolts, a connection with a hexagonal-type bolt pattern
was proposed in this study. This pattern was adjusted to allow for a balanced demand
distribution amongst the bolts without the need for a triangular stiffener. Through finite
element analysis utilizing trial and error convergence until optimized performance was

38
found, an angle of 37.5 between the top and bottom layers of bolts was found (See Figure
18a). This angle allows equal participation amongst the bolts without the need of a triangular
stiffener. The eight bolt extended end plate connection with this modified configuration of
the bolt was analyzed under the simulated earthquake loading history to investigate its
seismic performance and compare with the response of the stiffened connection. The FE
mesh and the equivalent plastic strain contour of the connection is shown in Figure 18 which
shows ductile behavior of the connection by formation of plastic hinge at a distance of 286
mm away from the face of the column. The connection showed almost the same response as
the stiffened connection. The moment-rotation plot of the connection (shown in Figure 19a)
showed initiation of strength degradation in the 2nd cycle of 4% interstory drift angle which
became significant (34%) during loading cycles for 5% interstory drift angle. As the loading
cycles progressed, the strength of the connection dropped by 56% at 6% interstory drift
angle, because of the instability created due to the lateral buckling of the beam.

Figure 19 Cyclic response of 8EM configuration of connection. (a) Moment-rotation


hysteresis response for cyclic loading upto 6% interstory drift angle; (b) Contribution of
different components of the connection in the total rotation.

39
The contributions of the beam, panel zone, column and end plate to the maximum
interstory drift angle of 6% were: b=5.55%, pz=0.26% c=0.1% ep=0.08% (shown in
Figure 19b). This is almost similar type of response that was obserevd for 8ES connection.
The beam took most of the inelastic action where more than 90% of the rotation capacity
came from the beam itself. On the other hand, the participation of the end plate and panel
zone increased in compariosn to 8ES connection. It is to be noted here that the contributions
from panel-zone, column and end plate in the rotation capacity of the connection started to
decrease when the beam started to buckle locally, indicating an increase in the beams
participation in the rotation capacity of the connection.
Moving on to the bolt strains, the hexagonal pattern of bolt arrangement was successful
in distributing the flange forces uniformly to the connecting bolts. As seen in Figure 20, the
strains in all the connecting bolts showed similar trend during the loading cycles. The bolt
strains remained in the elastic range and the strains cycled within the pretension strain range
with a gradual decrease of pretension strain with progression of loading cycles. For 8E and
4W configurations of the connections, the weld strain demand increased significantly due to
the absence of the end plate stiffener. However, for 8EM configuration of the connection, the
weld strain remained less than 2% during the loading cycles. The axial strain profile along
the width of the beam bottom flange at the weld toe for 8EM configuration of the connection
is shown Figure 22a for loading cycles upto 4% interstory drift angle which shows lower
strain demands on the welds unlike 8E or 4W connections. After 4% interstory drift angle,
the beam experienced local buckling which reduced the strain demand at the weld toe.
A comparison of the moment-rotation envelopes of the extended end plate connections
analyzed is presented in Figure 21. It can be observed that the connection with the modified
bolt arrangement showed better performance than the other configurations of the connection
in terms of strength degradation and energy dissipation. For 8EM configuration, the
connection started to loose strength after 4% interstory drift angle, whereas the connection
with stiffener (8ES) started to loose strength after 3% interstory drift angle. The 8E and 4W
connections also showed good performance in terms of strength degradation, but they
suffered with higher weld strains and non-uniform bolt strains which eventually made their

40
performance undesirable for seismic application. Upto 5% interstory drift angle, 8ES, 8E and
8EM showed similar strength; however, the degradation of 8EM was higher than the others
for loading at 6% interstory drift angle.

Figure 20 Bolt strains for 8EM configuration during loading cycles at the top flange of the
beam.

Figure 22b shows the axial strain profile along the width of beam bottom flange at the
location of plastic hinge for different configurations of the connection at the end of loading
cycles for 6% interstory drift angle. It is clearly visible that for 8ES configuration, there is a

41
significant increase in the strain demand near the center of the beam flange at the plastic
hinge location as a result of high stress concentration due to the presence of end plate
stiffener. This higher demand on the beam flanges were attenuated for 8EM configuration by
removing the stiffener which reduced the strain demand to 23% from 57% for 8ES
connection. This will allow the connection to sustain more story drift levels without fracture
of the flange. On the contrary, the maximum strain demand at the weld toe for both 8ES and
8EM configuration of the connection was low compared to the other configuration of the
connection (Figure 22a). In addition, the 8EM configuration increased the contributions of
end plate and panel-zone in the total rotation capacity of the connection compared to 8ES
configuration as shown in Figure 23. This will help in dissipating more energy.

Figure 21 Moment-rotation envelopes of the extended end plate connections for different
bolt arrangements.

To conclude, the modified bolt arrangement can be a viable alternative to the


conventional eight bolt stiffened extended end plate connection which eliminates the stress
concentration on the beam flanges producing better seismic response than the stiffened
connection. Moreover, it ensures better utilization of the floor space and reduces the overall

42
construction cost by excluding the weld works for the end plate stiffeners. However, the
analysis results presented to this point was based on only one size of the beam (W3099) that
was used by Sumner et al. [10]. More analyses with different sizes of the beam need to be
investigated to generalize the use of the hexagonal pattern of the bolt. The analysis results
presented hitherto manifests that the stiffener in the eight bolt stiffened extended end plate
connection can be removed with proper arrangement of the bolts to generate a favorable
loading path so that the flange forces are distributed evenly to all the connecting bolts,
thereby ensuring a ductile mode of failure through inelastic buckling of the beam.

Figure 22 Axial strain profile along the width of the beam bottom flange for different
arrangements of the bolts. (a) At weld toe during loading cycles at 4% interstory drift angle;
(b) At plastic hinge location during loading cycles at 6% interstory drift angle.

5. Further Development of the Modified Design for Extended End plate Connection
In order to investigate the applicability of the modified bolt arrangement for larger beam
section, finite element analysis was performed on eight bolt extended end plate connection
with modified bolt arrangement, where a W30148 beam section was connected to a
W14257 column section. The connection was designed for 110% plastic moment capacity

43
of the beam. The stiffener, bolt and end plate were designed based on the requirements of
AISC [45] for seismic application. The beam, column material used was ASTM A992 and
the end plate material used was ASTM A36. The design required an end plate thickness of
around 47.625 mm which was bolted to the flange of the column with 38.1 mm diameter
A490 bolts. The bolts were given a pretension of 658 kN as specified in the AISC LRFD
specification. The column was reinforced with 25.4 mm continuity plates and a 25.4 mm
doubler plate. This connection was analyzed under the simulated seismic loading as shown in
Figure 5.

Figure 23 Participation of end plate and panel zone in the total rotation capacity at 6%
interstory drift angle for different configurations of bolt. (a) End plate; (b) Panel zone.

The equivalent plastic strain contour and the moment-rotation response for the eight bolt
modified extended end plate connection (8EM) with W30148 beam section is shown in
Figure 24a. The connection showed local and global buckling of the beam flanges and web
leading to formation of plastic hinge at a distance of 145 mm away from the end plate. As
expected the bolt forces were uniformly distributed as a result of the modified bolt
arrangement. However, since the plastic hinge formed very close to the weld connection, the

44
strain demand on the welds of beam flanges were very high. It was observed that the
maximum weld strain was around 5% for loading cycles at 4% interstory drift angle which
crossed 10% for loading cycles at 5% and 6% interstory drift angles. Hence, this
configuration of the connection might show undesirable mode of brittle failure at the weld
between the end plate and the beam flanges.

(a)

Figure 24 8EM configuration of the connection with W30148 beam section. (a) Equivalent
plastic strain contour showing formation of plastic hinge; (b) Axial strain profile along the
width of the beam flange for loading cycles upto 6% interstory drift angle.

From the results of the analysis, it is apparent that for larger size beam sections, the
rearrangement of the bolts in hexagonal pattern alone may not be sufficient to ensure ductile
mode of failure. Hence, the flanges of the beam section were heat treated at a specific
distance away from the face of the column to reduce the strength of the beam. This
innovative seismic performance enhancement technique (known as heat treated beam section
or HBS) has been developed by Morrison et al. [31] and it has been implemented for welded
unreinforced flange welded web (WUF-W) moment resisting connection successfully as a
part of the NEES (Network for Earthquake Engineering Simulation) project. In this type of
connection, the flanges of the beam near the connection region are heat treated (highlighted

45
red in Figure 25a) to a peak temperature (Tp) of 1050C; maintain the peak temperature for
15 minutes (th) before slow air cooling as shown in Figure 25b. For detail of the heat
treatment technique and the seismic performance of WUF-W connections with heat treated
beam section (HBS), the readers are referred to Morrison et al. [31]. However, it was shown
that the heat treatment reduced the strength of steel in the heat treated areas of the flange and
consequently, under seismic loading plastic hinge developed at the heat treated region of the
beam flanges and required ductility capacity of connections was achieved.

HBS

(a) Exterior connection with heat treated section (b) Heat treatment temperature

Figure 25 Heat treatment of modified extended end plate connection for seismic
performance enhancement.

During heat treating the beam flanges, it was observed that the temperature was not
uniform in the HBS region. The temperature was relatively uniform in the center of the HBS
region (1050C), whereas at the edges of the HBS region the temperature was found to be
around 850C. On the other hand, just outside of the HBS region, the temperature was found
to be in the range of 500-600C. A study by Morrison [46] on the seismic response
simulation of WUF-W HBS connections showed that the response of the connection during
inelastic loading cycles do not show significant difference whether temperature non-
uniformity in the HBS region is considered or not. Similar observations were made in this
study. Material tests results [47] also showed that the stress-strain response of ASTM A992

46
steel upto 600C is very close to the stress-strain response at room temperature. Hence,
instead of taking multiple material properties of ASTM A992 steel for the HBS region,
material properties of ASTM A992 steel at 1050C was used for HBS region, whereas room
temperature ASTM steel properties were used for the rest of the connection region and other
components of the connection. This reduces the material parameters to be determined for
each temperature range as well as the complexity of the model. The readers are referred to
[47] for detailed material response of ASTM A992 at different heat treatment temperature.

Figure 26 (a) Monotonic stress-strain response for unheat treated and heat treated ASTM
A992 specimens; (b) Experimental and simulated response of strain controlled hysteresis
loop with Chaboche parameters for heat treated ASTM A992 steel.

To implement heat treated beam flange in the finite element modeling, the heat treated
material properties of ASTM A992 steel was incorporated in the specified location of the
beam flanges. For determining material properties for heat treated ASTM A992 steel, a set of
monotonic tension and cyclic tests of beam flange steel coupons (ASTM A992), heat treated
by 1050C with a 15 minutes hold were conducted. The stress-strain response from one set of
the monotonic tension tests are shown in Figure 26a. It was observed that with heat treatment
both the yield and ultimate stresses of the beam steel decreased by downward shift of the

47
stress-strain curves. The reduction in yield stress by heat treatment with maximum

temperature 1050 C was about 40%. It is also important to note in Figure 26a that the elastic

modulus of the beam flange steel was not altered much by the heat treatment process. For
Chaboche model parameter determination for heat treated ASTM A992 steel (see Table 1),
the upgoing curve of the stable hysteresis loop obtained from single amplitude, strain-
controlled cycle was fitted. Simulations were made at the material level with obtained
parameters for validation and as shown in Figure 26b, the simulated responses matched very
well with the experimental responses.

(a)

Figure 27 Cyclic response of 8EMH configuration of the connection with W30148 beam
section. (a) Equivalent plastic strain contour showing formation of plastic hinge in the heat
treated region of the beam flanges; (b) Moment-rotation response upto 6% interstory drift
angle.

Once the material properties were determined, the 8EM connection was analyzed under
the simulated seismic loading with heat treated beam flanges. This configuration of the
connection has been denoted as 8EMH connection. The heat treatment was performed at a
distance of 127 mm away for the end plate for a length of 635 mm along the beam length.

48
The equivalent plastic strain contour along with the moment-rotation response of the
connection is shown in Figure 27. It can be observed that the plastic hinge formed right at the
heat treated beam flanges. The connections strength was lowered because of the heat
treatment which reduced the strength of the material. The strength of the connection started
to degrade at loading cycles for 4% interstory drift angle, where significant buckling of the
beam flanges were observed in addition to small web buckling. For loading cycles at 5% and
6% interstory drift angles, severe buckling of the beam flanges and web led to lateral
torsional buckling of the beam and the strength of the connection deteriorated by around 40%
at the end of 6% interstory drift angle. Most of the inelastic action occurred on the beam
flanges in heat treated region. In the total rotation capacity, beam played the most dominant
role as anticipated with small contributions from the end plate and panel zone as shown in
Figure 28b. It is very interesting to note here that the heat treatment reduced the strain
demand on the weld significantly.

Figure 28 Cyclic response of the connection for 8EMH configuration of the bolts. (a) Axial
strain profile along the beam flange at the weld toe; (b) Participation of different components
of the connection in the total rotation capacity of the connection.

49
Figure 29 Bolt strains for 8EMH configuration of the connection with W30148 beam
section.

The axial strains at the weld toe never crossed 2% upto interstory drift angle of 6% as
shown in Figure 28a. Moreover, both the outside and inside bolts showed equal and similar
pattern of strains indicating a better distribution of flange forces to the connecting bolts
(Figure 29). Overall, the required ductility and the energy dissipation capacity of the
connection was achieved. A comparison of the moment-rotation envelope of the peak
moments in each loading cycle for 8ES and 8EMH connection (Figure 34a) showed that the
strength degradation for 8ES connection was very steep compared to 8EMH connection,

50
which indicates a more ductile and stable response of the connection for 8EMH
configuration. Since the plastic hinge location with the modified bolt arrangement varies with
the size of the beam section, heat treatment of the beam flanges ensures formation of plastic
hinge at a specified location regardless of the size of the beam. In this way the proposed
modified connection design with hexagonal bolt arrangement and HBS can be implemented
as a unified design for all the beam sections available in the AISC LRFD design manual.
Although 8EMH connection showed improvement in the ductility and energy dissipation
capacity of the connection compared to 8ES connection, no significant improvement was
observed in terms of strength degradation. As can be noticed from Figure 34b that for both
the 8ES and 8EMH connections, the strength degradation initiated at the same interstory drift
angle, although the strength degradation was much slower for 8EMH connection compared
to 8ES connection. Hence, the strength degradation mechanism of the connection was
investigated to study whether the seismic performance of the modified connection can be
further improved by delaying the onset of strength degradation. The local flange and web
buckling were plotted for each cycle of loading to explore the mechanism involved in
strength degradation. The local buckling modes during the onset of strength degradation are
shown in Figure 30.

Figure 30 Local buckling of beam web and flanges at different loading cycles for 8EMH
configuration of the connection with W30148 beam section.

51
It was observed that beam web started to buckle at the first cycle of 3% story drift
followed by beam flange buckling, but no strength degradation was observed. At the second
cycle of 3% storey drift, the web of the beam started to twist which initiated lateral-torsional
buckling of the beam and this in turn initiated strength degradation of the connection. As the
load amplitude was further increased, the lateral torsional buckling became more pronounced
and strength degradation increased significantly. Similar observations were also made by [16,
18, 48-52] where they observed that strength degradation is caused by either lateral buckling
or lateral torsional buckling of the beam. Since lateral torsional buckling for the 8EMH
connection under consideration was initiated by the web buckling of the beam, incorporating
a web stiffener may delay the strength degradation of the connection by delaying the
initiation of web buckling as well as lateral torsional buckling.

(a) HBS (b)

Web Stiffener

Figure 31 (a) Modified extended end plate connection (8EMH-W) with heat treatment and
web stiffener; (b) Formation of plastic hinge in the heat treated region of beam flanges under
simulated seismic loading with 8EMH-W configuration of the connection.

A finite element analysis was performed on the 8EMH connection by incorporating a


web stiffener on each side of the beam web. This connection configuration is denoted as
8EMH-W. The web stiffener consisted of a 6.35 mm thick plate welded to the beam web and
extending from the face of the end plate to the end of the heat treated region as shown in

52
Figure 31a. It was observed from the analysis responses that the plastic hinge still formed at
the heat treated beam flanges (Figure 31b) and there was no noticeable strength degradation
upto 6% interstory drift angle (Figure 32a). The axial strain at the weld toe was less than 2%
during the loading cycles (Figure 32b) since most of the inelastic action was taking place at
the heat treated beam region. Moreover, the connecting bolts were equally effective in
distributing the flange forces as a result of the modified bolt arrangement (Figure 33).

Figure 32 Cyclic response of 8EMH-W configuration of the connection upto 6% interstory


drift angle. (a) Moment-rotation hysteresis response; (b) Axial strain profile along beam
flange at the weld toe.

A comparison of the moment-rotation envelope of the 8ES, 8EMH and 8EMH-W


connection with W30148 beam section is shown in Figure 34a. It is evident that 8EMH-W
connection shows a marked improvement in the ductility and energy dissipation capacity
compared to the 8ES and 8EMH connections. However, the peak strength of the connection
was lower than the 8ES connection because of the reduced strength of the material due to
heat treatment, whereas the peak strength of the connection is higher compared to the 8EMH
connection because of the presence of the web stiffener.

53
Figure 33 Bolt strains for 8EMH-W configuration of the connection with W30148 beam
section.

The modified design of the extended end plate connection with heat treatment and web
stiffener was also investigated for W3099 beam section (Figure 34b) and similar
observations were made where 8EMH-W connection showed no significant degradation in
strength upto 6% interstory drift angle, still plastic hinge formed at the heat treated region of
the beam flanges. Based on the observations from the finite element investigation performed
in this study, it is apparent that the proposed design of extended end plate connection with
hexagonal bolt arrangement, heat treated beam section and web stiffener can improve the

54
seismic resilience of the extended end plate connection significantly compared to the existing
extended end plate connection. The novel connection design has been experimentally
validated and will be presented in a separate study.

Figure 34 Moment-rotation envelopes of different configurations of the extended end plate


connections. (a) W30148 beam section; (b) W3099 beam section.

6. Conclusions
The stiffeners in the eight bolt stiffened extended end plate connection causes disparaging
failure mechanisms due high stress concentration. In addition to the stress concentration, the
stiffeners reduce the usable floor area as well as increase cost of construction for the weld
works associated with the stiffener. This study proposed a modified design of extended end
plate moment connection without end plate stiffeners based on a detailed finite element
study.The modified design consists of rearranging the bolts into a hexagonal pattern to
distribute the flange forces evenly to the connecting bolts, heat treating the beam flanges to
ensure formation of plastic hinge away from the connection region, and welding stiffeners on
both the sides of the beam web to delay the onset of strength degradation. The performance
of the proposed connection has been evaluated under simulated seismic loading and

55
compared with the stiffened connection. The connection showed significant improvement
over the existing connection in terms of strength, ductility and energy dissipation capacity.
To conclude, the modified design of extended end plate connection demonstrates a great
potential to be used as a viable alternative to the conventional stiffened extended end plate
connection where the uncertainty and unfavorable failure mechanisms associated with the
end plate stiffener can be eliminated. The novel connection design has been validated
through full scale experimental testing of extended end plate moment connections and has
demonstrated excellent performance under simulated seismic loading, which will be
presented in a separate study.

7. References
[1] FEMA-350. Recommendation and seismic design criteria for new steel moment-frame
buildings. Prepared by the SAC Joint Venture for the Federal Emergency Management
Agency, Washington, DC, 2000.

[2] FEMA-353. Recommended specifications and quality assurance guidelines for steel
moment- frame construction for seismic applications. Prepared by the SAC Joint Venture
for the Federal Emergency Management Agency, Washington, DC, 2000.

[3] FEMA-355B. State of the art report on welding and inspection. Prepared by the SAC
Joint Venture for the Federal Emergency Management Agency, Washington, DC, 2000.

[4] FEMA-355D. State of the art report on connection performance. Prepared by the SAC
Joint Venture for the Federal Emergency Management Agency, Washington, DC, 2000.

[5] Stojadinovic B, Goel SC, Lee KH, Margarian AG, Choi JH. Parametric tests on
unreinforced steel moment connections. Journal of Structural Engineering 2000;126:40-
49.

56
[6] SAC/BD-00/01. Parametric tests on unreinforced connections. by Lee, K.H.,
Stojandinovic, B., Goel, S.C., Margarian, A.G., Choi, J., Wongkaew, A., Reyher, B.P.,
and Lee, D.-Y.

[7] SAC/BD-00/24. Development and evaluation of improved details for ductile welded
unreinforced flange connections. by Ricles, J.M., Mao, C., Lu, L.-W., Fisher, J.W.

[8] Lee D, Cotton SC, Hajjar JF, Dexter RJ. Cyclic behavior of steel moment-resisting
connections reinforced by alternative column stiffener details, I. Connection performance
and continuity plate detailing. Engineering Journal 2005;4th Quarter:189-213.

[9] AISC (2010a). Prequalified connections for special and intermediate steel moment frames
for seismic applications. American Institute of Steel Construction 358, AISC 358-10.

[10] Sumner EA, Mays TW, Murray TM. Cyclic testing of bolted moment end-plate
connections. Research Report No. CE/VPI-ST-00/03, SAC Report No. SAC/BD-00/21,
submitted to the SAC Joint Venture, 2000, 327 pages.

[11] Sumner EA, Murray TM. Experimental investigation of four bolts wide extended end-
plate moment connections. Research Report No. CE/VPI-ST- 01/15, submitted to Star
Building Systems, Inc., Oklahoma City, Oklahoma, 2001a, 114 pages.

[12] Sumner EA, Murray TM. Experimental investigation of the mre 1/2 end-plate moment
connection. Research Report No. CE/VPI-ST-01/14, submitted to Metal Building
Manufacturers Association, Cleveland, Ohio, 2001b, 91 pages.

[13] Sumner EA, Murray TM. Behavior of extended end-plate moment connections subject
to cyclic loading. Journal of Structural Engineering 2002;128(4):501-508.

57
[14] SAC/BD-97/02. Protocol for fabrication, inspection, testing and documentation of
beam-column connection tests and other experimental specimens. by Clark, P., Frank, K.,
Krawinkler, H., and Shaw, R.

[15] Tsai KC, Popov EP. Cyclic behavior of end-plate moment connections. Journal of
Structural Engineering 1990;116(11):2917-2930.

[16] Korol RM, Ghobarah A, Osman A. Extended end-plate connections under cyclic
loading: behavior and design. Journal of Constructional Steel Research 1990;16(4):253-
279.

[17] Ghobarah A, Korol RM, Osman A. Cyclic behavior of extended end-plate joints. Journal
of Structural Engineering 1992;118(5):1333-1353.

[18] Ghobarah A, Osman A, Korol RM. Behavior of extended end-plate connections under
cyclic loading. Engineering Structures 1990;12:15-26.

[19] Castellani A, Castiglioni CA, Chesi C, Plumier A. A European research program on the
cyclic behavior of welded beam to column connections. Proceedings of the NEHRP
Conference and Workshop on Research on the Northridge, California Earthquake of
January 17, 1994, National Earthquake Hazards Reduction Program 1998;Vol. III-B:510-
517.

[20] Coons RG. Seismic design and database of end plate and t-stub connections. M.S.
Thesis, University of Washington, Seattle, Washington, 1999.

[21] Adey BT, Grondin GY, Cheng JJR. Cyclic loading of end plate moment connections.
Canadian Journal of Civil Engineering 2000;27:683-701.

58
[22] Adey BT, Grondin GY, Cheng JJR. Extended end plate moment connections under
cyclic loading. Structural Engineering Report No. 216, Department of Civil and
Environmental Engineering, University of Alberta, Alberta, Canada, 1997.

[23] Adey BT, Grondin GY, Cheng JJR. Extended end plate moment connections under
cyclic loading. Journal of Constructional Steel Research 1998;46(1-3),Paper No. 133.

[24] Guo B, Gu Q, Liu F. Experimental behavior of stiffened and unstiffened end-plate


connections under cyclic loading. Journal of Structural Engineering 2006;132(9):1352-
1357.

[25] Shi G, Shi Y, Wang Y. Behavior of end-plate moment connections under earthquake
loading. Engineering Structures 2007;29:703-716.

[26] Murray TM, Kukreti AR. Design of 8-bolt stiffened moment end plates. Engineering
Journal 1988;2nd Quarter:45-52.

[27] Borgsmiller JT, Sumner EA, Murray TM. Tests of multi-row moment end-plate
connections having large inner pitch distances. Research Report No. CE/VPI-ST-95/01,
submitted to Kawada Industries, Costa Mesa, California, 1995.

[28] Meng RL. Design of moment end-plate connections for seismic loading. Ph.D.
Dissertation, Virginia Polytechnic Institute and State University, Virginia Polytechnic
Institute and State University, Blacksburg, VA, 1996.

[29] Mays TW. Application of the finite element method to the seismic design and analysis
of large moment end-plate connections. Ph.D. Dissertation, Virginia Polytechnic Institute
and State University, Blacksburg, VA, 2000.

59
[30] Gerami M, Saberi H, Saberi V, Daryan AS. Cyclic behavior of bolted connections with
different arrangement of bolts. Journal of Constructional Steel Research 2011;67:690-
705.

[31] Morrison M, Schweizer D, Hassan T. An innovative seismic performance enhancement


technique for steel building moment resisting connections. Submitted to Journal of
Constructional Steel Research 2014.

[32] Kiamanesh R, Abolmaali A, Razavi M. Effect of circular bolt pattern on behavior of


extended end-plate connection. Journal of Structural Engineering 2012; 139(11):1833-
1841.

[33] Bahaari MR, Sherbourne AN. Computer modeling of an extended end-plate bolted
connection. Computers and Structures 1994;52:879-893.

[34] Choi CK, Chung GT. Refined three-dimensional finite element model for end-plate
connection. Journal of Structural Engineering 1996;122(11):1307-1316.

[35] Bose B, Wang ZM, Sarkar S. Finite-element analysis of unstiffened flush end-plate
bolted joints. Journal of Structural Engineering 1997;123:1614-1621.

[36] Bursi OS, Jaspart JP. Basic issues in the finite element simulation of extended end plate
connections. Computers and Structures 1998;69(3):361-382.

[37] Maggi YI, Goncalves RM, Leon RT, Ribeiro LFL. Parametric analysis of steel bolted
end plate connections using finite element modeling. Journal of Constructional Steel
Research 2005;61:689-708.

[38] Daz C, Victoria M, Mart P, Querin OM. FE model of beam-to-column extended end-
plate joints. Journal of Constructional Steel Research 2011;67(10):1578-1590.

60
[39] Lim C, Choi W, Sumner EA. Low cycle fatigue life prediction using a four-bolt
extended unstiffened end plate moment connection. Engineering Structures 2012;41:373-
384.

[40] Kim J, Yoon JC, Kang BS. Finite element analysis and modeling of structure with
bolted joints. Applied Mathematical Modeling 2007;31:895-911.

[41] AISC (2006). Manual of steel construction load and resistance factor design. 13th ed.
Chicago, IL.

[42] Chaboche JL. Constitutive equations for cyclic plasticity and cyclic viscoplasticity.
International Journal of Plasticity 1989;5(3):247-302.

[43] Kaufmann EJ, Metrovich BR, Pense AW. Characterization of cyclic inelastic strain
behavior on properties of A572 Gr. 50 and A913 Gr. 50 rolled sections. ATLSS Report
No. 01-13, Lehigh University, 2001.

[44] Royster PB. Localized low cycle fatigue failure of welded steel moment connections.
MS Thesis, North Carolina State University, Raleigh, NC, 2007.

[45] AISC (2010b). Seismic provisions for structural steel buildings. American Institute of
Steel Construction 341, AISC 341-10.

[46] Morrison M. An innovative seismic performance enhancement techniques for welded


steel moment connections. PhD Dissertation Research, North Carolina State University,
Raleigh, NC, 2014.

[47] Schweizer DQ. Experimental investigation of innovative seismic performance


enhancement techniques for steel building beam to column moment connections. MS
Thesis, North Carolina State University, Raleigh, NC, 2013.

61
[48] Ellison MS, Corona E. Buckling of T-beams under cyclic bending. International Journal
of Mechanical Sciences 1998;40(9):835-855.

[49] Stojadinovic B. Stability and low-cycle fatigue limits of moment connection rotation
capacity. Engineering Structures 2003;25:691-700.

[50] Nakashima M, Liu D, Kanao I. Lateraltorsional and local instability of steel beams
subjected to large cyclic loading. International Journal Steel Structures 2003;3(3):179-
189.

[51] Yin S, Corona E, Ellison MS. Degradation and buckling of I-beams under cyclic pure
bending. Journal of Engineering Mechanics 2004;130(7):809-817.

[52] Okazaki T, Liu D, Nakashima M, Engelhardt MD. Stability requirements for beams in
seismic steel moment frames. Journal of Structural Engineering 2006;132(9):1334-1342.

62
CHAPTER 3: AN IMPROVED UNSTIFFENED 8 BOLT EXTENDED
END PLATE MOMENT CONNECTION FOR SPECIAL AND
INTERMEDIATE MOMENT FRAMES
PART II: EXPERIMENTAL VALIDATION AND ADDRESSING
FURTHER DEVELOPMENT

Abstract
Extended end plate (EEP) moment resisting connections provide the advantage of
eliminating field welding and by virtue of this, facilitate fast field erection of building
frames. Currently, EEP connections have been prequalified for use in the 2010ANSI/AISC
358 standard. For seismic applications rolled wide flange sections with a depth greater than
24 necessitate the use of a stiffened 8 bolt configuration. In this connection the stiffener
serves the purpose of strengthening the end plate, distributing flange forces uniformly
amongst the bolt group and promoting the formation of a plastic hinge away from the welds
connecting the beam flange to the end plate. The study reported herein proposes an
unstiffened 8 bolted end plate connection in which the bolt arrangement is modified so as to
promote uniform distribution of flange force among the bolt group and plastic hinge
formation is promoted by reducing the strength of steel over specified regions of the beam
flanges by exposing the regions to high temperatures followed by slow cooling. Finally,
strength degradation due to local web and flange buckling in the plastic hinge region is
delayed via the addition of a web stiffener. Development of the proposed connection through
detailed finite element studies was presented in part 1 of the study. Here, the results of full
scale testing are presented and analyzed to demonstrate the seismic performance of the
proposed connection. Connection behavior is further analyzed through post experiment finite
element analysis. Future analytical as well as experimental needs for further development of
the proposed connection are discussed.

63
Keywords: Steel moment connection, Extended end plate connection, Seismic performance
enhancement, Beam plastic hinge, Reduced beam section

1. Introduction
In bolted extended end plate moment connections (BEEP) a steel plate is shop welded to
the end of a beam which is then field bolted to the connecting members. Though shop
fabrication of BEEP connections may be more costly than field welded connections, they
offer the advantage of eliminating the difficulties associated with field welding and may
provide rapid erection of moment frames [1]. Simulated seismic testing of BEEP connections
have also shown them to be capable of providing significant ductility and seismic resilience
and as a result, BEEP connections have prequalified for use in special moment frames (SMF)
in the 2010ANSI/AISC 358 [2] prequalified connections for special and intermediate
moment frames for seismic applications. The 2010ANSI/AISC 358 prequalifies 3 types of
BEEP connections, the four bolt unstiffened, four bolt stiffened and the eight bolt stiffened
extended end plate connections. In the stiffened connections a gusset plate is welded to the
extended portion of the end plate and to the outer surface of the beam flange. This stiffener
has been found to strengthen and stiffen the extended portion of the end plate usually
resulting in thinner end plates in current design practice [3]. The stiffener also promotes
yielding and plastic hinge formation at the base of the stiffener away from the beam flange to
end plate CJP welds [4]. Investigations on 8 bolted and 4 bolted stiffened extended end plate
connections [5-7] have shown ductile performance and good energy dissipation. However, in
other studies [7-11] premature failure of the weld between the beam flange and the stiffener
or fractures initiating at beam flange to end plate CJP welds followed by gross section
fractures of the stiffener have been observed. In one study [12] fracture was reported to have
initiated at the toe of the weld between the end-plate stiffener and the beam flange and was
attributed to the stress concentration in this region. This stress concentration results from the
geometric discontinuity introduced by the stiffener and may be exacerbated by the presence
of weld defects some of which may not be detected by welding inspection practice [13].
Careful detailing, fabrication and inspection of stiffened BEEP connections may reduce the

64
likelihood of failures arising from the stress concentration induced by the flange stiffener,
however the seismic performance uncertainty and added complexity to design, fabrication
and inspection associated with stiffened BEEP connections has motivated the development of
an 8 bolt unstiffened extended end plate connection in which 3 novel seismic performance
enhancing techniques have been introduced. This connection was developed through detailed
finite element simulations in a companion paper by Quayyum and Hassan [14]. In that paper,
the 8 bolted extended end plate moment connection was studied in detail to access the
influence of the flange stiffeners, bolt arrangement, plastic hinge formation and other
parameters on connection behavior and ductility. The study presented herein describes the
results of full scale testing of the connection proposed by Quayyum and Hassan [14].
Experimental development as well as measured global and local experimental responses will
be discussed. Connection performance is further analyzed through post-test finite element
simulations; also, analytical as well as experimental needs for further development of this
connection will be addressed.

2. Review of Proposed Novel Performance Enhancing Techniques


In the companion paper by Quayyum and Hassan [14] details of the conceptual and finite
element development of the proposed modified 8 bolt extended end plate connection are
presented, however they are briefly summarized here as follows. The first of the 3
performance enhancing modifications is the removal of the end plate stiffener and
rearrangement of the bolts in a hexagonal pattern as shown in Figure 1 and Figure 4d. Finite
element analysis (FEA) showed that the end plate stiffener creates a stress concentration at
the toe of the weld between the stiffener and the beam flange. To avoid this stress
concentration the stiffener was removed, however in order to ensure uniform bolt force
distribution rearrangement of the bolt group was necessary. The proposed heaxagonal pattern
was found to be effective in promoting uniform force distribution amongst the bolt group
reducing the likelihood of non-ductile failure modes such as bolt rupture and end plate
fractures. Such failures were observed in 8 bolt-4 wide unstiffened extended end plate
connections tested by Sumner et al. [7].

65
Hexagonal bolt
arrangement 600

(MPa)
HBS
400

200 A992
Webstiffener
Heat-treated
A992

0
0 0.1 0.2 0.3 0.4 0.5

Figure 1 Sketch of Modified BEEP Figure 2 Engineering stress-strain


Connection. response of A992 and heat-treated A992
steel.

To further enhance performance, high temperature heat treatment of the beam flanges
was added to the proposed connection. This heat treatment involves subjecting selected
regions of the beam flange to high temperatures followed by slow cooling and by so doing
reduces the strength of steel in the heat-treated beam section (HBS). Details of the heat
treatment and its effect upon a wide range of material properties are presented elsewhere [15]
however, the uniaxial stress strain response comparison of an ASTM A992 steel coupon
before and after heat treatment is shown in Figure 2. As can be seen from this figure the yield
and tensile strength of the material is reduced by 38% and 25% respectively while the elastic
modulus and ductility remain unaffected. This reduction in strength promotes yielding and
plastic hinging within the HBS in a manner similar to the reduced beam section (RBS).
However, a connection modified with the HBS does not sacrifice elastic stiffness or buckling
resistance as does the RBS. The HBS technique was previously validated on welded flange-
welded web (all welded) moment connections by Morrison et al. [15].

66
Finally, a technique for delaying the onset of strength degradation was included in the
proposed connection. This technique involves welding a steel plate to the beam web and end
plate in the expected plastic hinge region which stiffens the beam web and delays the onset of
web local buckling. In the companion study web local buckling was observed to be the first
mode of instability that eventually led to lateral torsional buckling and strength degradation.
The addition of the proposed web stiffener was found to be effective in delaying the onset
and slowing the rate of strength degradation [14]. The connection which was proposed in the
companion paper was called 8EMH-W [14], however the testing of this connection was
included as a part of a broad experimental program in which the test specimen was assigned
the name HBS 4 and will be referred to as such in the following.

(a)
(b)

Figure 3 HBS 4 Test setup. (a) Sketch of test setup, (b) photograph showing test setup.

3. Connection Design and Experimental Setup


One large scale specimen (HBS 4) was tested to evaluate all three performance enhancing
modifications. The test setup and details of the test specimen are shown in Figures 3 and 4.
The connection was fabricated by an AISC certified commercial fabricator. All welds were
made in the fabrication shop in the down hand position and welding was accomplished using

67
gas shielded metal arc welding (GMAW) with E71T-1C-H8 electrodes (Lincoln ultra core
71C) which is specified to meet all the requirements of AWS D1.8. The beam flange was
welded to the end plate with complete joint penetration welds (CJP) similar to the AWS
prequalified TC-U4b-GF weld. These welds were made without an access hole and backing
bar, however an 8 mm (5/16 inch) backing fillet weld was provided.

(a) (b)

(c) (d)

Figure 4 HBS 4 connection details. (a) Elevation of connection, (b) photograph of HBS 4
connection prior to testing, (c) section A-A showing connection of beam web and web
stiffeners to end plate, (d) elevation of end plate showing modified bolt arrangement.

68
This weld detail is similar to that used in other BEEP studies [3] and results in a small
region of the CJP weld (directly above the beam web) that is uninspected. The beam web was
welded using a CJP weld, which was used in lieu of fillet welds to facilitate the fit up and
welding of the web stiffeners as shown in Figures 4a and 4c. As may be inferred from Figure
4c the web stiffener plate adjacent to the root of the bevel was used as a backer for the beam
web CJP weld. Fillet welds were used to connect the web stiffeners to the end plate and to
the beam web as shown in Figure 4a. These welds resulted in distortion of the web stiffener
plates which adversely affected their effectiveness in delaying the onset and retarding the
progression of local buckling as will be discussed later. Finally, the column was reinforced
with two-12.7 mm (1/2 inch) doubler plates which were designed to promote strong panel
zone behavior in order to ensure that most of the inelastic action was confined to the beam.

(a) (b)

Figure 5 Heat-treatment setup. (a) Photograph showing electric surface heating pads during
installation on a beam flange, (b) photograph showing insulation of beam flange for well
controlled heating and cooling.

Upon completion of fabrication, selected regions of the beam flanges were heat treated.
The locations of the heat treated regions are shown in Figure 4a and were decided upon
numerically through finite element simulations [14-15]. Heat treatment was performed using
electrical resistance ceramic mat heating pads as shown in Figure 5a. Heating pads were

69
sized according to the required dimensions of the HBS and were installed on the inner and
outer surface of the beam flanges. The heating pads were connected to a power supply and
type K thermocouples were used to monitor temperatures and provide continuous feed back
to the power supply. Two (2) layers of 50 mm (2 inch) high density ceramic fiber insulation
blankets were wrapped around the beam flanges as shown in Figure 5b to provide well
controlled heating and cooling. More details about the heat treatment setup and procedure are
provided in Morrison et al. [15]. The assembly of the beam and column was made with the
column in the vertical upright position. Bolts were installed and tightened using the turn-
of-nut method. The turn-of-nut procedure was first verified by tightening bolts from the same
batch of those used in the test specimen in a Skidmore Wilhelm bolt tension calibrator to the
minimum pretension specified by the AISC LRFD Specification [16].

4. Instrumentation
The test specimen was equipped with strain gauges along the beam flanges to monitor
longitudinal flange strains at various locations including the beam flange to end plate weld
toe and HBS region. Bolt response was of particular interest; as such, 4 of the 16 bolts were
instrumented by installing a strain gauge in the center of the unthreaded portion of the bolt
shank. The instrumented bolts are highlighted in blue in Figure 4d. String and linear
potentiometers were used to monitor displacements in the column, beam and panel zone. A
calibrated load cell in the hydraulic actuator provided readings of force response during the
experiment. All specimens were painted with hydrated lime prior to testing to visually
indicate regions of yielding. The Optotrak Certus HD three-dimensional (3D) position system
produced by Northern Digital Incorporated was used to capture the positions of markers
placed along the beam flanges and web as shown in Figures 6a and 6b. Two Optotrak
cameras were used which were able to capture the motion of markers placed on the top and
sides of the beam top flange, the beam web and the side of the beam bottom flange. Position
time history data obtained from this system was post processed to calculate displacements
and strains in areas of interest. The Optotrak system provides the advantage of being able to

70
record large cyclic inelastic strains while electrical resistance strain gages may either exhibit
gradual strain drift or fail to remain adhered.

(a) (b)

Figure 6 Instrumentation of HBS 4. (a) LED markers placed on beam top flange, beam web
and web stiffeners, (b) LED markers placed on the sides of beam top and bottom flange,
beam web and webstiffners.

5. Test Results
5.1 Global Response of Modified BEEP Connection
Testing was carried out at the North Carolina State University Constructed Facilities
Laboratory. Loads were applied at the beam tip in accordance with the 2010 ANSI/AISC 341
[17] seismic provision Appendix S loading protocol consisting of quasi-static increasing
displacement cycles. Figure 7a shows the moment-rotation response of HBS 4. The global
response is characterized by wide hysteretic loops indicating good energy dissipation. The
specimen exceeded the 2010 AISC Seismic Provisions [17] special moment frame (SMF)
qualifying 4% interstory drift angle without significant strength loss. Slight strength
degradation began during the 1st loading cycle at 4% interstory drift and progressed during
later loading cycles at 5% and 6% interstory drift. Loading was terminated during the second
cycle at 6% story drift due to fracture in the location of significant beam flange buckling (see
Figure 13b). The response of HBS 4 is further analyzed by plotting the total rotation
contributions and the moment-plastic rotation curves for each component of the connection
assembly which were calculated using data measured by instrumentation during the test and
equations proposed by Popov et al. [18]. These graphs as shown in Figure 7b and Figure 8a-d

71
indicate that panel zone shear and column flexural deformations contributed only to the
elastic behavior of the sub-assemblage and as a result, inelastic action was entirely confined
to the beam. It is of note that end plate separation was not recorded, however given the strong
plate approach adopted in this modified BEEP connection, end plate inelastic flexural
deformations and end plate separation are expected to be negligible.

4
(a)
100 (b)
M % of
(MN-m)
80
2 Beam
Rotation
60
Panel Zone
Rotation
0 40
Column
Rotation
20
-2
0

-4
-0.08 -0.04 0 0.04 0.08 Interstory Drift (%)
(rad)
Figure 7 Experimental response of HBS 4. (a) Moment-rotation hysteresis response,
(b) total rotation contributions.

5.2 Plastic Hinge Formation in the Modified BEEP Connection


Figure 9a-d illustrates the progression of inelastic action along the beam flange via bar graphs
in which the distribution of longitudinal tensile strains (normalized by the yield strain) along the
centerline of the beam flange at various stages of the loading history are plotted. Bars highlighted
in red represent the strains in the heat-treated (weakened) regions. Strains were calculated by post
processing data obtained from 3D noncontact spatial displacement measurement sensors placed
along the beam flange as shown in Figure 6a. These plots illustrate the influence of the HBS in
promoting majority of the inelastic action away from the beam flange to end plate connection.
Similar observations have made in all welded connections modified with HBS [15]. The strain
profiles indicate that flexural yielding initiates in the heat treated areas at 1% drift. As loading is
continued longitudinal strains increase more in the heat treated areas than in the unheat-treated

72
regions adjacent to the beam flange to end plate welds. It is important to note that the beam length
of 3.4 m (134 in) used in this test corresponds to a moment frame with a clear span of 6.8 m (22
feet). This relatively short span results in a large moment gradient which is reflected in the strain
profiles. As drift angle increases from 3% to 4% longitudinal strain in the heat treated regions
increase significantly while strains outside of the heat treated areas grow only slightly. This
demonstrates that the large displacements imposed at the beam tip are mostly accommodated by
flexural deformation in the HBS.

4 4
(a) (b)
M
(Mn-m)
2 2

0 0

-2 -2

-4 -4
-0.08 -0.04 0 0.04 0.08 -0.08 -0.04 0 0.04 0.08
Total Plastic Rotation (rad)
Panel Zone Plastic Rotation (rad)

4 4
M
(c) (d)
(MN-m)
2 2

0 0

-2 -2

-4 -4
-0.08 -0.04 0 0.04 0.08 -0.08 -0.04 0 0.04 0.08
Column Plastic Rotation (rad) Beam Plastic Rotation (rad)

Figure 8 Plastic rotation contributions by components of HBS 4. (a) Total connection plastic
rotation, (b) panel zone shear plastic rotation, (c) column flexural plastic rotation, and (d) beam
plastic rotation.

73
18 18

15 (a) 15 (b)
y y
12 1% 12 2%
9 9

6 6

3 3
0 0
149 352 556 759 149 352 556 759
Dist. from column flange (mm) Dist. from column flange (mm)

18 18
y 15
(c) 15
(d)
y
12 3% 12 4%
9 9
6 6
3 3
0
0
149 352 556 759
149 352 556 759
Dist. from column flange (mm) Dist. from column flange (mm)

Figure 9 Recorded longitudinal strain along the center of top flange of the beam from HBS4.

(a) (b)

(c) (d)

Figure 10 Progression of yielding and plastic hinge formation in HBS 4. (a) 1% drift, (b) 2%
drift, (c) 3% drift, and (d) 4% drift.

74
The progression of the plastic hinge formation is also qualitatively reflected in the
photographs shown in Figure 10a-d where it is observed that flaking of the white wash takes
place away from the beam flange to end plate junction during loading cycles from 1% to 4%
interstory drift. The beam tip load vs. longitudinal strains recorded from strain gages placed
on the beam bottom flange at the weld toe and in the HBS region are shown in Figure 11.
This figure further highlights the performance of the HBS in shifting inelastic action away
from the beam welds. Here it can be inferred that the HBS acts like a damper for the
connection providing energy dissipation in a location free of stress and strain concentrations
(prior to the onset of local buckling) and resistant to fracture. As a result, locations which are
relatively less ductile or fatigue resilient (e.g., the beam flange to end plate welds or beam
flange to column flange welds in the case of all welded connections) are exposed to lower
strain demands providing enhanced connection performance.

1
Beam tip force (MN)

-1

SG 1- Weld Toe SG 2- HBS


-2
-20 -10 0 10 20
Strain (m )
Figure 11 Measured flexural strains from strain gages placed on the bottom flange of HBS 4.
Strains are shown from the beginning of test to the second cycle of 3% drift.

75
5.3 Bolt Response in the Modified BEEP Connection
The measured axial strains vs column centerline bending moment is shown in Figure 12
for the 4 bolts which were instrumented with strain gages. In general bolt axial strain
behavior is characterized by increased stains during tensile excursions (i.e. negative bending
of the beam) followed by a reduction in strain when the loading is reversed. In all cases a
gradual relaxation of the bolt strains was observed with progressive loading. Note that the
recorded tensile strains from the initial tightening of the bolts differ which reflects the
variability in pretension forces that may be encountered when using the turn of nut method.
Comparison of the inner and outer bolts (bolts 1 and 2) located above the top flange shows
approximately a 10% difference in the peak tensile strain response of these bolts throughout
the loading history (Figure 12a and 12b). Both bolts relaxed to approximately 75% of their
pre-tension strain at the end of the test. The bolts located below the flange (Figure 12c and
12d) displayed larger strains (especially bolt 4) indicating that these bolts resisted a larger
portion of the tensile forces. The difference in peak tensile strains between the inner and
outer bolts (bolts 3 and 4) was approximately 10%. In particular bolt 4 showed a sharp
increase in strain with applied moment. This was expected due to the relative proximity of
this bolt to the stiff beam web to flange junction. The welding of the beam web, web
stiffeners and beam flange in this region stiffens the end plate and attracts larger forces to the
bolts nearest to this area. This behavior was captured by pretest finite element analyses [14]
and though bolt 4 did experience slightly larger strains than the other instrumented bolts, no
permanent set in bolt strains was observed nor was this behavior found to be detrimental to
connection performance.

5.4 Failure Mechanism


Loading of the HBS 4 was terminated due to a fracture in the beam flange at the location of
significant local buckling as shown in Figure 13a and 13b. This flange buckling mechanism
appeared to have initiated during loading cycles at 3% drift and strength of the connection
progressively deteriorated during loading cycles at 4%, 5% and 6% drift. The initiation of
this strength degradation mechanism at 3% drift was unexpected based on pre-test FEA

76
(Figure15a). As mentioned before the edges of the web stiffener were continuously
connected to the beam web and end plate with (4.76mm) fillet welds.

10 10
(a) (b)
Bolt 1 Bolt 2
7.5 7.5

Pre-tension
5 5
Pre-tension

2.5 2.5

0 0
-4 -2 0 2 4 -4 -2 0 2 4
Column centerline moment (MN-m) Column centerline moment (MN-m)

10 10
(c) (d)
m m Bolt 4
Bolt 3
7.5 7.5

Pre-tension Pre-tension
5 5

2.5 2.5

0 0
-4 -2 0 2 4 -4 -2 0 2 4
Column centerline moment (MN-m) Column centerline moment (MN-m)

Figure 12 Recorded axial bolt strain vs column centerline moment.

These welds resulted in distortion (bowing outward) of the web stiffener plates. Based on
pretest measurements obtained from the optotrak markers placed on the beam web and on the
web stiffener plates it is estimated that at the center of the plates there was a maximum 19
mm gap between the beam web and the web stiffeners. During loading cycles at 1.5 % drift,
loud noises similar to those created during a fast fracture were heard. It is believed that these

77
noises were created from impact occurring between the web stiffeners and the beam web.
This impact was due to the buckling of the web stiffeners which was visually observed
during loading cycles at 2% drift and continued for the remainder of the test. This early
initiation of buckling limited the effectiveness of the web stiffeners in delaying the onset of
beam local web buckling and strength degradation and will be discussed further in the
following section.

(a) (b)

Figure 13 Local buckling failure of HBS 4. (a) Post test photograph showing flange and web
local buckling, (b) photograph showing fracture of beam flange in the crest of flange buckle.

6. Post-Test FEA and Discussion


In order to more closely analyze the behavior of HBS 4, post-test FEA was conducted
using general purpose finite element software ANSYS. The features of the finite element
model including the element type, material model and material parameters were kept same as
those presented in the pre-test numerical study [14]. However, an attempt was made in the
post-test FEA to model the welding induced initial distortions in the web stiffener plates as
shown in figure 14a. Estimates of geometry of the web stiffener plates were obtained from
the pre-test displacement fields recorded by the Optotrak system. As noted previously
Optotrak sensors were placed on the web stiffener plates and on the beam web. By assuming
a nominal thickness of the web stiffener plate and no distortion in the beam web, estimates of
the initial gap between the beam web and the web stiffener as well as the distorted shape of

78
the web stiffener could be made and included in the post-test FEA. Welding induced residual
stresses and strains were not included in the model. Also, no geometric imperfections were
included in the modeling of the beam and column. A comparison between the FEA
prediction and the observed moment-rotation response is shown in Figure 14b. The FEA
prediction is in good overall agreement with the experiment response both in terms of peak
strength prediction and the rate of strength degradation. The predicted deflected shape is
compared with a photograph of the test specimen during cycles at 6% drift in Figure 14c and
14d. Predictions of the top flange buckling amplitudes are also compared with the measured
top flange deflections Figure 14e and 14f. These comparisons show that local deformations
including buckling modes are reasonably predicted by the FEA model.
The finite element model was used to analyze the effect of the initial distortion of the web
stiffener plates on the initiation and rate of strength degradation of HBS 4. Presented in
Figure 15a is a comparison of the moment vs. interstory drift envelop in negative bending for
a pre-test simulation including perfect geometry of the web stiffener plates (full initial
contact between the web stiffener and the beam web), a post-test simulation including
welding induced distortion of the web stiffener plates, and a post-test simulation with no web
stiffener plates. As expected these results show that the welding induced distortions limited
the effectiveness of the web stiffener plates.
From comparison of both post-test simulations (with and without the web stiffener) it is
observed that web stiffener increases the moment capacity and slows the initial rate of
strength degradation between 3% and 4% interstory drift. However, the rate of strength
degradation between 4% and 6% is unaltered resulting in a strength loss of approximately
27% during cycles at 6% drift. This is approximately twice the amount of strength loss
predicted by the pre-test simulations. This accelerated strength loss is accompanied by larger
amplitudes of local flange buckling and larger plastic strains as shown in Figure 15b and 15c
which compares the equivalent plastic strains for pre and post-test simulations of HBS 4.

79
Webstiffener (a) 4
Experiment
w/ initial M (b)
(MN-m) Simulation
distortion
2

-2

-4
-0.08 -0.04 0 0.04 0.08
(rad)

(c) (d)

80
(e)
(f)
40
Vertical Position (mm)

0
-1000 -600 -200
1% Drift- Exp.
1% Drift-Sim. -40
3% Drift-Exp.
3% Drift-Sim.
4% Drift-Exp. -80
4% Drift-Sim.
5% Drift-Exp.
5% Drift-Sim.
6% Drift-Exp. -120
6% Drift-Sim.
Horizontal Position from Column Face (mm)

Figure 14 Validation of post-test FEA of HBS 4. (a) FEA discretization including initial
distortion of web stiffener plates, (b) comparison between FEA prediction and the recorded
moment-rotation response, (c) photograph of HBS 4 during loading cycles at 6% drift, (d)
FEA prediction of deformation at 6% drift, (e) comparison between FEA predictions and
recorded top flange vertical displacements during various stages of loading history, (f)
photograph of HBS 4 top flange during loading at 4% drift.

80
4
(a)
M
(MN-m)

2
Simulation w/o intial webstiffener distortion (pre-
test)
Experiment

Simulation w/ intial webstiffener distortion (post-


1 test)
Simulation w/o webstiffener

0
0 2 4 6 8
Interstory Drift (%)

(b) (c)

Figure 15 Post-test FEA simulation results showing the effect of initial web stiffener plate
distortion. (a) Moment-rotation envelop, (b) plastic equivalent strain prediction at 5% drift
for HBS 4 (post-test simulation with initial web stiffener distortion), (c) plastic equivalent
strain prediction at 5% drift for HBS 4 (pre-test simulation without initial web stiffener
distortion).

81
7. Future development of Modified BEEP Connections
Due to limited resources only 1 test specimen was evaluated in this study. This test
specimen included 3 modifications all of which were judged to have had an influence on
overall the performance of the connection. However, it is practically impossible to accurately
access and quantify the effect of each modification on various performance parameters
including ductility, strength, buckling resilience, reliability etc. from the result of one full
scale test. As such future development of the modified BEEP connection should involve
more experiments in which each performance enhancing modification can be studied
individually to more narrowly quantify and analyze its effect. For example, in the FEA
development of the modified BEEP connection it was found that the combination of a
strong end plate with the arrangement of bolts in the proposed hexagonal pattern
significantly reduced strain demands in the bolts and end plate, however plastic hinging of
this connection forms at the beam end and as a result high strain demands are placed on the
beam flange to end plate welds. This motivated the use of the HBS as opposed to the
currently used flange stiffener to shift plastic hinging away from the beam end without
introducing stress or strain risers. Shifting the plastic hinge away from the beam end in such
a manner reduces strain demands at the welded joint and so it is believed that this provides
more reliable connection performance.
However, experimental evaluation of the modified unstiffened BEEP connection without
the HBS may demonstrate that this connection can meet or exceed the performance
requirements for use in SMFs outlined in the AISC seismic provisions. This assertion is
made based on the following factors which may contribute to enhanced ductility of EEP
connections:
1. EEP connections are shop welded in more favorable conditions (year round) when
compared to field welded connections, thus eliminating some problems related with
field welding. In particular, shop welding ensures adequate access to perform welding
and also to make sound visual inspections during and after welding. It is also possible
to make all welds in the preferred down hand position which is not possible in field
welded connections such as the WUF-W or RBS.

82
2. Due to the fact that EEP connections are shop welded, no weld access holes are
required to make flange to end plate welds. This reduces the inevitable stress and
strain concentrations introduced by access hole cuts. It must be noted that as a
consequence of the elimination of the access hole, the beam flange to end plate weld
directly above the beam web is a partial-joint-penetration weld. However, this has not
proved to be detrimental to connection performance in published experimental
studies.
The combination of these factors with the use of notch-tough low hydrogen welding
electrodes with specified minimum Charpy V-notch toughness, strict quality control practices
such as the careful cleaning and preparation of the weld root opening, careful visual
inspection of root and subsequent weld passes, removal of weld run off tabs, compliance with
AWS specified minimum preheat temperatures etc., may provide adequate ductility to
modified 8 bolt EEP connections precluding the need for either reinforcing the connection or
weakening the beam. Such a connection may prove to be more economical than the currently
prequalified 8 bolt stiffened extended end plate connection and as such, future studies to
evaluate its seismic performance are warranted.

7.1 Future work on web stiffener


In the current study the web stiffener plates seemed to provide some resistance to strength
degradation but not to the extent that was predicted by pre-test analysis. It is noted here that
the web stiffener concept may be extended to field welded connections in which a fuse is
provided, for example the RBS and the recently validated HBS [15] connection. However,
given the lessons learned in this current study, more consideration should be given to the
attachment of the web stiffener. Further FEA studies evaluating various attachment details
for e.g. use of plug welds and slot welds to provide more attachment points between the
beam web and the web stiffener plates, thus improving the overall stiffness of the assembly
may be useful towards more successful implementation of this technique. Thermo-
mechanical simulations to include the effect of welding induced distortions and residual
stresses may provide useful insight in this process. In addition, compact wide flange sections

83
readily achieve rotation capacities of 3 [19] or greater before the onset of significant local
buckling and resulting strength loss. As such, carefully constructed moment connections
utilizing these sections readily achieve the 2010 AISC Seismic Provisions [17] special
moment frame (SMF) qualifying 4% interstory drift angle without more that 20% loss of
nominal moment capacity. Therefore, the web stiffener technique may provide more added
value to built-up beams with slender webs in which rotational capacities are reduced.

7.2 Design Development


Given that the proposed modified EEP connection deviates significantly from the
configuration of existing prequalified EEP connections, design methodologies for the
practical use of modified EEP connections in moment frames also need to be established, in
particular the limits of beam and column size including flange thickness, depth etc. Also
design equations or tables to proportion and configure the modified end plate are needed. In
the current study, design of the end plate was partly guided by using equations based on yield
line analysis to select trial end plate sizes. FEA was then used to optimize the design of the
end plate and bolt arrangement. This type of exercise is not well suited for design practice
and therefore future studies need to be conducted to develop prescriptive design procedures
for the proposed modified 8 bolt unreinforced EEP connection.

8. Conclusions
The test results of a modified 8 bolted unstiffened extended end plate connection has
been presented. Three (3) performance enhancement techniques were utilized in this
connection. These include the removal of the flange stiffener and rearrangement of the bolts
in a hexagonal pattern, heat treatment of the beam flanges in specified regions adjacent to the
connection, and the welding of steel plates called web stiffeners to the beam web. The
modified BEEP connection displayed ductile seismic response which exceeded the 2010
AISC Seismic Provisions [17] SMF qualifying 4% interstory drift angle without much
strength loss. Test data shows that the use of a strong end plate with a modified bolt
arrangement performed well in distributing bolt forces uniformly amongst the bolt group

84
while the HBS was successful in shifting the majority of inelastic action away from the beam
flange to end plate welds. The observations are consistent with pre-test FEA predictions [14].
The web stiffener, did not perform as predicted by pre-test FEA, however this discrepancy
was shown by post-test FEA to be attributable to the presence of welding induced distortions
of the web stiffener plates. Though the excellent seismic performance of HBS 4 shows the
promise of the modified BEEP connection, future effort is needed to optimize this connection
for both performance and economy. In addition, prescriptive design procedures for seamless
inclusion of this connection in current seismic design practice and for prequalification in the
ANSI/AISC 358 standard are required.

9. References
[1] Murray TM, Sumner EA. AISC design guide series 4, extended end-plate moment
connections. Chicago, 2003.

[2] ANSI/AISC 358-10. Prequalified connections for special and intermediate steel
moment frames for seismic applications. American Institute of Steel Construction 358,
2010.

[3] Sumner EA. Unified design of extended end plate moment connections subject to
cyclic loading. Ph.D. Dissertation, Virginia Polytechnic Institute and State University,
Blacksburg, VA, 2000.

[4] Mays TW. Application of the finite element method to the seismic design and analysis
of large moment end-plate connections. Ph.D. Dissertation, Virginia Polytechnic
Institute and State University, Blacksburg, VA, 2000.

[5] Tsai K, Popov EP. Cyclic behavior of end-plate moment connections. Journal of
Structural Engineering 1990;116(11):2917-2930.

[6] Sumner EA, Murray TM. Behavior of extended end-plate moment connections
subject to cyclic loading. Journal of Structural Engineering 2002;128(4):501-508.

85
[7] SAC/BD-00/21. Cyclic testing of bolted moment end-plate connections. by Sumner,
E.A., Mays, T., Murray, T.M.

[8] Korol RM, Ghobarah A, Osman A. Extended end-plate connections under cyclic
loading: behavior and design. Journal of Constructional Steel Research
1990;16(4):253-279.

[9] Ghobarah A, Osman A, Korol RM. Behavior of extended end-plate connections under
cyclic loading. Engineering Structures 1990;12(1):15-27.

[10] Adey BT, Grondin GY, Cheng JJR. Cyclic loading of end plate moment connections.
Canadian Journal of Civil Engineering 2000;27:683-701.

[11] Shi G, Shi Y, Wang Y. Behavior of end-plate moment connections under earthquake
loading. Engineering structures 2007;29(5):703-716.

[12] Guo B, Gu Q, Liu F. Experimental behavior of stiffened and unstiffened end-plate


connections under cyclic loading. Journal of Structural Engineering
2006;132(9):1352-1357.

[13] FEMA-350, 2000, State of the Art Report on Welding and Inspection, prepared by the
SAC Joint Venture for the Federal Emergency Management Agency, Washington, DC

[14] Quayyum S, Hassan T. An improved unstiffened 8 bolt extended end-plate moment


connection for special and intermediate moment frames, part I: numerical modelling
and development. To be submitted to Journal of Constructional Steel Research 2014.

[15] Morrison ML, Schweizer DQ, Hassan T. An innovative seismic performance


enhancement technique for steel building moment resisting connections. Submitted to
Journal of Constructional Steel Research 2014.

86
[16] AISC. Manual of steel construction load and resistance factor design. 13th ed.
Chicago, IL, 2006.

[17] ANSI/AISC 341-10. Seismic provisions for structural steel buildings. American
Institute of Steel Construction 341, 2010.

[18] Popov EP, Amin N R, Louie JC, Stephen RM. Cyclic behavior of large beam-column
assemblies. Engineering Journal 1986;23(1):9-23.

[19] Bruneau M, Uang CM, Whittaker A. Ductile design of steel structures. New York:
McGraw Hill Companies, 1998.

87
CHAPTER 4: INFLUENCE OF WELD SEQUENCE ON THE SEISMIC
RESPONSE OF MOMENT RESISTING CONNECTIONS

Abstract
During the 1994 Northridge earthquake, moment resisting connections suffered from
low-cycle fatigue failure near the beam-to-column welded joints and around weld access hole
region. Two decades of research after the 1994 Northridge earthquake brought new and
improved design of moment connections. However, still low-cycle fatigue failures have been
observed in the post Northridge moment connections near the welded region and research
studies indicate that welding induced residual stresses is one of the contributing factors to
low-cycle fatigue damage in the connections. This study investigates the influence of
welding sequence on the localized fatigue failure mechanisms of moment resisting
connections. Two exterior sub-assemblages of welded unreinforced flange bolted web
(WUF-B) connections were tested under displacement controlled constant amplitude cyclic
loading, where the beam flanges were welded to the column flange with two different
welding sequences. In both the connections tested, axial strain accumulation (known as
ratcheting) was observed very close to the welded joints and around the weld access hole.
Ratcheting is known to occur in metallic materials as they undergo inelastic cyclic loading,
along the directions of force-controlled loading. Hence, it is anticipated that the occurrence
of ratcheting at the welded joint under displacement-controlled loading could be influenced
by residual stresses near the welded joints. Moreover, variability in the ratcheting rate was
observed with two different weld sequences which led to different fatigue lives of the
connections. The results of the experimental study demonstrated the effect of welding
procedure as well as welding induced residual stresses on the low-cycle fatigue failure of the
connections.

Keywords: Moment connections, weld sequence, residual stress, low-cycle fatigue,


ratcheting.

88
1. Introduction
The design guidelines for moment resisting connections have been proposed in Steel
Construction Manuals [1-3] considering the global performance of the structure with no
emphasis on the basic correlation between localized member behavior and global structural
performance. However, since the current engineering trend is moving towards performance
based design which emphasizes that the structural design of the structure should meet the
inelastic seismic demands with a reliable global and local performance [4], it is essential that
the cause and mechanism of local failure of welded steel moment frames (WSMFs) be
understood along with the global failure mechanism to allow WSMFs effectively and reliably
meet the seismic demands. Although the local condition at one location will not dictate the
performance of a structure; the local condition of many structural members does dictate the
performance of the entire structure [5]. While many design procedures can predict the
performance of a global structural parameter under prescribed load conditions, those same
procedures do not attempt to correlate these quantities with local structural performance. This
is not a shortcoming of the current guidelines and design procedures; however it is a
limitation on our knowledge of local failure conditions. Post-Northridge researchers have
mainly addressed the global level mitigation of the failure risk; however the fundamental
reasons for the brittle failures at the local level were not uncovered. More than ten years of
research after the Northridge earthquake have developed modified designs of welded steel
moment frames (WSMFs) [6-8]. While test results on the new connections demonstrated
significant improvement with regard to ductility, many post-Northridge welded steel moment
connections (WSMCs) exhibited failure initiated from low-cycle fatigue cracks near the
beam flange weld toe [9-19]. The weld toe fatigue crack initiation motivated several
investigations on weld material and the outcome was the recommendation of using high
toughness weld material. However, the reason for the weld toe fatigue crack initiation in the
post-Northridge design has yet to be investigated in a rational manner. The Structural
Engineers Association of California (SEAOC) Seismology Committee emphasizes the
importance of localized failures, as indicated in their commentary and recommendation on
FEMA 350 [20]. The SEAOC committee identified research needs on weld interface field

89
conditions and low-cycle fatigue failure mechanisms towards developing fully rational
design guidelines. Another important aspect of characterizing the local failure mechanism is
that a steel structure that did not crack during an earthquake even may not always be safe for
continued operation. It may however accumulate enough fatigue damage due to repeated
yielding and may fail subsequently under service loading or another event.
Research programs at NC State University demonstrated that the welding sequence,
thereby the welding residual stress, has significant influence on fatigue crack initiation at the
weld toe of piping joints [21-23]. These studies demonstrated the effect of welding residual
stress on strain ratcheting leading to crack initiation in the piping joint. Welding sequence
thus can have a significant effect on the fatigue life of welded joints as demonstrated by these
studies. Ratcheting is known to be demonstrated by metallic materials under inelastic cyclic
loading along the directions of force-controlled loadingsteady or fluctuating [24]. Hence, it
was anticipated that the occurrence of ratcheting at the welded piping joints under
displacement-controlled loading could be influenced by its residual stresses. The influence of
residual stress on fatigue crack initiation at the weld toe in WSMCs was speculated [25], but
the failure mechanism is yet to be investigated. Due to the complexity involved in residual
stress calculation, most analyses either do not consider residual stresses or use simplified
methods of analysis [26-27]. Dong and his coworkers [28-29] performed residual stress
analysis in their investigation of WSMC failures. Their calculations demonstrated that the
magnitude of residual stresses in some locations can be greater than the yield stress of the
material. They studied crack propagation of preexisting cracks at the backing bars and
demonstrated that the triaxial residual stress greatly reduced the plastic deformation and thus
induced brittle fracture. Suita et al. [9] demonstrated a significant increase in the
accumulated axial strain in the vicinity of beam-column welded connection during cyclic
loading. The amount of strain accumulated decreased with distance from the weld interface.
They reported both brittle mode fracture and progressive local plastic strain accumulation.
More recent studies at NC State University by Royster [30] on bottom beam flange tee joint
showed the influence of welding sequence on the fatigue failure of the connections. It was
observed that different welding sequences induced different low-cycle fatigue responses of

90
the connections. Significant amount of ratcheting was observed in these experiments near the
welded joints, which was thought to be the contributing factors in the premature fatigue
failure of the connections. Moreover, variability of ratcheting response was observed for
different welding sequences. Based on the observations from the study it was concluded that
the ratcheting phenomena under displacement controlled loading could be attributed to the
different residual stresses induced by the different welding sequences. From the literature it is
apparent that low-cycle fatigue response of WSMCs can be influenced by the presence of
residual stresses and the sequence at which the welding was performed. A detailed
investigation on the WSMC residual stress and welding sequence is essential to explore its
low-cycle fatigue failure response. If the influence of residual stress on fatigue damage
accumulation in WSMC is revealed, fatigue life can be enhanced through optimizing the
weld sequence. The SAC Joint Venture did not investigate any aspect of the welding
procedure (number and size of weld beads and their sequence and the weld speed) and its
influence on strain accumulation near the weld toe and base metals due to high temperature
cycles during welding. Thus, a study is needed to capture the influence of welding procedure
on failures of welded steel moment connections. The objective of this research is to
investigate the influence of weld sequence on localized low-cycle fatigue failure mechanisms
observed in welded steel moment connections. Two exterior beam-column sub-assemblages
of welded unreinforced flange bolted web (WUF-B) moment connections were tested under
displacement controlled constant amplitude cyclic loading. The beam-to-column flange
groove welds were filled with different weld sequences in an attempt to investigate the effect
of weld sequence on the low-cycle fatigue response of the connections.

2. Cyclic Testing of WSMCs


The cyclic testing of welded steel moment connections (WSMCs) consisted of two full
scale exterior welded unreinforced flange bolted web (WUF-B) moment connections
(referred to as specimens MC1 and MC2) where a single W18 55 ASTM A992 beam was
attached to a W14 74 ASTM A992 column. The beam-to-column connection was designed
and detailed in accordance with the AISC Seismic Design Provisions [1, 3, 6], and AWS

91
D1.1 [32]. The flanges of the beam were welded to that of the column flange along the strong
axis of the column using a complete joint penetration (CJP) groove weld in accordance with
the post-Northridge design recommendations [6]. The web of the beam was connected to
column flange with four ASTM A325 high strength bolts connected through a 7.9375 mm
(5/16 in) shear tab ASTM A36 steel which was shop welded on the column. Also, the column
was reinforced with 15.875 mm (5/8 in) continuity plates that coincide with the top and
bottom beam flanges and a 12.7 mm (1/2 in) doubler plate in the column panel zone between
the continuity plates. The connection configuration is as shown in Figure 1.

15.875 mm
continuity plate

(4) 19.05 mm
12.7 mm
A325 Bolts
doubler plate

Figure 1 Sketch of the full-scale exterior WUF-B connection tested in this study.

2.1.Welding Procedure & Welding Details


The top and bottom flange of the beam were beveled as per AWS D1.1 [32] and the weld
access hole geometry is designed as per the recommendations in AISC Specification Section
J1.6 [1]. The welds were made using the Shielded Metal Arc Welding (SMAW) process and

92
E70T welding electrodes. While welding the bottom beam flange to the column flange the
web of the beam hinders the completion of a single pass across the width of the beam.
Whereas the top beam flange does not have any obstacle like the bottom beam flange.
Therefore, weld is not uniform in the bottom beam flange when compared to the top flange.
Further, the start and stop points in the bottom flange weld may cause discontinuities or
higher residual stresses in the joint that lead to low cycle fatigue failure. To gain an
understanding of the effect of the weld procedure on connection performance, welding was
closely monitored. Weld sequence, start and stop locations, bead size and length of the weld,
weld speed, weld direction, temperature cycles, voltage, and current were recorded. The
diameter of the electrode used in welding was 3.175 mm (1/8 inch). The voltage and current
measured at the arc were 28 volts and 160 amperes respectively. The polarity of welding
machine was set such that the electrode was negatively charged. The average velocity with
which the weld was laid was 2 mm/s.

Figure 2 Sketch of weld sequences at the beam top and bottom flange for specimen MC1
showing number of welding passes in order of completion.

93
There are numerous ways of filling the CJP groove welds of the bottom beam flange to
column flange when using a weld access hole. In the current research two such weld
sequence options were studied in order to investigate the effect of welding sequence on the
residual stress distribution as well as on the low-cycle fatigue response of WSMCs. These
two welding sequences were denoted as weld sequence A (for specimen MC1) and B (for
specimen MC2). The details of the bead size of the weld and the sequence in which the
passes and beads were placed on the beam bottom flange and beam top flange for specimens
MC1 and MC2 are presented in Table 1 and Table 2, and Figure 2 and Figure 3. As can be
observed, the top flange welds were deposited in a similar fashion (from one end of the
flange to the other end) for both the specimens, only the number of beads and the size of the
beads were different. On the other hand, the bottom beam flange welds were laid in a
completely different manner for the two specimens.

Table 1 Details of weld sequence for specimen MC1


Beam Bottom Flange
Direction of weld: Right to Left Direction of weld: Left to Right
Bead# Pass Length Depth Width Time Length Depth Width Time
Pass#
# (mm) (mm) (mm) (s) (mm) (mm) (mm) (s)
1 1 150 5 8 90 10 65 5 8 30
2 2 135 5 12 70 11 70 5 12 40
3 3 125 5 15 60 12 75 5 15 40
4 4 115 5 11 60 14 75 5 10 40
5 5 110 5 14 60 13 80 5 6 40
6 6 120 5 10 60 15 80 5 10 40
7 7 115 5 15 60 17 80 5 12 40
8 8 115 5 10 60 16 85 5 10 40
9 9 110 4 8 60 18 85 5 12 40
Beam Top Flange
Direction of Welding: Right to Left Direction of Welding: Left to Right
Bead# Length Depth Width Time Length Depth Width Time
Pass# Pass#
(mm) (mm) (mm) (s) (mm) (mm) (mm) (s)
1 1 195 5 10 130
2 2 195 5 6 130
3 3 193 5 10 130
4 4 195 5 6 130
5 5 193 5 15 130
6 6 195 5 6 130
7 7 190 5 8 130
8 8 195 5 10 130
9 9 195 5 8 130 10 195 5 12 130

94
In specimen MC1 with weld Sequence A, the welder makes all the passes in different
layers with the welding material, required to fill the groove starting from one side of the
beam web (as shown in Figure 2). After the root pass is made, the passes in the next layer are
stepped back slightly than the previous layer in order to avoid the start and stop points falling
on top of each other. Also, care is taken to prevent placing of the start and stop points directly
under the beam web. Once one side of the web is completed, the welder finishes the vacant
groove on the opposite side of the beam web (as shown in Figure 2). As a result of this weld
sequence, the average temperature on side of the beam, where the groove is filled first, is
higher than the average temperature on opposite side of the beam, where welding is done
later. This creates a non-uniform temperature gradient along the width of the beam flange.
When the weld is allowed to cool, the nonuniform cooling distorts the specimen.

Figure 3 Sketch of weld sequences at the beam top and bottom flanges for specimen MC2
showing number of welding passes in order of completion.

On the contrary, for specimen MC2 with weld Sequence B, the welder alternates sides of
the web after placing each weld bead. This sequence fills the groove evenly from the bottom
to the top as opposed to filling one side of the groove at a time. In this sequence, the welder

95
lays the first bead from one side of the beam web, and then completes the bead by moving to
the other side of the beam (Figure 3). All the beads in a particular layer of weld placed on
one side and are completed from the other side of the beam in the same order in which they
were laid first. Each layer is completed across the full width of the beam flange before
beginning of the next layer.

Table 2 Details of weld sequence for specimen MC2


Beam Bottom Flange
Bead Direction of weld: Right to Left Direction of weld: Left to Right
# Pass Length Depth Width Time Length Depth Width Time
Pass#
# (mm) (mm) (mm) (s) (mm) (mm) (mm) (s)
1 2 155 4 9 90 1 145 4 90
2 3 120 5 4 60 5 83 5 6 40
3 4 110 5 7 60 6 86 5 10 45
4 9 85 5 4 40 7 124 5 6 60
5 10 100 5 12 60 8 110 5 10 65
6 11 90 5 6 50
7 12 95 5 7 55
8 14 80 4 5 40 13 115 5 5 65
9 15 100 4 5 60 17 95 4 5 50
10 16 90 5 5 50 18 105 5 7 60
Beam Top Flange
Bead Direction of Welding: Right to Left Direction of Welding: Left to Right
# Length Depth Width Time Length Depth Width Ti
Pass# Pass#
(mm) (mm) (mm) (s) (mm) (mm) (mm) m
1 2 195 8 7 125 1 195 6 7 13
2
3 4 190 8 8 125 3 195 6 6 13
4
5 6 195 8 8 125 5 195 8 6 13
6
7 7 195 8 8 13
8 8 195 5 5 13
9 9 195 5 5 13
10 10 195 5 8 13

The next layer is then started from the same side as that the last bead was finished from
and pulled from the opposite side of the web to the outer edge of the flange. This bead is then
completed by switching sides of the beam and completing the pass (Figure 3). For each layer
the weld start and stop points are on the opposite side of the beam web when compared to the
previous layer. These two weld procedures may cause different residual stress distributions in

96
the bottom beam flange and the column flange. One of these weld sequence options can
adversely affect the fatigue life of the connection. Due to similar reasons, weld bead size can
also cause differences in fatigue life of the welded steel moment connection. None of these
factors are taken into consideration in the design of the connection, yet they can greatly
influence the service life of the joint.

2.2. Test Setup


As already mentioned, the beam-to-column sub-assembly was modeled as an exterior
beam-to column WUF-B connection. The dimensions and the support conditions of the
specimens were chosen to represent the theoretical inflection points of the beam and the
column in a typical frame building. The specimen supports were designed to mimic the
support conditions of hinge support. Figure 4 shows a schematic view of the test setup. Test
specimen was treated as a cantilever beam connected to a column flange. The specimen was
held vertically during testing to represent their position in a building. Thus, the column was
vertical and parallel to the strong wall. This position of the specimen is optimal because it
enables unobstructed access to both sides of the connection, provides for lateral bracing, and
keeps the expected failure zone convenient for necessary measurement. The far end of the
beam was connected to the actuator which was suspended from a large loading frame. Lateral
supports for the specimen were provided on either side of the beam using I-beams bolted
firmly to the loading frame. Lateral supports were provided for the beam at a distance of 2 m
57.4 mm (6 ft. 9 in.) from the centerline of the column. No lateral supports were provided for
the column. The loading is applied to the beam at a distance of 2 m 641.6 mm (8 ft. 8 in.)
from the face of the column. Great care was taken to minimize slipping between the test
setup elements and the laboratory floor and between the specimen and its supports.
The specimens were extensively instrumented in order to acquire the global and local
responses under the prescribed loading. The instrumentation pattern for the test includes a
combination of linear potentiometers (POTs), string POTs and LVDTs to monitor the global
response of the specimen: beam end displacement, joint rotation, panel zone shear
deformation and possible movement of the support. A variety of strain gages were used to

97
investigate the local response of the beam flanges, column panel zone and the continuity
plates. The strain gages on the flanges of the beam were moved closer to the weld toe to get a
better insight of the local strain response in the heat affected zone as this was the main area
of interest and the research is focused on the local failure mechanism near the weld. To
monitor the displacements along the length of the beam while testing, Optotrack was used.
The loading unit comprised of an actuator and a controller. The actuator used has a push-pull
hydraulic actuator manufactured by MTS Systems Corporation with 508 mm (20 inch) stroke
maximum capacity of 489 kN (110 kips) both in tension and compression. The actuator was
mounted vertically between the free end of the beam and the loading frame.

1.88 m

3.35 m

2.64 m
3.20 m

Figure 4 Schematic of the test setup used for cyclic testing of WUF-B Connections.

2.3. Loading History


In order to investigate the influence of welding sequence on the localized failure
mechanisms of WUF-B connections, the specimens were subjected to low-cycle fatigue

98
loading. Traditionally, cyclic tests are carried out with reference to standard procedures such
as [33-35] which use loading histories consisting of groups of cycles at increasing amplitudes
as shown in Figure 5a. Such loading histories present the advantage of allowing on a single
specimen a satisfactory appraisal of the cyclic performance of the component. However, they
appear inadequate for the development of cumulative damage models useful not only for the
seismic design of new structures but also for the assessment of safety and reliability of
damaged structures and for the development of adequate repair procedures. Hence it has been
proposed by various authors both in Europe and U.S. [34, 36-40] to perform tests with cycles
at a constant amplitude. Such a procedure presents the advantage of allowing a clear
understanding of the damage accumulation process as well as of the key parameters
governing it [38].

Figure 5 (a) SAC loading protocol [35], (b) non-standard loading history prescribed in the
experiments.

In this study a non-standard loading history was prescribed which was primarily
comprised of constant amplitude loading cycles in the inelastic range (Figure 5b). Initially
12 cycles were prescribed in the elastic range, six cycles each corresponding to a rotation of
0.375% and 0.5% in the elastic phase (denoted as phase 1 and 2) analogous to the elastic

99
phase of SAC loading protocol. Later in the next phase (phase 3), constant cycles equivalent
to a rotation of 1.5% was chosen till the crack initiation. For further crack propagation and
strength deterioration of the specimen, two cycles each of rotations corresponding to 2%, 3%
so and so forth until failure, were applied. It is interesting to note here that the amplitude of
constant loading cycles were kept slightly in the inelastic range, as higher amplitude of
loading in the inelastic range may produce local buckling of the beam flange and web which
hinders accumulation of damage due to low-cycle fatigue.

3. Cyclic Response of WUF-B Connections


The WUF-B specimens with two weld sequences showed ominously different low-cycle
fatigue responses during experimental investigation although they had similar test setup and
beam-column configurations. It is anticipated that the fatigue lives of the specimens were
different because of the different weld sequences in the CJP groove welds between beam and
column flanges. The results of the experimental investigation are presented in the following
articles in terms of global and local responses.

3.1. Global Response


Specimen MC1
The response of the specimen MC1 is plotted in terms of moment-total rotation plot as
shown in Figure 6a. The joint remained elastic during phase 1 and 2 of the loading which
corresponded to rotation of 0.375% and 0.5% respectively. With the onset of phase 3 loading
which corresponded to 1.5% rotation with constant amplitude loading cycles, yielding
occurred in beam web and beam flange. In addition small yielding occurred in the column
panel zone at the beginning. The response of the specimen was stable hysteretic closed loops
with no indication of drop in load. The failure was a sudden and abrupt fracture of the top
beam flange with a bang sound at 29th cycle during phase 3 loading. The failure could not be
assessed from the global response of the specimen. No sign or indication of distress was seen
in the specimen. Neither there was any crack initiation observed in the complete joint
penetration weld nor in the reinforcing fillet weld. The rupture in the top flange propagated

100
suddenly almost across the top flange as shown in Figure 6b. Another interesting observation
made during the test was that considerable amount of necking in the beam flanges were
observed very close to the weld toe which is a sign of significant plastic damage
accumulation. The panel zone and the column responded elastically and their contributions
towards the total rotation or the plastic rotation of the joint were low. Most of the plastic
rotation contribution to the joint was observed from beam. A review of the local response of
the beam flanges is necessary to assess and understand the sudden failure observed in this
test.

Figure 6 Response of specimen MC1. (a) moment-rotation hysteresis response, (b) rupture in
the beam top flange.

Specimen MC2
For the specimen MC2, elastic response was observed during the first two phases of
loading. When the phase 3 loading started stable hysteresis response was observed as shown
in Figure 7a. With the first cycle of 1.5% rotation, beam flange yielded with very little
yielding in column panel zone. No further yielding was observed in the column panel zone
indicating that the panel zone remained elastic. Similarly, the column contribution towards
the total rotation or the plastic rotation of the joint was very small and the column also

101
remained elastic. Hence most of the total rotation as well as the plastic rotation of the joint
were seen from the beam rotation, indicating higher distress in the welded joint. Visible
cracks were observed at the weld toe of reinforcing fillet weld in the bottom beam flange
(Figure 7b) and near the complete joint penetration groove weld along with the base metal-
weld metal interface of the top beam flange (Figure 8b) at 59th cycle during phase-3 of
loading cycle, which corresponds to a total rotation of 1.5%. The cracks propagated during
the phase-4 (2% rotation) of loading at both bottom and top beam flanges weld toe. The
cracks were visible at the edges of the flanges and the outside of the beam flange near the
weld.

Figure 7 Response of specimen MC2. (a) moment-rotation hysteresis response, (b) cracks in
the reinforcing fillet weld of bottom beam flange.

The specimen failed in a ductile manner due to gradual tear in the beam top flange,
midway of the weld access hole during phase-5 (rotation 3%) of the loading (Figure 8b). The
rupture in the top flange was initiated at the flattened part of the weld access hole in the top
flange as shown in Figure 8a. The specimen sustained a total rotation of 3% at a ductility
level equal to 4 prior to the top flange failure. The ductile type failure can also be assessed
from the drop in the load seen during testing (Figure 7a). Neither local buckling of the beam

102
flanges nor the plastic hinge formation took place in the specimen during testing. As the
beam was securely braced, lateral torsional buckling was insignificant during testing of the
specimen. Prior to failure, the beam flanges yielded during first few cycles of phase 3 of the
loading up to a distance of half the beam depth from the column face.

Figure 8 Response of specimen MC2. (a) rupture of the top flange initiated at the flattened
portion of the weld access hole, (b) cracks in the weld-base metal interface of top beam
flange.

From the experimental responses, it is apparent that both the specimens showed localized
low-cycle fatigue failure occurring close to the weld toe of the beam top flange. Similar type
of failure was also observed during the failure analysis of the post-Northridge connection
tests conducted at the University of Michigan and Lehigh University. These studies
demonstrated that fatigue cracks were initiated either near the bottom weld access hole or the
beam flange weld toe (SAC/BD-99/23, SAC/BD-00/24). In the University of Michigan tests
89% of the connections showed crack initiation near the weld access hole and 50% showed
crack initiation at the weld toe. In the Lehigh tests, 80% of the fatigue cracks were initiated at
the weld toe. Based on the responses of the tests conducted and from the literature it is
evident that the connection regions around the beam flange weld toe and the weld access hole
are prone to fatigue crack initiation during low-cycle fatigue loading, and hence, it is

103
essential to investigate the cause and mechanism of such mode of failure. Although both the
specimens tested showed low-cycle fatigue failure close to the weld toe, the fatigue lives of
the specimens were different, where specimen with weld sequence A withstood only 29
loading cycles prior to failure, whereas, specimen with weld sequence B survived 61 loading
cycles. However, it is interesting to note that the specimens resembled each other in every
aspect, except the welding procedure, during the design and test setup for the connections.
From the global response of the connections tested, it is apparent that the global behavior was
not influenced by the welding procedure, hence the difference in the fatigue lives as well as
the mode of failure for two different welding sequences might be attributed to the different
localized failure mechanisms induced by different welding procedures. Hence, it is essential
that the local response of the connections be investigated to understand the localized failure
mechanisms involved.

3.2. Local Response


In order to investigate the local response of the connections, strain gage measurements
were taken at different locations of the connections. The strain gage measurements were
presented as mean strain and strain amplitude to elaborate the local response of the
connections. Mean strain gives an indication of plastic damage accumulation with the
progression of loading cycles, whereas strain amplitude gives an indication of the magnitude
of imposed loading. Although several strain gages were used in the experiments, results of
only few of the strain gages are presented in this paper. The axial mean strain and strain
amplitude responses for beam top flange, beam bottom flange and column flange for both
weld sequences are presented in Figure 9 through Figure 12, where the locations of the strain
gages are clearly indicated in the insets. In each of the plots, the dark strain gage indicates the
corresponding strain measurement location. For both the specimens, larger strain has
developed in the beam flanges and column flange close to the weld toe compared to the other
locations of the connection. Large amount of axial strain was also observed around the weld
access hole region. Despite the symmetrical total response due to the symmetrical imposed

104
displacement a lack of symmetry was observed in strain gage responses close to the weld toe
and weld access hole.

Figure 9 Mean axial strain and amplitude plotted against number of loading cycles recorded
at the top flange (inside surface) of the connections.

It was observed that for both the specimens, the axial mean strain and amplitude were
increasing slowly with number of loading cycles at every strain gage location during loading
phases 1 and 2 i.e. during elastic loading cycles. As the loading amplitude was increased

105
from elastic to inelastic cycles during loading phase 3, there was a sharp jump in the axial
mean strain and amplitude, and after that jump, the axial mean strain continued to increase
gradually with number of loading cycles, whereas the axial strain amplitude either decreased
or remained constant. As a result, axial mean strain accumulated close to the weld toe and
around the weld access hole both at the top (Figure 9) and the bottom beam flanges (Figure
10 and Figure 11).

Figure 10 Mean axial strain and amplitude plotted against number of loading cycles recorded
at the beam bottom flange (inside surface) of the connections.

106
Considerable amount of axial mean strain accumulation was also observed in the column
flange close to the weld toe (Figure 12). Since the cracks were initiated near the weld toe
region, it is anticipated that the axial strain accumulation with number of loading cycles
might have led to fatigue crack initiation near weld toe of the beam flanges of the
connections. Because of the axial strain accumulation near the weld toe region, necking of
the beam flanges near the weld toe was observed for the both the specimens tested which is a
sign of large plastic damage accumulation (Figure 7b, Figure 8b). This phenomenon of axial
strain accumulation with number of loading cycles is well known as ratcheting or sometimes
referred to as cyclic creep. Ratcheting is known to occur in metallic materials as they
undergo inelastic cyclic loading, along the directions of force-controlled loading [24, 41-42].
It is quite interesting to find that the ratcheting in the tests reported herein has occurred under
displacement-controlled loading which is kinematically inadmissible. Ratcheting
phenomenon under displacement controlled loading is only possible if there are prior mean
stresses that work as a driving force for ratcheting. Similar type of ratcheting response near
the weld toe of the SS304 piping joints has been demonstrated by Lu [21] and Cheng [23]
and it was concluded that residual stresses at the toe of the weld might be a key factor for the
occurrence of ratcheting under displacement controlled loading. Castiglioni [38] also
demonstrated presence of ratcheting near welded joint in the steel beam-column connection
under constant amplitude loading cycles and a decreasing trend in the amount of ratcheting
was observed as the distance from the weld toe was increased. Similar response was also
observed in the tests of this study, where the amount ratcheting was decreasing as the
distance from the weld toe increased. From the observations of the study it is evident that the
stresses that is working as driving force for ratcheting is not prevailing away from the weld
toe. Hence it is anticipated that the axial residual stress at the weld toe acts as a mean stress
to the stress cycle in the axial direction and causes ratcheting in this direction. When the
strain responses at different strain gage locations were compared between the two specimens
with different weld sequences, the lower fatigue life of the specimen MC1 is justified. The
specimen MC1 showed higher mean axial strain compared to the specimen MC2 close to the
weld toe of the top flange of the beam as shown in Figure 9, indicating higher axial strain

107
accumulation at the weld region of specimen MC1 compared to MC2 (Figure 9). This
demonstrates that specimen MC1 might have induced more residual stresses than specimen
MC2 at the beam top flange.

Figure 11 Mean axial strain and amplitude plotted against number of loading cycles recorded
at the beam bottom flange (outside surface) of the connections.

Similarly, the strain responses at most of the locations close to the weld toe of the beam
bottom flange showed higher mean axial strain for specimen MC1 with weld sequence A

108
compared to specimen MC2 with weld sequence B (Figure 10 b, c, d and Figure 11 a, b),
whereas there were few locations where mean axial strain response of specimen MC1 was
lower than that of specimen MC2. Hence, it is evident that the mean axial strain or axial
strain accumulation is significantly influenced by the welding procedure. On the other hand,
the influence of welding procedure on the axial strain amplitude response was inconclusive.
There was no specific trend observed in the axial strain amplitudes for both the specimens
based on the weld sequence. There were several locations at beam top and bottom flanges
where specimen MC1 showed higher axial strain amplitude compared to that of specimen
MC2 (Figure 10b, Figure 11), whereas there were several other locations where specimen
MC2 exhibited larger axial strain amplitude compared to that of specimen MC2 (Figure 9c,
Figure 10a). The local strain responses in the column flange near the weld toe showed that
the mean axial strain and amplitudes were comparable for the two specimens near the beam
top flange, whereas specimen MC1 showed higher mean and lower amplitude near beam
bottom flange (Figure 12).

Figure 12 Mean axial strain and amplitude plotted against number of loading cycles recorded
at the column flange of the connections.

109
Therefore, based on the local strain responses obtained from this study it is apparent that
mean axial strain plays more important role on the localized fatigue failure mechanism of the
connection compared to axial strain amplitude. For specimen MC1, axial strain accumulation
was higher in the welded region of the connection compared to specimen MC2, and hence,
specimen MC1 exhibited lower fatigue life than that of specimen MC2. From the
experimental observations, it is believed that the mechanism that could have influenced the
axial strain accumulation/ratcheting as well as the fatigue lives of the specimens is the
residual stresses induced from the welding procedure. Therefore, it is anticipated that the
welding procedure might have significant influence on the magnitude and distribution of
residual stresses in the beam-column connections, which eventually influences the low-cycle
fatigue response of the connections.

4. Conclusions
Two full-scale WUF-B connections were tested under displacement controlled constant
amplitude cyclic loading to investigate the effect of weld sequence on the low-cycle fatigue
failure mechanism of the connections. The results of the study led to the following
conclusions:
Welding procedure does not influence the global response of the connections and
hence, it is essential that the local response of the connections be investigated along
with the global response to understand the failure mechanism of the connections.
Axial strain accumulation/ratcheting was observed close to the weld toe for both the
specimens under investigation. Axial strain ratcheting under displacement controlled
loading is kinematically inadmissible, and hence it is anticipated that welding
induced residual stresses might be working as mean stress to the stress cycle in the
axial direction and causing ratcheting in this direction.
Different welding sequences might have induced different amount and distribution
of residual stresses in the connection regions. Hence the ratcheting phenomena were
different based on the welding procedures which eventually led to variability in the
fatigue lives of the connections.

110
The weld sequence where the welder switches sides of the beam web before welding
each bead showed better performance compared to the one where the welder
finishes all welding beads on one side of the beam web before switching to the other
side.
Welding induced residual stresses and the procedure of welding has significant
effect on the low-cycle fatigue response of the welded steel moment connections. It
is believed that based on the analysis of a large number of welding sequences, an
optimized weld sequence can be achieved which will reduce the residual stresses
significantly and thereby, increase the fatigue life of the connection.

5. References
[1] AISC. Manual of steel construction load and resistance factor design. 13th ed. Chicago,
IL, 2006.

[2] AISC. Prequalified connections for special and intermediate steel moment frames for
seismic applications. American Institute of Steel Construction 358, AISC 358-10, 2010a

[3] AISC. Seismic provisions for structural steel buildings. American Institute of Steel
Construction 341, AISC 341-10, 2010b.

[4] SEAOC. Recommended lateral force requirements and commentary 7th edition.
Structural Engineers Association of California Seismology Committee, 1999.

[5] Rodgers JE, Stephen AM. Effects of connection fractures on global behavior of steel
moment frames subjected to earthquakes. Journal of Structural
Engineering2006;132(1):78-88.

[6] FEMA-350. Recommendation and seismic design criteria for new steel moment-frame
buildings. prepared by the SAC Joint Venture for the Federal Emergency Management
Agency, Washington, DC, 2000.

111
[7] FEMA-353. Recommended specifications and quality assurance guidelines for steel
moment- frame construction for seismic applications. prepared by the SAC Joint Venture
for the Federal Emergency Management Agency, Washington, DC, 2000.

[8] FEMA-355D. State of the art report on connection performance. prepared by the SAC
Joint Venture for the Federal Emergency Management Agency, Washington, DC, 2000.

[9] Suita K, Nakashima M, Morisako K. Tests of welded beam-column subassemblies II:


Detailed behavior. Journal of Structural Engineering 1998;124(11):1245-1252.

[10] Nakashima M, Suita K, Marisako K, Maruoka Y. Tests of welded beam-column


subassemblies. I. Global behavior. Journal of Structural Engineering 1998;124:1236-
1244.

[11] SAC/BD-99/23. Failure analysis of welded beam to column connections. by Barsom, JM,
Pellegrino JV.

[12] Mazzolani FM. (Editor). Moment resistant connections of steel frames in seismic areas:
design and reliability. E&FN Spoon, 2000.

[13] Stojadinovic B, Goel SC, Lee KH, Margarian AG, Choi JH. Parametric tests on
unreinforced steel moment connections. Journal of Structural Engineering 2000;126:40-
49.

[14] SAC/BD-00/24. Development and evaluation of improved details for ductile welded
unreinforced flange connections. by Ricles JM, Mao C, Lu LW, Fisher JW.

[15] SAC/BD-00/22. Cyclic response of RBS moment connections: loading sequence and
lateral bracing effects. by Yu S, Gilton C, Uang CM.

112
[16] Uang CM, Yu QS, Noel S, Gross J. Cyclic testing of steel moment connections
rehabilitated with rbs or welded haunch. Journal of Structural Engineering 2000;126:57-
68.

[17] Ricles JM, Fisher JW, Lu LW, Kaufmann EJ. Development of improved welded moment
connections for earthquake-resistant design. Journal of Constructional Steel Research
2002;58:565-604.

[18] Stojadinovic B. Stability and low-cycle fatigue limits of moment connection rotation
capacity. Engineering Structures 2003;25:691-700.

[19] Lee D, Cotton SC, Hajjar JF, Dexter RJ. Cyclic behavior of steel moment-resisting
connections reinforced by alternative column stiffener details, I. Connection performance
and continuity plate detailing. Engineering Journal 2005;4th Quarter:189-213.

[20] SEAOC. Commentary and recommendations on FEMA 350. Structural Engineers


Association of California Seismology Committee, FEMA 350 Task Group, 2002.

[21] Lu X. Influence of residual stress on fatigue failure of welded joints. Ph.D. Dissertation,
North Carolina State University, Raleigh, NC, 2002.

[22] Humphreys AE. Influence of residual stress on the initiation of fatigue cracks at welded
piping joints. MS Thesis, North Carolina State University, Raleigh, NC, 2004.

[23] Cheng PY. Influence of residual stress and heat affected zone on fatigue failure of welded
piping joints. Ph.D. Dissertation, North Carolina State University, Raleigh, NC, 2009.

[24] Hassan T, Kyriakides S. Ratcheting of cyclically hardening and softening materials: I.


Uniaxial behavior. International Journal of Plasticity 1994;10:149-184.

[25] SAC/BD-99/24. Weld acceptance criteria for seismically-loaded welded connections. by


Mohr W.

113
[26] SAC/BD-97/05. Finite element fracture mechanics investigation of welded beam-to-
column connections. by Chi WM, Deierlein GG, Ingraffea A.

[27] Matos CG, Dodds Jr RH. Probabilistic modeling of weld fracture in steel frame
connections part ii: seismic loading. Engineering structures 2002;24(6):687-705.

[28] SAC/BD-99/04. Effects of strength/toughness mismatch on structural and fracture


behaviors in weldments. by Dong P, Kilinski T, Zhang J, Brust FW.

[29] Zhang J, Dong P. Residual stresses in welded moment frames and implications for
structural performance. Journal of Structural Engineering 2000;126(3):306-315.

[30] Royster PB. Localized low cycle fatigue failure of welded steel moment connections. MS
Thesis, North Carolina State University, Raleigh, NC, 2007.

[31] Syed S. Influence of weld sequence on the seismic failure of welded steel moment
connections in building structures. MS Thesis, North Carolina State University, Raleigh,
NC, 2009.

[32] ANSI/AWS D1.1. Structural welding code-steel. The American Welding Society, 2008.

[33] European Convention for Constructional Steelworks, ECCS. Recommended testing


procedure for assessing the behavior of structural elements under cyclic loads. TWG
1.3Seismic Design, ECCS Publication No. 45, Technical Committee 1, 1986.

[34] ATC. Guidelines for cyclic seismic testing of components of steel structures. ATC-24,
Redwood City, Calif., 1992.

[35] SAC/BD-97/02. Protocol for fabrication, inspection, testing and documentation of beam-
column connection tests and other experimental specimens. by Clark P, Frank K,
Krawinkler H, Shaw R.

114
[36] Krawinkler H, Zohrei M. Cumulative damage in steel structures subjected to earthquake
ground motions. Computers and Structures 1983;16:531541.

[37] Ballio G, Castiglioni CA. Seismic behavior of steel sections. Journal of Constructional
Steel Research 1994, 29:21-54.

[38] Castiglioni CA. Effects of the loading historey on the local buckling behavior and failure
mode of welded beam-to-column joints in moment-resisting steel frames. Journal of
Engineering Mechanics 2005;131:568-585.

[39] Mander JB, Chen SS, Pekan G. Low-cycle variable amplitude fatigue modelling of top-
and-seat angle connections. Engineering Journal 1994a;2:5462.

[40] Mander JB, Chen SS, Pekan G. Low-cycle fatigue behavior of semi-rigid top-and-seat
angle connections. Engineering Journal 1994b;3:111122.

[41] Corona E, Hassan T, Kyiakides S. On the performance of kinematic hardening rules in


predicting a class of biaxial ratcheting histories. International Journal of Plasticity
1996;12(1):117-145.

[42] Hassan T, Kyriakides S. Ratcheting of cyclically hardening and softening materials: II.
multiaxial behavior. International Journal of Plasticity 1994;10:185-212.

115
CHAPTER 5: SIMULATION OF INITIAL RESIDUAL STRESSES IN
HOT-ROLLED WIDE FLANGE SHAPES

Abstract
After the wide flange members are hot-rolled to their final shape they are air cooled
during which residual stresses are developed due to nonuniform cooling. No numerical
methods are available in the literature to simulate the residual stresses developed in the W
shapes. Because of the complexity involved in the residual stress calculation, most of the
works in the literature either ignore the presence of residual stresses or assume simplified
distribution of residual stresses. However, the residual stresses developed due to uneven
cooling (denoted as initial residual stresses) can be as high as one-half of the yield stress and
can affect significantly the load carrying capacity and deformation of structural members
especially under fatigue loading. Hence, it is essential to accurately predict the magnitude
and distribution of initial residual stresses to investigate its influence on the structural
performance. This study developed a numerical scheme to simulate the real time residual
stress distribution in hot-rolled W shapes. The numerical technique developed for simulating
residual stress distribution for different size W shapes was validated against the initial
residual stress data in the literature. Good conformity was observed between the simulated
and experimental responses. The simple numerical scheme developed in this study provides
experimentally validated technique for simulating the initial residual stresses in hot-rolled W
shapes for structural analysis and design of moment resisting connections.

Keywords: Finite element, W shape, uneven cooling, hot-rolled, residual stress

1. Introduction
Internal residual stresses develop in hot-rolled W shape sections during manufacturing
processes of hot-rolling and subsequent cooling. Such self-equilibrated residual stresses are
developed because of nonuniform cooling and nonuniform geometry. When W shape section

116
with initial residual stresses is loaded, the total stresses due to the external forces are
superimposed on initial residual stresses and consequently certain portions of the cross-
section may be stressed beyond the yield point [1-4]. Furthermore, the fatigue life will be
shortened due to the existence of residual stresses when the member is subjected to dynamic
loads such as seismic or wind loads. The effect of initial residual stresses on the performance
of steel structures has been addressed in quite a few literature. It has been shown that the
ultimate strength of beams is not affected by the presence of residual stress, although
deflections are increased near the ultimate [5-6]. For beam with intermediate slenderness
ratio, the influence of residual stresses was found significant in the lateral-torsional buckling
of the beams [6]. In columns, residual stresses lower the ultimate strength and for this reason
their evaluation is of great importance [6, 8-10]. Beedle and Tall [10] observed 30%
reductions of the compressive strength of structural steel columns made of W shapes or built-
up steel members. The numerical study of Lamarche and Tremblay [11] showed that residual
stresses reduce the compressive resistance of columns, especially at the first buckling.
Residual stress effects tend to gradually diminish in the post-buckling range, which suggests
that they may not be critical for prediction of the collapse of structures once the column
capacity is exceeded. Similar observations were also made in the studies by Okazaki et al.
[12], Lu and MacRae [13] and Mathur et al. [14] where the effects of initial residual stresses
were evaluated under cyclic loading response of steel framed structures. These studies
showed a decrease in the lateral strength of the structure when high axial compressive loads
were considered. However, the effects of residual stresses on cyclic stability behavior were
noticeable only after the peak lateral strength has been reached and afterwards, the effect of
initial residual stresses seemed to diminish with the progression of large inelastic cycles.
From the literature, it is apparent that initial residual stresses can be an important factor in
buckling [15], fatigue [16], stress corrosion [17] and brittle fracture [18] of structural
members. Therefore, the actual magnitude and distribution of residual stresses as well as the
factors which influence the formation of residual stresses are of considerable significance.
An investigation of residual stresses can be experimental, analytical or numerical.
However, the actual distribution of residual stresses in a single specimen can be verified only

117
by experimental measurements. The accurate measurements of residual stresses through
experiments are very tedious and expensive, and it is practically impossible to carry out
measurements on more than a few specimens out of all existing types of different shapes.
Moreover, the residual stress distribution may vary from specimen to specimen, implying
several measurements to be made on each shape to obtain a statistically significant result
[19]. On the other hand, analytical attempts have been made in some studies to calculate the
residual stresses which resulted in too simplified distribution of residual stresses and are not
widely applicable [2-3, 20]. Considering these facts, numerical methods appear to be useful
for the investigation of residual stresses and the variables which influence their formation
provided that the initial material characteristics and boundary conditions are properly
assumed. Several researchers showed that the finite element method could be used to
simulate the real processes by comparing the results with some experimental data.
Kamamoto et al. [20] have analyzed residual stresses and distortion of large steel shafts due
to quenching. The results showed that residual stresses are strongly related to the
transformational behavior. Toparli and Aksoy [21] analyzed residual stresses during water
quenching of cylindrical solid steel bars of various diameters by using finite element
technique. Jahanian [22] modeled heat treatment (quenching) and calculated the residual
stress in a long solid cylinder by using theoretical and numerical methods with different
cooling speeds. Fa-rong and Shang-li [23] used finite element program to analyze the
transient temperature and residual stress fields for a metal specimen during quenching.
Yamada [24] presented a method of solving uncoupled quasi-static thermoelastic problems in
perforated plates. In their analysis, a transient thermal stress problem was solved for an
infinite plate containing two elliptic holes with prescribed temperature. In al1 these models,
many assumptions were made to simplify the actual process. That is obviously too complex
to simulate all the parameters in details. However, reasonably good models [21, 24-25] were
developed in these past investigations by considering the key factors that affect the formation
of residual stresses. Many of these models produced results that had excellent agreement with
the results obtained from the experiments. However, no numerical studies were found in the
literature that dealt with the distribution of residual stresses developed during cooling of hot-

118
rolled steel sections. This necessitates the development of a numerical scheme to accurately
predict the formation and distribution of residual stresses due to uneven cooling of hot-rolled
steel sections which is the subject of this study. This study developed a scheme to
numerically simulate the initial residual stresses (IRS) of structural W shapes during the
cooling process after hot-rolling through advanced finite element modeling techniques. The
method is based on a calculation of the temperature and the resulting thermal stress
distribution throughout the cooling process, and simulating the actual cooling conditions,
including variable material coefficients. A sequentially coupled nonlinear transient thermo-
mechanical analysis was performed to calculate temperature history and residual stresses.
First a transient thermal analysis was performed during which the time-dependent
temperature distribution was determined during the cooling phase of the hot-rolled W shapes.
The temperature field at each time step was then used as an input for evaluating the stress
and strain field in the W shapes. The residual stresses obtained from the finite element
simulation was validated against the experimental data available in the literature.

2. Manufacturing Process of Structural Shapes and Formation of Initial Residual Stress


The schematic diagram of Figure shows the production process of structural steel shapes,
which consists of three main operations: melting, casting and rolling. To begin production,
the scrap material is melted in highly efficient and environmentally friendly electric arc
furnaces. The steel is then refined in a ladle furnace, and its temperature and chemistry is
regulated before the casting process. In a continuous caster, the steel is solidified to a semi-
finished product called a beam blank which is similar in shape to a steel beam. This is done
through a hollow interior of an oscillating mold (shown in Figure 2a) that have inside
dimensions corresponding to the width and thickness of the billets, beam blanks, and near-net
shape profile that is being cast. After casting, the beam blanks are directly reheated to rolling
temperature or stocked for later rolling. Depending upon the temperature of the entering
beam blank it may be necessary to reheat it in a furnace to a temperature of 1200-1300C.
The beam blank is then rolled in a number of passes in two or more rolling stands. In each
pass the cross section is reduced in a certain manner until the shape has reached its final cross

119
section in the finishing stand. After finish-rolling, the material is conveyed to a hot saw
where it is cut into lengths which can be conveniently transported. The member is then
allowed to cool down to ambient temperature on a cooling bed, that is, a metal structure
which permits the air to circulate around the rolled members. The detailed arrangement of the
shapes on the cooling bed varies with different mill practice and different shape of the
material. Residual stresses will result after the cooling because of the non-uniform
temperature distribution through the cross section during the cooling process. Figure 2b and
Figure 2c show sketches of the cross-sectional view of the semi-finished product after
continuous casting (beam blank) and finished product after all the manufacturing operations,
respectively.

Scrap/sub- Finished Products


materials

Electric Arc Furnace Inspection Bed

Ladle Furnace Roller Straighteners

Continuous
Casting Cooling Bed
Machine

Reheating
Bloom/Beam Blank Rolling Stands Hot Saw
Furnace

Figure 1 Schematic diagram showing the manufacturing process of structural steel shapes
[20, 26].

Due to uneven cooling conditions on the cooling bed, many members are crooked after
cooling. Therefore, most shapes have to be straightened in some manner. This is
accomplished in a roller-straightener where the member is passed through a number of rolls

120
which bend the shape in alternating directions about its weak axis. The straightening
operations are expected to influence the initial residual stress pattern; however, practical
measurements have indicated that residual stresses in delivered shapes often show the basic
pattern to be expected from the cooling process [19].

(a) (b) (c)


Figure 2 (a) Beam blank continuous caster mold, (b) cross-sectional view of beam blank, (c)
cross-sectional view of finished product after hot-rolling.

3. Finite Element Modeling of the Cooling Process


In order to simulate the residual stresses developed due to uneven cooling of hot rolled W
shapes, finite element method has been implemented in this study. Only the residual stresses
developed during the cooling phase of the W shapes was simulated, and the process of hot-
rolling was not accounted for. However, rolling would be important if initial material state
need to be calculated. In cooling process simulation, sequentially coupled, nonlinear,
transient, thermo-mechanical analyses were employed to calculate temperature history and
initial residual stresses. Initially a transient thermal analysis was performed on structural W
shapes to get a real time temperature distribution during the cooling process of reheated
rolled specimens. The temperature field at each time step is then used as an input for
evaluating the stress and strain field in the W shapes. In this approach the heat transfer
analysis is solved independently of the stress analysis. The link between the heat transfer
analysis and stress analysis is obtained through the temperature history, which is input as the

121
thermal load in the residual stress calculation. The details of the thermal and residual stress
analyses are presented next.

3.1. Thermal Analysis


In the thermal finite element analysis, 3D finite element models of structural W shapes
were generated by using nonlinear finite element software ANSYS. The W shape model was
discretized by twenty noded thermal solid elements with one degree of freedom, temperature,
at each node. Four different size W shapes were chosen for the finite element simulation, viz.
W1443, W1650, W27102, W36150. These sections were chosen since residual stress
measurements were available in the literature to validate the finite element simulation results
[1, 3, 20]. Figure 3 shows a typical finite element mesh developed for the thermal finite
element analysis of a 2.7m long W36150 beam. For the thermal finite element analysis,
temperature was given as input which was assumed to be constant and uniform all over the
body before the start of the cooling process. The initial uniform temperature that was
assigned to the W shape specimen was 1300C which falls within the range of the reheating
temperature (1200C-1300C) during rolling of the specimens. The temperature was kept
constant for 10 minutes before air cooling of the specimen. It is to be mentioned here that air
cooling was simulated in thermal finite element analysis by allowing free heat transfer
(convection, conduction, radiation into air) in all directions.

Figure 3 Finite element mesh for thermal finite element analysis of W36150 shape.

122
During heating and slow air cooling, the governing equations follow the differential
equations of three dimensional heat flow in a solid body under non-stationary conditions
which is given in Eq. 1. This is a generalized Fourier heat-conduction equation [26] and is
valid for the interior of the body.
T T T T E y
k k k Q cp x z
(1)
x x y y z z t 1 2 t t t

where k = thermal conductivity, Q = generated heat energy, = density, cp = specific heat, T


= temperature, E = modulus of elasticity, = coefficient of linear expansion, = poissons
ratio, = mechanical strain, t = time, x, y, z = Cartesian coordinates.
The boundary condition for solving Eq. 1 is given in Eq. 2 and is valid at the surface.
T
k h T Tatm (2)
n
where h = surface coefficient of heat transfer, n = coordinate normal to surface. However, for
solving the heat flow equation numerically, in the thermal finite element analysis it was
assumed that the material is homogeneous and isotropic throughout the cross section and the
material coefficients are known as functions of temperature.
For a body subjected to cooling in the atmosphere with no sharp variation in the time-
temperature history, the coupling effect of temperature and strain can be neglected [27].
Hence, the 2nd term in the right hand side of Eq. 1 can be neglected. Furthermore, Q=0 since
there is no heat generated in the body. These assumptions simplified Eq.1 as follows:
T T T T
k k k cp (3)
x x y y z z t
After discretization over the region, the heat-conductivity equation can be reduced to the
following matrix equation [28]:
C T K T Q (4)

Where {T} is the vector of nodal temperatures, [C] is the heat-capacity matrix, {K} is the
heat conductivity matrix, {Q} is the vector of heat energy which is zero for this case. Eq. 4 is
integrated numerically to get the temperature distribution during the entire cooling process of

123
the W shape specimen. The material coefficients that are required to solve Eq. (4) are
density, thermal conductivity, specific heat and surface coefficient of heat transfer.

Figure 4 Material coefficients of structural steel as a function of temperature used in thermal


finite element analysis. (a) Specific heat [31], (b) thermal conductivity [31], (c) heat transfer
coefficient [32], (d) enthalpy [33].

Of the material coefficients, density was assumed constant, whereas other material
coefficients were made as functions of temperature in the thermal finite element analysis.
Latent heat/phase transformation was accounted for by defining enthalpy of the material as a

124
function of temperature [30-31]. Figure 4 shows the material coefficients used in the thermal
finite element analysis. The discontinuity of specific heat at 650-760C is a result of the
phase transformation from ferrite ( -Fe) to austenite ( -Fe). The surface coefficient of heat
transfer includes the effect of convection, conduction and radiation. Using these material
coefficients along with the method outlined above, thermal finite element analyses were
performed on the aforementioned four different sizes of W shape specimens to get the real
time temperature distribution during the cooling process after hot-rolling.

(a) (b) (c)

Figure 5 Contour showing temperature distribution (in K) during the cooling process of hot
rolled W36150 section from initial 1300C reheat temperature. (a) after 15 minutes, (b) 30
minutes, (c) 45 minutes.

It was interesting to note that the general trend of temperature distribution for all the W
sections was similar during the cooling process, although variations were observed in the
magnitudes and gradients of temperature at different locations for different size W shapes.
However, all the sections cooled down to room temperature from initial 1300C reheat
temperature within two hours regardless of the section size. Figure 5 shows the temperature
distribution for a W36150 section at different stages of cooling upto 45 minutes. In the
beginning of the cooling process, the flange tips and web cooled much faster than the web-
flange junction and neighboring region. Also the cooling process was much faster in the edge
of the beam flanges and mid-depth of the beam web compared to the mid-section flanges of
the beam. When the coolest part reached the phase transformation temperature (650-760C),
the cooling curves tend to level out due to the transformation heat (Figure 6a). After the

125
transformation was finished, the cooling curves then proceeded with a slope almost as if they
were a continuation of the curves before the transformation, the cooling only being delayed
for a certain time. The curves continue smoothly until the temperature of the whole cross
section eventually approached the ambient temperature. In Figure 6b, it can be seen that there
is a fairly large temperature gradient through the cross section. These temperature differences
were the cause of residual stresses to be formed during cooling. The temperature differences
increased at first very rapidly, reach a maximum and then start to decrease indicating that the
material transformation has started in the coolest point. After a time the curves increased
again (for the flange points a very small increase) corresponding to the situation when
transformation has started in the hotter point while the transformation is finished in the other
point.

Figure 6 (a) Cooling curves at three different locations over the cross section of a W36150
section, (b) temperature difference as a function of cooling time over the cross section of a
W36150 section.

A new maximum is then reached, marking the situation where the transformation is
finished in both points, and after that the temperature differences decrease gradually to zero
as ambient temperature is reached. These observations from the finite element analysis are in

126
well agreement with the experimental results reported by Alpsten [19] for European shape
HE 200 B. This validates the fidelity of the developed thermal finite element analysis. This
demonstrates that the implemented analysis technique provides the time-temperature history
during cooling process with a reasonable accuracy. This, however, requires the thermal
parameters in Figure 4 to be determined accurately and the actual 3D geometry of the model
needs to be accurate.

3.2. Residual Stress Analysis


The residual stress analysis was conducted after the thermal analysis was finished and the
time-temperature history during the cooling process of the structural W shape was
determined. The temperature solution obtained from the thermal analysis, which normally
depends on position and time, was input into the stress analysis as a predefined field i.e. the
nodal temperature field at each time step from the thermal analysis was applied as thermal
load to the corresponding node in the structural model. As a result, the node labels for the
structural mesh must match those from the heat transfer mesh. This was accomplished by
changing the element types in the thermal model from thermal solids to structural solids by
using ETCHG command available in ANSYS. Once the real time temperature fields were fed
into the structural model for each of the nodes at each time step, then residual stresses were
calculated based on the temperature dependent material properties of structural steel and
relevant constitutive equations. Since the experimental residual stress data available in the
literature was based on ASTM A36 steel, this steel grade was chosen for validation of the
residual stress simulation technique. It is to be mentioned here that in the finite element
analysis for residual stress calculation both material and geometric nonlinearity was
incorporated in the finite element models.
The material properties used for the residual stress analysis in finite element simulation
were density, initial yield stress, elastic modulus, Poissons ratio and thermal expansion
coefficient for structural steel. The density was assumed constant, whereas all other material
parameters were used as a function of temperature. Figure 7 shows the variation of Poissons
ratio and thermal expansion coefficient for structural steel as a function of temperature. The

127
initial yield stress and the elastic modulus were obtained from the elevated temperature
stress-strain response of ASTM A36 steel under monotonic tension loading. The monotonic
stress-strain response of ASTM A36 steel was obtained from Harmathy and Stanzak [34]
upto 650C Figure 8a. For temperatures over 650C, no stress-strain data was available in the
literature for ASTM A36 steel.

Figure 7 (a) Poissons ratio [36] and (b) thermal expansion coefficient [37] of structural steel
plotted as a function of temperature.

However, monotonic stress-strain data was available for ASTM A992 steel at
temperatures over 650C [35] as shown in Figure 8b, and this data was scaled to obtain the
stress-strain response of ASTM A36 steel at temperatures over 650C as shown in Figure 8a.
The elevated temperature stress-strain response of ASTM A36 steel was implemented in the
finite element simulation by using multilinear kinematic hardening model available in
ANSYS which incorporates von-Mises yield criterion and associative flow rule as given by
Eq. 5 and Eq. 6 respectively.
1/ 2
3
f( ) s a s a 0 0 (5)
2

p f
d d (6)

128
where is the stress tensor, s is the deviatoric stress tensor, a is the current center of the yield
surface in the deviatoric stress space, 0
is the initial size of the yield surface, is plastic

multiplier which corresponds to the equivalent plastic strain increment as


2 p p
dp d d (7)
3

Figure 8 (a) Monotonic stress-strain response of ASTM A36 steel at different temperatures,
(b) monotonic stress-strain response of ASTM A992 steel at different temperatures [35].

In the solution of residual stress fields, it is usually assumed that the total strain increment
can be decomposed as
e p T
d d d d (8)
e p T
where d ,d and d are the elastic, plastic and thermal strain increment respectively.
Based on the von mises yield criterion, the constitutive relation can be written as
d D d C th dT (9)

where d is the stress increment, D is consistent tangent modulus matrix, d is strain

increment, Cth is thermal stiffness matrix, and dT is temperature increment.

129
4. Results and Discussion
The initial residual stress simulations were performed for four different W sections
(W1443, W1650, W27102, W36150) to validate the fidelity of the finite element
models developed. Figure 9 shows longitudinal residual stress contours for a W36150
section at different stages of the cooling process. It was evident from the thermal analysis
that the beam flange tips and web cooled faster than the beam flange-web junction. Thermal
contraction of the junction between the beam flange and web would be greater than that of
the cooler flange tips and web. However, due to the symmetry, the compatibility conditions
require that the contraction is equal throughout the cross section [19]. This will introduce
tensile stresses in the central flange-web junction, balanced by compressive stresses in the
regions which were cooling faster i.e. flange tips and web. The resulting residual stress
distribution after cooling to ambient temperature is shown in Figure 10a,b for a W36150
section. The results exemplify the rule of thumb that the portions of the cross section to cool
faster (flange tips and web) will be left in a state of compressive residual stress.

(a) (b) (c)

Figure 9 Contour showing longitudinal stress distribution during the cooling process of hot
rolled W36150 section. (a) after 15 minutes, (b) 1 hour, (c) 2 hours.

The simulated residual stresses were compared to the measured residual stresses
(reported in [3]) for the flange and web of the section at the centerline of the beam length at
the end of the cooling process and the simulated responses resembled the experimental
responses very well as shown in Figure 10a,b. The longitudinal residual stress distribution
and the peak stresses were predicted accurately by the finite element analysis which validates

130
the fidelity of the finite element technique implemented. The distribution of the residual
stress across the flanges was of a general parabolic form and had a tensile residual stress
value at the flange center (Figure 10a). The residual peak stress in the beam flange was seen
to be as high as one-third of the yield stress of the material. On the contrary, the residual peak
stress in the web of the beam was in between one-third to one-half of the yield stress (Figure
10b) and the distribution of the residual stress was also of a general parabolic form. In
addition, it was observed that initial residual stresses showed an increasing trend along the
length of the beam with peak stresses obtained at the center of the beam (Figure 10c). The
variation of longitudinal stresses along the length of the beam was observed for all the
locations across the cross section i.e. flange tip, web and flange-web junction. However, at
the ends of the beam section, the longitudinal residual stresses had very small value for a
distance of about equal to the depth of the beam (Figure 10c). Similar observations were also
seen in the measured trend of longitudinal residual stress by Alpsten [19]. Hence, it is
apparent that the influence of initial residual stresses would be dominant from a distance of
equal to the beam depth of the beam. When the residual stresses at different locations of the
beam cross section were plotted against the time of cooling, it was noticeable that the fastest
cooling part (flange tips and web) developed tensile stresses initially, whereas slowest
cooling part (flange-web junction) developed compressive stresses.
As the cooling progressed, the tensile stresses in the fastest cooling part turned finally to
compressive stresses and the compressive stresses in the slowest cooling part turned to
tensile stresses. Another interesting thing to note here that after about two hours of cooling,
the initial residual stresses stabilized to a steady state value, and no change was observed in
the initial residual stress distribution. To further validate the finite element modeling features
adopted in this study, residual stress simulations were also compared to the experimental
measured values for W1443, W1650, W27102 sections as shown in Figure 11 through
Figure 13, where flange and web longitudinal residual stresses are plotted at the center of the
beam length. Very good conformity between the measured and simulated responses was
observed for all the W sections investigated. The validation of the simulated responses

131
against experimental measurements for a wide range of W sections shows the robustness of
the finite element technique implemented in this study.

Figure 10 Residual stress analysis results for a W36150 section. (a), (b) Measured [3] and
simulated longitudinal residual stresses at the centerline along the beam length, (c) simulated
longitudinal residual stress distribution along the length of beam, (d) simulated longitudinal
residual stresses plotted against cooling time at different locations of the beam section.

While it cannot be expected that all experimental results will correlate well with finite
element simulation based on nominal shape dimensions, average material properties and

132
certain assumptions on the heat transfer in cooling, the comparisons between computed and
measured initial residual stresses (IRS) made in this study gives an idea of the applicability
of the method which can serve as a guide for analysis of other shapes. Moreover, since
cooling conditions can be variable in the cooling bed because of stacking side by side,
residual stresses may have different distributions and magnitudes even in the same shape
depending on the imposed condition of cooling. The simulations made in this study pertains
to the assumption of free heat transfer, and any change in the cooling condition other than
free cooling condition must be accounted for in order to predict the actual scenario.

Figure 11 Measured [1] and simulated longitudinal residual stresses at the centerline along
the beam length for a W1443 section.

5. Influence of IRS on Buckling Response of Wide Flange Steel Columns


With the developed simulation technique, the actual distribution of the initial residual
stresses can be simulated for all the structural shapes commonly used as beams and columns
in moment resisting building frames, and it can be included in the analysis of moment
resisting frames under different types of loading conditions such as seismic analysis, fire
performance evaluation, buckling resistance evaluation for columns, and wind damage
analysis. The presence of initial residual stresses (IRS) is more crucial for compression

133
members such as column, where it is expected that IRS will reduce the ultimate strength of
the column and thereby reduce the buckling strength. In literature most of the works related
to the analysis of steel building frames either did not consider initial residual stresses or
assumed a simplified linear distribution due to the complexity involved in IRS calculation.

Figure 12 Measured [19] and simulated longitudinal residual stresses at the centerline along
the beam length for a W1650 section.

Figure 13 Measured [19] and simulated longitudinal residual stresses at the centerline along
the beam length for a W27102 section.

134
Hence there is a lack of understanding of the influence of initial residual stresses on the
cyclic response of wide flange steel columns. With the implemented IRS simulation
technique, the actual distribution of IRS can be obtained and accounted for in the analysis
and design of steel structures. A study on the influence of initial residual stresses on the
cyclic response of wide flange steel columns has been conducted in this study to demonstrate
the applicability of the simulation technique. Wide flange steel columns have widespread
applications in moment resisting frames and bridge piers. However, due to the relative large
slenderness ratios of the columns, structures of this type may be susceptible to damage
caused by local buckling under cosntant axial load and cyclic lateral forces. To provide a
basis for performance evaluation of columns under cyclic lateral loads with residual stresses,
wide flange steel columns have been subjected to numerical investigation in this study.
Firstly, the finite element model and the technique of residual stress calculation has been
validated against the experimental data from Lamarche and Tremblay [11] where they
studied the response of wide flange steel columns under axial compression.

(a)

Figure 14 (a) Equivalent plastic strain contour of W1287 column under axial compression,
(b) simulated and experimental [11] axial load-displacement response, (c) simulated and
experimental [11] out-of-plane deflection response.

135
A finite element model was developed for an ASTM A992 W1287 (W310129 in mm)
wide flange steel column having 3725 mm length and its performance was evaluated under
axial compression. The loading was displacement controlled and the analysis was performed
with and without the presence of IRS. As can be obserevd in Figure 14 that the simulated
response from the finite element analysis with IRS predicted the experimental responses
better than without IRS. The axial load-displacement response (Figure 14b) as well as the
out-of-plane deflection (Figure 14c) was predicted very well by the simulation with IRS.
Although the simulated response without IRS was able to predict the experimental trend
reasonably well, but the peak strength was over predicted. This validates the finite element
modeling and the implemented technique of residual stress calculation in predicting the
response of wide flange steel columns.

P
Loaded End:
Ux = 0
Rx = Ry = Rz = 0

Fixed End:
z
Ux = Uy= Uz = 0
Rx = Ry = Rz = 0
y
x (b)
(a)

Figure 15 (a) Finite element mesh used in the cyclic response evaluation of wide flange
column sections, (b) displacement-based simulated earthquake loading history [41].

After validation of finite element technique, a detailed finite element analysis was
performed on wide flange steel columns to investigate the influence of initial residual
stresses on the cyclic response. Since the influence of IRS on the cyclic response of wide

136
flange steel columns can be a function of depth of the beam or the width-thickness ratio [39-
40], four different sizes of wide flange steel columns having different width-thickness ratio
were investigated in the finite element study. The column sections chosen for the analysis
consisted of W14132, W24131, W24250, W36230 and their performance was
evaluated under combined axial compression and cyclic loading for both the cases with and
without IRS. Table 1 shows the width-thickness ratios for different column sections used in
this study. In the finite element modeling, the experimental setup of Newell and Uang [38]
was simulated where 4572 mm long wide flange steel columns were tested under combined
axial and lateral loads. The wide flange steel columns were discretized by twenty noded solid
brick elements and the boundary conditions simulated those used for experimental testing of
the steel column specimens [38]. Figure 15a shows the column model boundary conditions
and finite element mesh.

Table 1 Width-thickness ratio for different column sections

Column Section h/tw


W14132 17.7
W24131 35.6
W24250 20.7
W36230 42.2

At the fixed end of the model all six translational and rotational degrees of freedom were
constrained to be zero. At the loaded end of the model the out-of-plane translational (Ux as
shown in Figure 15a) and all three rotational degrees of freedom were constrained to be zero.
Lateral drift displacement was applied in the Uy direction and constant axial load was applied
in the Uz direction. A rigid constraint was imposed on edges at the column ends to prevent
stress concentrations at the loading and reaction points. A compressive load equal to the 30%
of the nominal axial load capacity [40] of the column was applied at the loaded end of the
column and a displacement-based artificial loading protocol offered by SAC/BD-97/02 [41]
was applied at the tip of the column in the lateral direction as the cyclic load. The
displacement based time history was based on the SAC Test Protocol and consisted of

137
imposing a series of prescribed quasi-static cyclic displacements to the end of the column
[41]. The prescribed displacements included six cycles of 0.375, 0.5 and 0.75% interstory
drift, followed by four cycles of 1.0% interstory drift and two cycles each of 1.5, 2, 3, 4, 5,
6% and so forth interstory drift. A schematic of the loading protocol is shown in Figure 15b.

(a)

Figure 16 Cyclic response of W24131 column without considering IRS. (a) Equivalent
plastic strain contour, and (b) moment-rotation hysteresis response at the end of 6%
interstory drift angle.

The finite element results revealed that under cyclic loading, all the columns developed
plastic hinge near the fixed end of the column by inelastic local buckling of the flanges and
web regardless of the presence of IRS. Figure 16 and Figure 17 show the equivalent plastic
strain contours and moment-rotation hysteresis plots of a W24131 column without and with
initial residual stresses respectively under combined axial compression and cyclic lateral
loading. The moment-rotation envelopes of peak moments at each loading cycle for all the
column sections are shown in Figure 18. It can be observed that for all the column sections,
when IRS was included in the analysis, the strength of the connection started to degrade
earlier compared to when IRS was not considered in the analysis. For column section
W14132, there was almost no strength degrdadation for cyclic loading upto 6% interstory
drift angle when IRS was not included in the analysis, whereas the strength of the column

138
started to degrade at 5% interstory drift angle when IRS was included in the analysis, and the
strength degraded by around 37% at the end of 6% interstory drift angle (Figure 18a).
Similarly for W24131 and W36230 column sections, the strength degradation started at
4% interstory drift angle when IRS was not considered in the analysis, whereas the strength
degrdaded at 3% and 2% interstory drift angle for W24131 and W36230 column sections
respectively, when IRS was considered (Figure 18b, d). This demonstates that presence of
initial residual stresses initiates inelastic buckling of the columns earlier and hence, the
strength of the column will be overpredicted without considering the effect of IRS.

(a)

Figure 17 Cyclic response of W24131 column considering IRS. (a) Equivalent plastic
strain contour, and (b) moment-rotation hysteresis response at the end of 6% interstory drift
angle.

It was also noticed that as the slenderness of the column section increases (Table 1), there
is more degradation in the strength of the column (Figure 18). This indicates that the
influence of IRS is more protuberant for columns with high slenderness ratio compared to
more compact column section such as W14132. This is more evident when the cyclic
response of W24250 column was compared with W24131 column (Figure 18c). Although
both the sections depth was similar, their width-thickness ratio was significantly different.

139
W24250 had a width-thickness ratio almost twice that of W24131. As a result, the
influence of IRS was insignificant on the cyclic response of W24250 column upto 6%
interstory drift angle.

Figure 18 Moment-rotation envelope showing peak moment at each loading cycle for
different column sections with and without IRS. (a) W14132, (b) W24131, (c) W24250,
(d) W36230.

The moment-rotation envelope of peak moments at each loading cycle (Figure 18c)
shows that the column showed similar response for both with and without IRS upto 6%
interstory drift angle. Hence it can be concluded that the cyclic response of wide flange
column is directly influenecd by the presence of the IRS and the influence of IRS increases

140
with the width-thickness ratio of the member. Another interesting phenomenon that was
observed for all column sections is that for both the cases with and without IRS, plastic hinge
formed at the fixed end of the column and when IRS was included in the analysis the plastic
hinge moved more towards the fixed end of the column indicating a reduction in the strength
of the column. Based on the observations from the finite element analysis, it is evident that
the presence of initial residual stresses significantly influences the cyclic response of wide
flange steel columns, especially columns with high width-thickness ratio and henceforth,
need to be addressed during the design calculation for steel building moment resisting
frames. The IRS simulation technique presented in this study can be beneficial in order to
account for the actual distribution of initial residual stresses in wide flange steel members
and thereby, provide a conservative estimate of the strength.

6. Conclusions
An advanced numerical technique has been developed to simulate initial residual stresses
developed during the cooling process of hot-rolled wide flange steel members based on a
sequentially coupled transient thermo-mechanical analysis. The simulated residual stresses
for different size of W shapes were evaluated against the experimental data available in the
literature which validates the robustness of numerical modeling scheme developed. An
example of the applicability of the method is presented in this study on the cyclic response
simulation of wide flange steel columns which demonstrated the importance of considering
initial residual stresses in the design of steel building moment resisting frames. The
measurements of residual stresses based on developed simulation technique should be
beneficial for an estimate of the magnitude and distribution of residual stresses likely to be
encountered, which can be used in the analysis of moment resisting building frames under
different types of loading conditions and can provide a conservative estimate of the strength
of members or connections.

141
7. References
[1] Fujita Y. The magnitude and distribution of residual stress. Fritz Engineering Laboratory
Report No. 220A.20, Lehigh University, Bethlehem, PA, 1955.

[2] Ketter RL, Kaminsky FL, Beedle LS. Plastic deformation of wide-flange beam columns.
ASCE Trans. 1955;120, 1019, Reprint No. 91 and 101.

[3] Huber AW. The influence of residual stress on the instability of columns. Ph.D.
Dissertation, University Microfilms, Inc., Ann Arbor, Michigan, 1956.

[4] Ketter RL. The influence of residual stresses on the strength of structural members.
Welding Research Council Bulletin 44 1958; Reprint No. 132 (58-6).

[5] Luxion WW, Johnstone BG. Plastic behavior of wide flange beams. The Welding Journal
1948;27(11), Research Supplement, 538-s to 554-s.

[6] Yang CH, Beedle LS, Johnstone BG. Residual stress and the yield strength of steel
beams. The Welding Journal 1952;31(4), Research Supplement, 205-s to 229-s.

[7] Real PMMV, Cazeli R, Silva LSD, Santiago A, Piloto P. The effect of residual stresses in
the lateral-torsional buckling of steel I-beams at elevated temperature. Journal of
Constructioanl Steel Research 2004;60:783-793.

[8] Russel TF. Some mathematical considerations on the heating and cooling of steel. 1st
Report of the Alloy Steels Research Committee, Iron and Steel Instutute, London, 1936.

[9] Huber A, Beedle LS. Residual stress and the compressive strength of steel. The Welding
Journal 1954;33(12).

[10] Ding Y. Residual stresses in hot-rolled solid round steel bars and their effect on the
compressive resistance of members. MS Thesis, The University of Windsor, Windsor,
Ontario, Canada, 2000.

142
[11] Beedle LS, Tall L. Basic column strength. Journal of the Structural Division 1960;86
(ST7):13973.

[12] Lamarche CP, Tremblay R. Seismically induced cyclic buckling of steel columns
including residual-stress and strain-rate effects. Journal of Constructional Steel Research
2011;67:1401-1410.

[13] Okazaki TO, Parkolap M, Fahnestock LA. Interface of the direct analysis method and
seismic design. Proceedings, ASCE Structures Congress 2009, Austin, TX, May 2009.

[14] Lu AYC, MacRae GA. Residual stress effects on the seismic performance of low-rise
steel frames. Proceedings of the 9th Pacific Conference on Earthquake Engineering
Building an Earthquake-Resistant Society, Auckland, New Zealand, 2011.

[15] Mathur K, Fahnestock LA, Okazaki T, Parkolap MJ. Impact of residual stresses and
initial imperfections on the seismic response of steel moment frames. Journal of
Structural Engineering 2012;138 (7):942-951.

[16] Nishino F, Tall L. Residual stress and the local buckling strength of steel columns. Fritz
Engineering Laboratory Report No. 290.11, Lehigh University, Bethlehem, PA, 1967.

[17] Lammi CJ. The effects of processing residual stresses on the fatigue crack growth
behavior of structural materials. Doctoral dissertation, Worcester Polytechnic Institute,
2009.

[18] Toribio J. Residual stress effects in stress-corrosion cracking. Journal of materials


engineering and performance 1998;7(2):173-182.

[19] Yamashita Y, Minami F. Evaluation of residual stress effects on brittle fracture strength
based on weibull stress criterion. Journal of ASTM International 2008;5.

143
[20] Alpsten GA. Thermal residual stresses in hot-rolled steel members. Fritz Engineering
Laboratory Report No. 337.3, Lehigh University, Bethlehem, PA, 1968.

[21] Kamamoto S, Nihimori T, Kinoshita S. Analysis of residual stress and distortion resulting
from quenching in large low-alloy steel shafts. Journal of Material Science and
Technology 1985;1:798-804.

[22] Toparli M, Aksoy T. calculation of residual stresses in cylindrical steel bars quenched in
water from 600C. Proceedings of AMSE Conference 1991, 4, New Delhi, India, 93-104.

[23] Jahanian S. Residual and thermoelastoplastic stress distributions in a heat treated solid
cylinder. Materials at High Temperatures 1995;13(2):103-110.

[24] Fa-rong Y, Shang-li W. Transient-temperature and residual-stress fields in axisymmetric


metal components after hardening. Journal of Material Science and Technology
1985;1(10):851-856.

[25] Yamada K. Transient thermal stresses in an infinite plate with two elliptic holes. Journal
of Thermal Stresses 1988;11:367-379.

[26] Kozak B, Dzierzawski J. Continuous casting of steel: basic principles. Washington: AISI
Learning Center 2002;9.

[27] Fourier J. Analytical theory of heat. Dover Publications, New York, 1955.

[28] Boley BA, Weiner JH. Theory of thermal stress. John Wiley & Sons, Inc, 1960.

[29] Nyashin YI. Examination of residual stresses in hot-rolled flange beams and their effect
on beam stability. Strength of Materials 1982;14(2):258-261.

[30] Shamsundar N, Sparrow EM. Analysis of multidimensional conduction phase change via
the enthalpy model. Journal of heat transfer 1975;97(3):333-340.

144
[31] Comini G, Guidice D. Finite element solutions of non-linear heat conduction problems
with special reference to phase change. International Journal of Numerical Methods in
Engineering 1979;8:613-624.

[32] Eurocode 3. Design of steel structures, part 1.2, general rules, structural fire design.
EN1993-1-2 European Committee for Standardization, 2005.

[33] Brickstad B, Josefson BL. A parametric study of residual stresses in multi-pass butt-
welded stainless steel pipes. International Journal of Pressure Vessels and Piping
1998;75(1):11-25.

[34] Wickstrm U. TASEF-2: A computer program for temperature analysis of structures


exposed to fire. Report No. 79-2, Lund Institute of Technology, Lund, Sweden,
Department of Structural Mechanics, 1979.

[35] Harmathy TZ, Stanzak WW. Elevated-temperature tensile and creep properties of some
structural and prestressing steels. Fire Test Performance, ASTM STP 464, American
Society for Testing and Materials, 1970, 186-208.

[36] Hu G, Morovat MA, Lee J, Eschell E, Engelhardt M. elevated temperature properties of


ASTM A992 steel. In: Proceedings of ASCE structure congress 09, Austin, TX, USA,
2009.

[37] Wikander L, Karlsson L, Nasstrom M, Webster P. Finite element simulation and


measurement of welding residual stresses. Modelling and Simulation in Materials
Science and Engineering 1994;2(4):845.

[38] AISC. Manual of steel construction: load and resistance factor designVolume I:
Structural members, specifications and codes. 2nd Ed., American Institute of Steel
Construction, 1994.

145
[39] Newell JD, Uang CM. Cyclic behavior of steel wide-flange columns subjected to larger
drifts. Journal of Structural Engineering 2008;134(8):1334-1342.

[40] Newell JD. Cyclic behavior and design of steel columns subjected to large drift. Doctoral
Dissertation, University of California, San Diego, CA, 2008.

[41] Xiao Y, Li H, Zhou T. Seismic behavior of wide-flange steel column with confined
potential plastic hinge. Journal of Constructional Steel Research 2009;65:808-817.

[42] SAC/BD-97/02. Protocol for fabrication, inspection, testing and documentation of beam-
column connection tests and other experimental specimens. by Clark, P., Frank, K.,
Krawinkler, H., and Shaw, R.

146
CHAPTER 6: INFLUENCE OF WELDING ON THE SEISMIC
PERFORMANCE OF MOMENT RESISTING CONNECTION

Abstract
In an experimental study of the influence of welding on the seismic performance of
moment resisting connections, it was observed that that welding influences the localized low-
cycle fatigue failure response. Analysis of strain responses near the beam weld toe and weld
access hole region led to the conclusion that weld induced residual stresses influence the
localized fatigue responses. In order to get an understanding of the residual stresses and their
influence on the localized failure mechanism of welded steel moment resisting connections,
this study developed a numerical technique based on sequentially coupled transient nonlinear
thermo-mechanical analysis to simulate the welding residual stresses and their influence on
localized fatigue responses. The finite element simulation technique was validated against the
experimental temperature and strain measurements. The simulated response of residual
stresses showed development of very high residual stresses close to the weld region. The
welding induced residual stresses together with the initial residual stresses in wide flange
shapes were included for the seismic analysis of moment resisting connections, which
demonstrated the influence of the residual stresses on the low-cycle fatigue failure
mechanism of the connections. The results of the finite element investigation also explain the
variability of the low-cycle fatigue failure response for different weld sequences. Finally, a
method to reduce the magnitude of the residual stresses has been proposed based on the
concept of post weld heat treatment which showed improved performance under seismic type
of loading.

Keywords: Moment connection, Weld, Residual stress, Weld sequence, Low-cycle fatigue,
Seismic analysis

147
1. Introduction
The 1994 Northridge Earthquake demonstrated the inability of many moment
connections to meet the seismic demands, where a large number of steel buildings were
damaged by cracking near the weld region [1-3]. Many connections failed by brittle cracking
of the weld between the bottom beam flange and the column flange. Over the past two
decades, the root causes of the connection failures were learned, and many were addressed
through research. The lack of ductility displayed in the pre-Northridge welded steel moment
connections (WSMCs) prompted research by the Federal Emergency Management Agency
(FEMA) as well as the SAC Joint Venture. Weld metal matching, base metal and weld metal
toughness, backing bar effects, fillet weld reinforcement, residual stress effects, continuity
plates, panel zone deformations, reduced beam section details, weld access hole details, and
column flange through-thickness fractures have all been investigated and several
modifications were made in the exsisting design of moment connections [1-8]. While test
results on the post-Northridge modified connections demonstrated significant improvement
with regard to ductility and energy dissipation capacity; however, many localized failures
have also been demonstrated. Testing of post-Northridge welded steel moment connections
have shown that many failures are initiated from localized, low-cycle fatigue cracks near the
weld toe [4-5, 7-13]. The weld toe fatigue crack initiation motivated several investigations on
weld material and the outcome was the recommendation of using high toughness weld
material [28]. However, the reason for the weld toe fatigue crack initiation in the post-
Northridge design has yet to be investigated in a rational manner.
The low-cycle fatigue cracks near the weld toe has also been observed for piping joints
and research studies revealed that welding induced residual stress has significant influence on
fatigue crack initiation at the weld toe of piping joints [15-26]. Moreover, several studies on
piping joints addressed the influence of weld sequence on the amount and distribution of
residual stresses which led to the conclusion that the welding procedure can have a
significant effect on the residual stress and thus the fatigue life of a joint [27-32]. The
influence of weld sequence on the residual stress distribution was also addressed for welded
plate and T-joint specimens [33-39]. Thus literature suggests that welding induced residual

148
stresses can be one of the contributors to the low-cycle fatigue failure of the moment
resisting connections. The influence of residual stress on fatigue crack initiation at the weld
toe in WSMCs was speculated [40-46], but the failure mechanism is yet to be investigated.
However not much attention was paid to the influence of residual stress on the moment
resisting connections because of the common notion that the residual stress relaxes after a
couple of inelastic loading cycles. Due to the complexity involved in residual stress
calculation, most analyses either do not consider residual stresses or use simplified methods
of analysis [40, 45]. However, Dong and his coworkers [14, 47] did include detailed residual
stress analysis in their investigations. Their calculations demonstrated that the magnitude of
residual stresses in some locations can be greater than the yield stress of the material. They
performed fracture analysis for crack propagation of preexisting cracks at the backing bars
and demonstrated that the triaxial residual stress greatly reduced the plastic deformation
capacity of the welded joint and thus induced brittle fracture. In addition, an increase in
fracture driving force was observed in presence of high tensile residual stresses. However,
the analysis was two dimensional and hence, the influence of weld sequence was not
accounted for across the width of the beam flanges. Moreover, the influence of residual
stresses on the fracture behavior was investigated based on fracture mechanics analysis. No
light was shed on the low-cycle fatigue failure response of the moment connections in
presence of residual stresses. Nevertheless this study showed that even though the residual
stress relaxes after few inelastic loading cycles, it can influence the fracture behavior at the
welded joint of the moment connection. After that study, the influence of welding induced
residual stresses on the failure response of moment resisting connection was not studied in
detail and most of the literature either ignored or assumed simplified distribution of residual
stresses.
An experimental study on welded unreinforced flange bolted web (WUF-B) connections
demonstrated the influence of welding sequence on the fatigue failure of the connections
which was presented in Chapter 4. It was observed that different welding sequences induced
different strain responses of the connections. In addition, a phenomenon called strain
ratcheting was observed in these experiments near the weld toe, which can lead to crack

149
initiation. However, residual stress measurements could not be made in experiments, and
hence, the residual stress influence on the fatigue crack initiation could not be studied.
Hence, knowledge on WSMCs is required to understand the influence of residual stresses and
welding sequence on the fatigue response of the connections. In this study, a detailed 3D
finite element investigation was performed on welded steel moment connection considering
the effect of both initial (residual stresses due to uneven cooling of hot-rolled W shapes) and
welding induced residual stresses. The external sub-assemblage of beam-column connection
discussed in Chapter 4 was simulated in ANSYS to understand the local and global response
of the connection under constant amplitude loading with the presence of initial and welding
induced residual stresses. Advanced numerical schemes were developed to simulate welding
process, and hence the residual stress and subsequent fatigue responses of welded joints. In
welding process simulation, sequentially coupled, nonlinear, transient, thermo-mechanical
analyses were employed to calculate temperature history and residual stresses. Finally, the
fatigue responses of welded joints with the initial and welding residual stress fields were
simulated by prescribing displacement controlled cyclic loading. At the end, post weld heat
treatment technique was implemented to improve the fatigue resistance of the connections by
reducing the residual stresses.

2. Numerical Scheme for Weld Residual Stress Simulation


The thermo-physical and thermo-mechanical processes associated with weld residual
stress evolution during welding are extremely complex. Currently, no model is available to
realistically account for the arc physics, the weld pool phenomena, and finally, the
deformation and heat conduction in the solid metal [47]. In this study, the welding process
was simulated by employing sequentially coupled, nonlinear, transient, thermo-mechanical
analyses to calculate temperature history and subsequently the welding induced residual
stresses. First a transient thermal analysis was performed during which temperature history in
the connection from the heat input of the successive build-up of the welding passes was
determined. The temperature field at each time step was then used as an input into a transient
thermo-mechanical analysis for evaluating the residual stress and strain field in welded

150
joints. This analysis scheme was sequentially integrated to analysis of initial residual stress in
the wide flange beam of the WUF-W connection [48]. Finally, the cyclic loading responses
of welded joints with the residual stress fields were simulated by prescribing displacement
controlled cyclic loading. Figure shows a schematic of the sequentially coupled thermo-
mechanical analysis steps performed in this study.

Finite Element Mesh Heat Input Temperature Fields

Seismic Analysis Residual Stress & Strain


Fields
N

Figure 1 Steps performed in the thermo-mechanical analysis of WUF-W connection.

In order to study the development of weld residual stresses and its influence on the
seismic response of moment resisting connections, advanced 3D numerical schemes were
developed with finite element (FE) program ANSYS to simulate welding process, and hence
the residual stress and subsequent fatigue response of WSMCs. 3D FE models of the exterior

151
sub-assemblage of the WUF-B connections presented in Chapter 4 were simulated, where a
single W18 55 ASTM A992 beam was attached to a W14 74 ASTM A992 column.
However, in the experiments, the beam web was connected to column flange through bolted
shear tab, whereas in the simulation, the web was assumed welded to the column flange
which is the conventional WUF-W (welded unreinforced flange welded web) connection.
Since the main focus of this study was to study the development and influence of weld
residual stresses in the weld toe region of the connection, it was thought to be sufficient to
investigate WUF-W connection instead of WUF-B connection to avoid the complicated
contact geometries of the bolted web connection.

Figure 2 Finite element mesh of WUF-W connection used in the study.

Nevertheless, the weld access hole geometry and boundary conditions of the connections
were simulated exactly to mimic the experimental setup. The connection region was
discretized by using twenty noded solid elements with very fine mesh in order to capture the
behavior of the connection region precisely since most of the inelastic actions occur in the
connection region. Whereas the other parts of the connection which are expected to remain

152
elastic was discretized with eight noded solid elements with coarse mesh in order to reduce
the computational time and complexity. Figure 2 shows the FE mesh generated for the
analysis of the connection with the imposed boundary conditions. The bottom part of the
column was supported by a hinge whereas the top part of the column was supported by a
roller to represent the theoretical inflection points of a column in a typical floor system and
the free end of the beam was supported on a roller. No constraint was applied near the
connection region.

3. Thermal Analysis
Thermal analysis was performed in order to calculate transient temperature distribution
from the welding sequence and subsequent cooling. In thermal finite element analysis, heat
loss occurred from the material surface through both convection and radiation, which were
determined using the heat transfer coefficient in Fourier heat transfer equation [49]. The
temperature-dependent material properties (specific heat cp, thermal conductivity k, enthalpy,
heat transfer coefficient h) of structural steel used in the heat transfer analysis are shown in
Figure 3, where the material parameters were defined as a function of temperature. Latent
heat/phase transformation was accounted for by defining enthalpy of the material as a
function of temperature [52-53].
The principal parameter of the welding heat source for the temperature field is heat flow
Q (Joule/sec), which can be expressed as [52]:
Q UI (1)
where,
arc efficiency, which account for radiative and other losses from the arc to the ambient
environment and its value ranges between 0 and 1
U arc voltage (in Volt)
I arc current (in Amp)
The equivalent heat input can be assumed as the combination of both surface and body heat
flux components [53]. Hence the total heat input
Q Qs Qb (2)

153
where Qs and Qb indicate heat contents resulting from a specific surface heat flux and body
heat flux, respectively.

Figure 3 Material coefficients of structural steel as a function of temperature used in thermal


analysis. (a) Specific heat [54], (b) thermal conductivity [54], (c) heat transfer coefficient
[19], (d) enthalpy [33].

The specific surface heat flux (area-specific heat flow density) qs and specific body heat flux

(volume-specific heat flow density) qb can be expressed by,

Qs
qs (3)
S

154
Qb
qb (4)
V
where, S is the surface area, and V is the volume of each element.

Figure 4 Element birth and death technique in the FE simulation of welding process.

In order to simulate the multi-pass welding process, one needs to model the addition of
filler material as new weld passes are deposited and the geometry of the welded joint is built-
up. This was done through the element birth/death technique in ANSYS. The steps of using

155
this technique are depicted in Figure 4 for the beam-column welded joint of a WUF-W
connection. With this technique, a finite element mesh of the welded joint including the weld
beads is generated first. Subsequently, the element groups representing each weld pass are
deactivated (element death) before welding is included in the analysis. These elements are
reactivated (element birth) sequentially as the welding arc (heat source) advances along the
beam flange width direction. An example of the moving temperature field during the last
welding pass of the beam top flange groove weld for a welded steel moment connection is
demonstrated in Figure 5, where it can be seen that the high temperature field moves with the
welding arc (Figure 5a-e). The distribution of the temperature after cooling of the weld pass
is shown in Figure 5f. In the finite element simulation, it was assumed that the top and
bottom beam flange welds were made using the Shielded Metal Arc Welding (SMAW)
process and E70T welding electrodes as used in the experiments.

(a) (b) (c)

(d) (e) (f)

Figure 5 Illustration of moving temperature (in Kelvin) field during the final welding pass in
the top flange of a welded steel moment connection (a-e) and cooling (f).

156
While welding the beam bottom flange to the column flange the web of the beam hinders
the completion of a single pass across the width of the beam. Whereas the beam top flange
does not have any obstacle like the bottom beam flange. Therefore, weld is not uniform in the
bottom beam flange when compared to the top flange. Further, the start and stop points in the
bottom flange weld may cause discontinuities or higher residual stresses that may lead to
low-cycle fatigue failure of the connection. In order to investigate the development of
residual stresses due to welding, and the effect of welding sequence on the residual stress
distribution as well as on the low-cycle fatigue response of WSMCs, two specimens of
WUF-W connection (namely MC1 and MC2) were considered in the finite element analysis,
where each specimen had different weld sequence at the beam bottom flange, whereas beam
top flange was welded in a similar fashion for both the specimens. The welding details of the
two specimens are shown in Figure 6 and Figure 7.

Figure 6 Sketch of weld sequences in beam top and bottom flanges for specimen MC1
showing number of welding passes in order of completion.

These two specimens were used in the experimental investigation presented in Chapter 4
and were denoted by MC1 and MC2. The welding sequences used at the beam bottom flange

157
of specimens MC1 and MC2 were denoted by weld sequence A and weld sequence B
respectively. As can be observed from Figure 6 and Figure 7, the top flange welds were
deposited in a similar fashion for both the specimens, only the number of beads and the size
of the beads were different. On the other hand, the bottom beam flange welds were laid in a
completely different manner for the two specimens. In welding sequence A (specimen MC1),
the welder fills the groove with all the welding beads on one side of the beam web before he
switches to the other side of the beam web to complete the weld (Figure 6). This type of
welding sequence gives rise to nonuniform temperature gradient across the width of the beam
flange. On the contrary, in welding sequence B (specimen MC2), the welder alternates sides
of the beam web after placing each weld bead (Figure 7). This sequence fills the groove more
uniformly compared to sequence A.

Figure 7 Sketch of weld sequences in beam top and bottom flanges for specimen MC2
showing number of welding passes in order of completion.

All the welding sequences in beam top and bottom flange for specimens MC1 and MC2
were simulated accordingly in the finite element analysis by using the element birth and

158
death technique. Figure 8 shows comparison between the measured and simulated
temperature profiles during the welding process for both the specimens at the top flange of
the beam measured at a distance of 75 mm away from the weld toe and 50 mm away from
the edge of the beam flange. It was observed that the simulated responses resembled the
experimental temperature profiles closely for both the specimens. This validates the thermal
finite element model and it indicates that the element birth-death technique can simulate the
welding process well.

Figure 8 Measured and simulated temperature profiles at the beam top flange for the two
specimens. (a) Specimen MC1, (b) specimen MC2.

When temperature profiles of the two specimens were compared it was interesting to note
that the temperature profiles were different for different welding sequences for both the top
and bottom flanges measured at the same location. Although the top flange welds were laid
in a similar fashion for both the specimens, the temperature profiles varied (Figure 9a) as a
consequence of difference in the number of beads and bead sizes. However, in both the cases,
the mean temperature showed an increasing trend with time, and specimen MC1 produced
higher mean temperature than specimen MC2. On the other hand, the temperature profiles for
the bottom beam flange showed significant difference for the two welding sequences. As

159
anticipated, for welding sequence A in specimen MC1, the average temperature on one side
of the beam, where the groove was filled first, was higher than the average temperature on
opposite side of the beam, where welding was done later, whereas for welding sequence B in
specimen MC2 the average temperatures were similar on both sides of the beam web (Figure
9b). The contour plots in Figure 10 and Figure 11 show the temperature distribution in the
bottom flange of the beam during the first two welding passes for weld sequences A and B
respectively.

Figure 9 Simulated temperature profiles at the top and bottom flange of the beam showing
variation for two different welding sequences. (a) Top flange, (b) bottom flange.

It is clearly seen that the two welding passes for weld sequence A developed higher
temperatures only on one side of the beam bottom flange since both the passes were
completed only on one side of the beam (Figure 10) and hence, the temperatures were
distributed nonuniformly across the beam flange. On the contrary, the two passes for weld
sequence B were completed on the two sides of the beam bottom flange by switching sides
once one pass is completed (Figure 11), and hence, weld sequence B generated an even
distribution of temperature across the beam flange.

160
Figure 10 Temperature distribution (in Kelvin) at the end of weld passes 1 and 2 of welding
sequence A at the beam bottom flange of specimen MC1 showing nonuniform temperature
on both sides of the beam web.

Figure 11 Temperature distribution (in Kelvin) at the end of weld passes 1 and 2 of welding
sequence B at the beam bottom flange of specimen MC2 showing uniform temperature on
both sides of the beam web.

161
4. Residual Stress Analysis
The residual stress analysis was conducted after the thermal analysis was finished and the
time dependent temperature field during the welding process was determined. The
temperature solution obtained from the thermal analysis, which normally depends on position
and time, was read into the stress analysis as a predefined field i.e. the nodal temperature
field at each time step from the thermal analysis was applied as thermal load to the
corresponding node in the structural model. As a result, the node labels for the structural
mesh must match those from the heat transfer mesh. Similar to the thermal analysis, element
birth/death was used in the residual stress analysis. Once the real time temperature fields
were fed into the structural model for each of the nodes at each time step, then residual
stresses were calculated based on the temperature dependent material properties of structural
steel and relevant constitutive equations. The material properties used for the residual stress
analysis in finite element simulation were density , initial yield stress 0, elastic modulus E,
Poissons ratio and thermal expansion coefficient for structural steel as shown in Figure
12b,c and Figure 13. The elevated temperature stress-strain response of ASTM A992 steel
under monotonic tension loading was obtained from Hu et al. [56] as shown in Figure 12a.
The initial yield stress and elastic modulus of ASTM A992 steel obtained from these
experiments are plotted separately in Figure 12b and Figure 12c along with comparison with
AISC [57] and Eurocode 3 [54] prescribed values. The elevated temperature stress-strain
response of ASTM A992 steel was implemented in the finite element simulation by using
bilinear kinematic hardening model available in ANSYS which incorporates von-Mises yield
criterion and associative flow rule. The bilinear kinematic hardening model was chosen to
reduce the complexity of the model and computational time and cost. The bilinear kinematic
hardening model parameters were obtained through fitting the stress-strain response of A992
steel as shown in Figure 12a. Note that there was no material data available for the
temperature range 900-1400C, for which linear interpolation method was used to obtain the
model parameters.
During welding, the base material is exposed to high temperatures which will induce
microstructural changes in the material leading to change in the material properties. The

162
region which exhibits microstructural change due to high temperature exposure is referred to
as the heat affected zone (HAZ). Studies on stainless steel SS304 by Johnson and Ramirez
[60] showed five different microstructures depending on the temperature levels experienced
by the base metal during welding.

Figure 12 Monotonic tension test response of ASTM A992 steel at different temperatures
[56]. (a) Monotonic stress-strain response, (b) initial yield stress plotted as a function of
temperature, (c) elastic modulus plotted against temperature.

163
Figure 13 (a) Poissons ratio [58] and (b) thermal expansion coefficient of structural steel
plotted as a function of temperature [59].

These are referred to as course grained HAZ (1100C and above), fine grained HAZ
(875C-1100C), intercritical HAZ (700C-875C), subcritical HAZ (575C-700C) and the
base metal is unaffected below 575C. To properly simulate the welding induced residual
stresses, it is required to account for the heterogeneity in the material properties in the HAZ
[29]. To determine the material properties of ASTM A992 steel for the five HAZ discussed
above, monotonic tension tests were performed on steel coupons [61]. The coupons were
conditioned at different peak temperatures; hold the peak temperature for 15 minutes
followed by slow air cooling until it reaches room temperature. The stress-strain response of
one set of monotonic tension tests is shown in Figure 14a. It was observed that as the
maximum temperature increased both the yield and ultimate stresses of the steel gradually
decreased by downward shift of the stress-strain curves. The reduction in yield stress with
maximum temperature of 1200 C is about 40%. The response from the monotonic tension
tests demonstrated that the yield stress of ASTM A992 decreased with increasing
temperature, and upon removal of the temperature the yield stress never went back to the
room temperature value, rather it reduced from that of the room temperature value. The yield
stress values of unconditioned ASTM A992 coupons tested at different temperatures up to

164
900C was obtained from Hu et al. [56] as shown in Figure 12a. However, for steel coupons
conditioned at different high temperatures, no stress-strain data was available for higher
temperatures. Hence, least square method was adopted to establish a relationship between
temperature and yield stress (solid line in Figure 14b) using the data from Hu et al. [56] for
unconditioned A992 steel (solid circles in Figure 14b).

Figure 14 (a) Stress-strain response of ASTM A992 coupons conditioned at different peak
temperatures [61], (b) temperature dependent yield stress of ASTM A992 for five
temperature levels to account for the material heterogeneity.

The relation can be expressed as follows:

0 T 0RT for T 400C


(5)
c0 T 400
0 T 0RT e for T > 400C

where, 0RT is the yield stress for unconditioned A992 steel at room temperature (MPa), c0 =

0.0045 is determined by the least-square fit, and T is the current temperature (C). Eq. (5)
was modified for conditioned A992 steel using the yield stress reduction factor from Figure
14a and assuming that the evolution of the yield stress during cooling for different
conditioning temperature follow the same trend as the unconditioned case. The modification

165
of Eq. (5) led to the following generalized expression of yield stress as a function of current
and conditioning temperatures:

0 T 0RT for T 400C


1 (6)
2 2
Tc 25 ci T 400
i T ,Tc 0 RT 1 e for T > 400C
1650
1
2 2
0.0045 Tc 25 Tc 25
ln e 1
1650
where ci , Tc is conditioning temperature (C).
25 Tc

Five sets of temperature-dependent material properties were defined depending on the


temperature level to incorporate the material heterogeneity in the finite element modeling of
residual stresses as shown in Figure 14b, where solid circles represent test data for
unconditioned A992 steel [56], the open circles represent yield stresses for conditioned A992
steel at different temperatures from the experiments [61], the solid line represents the model
fit by using least squares for unconditioned case, whereas dashed lines represent the model fit
assumed for conditioned cases.
Initially, finite element analysis was performed considering only the welding induced
residual stresses. The residual stress distributions in longitudinal and transverse directions for
specimens MC1 and MC2 are shown in Figure 15 and Figure 16, respectively. It is evident
that significant amount of residual stresses developed close to the welded region of the
connection both in longitudinal and transverse direction. The magnitude of the residual
stresses at several locations near the weld toe were very close to the yield stress of the base
metal, and around the weld access hole the residual stresses were higher than the yield stress.
The residual stresses in the longitudinal direction were higher at the weld toe and weld access
hole, and it gradually decreased as the distance from the weld toe increased. However, the
magnitude of residual stresses in the transverse direction was only high near the welded
region, and there were barely any residual stresses away from the weld toe. This is attributed
to the severe structural restraint in the welded region which induced higher magnitudes of

166
transverse residual stresses, and away from the welded region, there was not much restraint
to induce transverse residual stresses.

Figure 15 Residual stress distribution (in Pa) for specimen MC1. (a) Longitudinal direction,
(b) transverse direction.

Figure 16 Residual stress distribution (in Pa) for specimen MC2. (a) Longitudinal direction,
(b) transverse direction.

167
Figure 17 Welding induced residual stress distribution in the beam top flange of specimen
MC1. (a) Across width, (b) across length.

Nevertheless, for seismic response of moment resisting connections, the longitudinal


direction is of particular interest, and hence, the further discussion on the residual stresses
will be addressed only for the longitudinal direction. The longitudinal residual stress
distribution across the width and length of beam top and bottom flanges are presented in
Figure 17through Figure 20. It was observed that significant amount of residual stresses
developed very close to the weld toe of the beam-column joints for both top and bottom
beam flanges. The peak stresses were around 300 MPa on both the top and bottom flanges,
which were about two-thirds of the yield stress of A992 steel. For top flange, the residual
stresses were tensile and almost constant throughout the width of the flange, only at the edges
of the flange compressive residual stresses developed. The peak residual stress was tensile
with 300 MPa value, whereas the compressive stresses varied between 100 to 200 MPa. On
the other hand, for bottom flange, significant amount of compressive stresses developed
across the width of the beam flange, however at stop and start locations of welding passes
tensile residual stresses also developed. The peak compressive stresses were 300 MPa,
whereas the tensile stresses varied between 0 to 100 MPa. It is interesting to note that for

168
both top and bottom flanges, the residual stress values showed a decreasing trend as the
distance from the weld toe increased with almost zero values after 200 mm away from the
weld toe. Based on the observations from the finite element study, it is evident that very high
values of residual stresses develop close to the weld toe and around the weld access hole for
both the specimens which may play a significant role under service conditions. Similar type
of residual stresses was also observed in the studies on piping joints by Lu [21] and Cheng
[29] and it was shown that these residual stresses worked as a driving force for axial strain
accumulation near weld toe.

Figure 18 Welding induced residual stress distribution in the beam bottom flange of
specimen MC1 for weld sequence A. (a) Across width, (b) across length.

To investigate the influence of the welding sequences on the residual stress distribution,
the residual stresses were plotted across the width of the beam flange at the location of peak
residual stresses as shown in Figure 21. It was observed that for the top flange, specimen
MC1 developed more residual stresses compared to specimen MC2. It is interesting to note
that the residual stresses at the beam top flange were almost constant along the width of the
flange except at the edges of the flange. This is expected as each of the welding beads in the
top flange weld was laid in one pass with no discontinuity during welding. For beam bottom

169
flange, there were several locations across the width of the beam flange where welding was
stopped and restarted, and hence depending on the temperature distribution, the two
specimens experienced different amount of residual stresses and distribution across the width
of the flange for the two weld sequences A and B. It was observed that welding sequence A
developed more residual stresses in the beam bottom flange of specimen MC1 than welding
sequence B in specimen MC2 as anticipated. The nonuniform temperature gradient produced
in the beam bottom flange during welding sequence A led to the development of higher
residual stresses compared to weld sequence B. It was observed that the residual stresses had
larger values at the start and stop location of the weld as can be seen from Figure 21b.

Figure 19 Welding induced residual stress distribution in the beam top flange of specimen
MC2. (a) Across width, (b) across length.

The results from the residual stress analysis clearly shows that significant amount of
residual stress can be developed from the welding procedure, and the distribution and the
peak values of the residual stresses is influenced by the welding sequences employed during
welding. No experimental measurements of the residual stresses were available from the
experiments and hence, the residual stress values and distribution obtained from the finite
element analysis (FEA) could not be validated. However, the results of the analysis

170
qualitatively demonstrates that residual stresses induced by welding can be as high as two-
thirds of the yield stress of the base metal at the weld toe and higher than the yield stress at
the weld access hole, and the welding sequence can significantly change the distribution and
peak values of the residual stresses.

Figure 20 Welding induced residual stress distribution in the beam bottom flange of
specimen MC2 for weld sequence B. (a) Across width, (b) across length.

Figure 21 Welding induced residual stress distribution across the width of beam flange near
weld toe for the two specimens. (a) Top flange, (b) bottom flange.

171
5. Influence of Initial Residual Stresses
When a wide flange section is hot-rolled followed by cooling, the final section will have
residual stresses. Welding during fabrication further induces residual stresses to the beam-
column connection as shown in this study. Studies by Ketter et al. [61], Huber [63], Ketter
[64], Alpsten [65] showed that the residual stresses during cooling of hot-rolled sections can
be as high as one-half of the yield stress. In their study they demonstrated the distribution of
residual stresses due to cooling and the mechanism involved in the development of residual
stresses have been discussed. In order to determine the residual stresses developed during
cooling of hot-rolled wide flange sections (denoted as initial residual stress in this study), an
advanced numerical scheme based on a sequentially coupled transient thermo-mechanical
analysis was developed in a study by Quayyum and Hassan [48]. In this analysis, a wide
flange section was heated up to 1300C and then air cooled until the section reached the room
temperature. The temperature transients obtained from the thermal analysis were then used as
input for the residual stress analysis similar to the analysis scheme used in this study. The
scheme was used to evaluate the initial residual stresses for a W1855 beam section, which is
the beam section used in the WUF-W connection in this study. It is to be mentioned here that
the initial residual stresses were not considered for the column sections since it was observed
that the magnitude and distribution of weld induced residual stresses at the weld toe and near
the welded region and weld access hole was not altered noticeably by the addition of column
initial residual stresses. Hence, in this study column initial residual stresses were not
considered to reduce the complexity in the FE model and computational time and cost.
In order to study the influence of initial residual stresses due to uneven cooling of wide
flange beam section on the overall distribution of residual stresses after the welding process,
the simulated initial residual stresses were incorporated along with the welding induced
residual stresses. In the finite element simulation, the initial residual stresses were
implemented as the initial state for the welding induced residual stress analysis. Figure 22
shows the residual stress distribution along the length of the beam up to 500 mm for
specimen MC1 considering both cases with and without initial residual stresses. After

172
analyzing the residual stress distribution, it was observed that the inclusion of initial residual
stresses did not increase the peak residual stress values close to the weld toe.

Figure 22 Residual stress distribution along the length of the beam for specimen MC1
showing the influence of initial residual stresses. (a) Beam top flange, (b) beam bottom
flange.

The distribution of the residual stresses was not influenced by the initial residual stresses
for a length of about 150 mm from the weld toe. This is due to the fact that initial residual
stress values were insignificant at the ends of the beam [48]. However, initial residual
stresses affected the stress values as the distance from the weld toe increased, especially after
a distance of about 150 mm from the weld toe. After 150 mm, the total residual stresses
became constant, which was predominantly contributed by the initial residual stresses.
Although the initial residual stresses of the beam section did not change the distribution of
the overall residual stresses near weld toe, it may not have significant influence on the low-
cycle fatigue failure of the connection. However, in this study both the initial residual
stresses (for the beam section) and weld residual stresses were considered for investigating
the seismic response of the WUF-W connections to represent the actual physics of residual
stress development in the connections.

173
6. Effect of Residual Stresses on the Seismic Performance of WSMCs
Since structural behavior under seismic events is usually characterized by small number
of cycles, only a few of which involve large excursions in the plastic range, the failure under
such conditions is conventionally referred as low-cycle fatigue failure [66]. Fatigue failure of
a structure usually results from localized crack initiation, subsequent crack propagation and
final fracture. However, fatigue design methods generally use global approaches. These
global approaches are based on nominal stresses on critical cross-sections and corresponding
fatigue lives obtained from experiments. The mechanisms of localized fatigue damage
accumulation and crack initiation in the presence of residual stresses are not considered in the
design approaches. Residual stresses which arise during welding processes are an important
factor in fatigue failure of welded structures [28, 30]. In the low-cycle fatigue range, the
influence of residual stresses is usually ignored because of the assumption that the residual
stresses relax to zero after only a few cycles. The finite element analysis results in this study
demonstrated that both tensile and compressive residual stresses in welded joints can be as
high as two-thirds of the yield stress at the weld toe and higher than yield stress at the weld
access hole. The development of the residual stresses can act as mean stresses to the fatigue
loading cycle and thus lead to degradation and failure of structures due to the accumulation
of strain. In order to investigate the influence of residual stresses on the seismic response of
WSMCs, the finite element models of the WUF-W connections were subjected to seismic
type of loading to investigate the localized failure mechanisms in presence of the residual
stresses. For the fatigue failure analysis, the constant amplitude loading history used in the
experiments presented in Chapter 4 was prescribed in order to make comparisons with the
experimental observations.
Traditionally, cyclic tests are carried out with reference to standard procedures such as
[68-70] which use loading histories consisting of groups of cycles at increasing amplitudes
(Figure 23a). Such loading histories present the advantage of allowing on a single specimen a
satisfactory appraisal of the cyclic performance of the component. However, they appear
inadequate for the development of cumulative damage models useful not only for the seismic
design of new structures but also for the assessment of safety and reliability of damaged

174
structures and for the development of adequate repair procedures. Hence it has been proposed
by various authors both in Europe and U.S. [69, 71-75] to perform tests with cycles at a
constant amplitude. Such a procedure presents the advantage of allowing a clear
understanding of the damage accumulation process as well as of the key parameters
governing it. The constant amplitude loading history used in the analysis is shown in Figure
23b.

Figure 23 (a) SAC loading protocol [34], (b) non-standard constant amplitude loading
history prescribed in this study.

The prediction of experimental responses of moment connections using finite element


simulation is highly sensitive to the material models adopted and the parameters of the
adopted material models, especially under cyclic loading. In the literature, most of the works
related to the finite element modeling and analysis of moment resisting connections used
either bilinear or multilinear kinematic hardening material model in the finite element
simulation of the cyclic behavior of moment connections. Although both bilinear and
multilinear material models obey von-Mises yield criterion and include Bauschinger effect,
both the models are incapable of simulating ratcheting or shakedown, which is essential for
cyclic response simulation of moment connections. Both multilinear and bilinear material
models have a linear kinematic hardening rule and hence, are not sufficient for simulating

175
ratcheting or shakedown. To simulate ratcheting and shakedown in a FE simulation, a metal
plasticity model is required with a nonlinear kinematic hardening rule. Nonlinear kinematic
hardening implies a shift (or movement) of the yield surface along a nonlinear path. The
Chaboche material model [75], available in ANSYS Mechanical software, offers such a
nonlinear kinematic hardening rule. Since this study deals with cyclic response simulation of
WUF-W connections, the advanced nonlinear kinematic hardening model of Chaboche [75]
has been adopted in this study to define the material characteristics. A brief description of the
Chaboche [75] model is given below to demonstrate different features of the model.
The yield criterion:
1/ 2
3
J2 ( ) s a s a 0 R ( p, q) , (7)
2
where is the stress tensor, is the current center of the yield surface in the total stress
space, s is the deviatoric stress tensor, a is the current yield surface center in the deviatoric
space, 0
is the initial size of the yield surface, and R represents the isotropic hardening
variable as a function of the accumulated plastic strain p and the size of the plastic strain
surface q. The rate-independent plastic strain increment will be calculated using the
associated flow rule:

p J2 3 s a
d d dp (8)
2 0 R
The superimposed kinematic hardening rule is given by:
4
da dai (9)
i 1

d ai 2 Ci d p
a i dp (10)
3 i

Each of the superposed kinematic hardening rules has a strain hardening term (1 st term in
Eq. 10) and a dynamic recovery term (2nd term in Eq. 10). These are important modeling
features for simulating cyclic responses of materials. Use of Chaboche model to simulate the
material responses requires that the stable hysteresis loop of the cyclic stressstrain response
be known. In this case, stable hysteresis loops were obtained from single amplitude strain

176
controlled cyclic test of ASTM A992 steel coupons [30]. Chaboche parameters were
determined from the up going strain controlled experiment of ASTM A992 steels. The
nonlinear kinematic hardening parameters were determined by using a genetic algorithm
based optimization method. Simulations were made at the material level with obtained
parameters for validation. As shown in Figure 24, the simulated responses matched very well
with the experimental responses obtained from [61]. It is to be noted here that the
heterogeneous material models used in the residual stress calculation was not accounted for
in the fatigue analysis of the connection. This is attributed to the unavailability of cyclic
response data of ASTM A992 steel at different conditioning temperature.

Figure 24 Experimental and simulated response of strain controlled hysteresis loop for
ASTM A992 steel with Chaboche parameters.

In the finite element study, firstly, the importance of considering the residual stresses will
be discussed by showing comparisons with the experimental results presented in Chapter 4,
and later the effect of the welding sequence on the low-cycle fatigue response will be
addressed. In the experiments, specimen MC2 showed ductile mode of failure by fracture of

177
the beam top flange in the mid-way of weld access hole. Visible cracks were observed at the
weld toe and around the weld access hole. On the contrary, specimen MC1 showed brittle
and sudden failure by rupturing of the beam top flange at a small distance away from the
weld toe. No cracks were visible at the weld toe or near weld access hole. However,
specimen MC1 showed much lower fatigue life than specimen MC2.

Figure 25 Plastic strain contour in axial direction showing accumulation of strains near weld
toe and weld access hole in presence of residual stresses.

Detailed investigation of the experimental results demonstrated accumulation of axial


strain (known as ratcheting) at the weld toe, weld access hole and also in the region of top
and bottom beam flanges close to the weld toe. Ratcheting is known to be demonstrated by
metallic materials under inelastic cyclic loading along the directions of force-controlled
loadingsteady or fluctuating [24]. Hence, it was anticipated that the occurrence of ratcheting
at the welded joint under displacement-controlled loading could be influenced by its residual
stresses. Similar to the experimental observations, axial strain accumulation was observed
very close to the weld toe and also around the weld access hole in the finite element
simulation (Figure 25).

178
Figure 26 Low-cycle fatigue response of the connection for specimen MC1. (a) Axial mean
strain and amplitude plotted against number of loading cycles, (b) axial stress-strain
hysteresis cycles showing ratcheting.

Figure 27 Low-cycle fatigue response of the connection for specimen MC2. (a) Axial mean
strain and amplitude plotted against number of loading cycles, (b) axial stress-strain
hysteresis cycles showing ratcheting.

179
A comparison of the axial mean strain and amplitude obtained from the finite element
analysis with that of the experimental results is shown in Figure 26a and Figure 27a for
specimen MC1 and MC2 respectively. It is interesting to note that finite element simulation
results predicted the axial strain ratcheting correctly when residual stresses were considered
in the analysis. The results of finite element analysis without residual stress showed very
small ratcheting phenomena. This was observed for both the specimens. In the experiments,
it was observed that the mean axial strain had a sharp increase when loading changed from
elastic to inelastic cycles, which was captured by the FE simulation when residual stresses
were considered. Although FE simulation without residual stresses showed small ratcheting,
it did not show this trend of sharp increase in the magnitude of axial mean strain. However,
the rate of ratcheting for both cases was comparable after the sharp increase in the mean axial
strain; still the magnitude of mean axial strain was much higher for the case with residual
stress as a result of the strain jump. The discrepancy in the FE simulation between the cases
with and without residual stresses as well as the sharp increase in the mean axial strain can be
clearly seen in Figure 26b and Figure 27b, where axial stress-strain hysteresis cycles shifted
in the direction of axial strain indicating axial strain accumulation when residual stresses
were considered in the analysis, whereas with no residual stress, the hysteresis loops were
cycling almost in the same location. Based on the experimental and simulated responses, it is
apparent that the increase in the axial strain accumulation observed in the experiments was
attributed to the presence of high residual stresses near the welded joint of the connection,
which can be simulated properly by considering residual stresses in the analysis. The
accumulation of axial strain near the connection can be very detrimental to the performance
of the connection which may lead to brittle low-cycle fatigue failure. Hence, analysis without
considering the residual stresses may not yield the actual failure response of the connection.
In order to investigate the region of influence of the residual stresses on the axial strain
accumulation, mean axial strains and strain amplitudes were plotted at different locations
across the flange width and length. One set of such plots for specimen MC1 is shown in
Figure 28. Accumulation of axial strain was observed across the flange width near weld toe

180
(Figure 28a). As already discussed, this can be attributed to the high residual stresses induced
by welding near the weld toe.

Figure 28 Low-cycle fatigue response of the connection at the beam bottom flange for
specimen MC1. (a) Axial mean strain, and (b) strain amplitude plotted against number of
loading cycles at different locations across the width near weld toe; (c) axial mean strain and
(d) strain amplitude across the length.

181
Variation in the axial strain amplitudes was observed along the flange width and there
was no specific trend in the variation of axial strain amplitude (Figure 28b). However, as the
distance from the weld toe increased, there was a decreasing trend in both the axial mean
strain and amplitude with almost no ratcheting after a distance of 108.8 mm from the weld
toe (Figure 28c,d). This is due to the fact that the magnitude residual stresses decreased as the
distance from the weld toe increased. The axial mean strain and amplitude plots near the
weld access hole (Figure 29a) shows very high amount of negative strain accumulation
which was also observed in the experiments. Without considering the residual stresses, the
finite element simulation did not show any axial strain accumulation near the weld access
hole. Another interesting phenomenon observed was the mean axial stress relaxation with
loading cycles near weld toe (Figure 29b). The mean axial stresses relaxed completely after
15-20 loading cycles, however, the axial strain continued to accumulate with number of
loading cycles. This unique feature of strain ratcheting even after relaxation of the mean axial
stresses was also observed by Lu [21] and Cheng [29] during experimental and numerical
study on cyclic response of welded piping joints. However, the reason for such type of
behavior was not explained. To find an explanation of the strain ratcheting phenomena even
after relaxation of the mean axial stresses, axial and von-Mises mean stresses were plotted in
Figure 30a and Figure 30b at two locations near (25 mm) and away (190 mm) from the weld
toe of the top flange for specimen MC1, respectively. The mean axial strains were also
plotted in the same figures to compare the responses. It was observed that although the mean
axial stresses relaxed with loading cycles, the von-Mises mean stresses did not relax, rather
remained constant throughout the loading cycles for simulation considering residual stresses
(Figure 30a). The magnitude of von-Mises mean stress was around 250 MPa for a location
close to the weld toe, whereas the mean axial stress peak was around 180 MPa (Figure 30a).
The mean axial stresses relaxed completely when the loading cycles were changed from
elastic to inelastic cycles, at the same time there was a sharp increase in the mean axial strain.
On the contrary, for simulation without considering residual stresses, the mean axial stress
was zero during the loading cycles, whereas von-Mises mean stress had a small value of 74
MPa at the beginning of the loading cycles. When the loading cycles had a transition from

182
elastic to inelastic phase, the von-Mises mean stress had a sharp increase and approached a
value close to the von-Mises mean stress value with residual stress case, and afterwards
remained constant. Similar observations were also made for a location away from the weld
toe, nonetheless, the magnitude of the stresses was lower compared to the location near the
weld toe (Figure 30b). The beam bottom flange also experienced analogous trend of stress
development, and hence not shown here.

Figure 29 (a) Axial mean stress plotted against number of loading cycles for specimen MC1
at different locations across the width of near weld toe showing stress relaxation, (d) axial
mean strain and amplitude near the top and bottom beam flange weld access hole for
specimen MC1.

Based on these observations, the ratcheting phenomenon or the sharp increase in the
mean axial strain can be attributed to either of the two following reasons or combination of
both. Firstly, the magnitude of mean axial stress was high at the beginning of the transition
phase between the elastic and inelastic loading cycles, which might have triggered the sharp
increase in the mean axial strain. Alternatively, at the transition phase between elastic and
inelastic loading cycles, the magnitude of von-Mises mean stress was large which might have

183
also influenced the sharp increase in the mean axial strain. After the sharp increase in the
mean axial strain both the analyses with and without residual stresses showed ratcheting. At
this point the mean axial stress had completely relaxed, and hence, it is apparent that at this
stage the ratcheting must have been influenced by the von-Mises mean stresses since both the
analyses with and without residual stress showed higher values of von-Mises mean stresses,
which remained constant after the sharp increase in the mean axial strain. Consequently the
ratcheting rate was constant after the sharp increase in the mean axial strain.

Figure 30 Axial mean stress and strain, and von-Mises mean stress plotted against number of
loading cycles at the top flange for specimen MC1. (a) 25 mm from weld toe, (b) 190 mm
from weld toe.

Next, results are compared for the two specimens under consideration to show the
influence of the welding sequence on the low-cycle fatigue response of the connection. It has
been already shown that the two specimens developed different peak residual stresses and
different distribution. The mean strain and strain amplitudes as well as the mean stress and
stress-strain hysteresis loops are plotted at different locations across the beam top and bottom
flanges to study the difference in the response of the connections. Figure 31 and Figure 32
show stress-strain hysteresis cycles along with axial strain mean and amplitude near the weld

184
toe for top and bottom beam flanges respectively. More plots of axial strain mean and
amplitude are presented in Figure 33.

Figure 31 Comparison of stress-strain response at the beam top flange for the two
specimens. (a) Stress-strain hysteresis response, (b) axial mean strain and amplitude plotted
against number of loading cycles.

It is apparent that the axial strain accumulation rate was completely different for the two
specimens with different weld sequences. In most of the locations across the weld toe and
weld access hole, specimen MC1 developed significantly higher strain ratcheting compared
to specimen MC2 for both beam top and bottom flanges. However, there were very few
locations where the axial strain ratcheting for specimen MC2 was either equal or slightly
higher than those for specimen MC1. Finally, the axial strain mean plots for the approximate
positions of fracture in the experiments (Figure 33d) evidently shows that the mean strain
values for specimen MC1 was almost two times higher than those for specimen MC2. This
observation clearly justifies the lower fatigue life that was observed for specimen MC1
during the experiments and the influence of weld sequence on the strain ratcheting.

185
Figure 32 Comparison of stress-strain response at the beam bottom flange for the two
specimens with different weld sequences. (a) Stress-strain hysteresis response, (b) axial mean
strain and amplitude plotted against number of loading cycles.

However, although the influence of weld sequence on the mean axial strain was
significant, the axial strain amplitude responses were not influenced by the weld sequence as
much as the mean axial strain. In several locations (Figure 31b, Figure 32b, Figure 33a)
specimen MC1 had higher strain amplitudes compared to specimen MC2 as anticipated.
Subsequently, there were several locations where specimen MC2 had equal and higher strain
amplitudes compared to specimen MC1 (Figure 33b,c,d). Hence, the direct influence of weld
sequence on the axial strain amplitude was not as clear as its influence on the mean axial
strain or ratcheting. Based on these observations, it can be concluded that the axial mean
strain has more influence on the low-cycle fatigue failure of the connections compared to
axial strain amplitude. Nonetheless, in both the specimens, the mean stresses relaxed within
10-20 cycles of loading, but still the ratcheting continued to increase with number of loading
cycles. High axial strain ratcheting rate was also observed around the weld access hole region
for both the specimens. As a result, the weld access holes are more prone to cracking during

186
low-cycle fatigue loading. The trend of axial strain accumulation near the weld access hole
was similar for both the specimens with different welding sequences.

Figure 33 Axial mean strain and amplitude plotted against number of loading cycles at
different locations of the beam flanges showing influence of weld sequence.

From the finite element investigation, it is apparent that the residual stress distribution
can be significantly influenced by the sequence of welding which inevitably may lead to

187
significantly different stress-strain response at different locations of WSMCs, especially near
the weld toe region of the beam flanges. It is believed that an optimized sequence of welding
can be obtained with a broad set of finite element analyses as presented in this study to
reduce the residual stresses near the weld toe and access hole region and thereby, increase the
fatigue life of the connections. Alternatively, the as-welded beam-column connection can be
heat treated based on the post weld heat treatment technique recommended by AWS D1.1
[78] to reduce the magnitude of residual stresses. This study dealt with the post weld heat
treatment technique to reduce the residual stresses which is presented later. Therefore, based
on the responses from the seismic analysis of WUF-W connections, it is apparent that the
residual stresses can be one of the contributing factors of the low-cycle fatigue failures that
have been observed in the cyclic testing of post-Northridge connections. The global
approaches of connection design ignore the influence of residual stresses on the seismic
performance of the connection because of the stress relaxation observed after few cycles of
loading. However, this study clearly showed that although the mean stresses relaxed after
few cycles of loading, the mean stresses worked as a driving force for axial strain
accumulation near the weld toe region and the weld access hole. Moreover, based on the
welding sequences, the axial strain accumulation rate was different at different locations near
the welded joint. This study manifests the necessity of incorporating welding induced
residual stresses in the analysis and design of moment connections during seismic loading,
where the connection may develop low-cycle fatigue damage leading to brittle failure of the
entire connection.

7. Post Weld Heat Treatment of WSMCs


The influence of post weld heat treatment (PWHT) on the residual stress relaxation at the
welded joints has been studied both experimentally and numerically and it has been observed
that the residual stresses can be significantly relieved with systematic post weld heat
treatment [22, 80-86]. This is the process commonly referred to as stress relief, so called
because it is carried out at temperatures at which the yield strength has fallen to a low value.
If the structure is heated uniformly, the yield strength of the material around the weld is

188
unable to support the initial level of the residual stresses, which are relieved by plastic
deformation [86]. Creep occurs at elevated temperatures and strain occurs by a diffusion
mechanism, relaxing the residual stresses even further. The extent to which the residual
stresses are relaxed depends on temperature and time for any given material and on material
for any given temperature.

Figure 34 A schematic of the time-temperature curve for the post weld heat treatment
specified by AWS D1.1 [78].

The stress distribution at the higher temperatures becomes more uniform and stress
reduces to a low level. After cooling, provided that it is carried out in a controlled manner,
the improved stress distribution is retained. Stress relieving heat treatments are generally
avoided unless specified as mandatory by Codes and/or Standards, because of the high cost
involved and potential adverse consequence of incorrect PWHT procedure [82]. However,
the influence of PWHT on relieving the residual stresses at the welded joint of moment
resisting connections has never been studied. Hence, in this study an attempt was made to
investigate the feasibility of the PWHT for moment resisting connections in increasing the
fatigue life by relieving the residual stresses.
AWS D1.1 [78] provides guidelines for post weld heat treatment of welded joints.
According to AWS D1.1 the welded assembly needs to be heated in the temperature range of
600-650C with a rate of heating as high as 220C/hr for heating temperatures above 315C.

189
Once the peak temperature is reached, it is kept constant for a certain amount of time based
on thickness of the connecting parts which may vary from 15 minutes to several hours. Then
the assembly is allowed to cool down in a controlled way with the rate of cooling being as
low as 55C/hr and as high as 260C/hr. A schematic of the time-temperature curve for the
post weld heat treatment proposed by AWS D1.1 is shown in Figure 34.

Case-I Case-II Case-III

Case-IV Case-V Case-VI

Figure 35 Post weld heat trreatment cases investigated in the study.

In order to incorporate the post weld heat treatment in the finite element simulation,
sequentially coupled transient thermo-mechanical analysis was performed similar to the
procedure described for weld residual stress simulation. First a heat transfer analysis was
performed on the WUF-W connection with six different cases of post weld heat treatment as

190
shown in Figure 35. These six different cases of PWHT were determined based on residual
stress distribution observed after the initial and welding process simulation. Since the
residual stress values were high up to a distance of 200 mm away from the weld toe, it was
decided to heat treat only 200 mm distance of the beam. In case-I of PWHT only beam
flanges were heat treated, whereas case-II involves heat treating both beam flanges and web.
Case-III, IV, V consist of heating the beam flanges, web and column flange close to the
welded joint, while case-VI involved heating the whole connection region. For the PWHT,
the peak temperature chosen was 600C with a heating rate of 200C/hr. Then the peak
temperature was hold constant for a time interval of 1 hour which was followed by slow
cooing at 55C/hr until the assembly reached the ambient temperature. The thermal analysis
of the six cases of PWHT provided the real time temperature distribution during the heating
and cooling phases which were then used as input in the residual stress calculation. However,
for the simulation of residual stresses after post weld heat treatment, both the initial and weld
residual stresses were fed as the initial state. The distribution of residual stresses after post
weld heat treatment for the six cases under investigation is shown in Figure 36 for specimen
MC1.
For cases I through V, the residual stresses were relieved reasonably, but the
redistribution of residual stresses as a result of post weld heat treatment led to development
of residual stresses higher or very close to the yield stress of the base material at several
locations near the weld toe and around the weld access hole. Since for cases I through V, the
entire connection region was not heated, the cooler adjacent areas led to the nonuniform
cooling rate at different parts of the heated welded joint. This different cooling rate
eventually led to the generation of residual stresses at the welded joints after PWHT.
However, case-VI where the whole connection region was heat treated showed significant
reduction in the residual stresses with no sign of higher stresses anywhere in the welded joint.
This was attributed to the uniform cooling of the welded joint as a result of heating the entire
connection region. The peak tensile and compressive residual stresses observed after post
weld heat treatment were 203 MPa and 186 MPa respectively which are less than half the
yield stress of the base material.

191
Case-I Case-II

Case-III Case-IV

Case-V Case-VI
Figure 36 Residual stress contours showing distribution of residual stresses after PWHT for
the six cases under investigation for specimen MC1.

192
Figure 37 Residual stress distribution in the beam top flange for specimen MC1 showing
effect of post weld heat treatment (Case-VI). (a) Across the width of the flange near the weld
toe, (b) along the length of the beam.

In order to investigate the residual stress redistribution at different locations in the


connection after post weld heat treatment for case-VI, residual stress distribution was plotted
at different locations of the connection as shown in Figure 37 and Figure 38.For specimen
MC1, the peak tensile residual stress at the weld toe across the width of the top flange was
around 300 MPa with compressive peaks at the edges of the flange as shown in Figure 37a.
After post weld heat treatment, the tensile stresses transformed to compressive stresses across
the width of the flange with a peak value less than 50 MPa indicating more than 83%
reduction in the peak value of the residual stresses. On the other hand, the compressive
stresses at the edges of the flange converted to tensile stresses with a peak value less than 75
MPa indicating 62% reduction in the peak value of residual stresses. The reduction of
residual stresses were significant around 200 mm length of the beam from the weld toe,
which was heat treated, however, the influence of heat treatment was visible up to 500 mm
distance along the length of the beam as seen in Figure 37b. Unlike beam top flange, the
residual stress distribution after welding was not uniform for beam bottom flange because of
the several start and stop locations of welding beads across the width of the flange. The

193
residual stress distribution at the weld toe across the width of the beam bottom flange is
shown Figure 38a, where there were several locations experiencing peak tensile (100 MPa)
and compressive residual stresses (300 MPa). However, after post weld heat treatment
significant reduction in the residual stresses was observed. Interestingly, the residual stresses
became completely compressive with almost a uniform distribution across the width of the
beam bottom flange.

Figure 38 Residual stress distribution in the beam bottom flange for specimen MC1 with
weld sequence A showing effect of post weld heat treatment (Case-VI). (a) Across the width
of the flange near the weld toe, (b) along the length of the beam.

The peak value of the residual stress observed after post weld heat treatment was around
100 MPa which indicates a 65% reduction in the peak value. Similar to the top flange of the
beam, the residual stress reduction was significant near the weld toe and the extent of
reduction decreased as the distance from the weld toe increased. However, unlike beam top
flange, there was no influence of post weld heat treatment in reducing the residual stresses
after 300 mm distance from the weld toe. One thing that must have been noted in both beam
top and bottom flanges is that after post weld heat treatment the distribution of residual

194
stresses near the weld region became almost uniform across the width of the flange and the
residual stresses developed were predominantly compressive. It has been concluded by many
studies that compressive residual stress is beneficial while tensile residual stress is
detrimental to fatigue strength [21, 27, 88-90].

Figure 39 Axial strain mean and amplitude plotted against number of loading cycles at (a)
beam top flange near weld toe, (b) bottom flange near weld toe, (c) around weld access hole
near beam top flange of the connection for specimen MC1 showing influence of PWHT
(Case-VI).

195
Hence, it is evident that systematic post weld heat treatment has a great potential in
increasing the fatigue life of moment resisting connections by relieving the residual stresses
near the weld region of the connection. The influence of case-VI of post weld heat treatment
on the low-cycle fatigue response of the WUF-W connections was investigated by
performing finite element analysis of the connections with PWHT under seismic type of
loading that was used in the previous analyses as shown in Figure 23b. As anticipated, the
axial mean strain accumulation was significantly reduced by the post weld heat treatment
(Figure 39) since the residual stresses were relieved significantly. As can be seen in Figure
39a and Figure 39b, the magnitude of axial mean strain in each loading cycle with PWHT
near the weld toe of beam top and bottom flanges was around half in magnitude compared to
the cases without PWHT.
On the contrary, around the weld access hole, the magnitude of axial mean strain without
PWHT was three times higher compared to the case with PWHT (Figure 39c). Hence, it is
evident that the axial strain ratcheting can be significantly reduced by incorporating post
weld heat treatment. Although the magnitude of ratcheting was lower with PWHT, the rate of
ratcheting was similar for both the cases with and without PWHT. The responses also
revealed that near the weld toe of the top and bottom beam flanges, the post weld heat
treatment did not have any influence on the axial strain amplitude (Figure 39a and Figure
39), however, around the weld access hole the axial strain amplitude showed significant
decrease similar to what was observed for axial mean strain (Figure 39c). The explanation for
such type of behavior in the axial strain amplitude was not very clear, and hence, the
influence of residual stresses and the post weld heat treatment on the axial strain amplitude
response needs further investigation. Moreover, further research is necessary to study the
influence of axial strain amplitude on the low-cycle fatigue response of moment connections.

8. Conclusions
This study demonstrated the development of initial and welding induced residual stresses
in moment resisting connections by using advanced numerical techniques which exhibited
good correlation with the experimental observations. Two WUF-W moment connections

196
with different welding sequences were simulated to investigate the influence of welding
sequence on the residual stress distribution, and subsequently on the low-cycle fatigue
response of moment resisting connections under seismic type of loading. The observations of
the study led to the following conclusions:
The development of initial and welding residual stresses can be properly simulated
by using advanced numerical techniques as described in this study, and three
dimensional modeling of the welding process is required in order to model the weld
sequence effect.
Significant amount of residual stresses develop close to the weld toe and weld access
hole of the connection because of the welding procedure, which gradually diminishes
with increase in the distance from the weld toe. The peak value of residual stresses
can be higher than the yield stress of the base material and it can be either tensile of
compressive or a combination of both.
Initial residual stress values are negligible around the edges of the beam, and it
increases with an increase in the distance from the edge with a peak value at the
centerline along the beam length. As a result, the residual stresses close to the weld
region are predominantly welding induced residual stresses.
The presence of residual stresses may lead to axial strain ratcheting near the weld toe
and weld access hole, and subsequently may lead to low-cycle fatigue failure of the
connection. Hence, the low-cycle fatigue failures that have been observed in the
cyclic testing of the post-Northridge connections can be attributed to the
development of damage accumulation near the weld toe and weld access hole as a
result of the axial strain ratcheting.
Different weld sequences induce different distribution and magnitude of residual
stresses around the welded joint, and hence, the ratcheting phenomenon would be
different for different weld sequences which subsequently influence the fatigue life
of the connections.

197
Post weld heat treatment as specified by AWS D1.1 [78] can be used systematically
to reduce the magnitude of residual stresses which is expected to increase the fatigue
life of moment resisting connections.

9. References
[1] FEMA-350. Recommendation and seismic design criteria for new steel moment-frame
buildings. prepared by the SAC Joint Venture for the Federal Emergency Management
Agency, Washington, DC, 2000.

[2] FEMA-353. Recommended specifications and quality assurance guidelines for steel
moment- frame construction for seismic applications. prepared by the SAC Joint Venture
for the Federal Emergency Management Agency, Washington, DC, 2000.

[3] FEMA-355D. State of the art report on connection performance. prepared by the SAC
Joint Venture for the Federal Emergency Management Agency, Washington, DC, 2000.

[4] SAC/BD-99/23. Failure analysis of welded beam to column connections. by Barsom, JM,
Pellegrino JV.

[5] Stojadinovic B, Goel SC, Lee KH, Margarian AG, Choi JH. Parametric tests on
unreinforced steel moment connections. Journal of Structural Engineering 2000;126:40-
49.

[6] SAC/BD-00/01. Parametric tests on unreinforced connections. by Lee KH, Stojandinovic


B, Goel SC, Margarian AG, Choi J, Wongkaew A, Reyher BP, Lee DY.

[7] SAC/BD-00/24. Development and evaluation of improved details for ductile welded
unreinforced flange connections. by Ricles JM, Mao C, Lu LW, Fisher JW.

198
[8] Lee D, Cotton SC, Hajjar JF, Dexter RJ. Cyclic behavior of steel moment-resisting
connections reinforced by alternative column stiffener details, I. Connection performance
and continuity plate detailing. Engineering Journal 2005;4th Quarter:189-213.

[9] Suita K, Nakashima M, Morisako K. Tests of welded beam-column subassemblies II:


Detailed behavior. Journal of Structural Engineering 1998;124(11):1245-1252.

[10] Nakashima M, Suita K, Marisako K, Maruoka Y. Tests of welded beam-column


subassemblies. I. Global behavior. Journal of Structural Engineering 1998;124:1236-
1244.

[11] Mazzolani FM. (Editor). Moment resistant connections of steel frames in seismic areas:
design and reliability. E&FN Spoon, 2000.

[12] SAC/BD-00/22. Cyclic response of RBS moment connections: loading sequence and
lateral bracing effects. by Yu S, Gilton C, Uang CM.

[13] Uang CM, Yu QS, Noel S, Gross J. Cyclic testing of steel moment connections
rehabilitated with rbs or welded haunch. Journal of Structural Engineering 2000;126:57-
68.

[14] SAC/BD-99/04. Effects of strength/toughness mismatch on structural and fracture


behaviors in weldments. by Dong P, Kilinski T, Zhang J, Brust FW.

[15] Smith JK. Vibrational fatigue failures in short cantilevered piping with socket-welded
fittings. ASME PVP-Vol. 338, Pressure Vessels and Piping Codes and Standards
1996;1:21-24.

[16] Vecchio RS. Fatigue evaluation of socket welded piping in a nuclear power plant. ASME
PVP-Vol. 338, Pressure Vessels and Piping Codes and Standards 1996;1:25-41.

199
[17] Brust FW, Zhang J, Dong P. Pipe and pressure vessel cracking: the role of weld induced
residual stresses and creep damage during repair. Trans. of the 14th Int. Conf. on Struct.
Mech. in Reactor Technol. (SMiRT 14), Lyon, France, 1, 297306, 1997a.

[18] Brust FW, Dong P, Zhang J. Influence of residual stresses and weld repairs on pipe
fracture. Approximate Methods in the Design and Analysis of Pressure Vessels and
Piping Components, W. J. Bees, ed., PVP, 347, 173191, 1997b.

[19] Zhang J, Dong P, Brust FW. Analysis of residual stress in a girth weld of a bwr core
shroud. Approximate Methods in the design and Analysis of Pressure Vessels and Piping
Components, W. J. Bees, ed., PVP, 347, 141156, 1997.

[20] Brickstad B, Josefson BL. A parametric study of residual stresses in multi-pass butt-
welded stainless steel pipes. International Journal of Pressure Vessels and Piping
1998;75(1):11-25.

[21] Higuchi M, Nakagawa A, Iida K, Hayashi M, Yamauchi T, Saito M, Sato M.


Experimental study on fatigue strength of small-diameter socket-welded pipe joints.
Journal of Pressure Vessel Technology 1998;120:149-156.

[22] Riccardella PC, Pan SH, Sullivan M, Schletz J, Gosselin SR. vibration fatigue testing of
socket welds. Pressure Vessel and Piping Codes and Standards 1998;360:453-463.

[23] Duranton P, Devaus J, Robin V, Gilles P, Bergheau JM. 3D modelling of multipass


welding of a 316L stainless steel pipe. Journal of Materials Processing Technology
2004;153-154:457-463.

[24] Joseph A, Rai SK, Jayakumar T, Murugan N. Evaluation of residual stresses in dissimilar
weld joints. International Journal of Pressure Vessels and Piping 2005;82:700-705.

200
[25] Gosselin S. Fatigue in operating nuclear power plants components after 60 years.
Workshop on U.S. nuclear power plant life extension research and development issues,
2008.

[26] Zhang YH, Maddox SJ. Fatigue testing of full scale girth welded pipes under variable
amplitude loading. Journal of Offshore Mechanics and Arctic Engineering
2014;136:021401-1- 021401-10.

[27] Mochizuki M, Hattori T, Nakakado K. Residual stress reduction and fatigue strength
improvement by controlling welding pass sequences. Journal of Engineering Materials
and Technology 2000;122:108-112.

[28] Lu X. Influence of residual stress on fatigue failure of welded joints. Ph.D. Dissertation,
North Carolina State University, Raleigh, NC, 2002.

[29] Humphreys AE. Influence of residual stress on the initiation of fatigue cracks at welded
piping joints. MS Thesis, North Carolina State University, Raleigh, NC, 2004.

[30] Cheng PY. Influence of residual stress and heat affected zone on fatigue failure of welded
piping joints. Ph.D. Dissertation, North Carolina State University, Raleigh, NC, 2009.

[31] Deng D, Kiyoshima S. FEM prediction of welding residual stresses in a SUS304 girth-
welded pipe with emphasis on stress distribution near weld start/end
location. Computational Materials Science 2010;50(2):612-621.

[32] Jiang W, Yahiaoui K. Effect of welding sequence on residual stress distribution in a


multipass welded piping branch junction. International Journal of Pressure Vessels and
Piping 2012;95:39-47.

[33] Teng TL, Chang PH, Tseng WC. Effect of welding sequences on residual
stresses. Computers & structures 2003;81(5):273-286.

201
[34] Ji SD, Fang HY, Liu XS, Meng QG. Influence of a welding sequence on the welding
residual stress of a thick plate. Modelling and Simulation in Materials Science and
Engineering 2005;13(4):553.

[35] Machold W, Staron P, Bayraktar FS, Riekehr S, Koak M, Schreyer A. Influence of the
welding sequence on residual stresses in laser welded t-joints of an airframe aluminium
alloy. Materials Science Forum 2008;571:375-380.

[36] Brust FW, Raju IS, Cheston D. Residual stresses and critical initial flaw size analyses of
welds. 50th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and
Materials Conference, Palm Springs, California, 2009.

[37] Gannon L, Liu Y, Pegg N, Smith M. Effect of welding sequence on residual stress and
distortion in flat-bar stiffened plates. Marine Structures 2010;23(3):385-404.

[38] Deng D. Influence of deposition sequence on welding residual stress and deformation in
an austenitic stainless steel j-groove welded joint. Materials & Design 2013;49:1022-
1033.

[39] Keivani R, Jahazi M, Pham T, Khodabandeh AR, Afshar MR. Predicting residual stresses
and distortion during multisequence welding of large size structures using fem. The
International Journal of Advanced Manufacturing Technology 2014;1-11.

[40] SAC/BD-97/05. Finite element fracture mechanics investigation of welded beam-to-


column connections. by Chi WM, Deierlein GG, Ingraffea A.

[41] SAC/BD-99/24. Weld acceptance criteria for seismically-loaded welded connections. by


Mohr W.

[42] Chi WM, Deierlein GG, Ingraffea A. Fracture toughness demands in welded beam-
column moment connections. Journal of structural engineering 2000;126(1):88-97.

202
[43] Matos CG, Dodds Jr RH. Modeling the effects of residual stresses on defects in welds of
steel frame connections. Engineering Structures 2000;22(9):1103-1120.

[44] Matos CG, Dodds Jr RH. Probabilistic modeling of weld fracture in steel frame
connections part i: quasi-static loading. Engineering structures 2001;23(8):1011-1030.

[45] Matos CG, Dodds Jr RH. Probabilistic modeling of weld fracture in steel frame
connections part ii: seismic loading. Engineering structures 2002;24(6):687-705.

[46] Righiniotis TD, Omer E, Elghazouli AY. A simplified crack model for weld fracture in
steel moment connections. Engineering structures 2002;24(9):1133-1140.

[47] Zhang J, Dong P. Residual stresses in welded moment frames and implications for
structural performance. Journal of Structural Engineering 2000;126(3):306-315.

[48] Lindgren LE. Finite element modeling and simulation of welding part 1: increased
complexity. Journal of Thermal Stresses 2001;24(2):141-192.

[49] Quayyum S., Hassan T. Simulation of initial residual stresses in hot-rolled wide flange
shapes. To be submitted to Journal of Constructional Steel Research, 2014.

[50] Fourier J. Analytical theory of heat. Dover Publications, New York, 1955.

[51] Shamsundar N, Sparrow EM. Analysis of multidimensional conduction phase change via
the enthalpy model. Journal of heat transfer 1975;97(3):333-340.

[52] Comini G, Guidice D. Finite element solutions of non-linear heat conduction problems
with special reference to phase change. International Journal of Numerical Methods in
Engineering 1979;8:613-624.

[53] Radaj D. Heat effects of welding. Springer-Verlag,1992, ISBN 3-540-54820-3.

203
[54] Hong JK, Tsai CL, Dong P. Assessment of numerical procedures for residual stress
analysis of multipass welds. Welding Research Supplement 1998;372s-382s.

[55] European Committee for Standardization EN1993-1-2. Eurocode 3: Design of steel


structures, part 1.2, general rules, structural fire design. European Committee for
Standardization, 2005.

[56] Wickstrm U. TASEF-2: A computer program for temperature analysis of structures


exposed to fire. Report No. 79-2, Lund Institute of Technology, Lund, Sweden,
Department of Structural Mechanics, 1979.

[57] Hu G, Morovat MA, Lee J, Eschell E, Engelhardt M. Elevated temperature properties of


astm a992 steel. In: Proceedings of ASCE structure congress 09, Austin, TX, USA, 2009.

[58] AISC. Specification for structural steel buildings. Chicago, IL, American Institute of
Steel Construction, 2005.

[59] Wikander L, Karlsson L, Nasstrom M, Webster P. Finite element simulation and


measurement of welding residual stresses. Modelling and Simulation in Materials
Science and Engineering 1994;2(4):845.

[60] AISC. Manual of steel construction: load and resistance factor designVolume I:
Structural members, specifications and codes. 2nd Ed., American Institute of Steel
Construction, 1994.

[61] SAC/BD-00/13. Preliminary evaluation of heat affected zone toughness in structural


shapes used in the construction of seismic moment frames. by Johnson MQ, Ramirez JE.

[62] Morrison M. An innovative seismic performance enhancement techniques for welded


steel moment connections. PhD Dissertation Research, North Carolina State University,
Raleigh, NC, 2014.

204
[63] Ketter RL, Kaminsky FL, Beedle LS. Plastic deformation of wide-flange beam columns.
ASCE Trans. 1955;120:1019, Reprint No. 91 and 101.

[64] Huber AW. Residual stresses in wide flange beams and columns. Fritz Engineering
Laboratory Report No. 220A.25, Lehigh University, Bethlehem, PA, 1958.

[65] Ketter RL. The influence of residual stresses on the strength of structural members.
Welding Research Council Bulletin 44, (November 1958), Reprint No. 132 (58-6).

[66] Alpsten GA. Thermal residual stresses in hot-rolled steel members. Fritz Engineering
Laboratory Report No. 337.3, Lehigh University, Bethlehem, PA, 1968.

[67] Bernuzzi C, Calado L, Castiglioni CA. Ductility and load carrying capacity prediction of
steel beam-to-column connections under cyclic reversal loading. Journal of Earthquake
Engineering 1997;1(2):401432.

[68] European Convention for Constructional Steelworks, ECCS. Recommended testing


procedure for assessing the behavior of structural elements under cyclic loads. TWG
1.3Seismic Design, ECCS Publication No. 45, Technical Committee 1, 1986.

[69] ATC. Guidelines for cyclic seismic testing of components of steel structures. ATC-24,
Redwood City, Calif., 1992.

[70] SAC/BD-97/02. Protocol for fabrication, inspection, testing and documentation of beam-
column connection tests and other experimental specimens. by Clark P, Frank K,
Krawinkler H, Shaw R.

[71] Krawinkler H, Zohrei M. Cumulative damage in steel structures subjected to earthquake


ground motions. Computers and Structures 1983;16:531541.

[72] Ballio G, Castiglioni CA. Seismic behavior of steel sections. Journal of Constructional
Steel Research 1994;29:2154.

205
[73] Mander JB, Chen SS, Pekan G. Low-cycle variable amplitude fatigue modelling of top-
and-seat angle connections. Engineering Journal 1994a;2:5462.

[74] Mander JB, Chen SS, Pekan G. Low-cycle fatigue behavior of semi-rigid top-and-seat
angle connections. Engineering Journal 1994b;3:111122.

[75] Castiglioni CA. Effects of the loading history on the local buckling behavior and failure
mode of welded beam-to-column joints in moment-resisting steel frames. Journal of
Engineering Mechanics 2005;131:568-585.

[76] Chaboche JL. Constitutive equations for cyclic plasticity and cyclic viscoplasticity.
International Journal of Plasticity 1989;5(3):247-302.

[77] Royster PB. Localized low cycle fatigue failure of welded steel moment connections. MS
Thesis, North Carolina State University, Raleigh, NC, 2007.

[78] Hassan T, Kyriakides S. Ratcheting of cyclically hardening and softening materials: I.


Uniaxial behavior. International Journal of Plasticity 1994;10:149-184.

[79] ANSI/AWS D1.1. Structural welding code-steel. The American Welding Society, 2008.

[80] Josefsson BL. Residual stresses and their redistribution during annealing of a girth-butt
welded thin walled pipe. Journal of Pressure Vessel Technology 1982;104:245250.

[81] Berglund D, Alberg H, Runnemalm H. Simulation of welding and stress relief heat
treatment of an aero engine component. Finite elements in analysis and design
2003;39(9):865-881.

[82] Sedek P, Brozda J, Wang L, Withers PJ. Residual stress relief in MAG welded joints of
dissimilar steels. International Journal of Pressure Vessels and Piping 2003;80:705-713.

206
[83] Paradowska A, Price JWH, Dayawansa P, Kerezsi B, Zhao XL, Ibrahim R. Measurement
of residual stress distribution in tubular joints considering post weld heat treatment.
Materials Forum 2006;30:30-38.

[84] Yaghi AH, Hyde TH, Becker AA, Sun W. Finite element simulation of welding and
residual stresses in a p91 steel pipe incorporating solid-state phase transformation and
post-weld heat treatment. The Journal of Strain Analysis for Engineering Design
2008;43(5):275-293.

[85] Wen Y, Geng L, Gong J, Wang Y, Tu S. Finite element analysis of welding residual
stress of the ultra-thick 13MnNiMoR steel cylinder of a large EO reactor made in China.
In ASME 2013 Pressure Vessels and Piping Conference, pp. V06BT06A071-
V06BT06A071. American Society of Mechanical Engineers, 2013.

[86] Akbarzadeh I, Attarha MJ, Salehi M, Sattari-Far I. Numerical and experimental study on
the effect of post-weld heat treatment parameters on the relaxation of welding residual
stresses. The Journal of Strain Analysis for Engineering Design 2011;46(2):79-86.

[87] Hosford WF. Mechanical behavior of materials. Cambridge University Press, New York,
2005.

[88] Chiarelli M, Lanciotti A, Sacchi M. Fatigue resistance of MAG welded steel elements.
International Journal of Fatigue 1999;21:1099-1110.

[89] Iida K, Yamamoto S, Takanashi M. Residual stress relaxation by reversed loading.


Welding in the World 1997;39(3):138-144.

[90] Teodosio JR, Fonseca MC, Pedrosa PD. Relaxation of residual stresses during fatigue
cycles in steels. Materials Science Forum 2003;426-432:3981-3988.

207
CHAPTER 7: SEISMIC PERFORMANCE OF A FIRE DAMAGED
MOMENT RESISTING FRAME

Abstract
If a steel building skeleton does not deform severely by a fire event, it is usually
rehabilitated for continued service. Material properties of structural steel undergo significant
change when exposed to high temperatures and subsequently cooled to room temperature.
During fire, a steel structure is exposed to very high temperatures with varying peak
temperatures at different locations of the structure. Hence, it is anticipated that the material
properties of the structural steel will be heterogeneous throughout the structure after the fire
event. It is not known how such a fire exposed steel building will perform under a seismic
event, if the extent of fire damage is low enough to rehabilitate for continued operation. In
this study, finite element analyses were performed on such type of fire damaged steel
moment resisting frames to investigate their seismic performance by considering the
heterogeneity of the material due to fire exposure. A seismic analysis was also performed on
a fire damaged isolated beam-column connection to characterize the influence of material
heterogeneity on the connections performance. The results of the study demonstrate the
vulnerability of fire damaged steel structures.

Keywords: Seismic analysis, Moment frame, Fire damage, Material heterogeneity, High
temperature

1. Introduction
Steel building structures may suffer various degrees of damages from fire exposure and
extinguishing, which can range from minor structural damage to complete collapse of the
building. To name few of the past major fires which exhibited severe damage to steel
structures but no collapse are the 1980 First Interstate Bank fire in Los Angeles, USA, the
1990 Broadgate Phase 8 fire in London, UK, the 1991 Credit Insurance Building fire in

208
Churchill, UK, the 1991 One Meridian Plaza fire in Philadelphia, USA, and the 2004 Parque
Central fire in Caracas, Venezuela [1-6]. All of these buildings, with the exception of One
Meridian Plaza, were repaired and returned to service. After extensive investigations and
studies, One Meridian Plaza was dismantled for economic reasons [6]. On the other hand, the
1970 One New York Plaza fire in New York, USA, the 1986 Alexis Nihon Plaza fire in
Montreal, Canada, the 2001 World Trade Center fire in New York, USA and the 2005
Windsor Building fire in Madrid, Spain are examples of catastrophic fire damage to steel
structures which led to complete collapse of the building [6-9]. The damages induced by fire
can be further exacerbated by another hazardous event, such as an earthquake. It is well
recognized that immediately after most major earthquakes in urban areas, a cluster of
building fires occurs. The post-earthquake fire losses can at times be similar to those made by
the ground movement alone. Under these circumstances, there is an expanded danger of fire
sources, diminished or strained firefighting capacities, the presence of structural damage, and
damage to the buildings fire protection systems. The 1906 San Francisco earthquake, the
1923 Tokyo earthquake, the 1989 Loma Prieta earthquake, the 1994 Northridge earthquake,
1995 Kobe earthquake are some of the examples of post-earthquake fire disasters. As
reported by Scawthorn et al. [10], 80% of 28000 buildings that were destroyed during the
1906 San Francisco earthquake and 77% of the 447000 houses that were destroyed by the
1923 Tokyo earthquake were caused by fire following the earthquakes. During the 1989
Loma Prieta Earthquake, San Francisco experienced 22 structural fires and over 500 reported
incidents during the seven hours after the earthquake began [6]. During the 1994 Northridge
earthquake, the Los Angeles Fire Department reported 476 fire incidents during almost a 20-
hour period following the earthquake [6, 11]. During the 1995 Kobe Earthquake
approximately 100 fires started within minutes of the quake, primarily in the densely
populated, low-rise residential areas of Kobe, Japan. It was reported that several large
conflagrations had developed within 1 to 2 hours, with a total of 142 fires and numerous
collapses and destruction [6, 11]. Accessible authentic records do not clearly demonstrate any
particular multi-story building failures because of post-earthquake fires. This may be the
consequence of an absence of fiasco information and exact representing the resultant building

209
damages or collapses between the seismic tremor and blaze causes. Nevertheless, post-
earthquake fires represent a serious danger to different varieties of development in their
conceivably harmed post-seismic tremor state, both regarding their diminished structural and
fire safety. While past experience does not give any immediate confirmation of such events
in multi-story structures, the likelihood for this consolidated extreme hazard from both
seismic tremor and fire exposures does exist.
Although significant loss to properties and human lives resulted during these fires, still
the knowledge-base on structural steel building during and after fire is relatively unknown;
especially when the structure involved in a fire survives. When a steel structure survives a
fire, the safety and reparability of the structure depends on the extent of structural damage
and the mechanical properties of structural steel of the connecting members [12-13]. On
many occasions fire affected steelwork shows little or no distortion resulting in considerable
uncertainty regarding its re-usability. An often quoted general rule for fire affected hot rolled
structural steels is that if the steel is straight and there are no visible distortions then it is
probably still fit for use. At 600C the yield strength and stiffness of steel is equal to about
40% and 30% respectively of their room temperature values; it follows therefore that any
steel still remaining straight after the fire and which had been carrying an appreciable load
was probably not heated beyond 600C, will not have undergone any metallurgical changes
and will probably be fit for re-use [12, 14-18].
During post-fire structural damage evaluation, the structural members can be categorized
into three types [12]- firstly, members that appear to be straight including slightly distorted
members, which is not easily detected by visual observations (within 4-5 times of ASTM A6
[19] rolling tolerances), secondly, members noticeably deformed but could be heat
straightened if economically justified, and finally, members severely deformed that only
under extreme circumstances would repair be given any consideration. When the member is
straight, it is generally assumed that the member has experienced temperatures less than
600C, and hence, the member can be accepted unconditionally for re-use assuming no
metallurgical change has occurred, whereas members with distortion may need to be
investigated properly to elucidate whether they can be accepted unconditionally, or there is

210
need for any rehabilitation measures such as heat straightening, strengthening with steel
plates or braces. On the other hand, for members with severe distortion, it is anticipated that
there is major metallurgical change which may affect the load carrying capacity of the
members. As a result, these types of members are usually replaced unless it is inappropriate
or impossible to remove the member. Only in those cases, repair or strengthening is allowed.
Test data on the fire performance of structural steel members led to the notion that over
600C the members start to deform noticeably, which become significant with increasing
temperature based on the size of the members. Studies by Saul [14], Dill [20], Smith [16],
Wright [21], Kirby et al. [22] showed that buckling and large deflection can occur at
temperatures well below 650C, whereas Kirby [23] and Li and Guo [24] observed
substantial distortion in a temperature range of 700C-900C. Another study by Wald et al.
[25] showed development of large buckle at temperatures over 1000C, and they showed that
the distortion of structural members not only depends on the peak temperature, but also on
the fire exposure time. However, in the literature, the experimental test data pertaining to the
fire performance evaluation of structural members consist of smaller beam and column
sections, and hence, most of the beam-column specimens distorted considerably at relatively
lower temperature range (600C-800C). Structures with large beam and column sections was
not investigated under fire scenario, and hence, there is a possibility that they may be able to
sustain very high temperatures (~1000C) without severe distortion. As a result, the member
will appear as if no or little distortion has taken place, nonetheless, there would be significant
degradation in the material strength as a result of very high temperature exposure and
subsequent cooling, which will be discussed later in this paper. Moreover, the peak
temperatures will vary along the length and depth of the members leading to material
heterogeneity in the fire exposed part of the structure. From the rehabilitation point of view,
the members might have been accepted unconditionally or with minor repairs since they look
nearly straight. Nobody knows how such type of fire damaged structure will perform if the
structure faces an earthquake after being exposed to fire, which is the subject of this study.
Although the post-fire material strength degradation has been addressed in few studies
[26-29], its influence on the seismic response of the structure has never been investigated. In

211
this study, seismic responses of special moment frames were investigated by performing
finite element analysis to account for the influence of material strength degradation and
heterogeneity due to fire exposure. A thermal analysis was performed to get the temperature
distribution in beam-column members during and after fire exposure. Once the temperature
distribution is obtained for each of the element in the finite element models, the material
properties or model parameters were determined. Subsequently, seismic performance of the
frames was investigated under simulated seismic loading history [30]. The results of the
finite element analysis were evaluated to explore the seismic responses of fire damaged steel
buildings.

2. Post-Fire Mechanical Properties of Steel


Since the objective of this study is to investigate the influence of post-fire change in the
mechanical properties of structural steel on the seismic response of moment frames, hence it
is essential to shed some light on the change in the mechanical properties of structural steel
due to fire exposure and subsequent cooling to room temperature. Limited past studies have
addressed the post-fire mechanical properties of structural steel [26-27]. However, the
available experimental data pertinent to post-fire evaluation of structural steel is limited.
Further, a review of the literature suggests that there is very little data available on the effects
of heating and cooling on the mechanical properties of ASTM A992 steel, which is currently
the most common structural steel used for rolled wide flange shapes in the U.S. Morrison
[29] studied the change in the mechanical properties of ASTM A992 steel due to heating at
elevated temperatures and subsequent cooling, where monotonic tension tests were
performed on ASTM A992 steel coupons. The coupons were conditioned at different peak
temperatures varying from 600-1200C; hold the peak temperature for 15 minutes followed
by air cooling until it reaches the room temperature. The stress-strain response of one set of
monotonic tension tests is shown in Figure . It was observed that as the maximum
temperature increased both the yield and ultimate stresses of the steel gradually decreased by
downward shift of the stress-strain curves. The influence of temperature on the reduction of
the yield stress was more noticeable in the temperature range of 800-1200C, since steel

212
undergoes a phase change near 730C from ferrite ( -Fe) to austenite ( -Fe). As seen in
Figure , up to 600C, the yield stress was close to the room temperature value, and a 40%
reduction in yield stress was observed for maximum temperature of 1200 C. Based on the
observations from the experiments it can be concluded that if ASTM A992 steel is heated
over 600C followed by air cooling, the yield stress never goes back to the room temperature
value upon removal of temperature, rather it reduces from that of the room temperature
value. Another important observation made in this study was that the elastic modulus does
not change with the peak temperature. A contemporary study by Lee and Engelhardt [28]
observed similar responses for ASTM A992 steel, where they performed monotonic tension
tests on ASTM A992 steel heated up to 1000C followed by three different cooling
conditions - air cooling, blanket cooling and water cooling. About 25% reduction in the yield
stress was observed under blanket and air cooling conditions, whereas water cooling
condition showed decrease in yield stress only at 800C after which the yield stress started to
increase. Moreover, the tensile strength increased with temperature under water cooling
condition, while for blanket and air cooling conditions, the tensile strength decreased with
temperature. It is obvious that the behavior of ASTM A992 steel is dependent on the peak
temperature exposure and the corresponding cooling condition.

Figure 1 Monotonic stress-strain response of ASTM A992 steel heated at different peak
temperatures followed by air cooling [29].

213
This study accounts for the post-fire change in the mechanical properties of structural
steel to investigate its influence on the seismic response of steel structures by performing
finite element analyses on steel moment resisting frames. The experimental data of Morrison
[29] was used to characterize the post-fire material properties of ASTM A992 steel.

3. Seismic Analysis of Moment Resisting Frames with Fire Damage


The study reported herein dealt with finite element investigation of special moment
frames under simulated earthquake loading history with and without fire exposure. The
influence of material heterogeneity on the seismic performance of moment frames was
investigated by performing nonlinear finite element analysis of a three story special moment
resisting frame. A three story special moment frame was adapted from the study by Adan
[31] to study the influence of post-fire material properties on the seismic response of the
frame. The special moment frame was designed and detailed in accordance with the
recommendations of AISC [32-33]. A schematic of the special moment frame adapted in this
study is shown in Figure 2.

(a) (b)
Figure 2 (a) A schematic of the special moment frame used in this study (adapted from Adan
[31]), (b) finite element mesh developed for the frame analysis.

The frame consisted of two bays of 9.14 m length each and three stories with a height of
4.26 m. The frame comprised of W36150 beams in the 2nd and 3rd levels with W24131

214
beams in the roof level. The exterior column size was W14257, while interior column size
was W14283. The connection between beam-to-column was designed as welded
unreinforced flange welded web (WUF-W) connection. The factored uniformly distributed
load on the beam was assumed to be 14.6 kN/m, whereas the factored axial load on the
column was assumed to be 444 kN as adopted by Adan [31].

3.1. Finite Element Modeling


Three dimensional finite element models were developed for the special moment frame
as shown in Figure 2a by using the finite element analysis software ANSYS. The frame was
discretized by 8-noded solid elements, and both geometric and material nonlinearities were
considered. The FE mesh developed for the special moment frame is shown in Figure 2b.
The frame assembly was assumed to be made of ASTM A992 steel, and both the room
temperature and post-fire material properties of ASTM A992 steel was obtained from the
study by Morrison [29]. The multilinear kinematic hardening model was used in the finite
element simulation to characterize the material behavior and the material model parameters
were obtained by fitting the stress-strain response of ASTM A992 at different temperatures
as shown in Figure . For the simplification of the analysis and to reduce the computational
time and cost, the weld access hole and the shear tab was excluded in the frame analysis.
Since the purpose of the study was to investigate the global performance of the frame, the
exclusion of weld access hole or shear tab would not be much influential. The finite element
frame models were investigated under a displacement-based artificial loading protocol
offered by SAC/BD-97/02 [30]. The displacement based time history was based on the SAC
Test Protocol and consisted of imposing a series of prescribed quasi-static cyclic
displacements to the top of the building frame. The prescribed displacements included six
cycles of 0.375, 0.5 and 0.75% interstory drift, followed by four cycles of 1% interstory drift
and two cycles each of 1.5, 2, 3, 4, 5, 6% and so forth interstory drift. A schematic of the
loading protocol is shown in Figure 3. The frame was analyzed for room temperature
condition followed by four fire case studies based on fire in different compartments. The four
fire cases are defined as follows:

215
Fire Case-I: Fire in left compartment at 1st story
Fire Case-II: Fire in both compartments at 1st story
Fire Case-III: Fire in left compartment at 2nd story
Fire Case-IV: Fire in both compartments at 2nd story
Before performing detail finite element investigation of the special moment frame
under different fire scenarios, it is important to define a time-temperature history to
characterize the fire event and a distortion analysis need to be performed on the moment
frame using the defined time-temperature history to ensure selection of a peak temperature
which will not cause severe deformation in the frame members.

Figure 3 Displacement based simulated earthquake loading history [30].

3.2. Distortion Analysis of a Single Story Moment Resisting Frame during Fire
In order to study the influence of material heterogeneity induced by fire exposure, the
distribution of temperature across the members need to be determined during a fire scenario
by performing a thermal heat-transfer analysis. Based on the peak temperature of each of the
elements in the FE model from the thermal analysis, the relevant mechanical properties of
structural steel will be assigned to each of the element in the seismic analysis. However, the
thermal analysis requires a specific time-temperature history to adequately represent a real

216
fire event. In this study a realistic fire time-temperature curve, which includes a cooling
phase, is used instead of the more traditional standard fire curves such as those presented in
ASTM E-119 [34] and ASTM E-1529 [35]. The selected curve is taken from a study by
Quiel and Garlock [36] and provides an accurate representation of a typical compartment fire
seen in office buildings. The realistic time-temperature curve used for this study can be seen
in Figure 4a compared to the standard fire curves presented in ASTM E-119 and ASTM E-
1529. It can be observed that the peak temperature of the prescribed time-temperature curve
in this study was around 1000C, and it is not known whether the beam and column sections
in the current study will be significantly distorted at this peak temperature level. Since the
main focus of this study is to investigate the influence of post-fire mechanical properties of
steel on the seismic performance of moment frames, it is assumed that the damage of the
frame members from fire exposure will be accepted unconditionally or with minor repairs
before being re-used. Hence, a coupled thermo-mechanical analysis was performed for a
single story of the special moment frame shown in Figure 2a, to investigate whether there is
any substantial distortion in the connecting members during fire exposure. The thermo-
mechanical analysis will enable to establish the peak temperature for the fire time-
temperature curve. The thermo-mechanical analysis was performed under gravity loading,
where a factored uniformly distributed load of 14.6 kN/m was assumed on the beam, whereas
a factored axial load of 444 kN was assumed on the column. Supports were provided in the
out-of-plane direction at every 1.83 m distance along the length of the beam to avoid the out-
of-plane bending of the beam section. Figure 5 shows the finite element mesh of the single
story two bay moment frame used for the thermo-mechanical analysis, where the frame was
discretized by 8-noded solid elements. To reduce the computational time and cost, a coarser
mesh was used. The distortion analysis was performed considering two fire scenarios-fire in
a single compartment, and fire in both compartments.
Thermal analysis was performed on the moment frame in order to get the real time
temperature distribution during the fire event and subsequent cooling. For the thermal finite
element analysis, temperature was given as input with time as presented in Figure 4a (red
line). The temperature was assigned to the bottom surface of the beam bottom flange and

217
inside surface of the column flanges in order to simulate the real compartment fire scenario.
The fire temperature initially increases with time up to 20 minutes when it reaches the peak
temperature of 1000C, where it is held constant for 30 minutes before the start of the air
cooling process. The air cooling was simulated in thermal finite element analysis by allowing
free heat transfer in all directions. Heat loss occurred from the material surface through both
convection and radiation, which were related to the heat transfer coefficient. The
temperature-dependent material properties (specific heat, thermal conductivity, enthalpy,
heat transfer coefficient) of structural steel were used in the heat transfer analysis, where the
material parameters were defined as a function of temperature. Latent heat/phase
transformation was accounted for by defining enthalpy of the material as a function of
temperature.

Figure 4 (a) Time-temperature curves to represent a fire scenario, (b) monotonic stress-strain
response of ASTM A992 steel at elevated temperatures [37].

Figure 6a shows the distribution of temperature in the moment frame after 50 minutes of
fire exposure in the left compartment only. It is observed that the temperature at beam bottom
flange and column inside flange was 1000C. However, the temperature at the web of the
beam located close to the bottom flange reached a temperature of 800-900C, and as the

218
distance increased along the depth of the beam there was a decrease in the temperature. It
was noted that after about one-third distance away from the beam bottom flange, the
temperature of the beam web was almost close to the room temperature. As a result, the
material properties of structural steel will be heterogeneous along the depth of the beam. The
beam bottom flange and the beam web close to the bottom flange will experience major
change in the material properties because of very high temperature exposure (800-1000C),
whereas the beam web one-third distance away from the bottom flange and the beam top
flange will experience no change in the material properties. Similar type of temperature
variation along the depth of the column was observed, where the inside flange of the column
experienced 1000C while the outside flange was at room temperature. The temperature
distribution in the column web was variable where the temperature reached room temperature
value about two-third distance away from the inside flange. This will induce material
heterogeneity along the depth of the columns as well. Similar observations were made in the
temperature distribution in the moment frame when both the compartments were subjected to
fire (Figure 7a). The beam and the exterior columns showed the same trend in the
temperature distribution as was observed for single compartment fire, however, unlike
exterior column, in the interior column both the flanges experience 1000C, and the
temperature varied from 800C-360C along the depth of the web.

Figure 5 Finite element mesh developed for thermo-mechanical analysis of a single story
two bay moment frame.

219
(a)

(b)

(c)

Figure 6 Response of a single story frame with fire in a single compartment. (a) Temperature
distribution (in Kelvin) at the end of peak temperature exposure (50 minutes), (b) vertical
deflection (in meter), and (c) out-of-plane deflection (in meter) at the end of the cooling
phase.

220
(a)

(b)

(c)

Figure 7 Response of a single story frame with fire in both compartments. (a) Temperature
distribution (in Kelvin) at the end of peak temperature exposure (50 minutes), (b) vertical
deflection (in meter), and (c) out-of-plane deflection (in meter) at the end of the cooling
phase.

221
Once the heat-transfer analysis was finished, structural analysis was performed under
gravity loading by changing the thermal solid elements to structural solid elements, where the
results from the solution of the thermal environment was coupled to the information provided
in the structural environment. In the structural analysis of the frame, temperature dependent
mechanical properties of ASTM A992 steel were used with relevant constitutive equations.
The mechanical properties used in the structural analysis were density, elastic modulus,
initial yield stress, Poissons ratio and coefficient of thermal expansion. The elevated
temperature stress-strain response of ASTM A992 steel under monotonic tension loading was
obtained from Hu et al. [37] as shown in Figure 4b. The elevated temperature stress-strain
response of ASTM A992 steel was implemented in the finite element simulation by using
multilinear kinematic hardening model available in ANSYS which incorporates von-Mises
yield criterion and associative flow rule. The multilinear kinematic hardening model
parameters were obtained through fitting the stress-strain response of A992 steel. Note that
there was no material data available for temperatures over 900C, for which linear
interpolation method was used to obtain the model parameters. Figure 6b and Figure 6c show
the deformation in the vertical and out-of-plane direction of the frame at the end of the
cooling phase after fire exposure in a single compartment. It can be observed that the
maximum vertical and out-of-plane deflections were 21.7 mm and 22.7 mm at the end of the
cooling phase, which is only twice the values permitted by ASTM A6 [19] for variation in
straightness of W shapes. Similarly when both the compartments were subjected to fire
(Figure 7b and Figure 7c), the maximum vertical and out-of-plane deformations observed
were 19.7 mm and 25.1 mm respectively, which are not significantly large compared to the
values permitted by ASTM A6 [19] for variation in straightness of W shapes. Hence, based
on the observed deformation values it is quite obvious that the frame did not incur major
structural distortion, which can be accepted unconditionally, or recovered by minor heat
straightening if necessary. Therefore, the members of the special moment frame adapted in
this study should not distort significantly under the prescribed fire condition, and hence,
could be used further to investigate the influence of material heterogeneity on the seismic
response.

222
(a)

(b)

Figure 8 Stress (in Pa) in longitudinal direction of a single story frame with fire in a single
compartment. (a) At the end of peak temperature exposure (50 minutes), (b) at the end of the
cooling phase.

It is important to note here that the out-of-plane deformation was predicted by the global
finite element models despite the fact that initial imperfections were not assigned to the
geometry of the model. Typically, finite element models require the inclusion of initial
imperfections to predict out-of-plane deformation; however, for the models considered in this
study out-of-plane deformation was predicted even though the initial prescribed geometry
was perfect. The out-of-plane deformation observed in the moment frame for both the cases

223
with single and two compartment fire was due to the thermal contraction and expansion of
the beam as a result of the fire exposure and subsequent cooling. The thermal expansion
during the fire exposure led to high compressive stresses in the bottom flange of the beam
and the lower part of the beam web, whereas tensile stresses developed at the top flange of
the beam and upper part of the beam web (Figure 8a). As a result, noticeable buckling was
observed at the beam bottom flange and beam web. During the cooling phase, due to thermal
contraction, the compressive stresses developed at the bottom flange and lower part of the
beam web transmitted to tensile stresses, whereas the tensile stresses developed at the top
flange and upper part of the beam web became compressive (Figure 8b). As a result, the
deformation on the beam bottom flange and lower part of the beam web decreased. This
decreased the out-of-plane deformation of the frame. This is clearly seen in Figure 9a where
the axial stresses at three different locations of beam cross-section is plotted against time at
the centerline along the beam length. The out-of-plane deformation at the same locations are
also plotted in Figure 9b.

Figure 9 (a) Stress in longitudinal direction, and (b) out-of-plane deformation at three
different locations across the beam cross section along the centerline of the beam length.

224
3.3. Seismic Response of Fire Damaged Moment Resisting Frames
As already mentioned, the special moment frame shown in Figure 2 was investigated for
room temperature and four different fire scenarios. In Fire Cases-I and III, a single
compartment was subjected to fire in 1st and 2nd story of the frame respectively, whereas in
Fire Cases-II and IV, both the compartments were subjected to fire in 1 st and 2nd story of the
frame respectively. The temperature distribution in the frame for Fire Cases-I and III were
obtained from the thermal analysis of single story frame with one compartment fire used in
the distortion analysis of the frame (Figure 6a). Similarly, the temperature distribution in the
frame for Fire Cases-II and IV were obtained from the thermal analysis of single story frame
with fire in both compartments used in the distortion analysis of the frame (Figure 7a). The
material properties of ASTM A992 steel were defined at different locations of the moment
frame based on the temperature distribution obtained from the thermal analysis.
To study the seismic behavior of the special moment frame at room temperature the
simulated earthquake loading history was applied at the roof level of the frame up to 6%
interstory drift angle. The frame was designed based on the strong column-weak beam
philosophy, and hence, all the connections in the frame showed development of plastic
hinges in the beam flanges as shown in Figure 10a, where the equivalent plastic strain
contour is plotted for the entire frame. Zoomed views of the exterior (left) and interior
connections at the 3rd level of the frame are also shown in Figure 10c and Figure 10d
respectively. The base shear-interstory drift angle response of the frame (Figure 10b) shows
that the frame started to loose strength at 4% interstory drift angle with a gradual decrease in
the strength of the frame with loading cycles. The reduction in the strength at 6% interstory
drift angle was 17% which indicates very stable response of the frame. The lateral and out-
of-plane displacement contours of the frame are plotted in Figure 11a, and b respectively.
The lateral displacement profile of the frame showed that in each story the drift was about the
same, and hence, the energy was dissipated over the entire height of the building which
resulted in lower maximum drift demands. On the other hand, the out-of-plane displacement
contour shows that the beam-column assemblies developed significant buckles in the beam
flanges and web, which resulted in out-of-plane deformation of the beam, and the maximum

225
value of the out-of-plane deformation obtained was around 92.5 mm for loading up to 6%
interstory drift angle. When fire damage was included in the finite element analysis by
incorporating post-fire material heterogeneity, it was observed that the strength of the frame
decreased regardless of the fire case.

(a) (b)

(c) (d)

Figure 10 Cyclic response of the frame at room temperature. (a) Equivalent plastic strain
contour of the frame at 6% interstory drift angle, (b) base shear plotted against interstory drift
angle showing hysteresis response of the frame, (c) equivalent plastic strain contour of the
exterior (left), and (d) interior connection at the 3rd level at 6% interstory drift angle.

226
(a) (b)

Figure 11 Displacement contours (in meter) of the frame at room temperature at 6%


interstory drift angle. (a) Lateral displacement, (b) out-of-plane displacement.

For Fire Cases-II and IV where both compartments were exposed to fire in a story, the
reduction in strength was more compared to Fire Cases-I and III, where only one
compartment was exposed to fire. The hysteresis responses of the frame for different fire
cases are shown in Figure 12, where it can be observed that the basic pattern of the hysteresis
response with fire damage was similar to the room temperature hysteresis response with
reduced peak forces at each loading cycles. The envelope of the positive and negative peak
lateral forces are compared in Figure 13a for different fire cases with room temperature
response. For Fire Case-I only 4% reduction in strength of the frame was observed, whereas
Fire Cases-II and III yielded 5% reduction in strength. On the other hand, Fire Case-IV
showed 8% decrease in the lateral strength. It was observed that the hysteresis responses
were symmetric for all the fire cases. However, although the material strength degradation in
the fire exposed members was not influential in the overall strength of the frame, the
interstory drift demands of the frame was significantly affected by the material strength
degradation and heterogeneity. For Fire Case-I, only the left compartment on 1st story was
involved in fire. As a result, the beams and columns in the compartment experienced high
temperature and degradation in material strength. Consequently, the 1st story possessed lower

227
lateral strength compared to the adjacent stories and hence, more lateral deformation took
place at the 2nd level indicating higher maximum drift demands at the 2 nd story level. As
shown in Figure 13b, nearly all of the lateral displacement occurred at the 2nd story level, and
the upper stories did not show much movement relative to the adjacent stories. Similar
observations were made for Fire Case-II where both the compartments at 1st story were
involved in fire.

Figure 12 Hysteresis response of the frame for different fire cases.

228
Figure 13 Comparison of frame responses at room temperature and different fire cases. (a)
Peak positive and negative lateral forces at each loading cycle plotted against interstory drift
angle, (b) interstory drift angles at each story level.

(a) Fire Case-III (b) Fire Case-IV

Figure 14 Formation of plastic hinge in the interior column at the 2nd level. (a) Fire Case-III,
(b) Fire Case-IV.

229
Since two compartments were exposed to fire, the lateral strength at the 1st story was
even less than Fire Case-I and as a result, the maximum drift demands at the 2nd level was
even higher compared to Fire Case-I. On the other hand, for Fire Cases-III and IV,
compartments at 2nd story were exposed to fire. Similar to Fire Cases-I and II, the maximum
drift demands now increased at the 3rd level indicating significant reduction in lateral strength
at the 2nd story. As shown in Figure 13b, for Fire Case-III, at the 2nd level, the interstory drift
angle was around 1.7%, whereas, at the 3rd level, the interstory drift angle was around 3.1%
indicating more than 50% of the total drift occurring at the 3rd level. For Fire Case-IV, where
two compartments at the 2nd story were exposed to fire, the maximum drift demands at the 3rd
level increased even more compared to Fire Case-III, In this case, 67% of the total story drift
was occurring at the 3rd story level. It was interesting to note here that for Fire Cases-III and
IV, where compartments at the 2nd story were exposed to fire, the interior connection at 2nd
level did not exhibit development of plastic hinges in the beam flanges, rather plastic hinges
started to develop at the column as shown in Figure 14. Since the 2nd story was weaker in the
lateral direction compared to 1st and 3rd stories, an inflection point developed between the
relatively stiff 1st story and weak 2nd story. This indicates a weak column-strong beam
mechanism, which is essentially undesirable mode for seismic applications. Based on the
responses, it is quite obvious that when compartments at any specific story are exposed to
fire, the story becomes weaker or more flexible than the other stories. Consequently, there is
significant reduction in the lateral strength at that story level which leads to very high drift
demands. This is a sign of development of single story mechanism commonly known as soft
story which may lead to catastrophic failure of the whole structure. The post-fire change in
the material properties not only influenced the lateral deformation of the frame, but also the
out-of-plane deformation. The out-of-plane displacement contours for the fire cases are
shown in Figure 15. The same plot for the frame at room temperature was shown in Figure
11b. It can be observed that for all fire cases as well as at room temperature, severe buckling
occurred in the beam flanges and web close to the connection region, except Fire Cases-III
and IV where the interior connection at the 2nd level did not show buckling in the beam.

230
(a) Fire Case-I (b) Fire Case-II

(c) Fire Case-III (d) Fire Case-IV

Figure 15 Out-of-plane displacement contours (in meter) of the frame for different fire cases.

This is a common phenomenon observed in large inelastic loading cycles during the
cyclic testing of beam-column assemblies which initiates instability in the connection
assemblies. Comparing the contour plots of Figure 11b and Figure 15, it can be observed that
post-fire material strength degradation increased the out-of-plane deformation in the beam
corresponding to the fire exposed compartment. However, for Fire Case-I, the out-of-plane

231
deformation was insensitive to the post-fire material strength, whereas for fire cases-II, III
and IV, the increase in the out-of-plane deformation were 20%, 37%, 72% respectively. This
manifests that the fire damaged structures will be susceptible to early instable behavior
compared to the structures without fire damage. Therefore, based on the responses from the
analyses performed in this study it is apparent that if a steel building does not experience
severe distortion during a fire event and is rehabilitated for continued service, it may be
vulnerable to earthquake loading due to the post-fire material strength degradation and
heterogeneity. Hence, it is essential that the post-fire material strength be investigated
regardless of the distortion level before making a decision on the re-usability or reparability
of the structure.

3.4. Seismic Response of an Exterior WUF-W Connection with Fire Damage


To characterize the influence of post fire material properties of the structural steel on the
seismic response of moment connections, an exterior welded unreinforced flange welded web
(WUF-W) moment connection was chosen from the reference special moment frame as
shown in Figure 16a. The finite element model developed for the exterior WUF-W
connection is shown in Figure 16b, where the beam and column were modeled and
discretized by 8-noded solid brick elements and the presence of weld access hole and shear
tab was included. A finer mesh was considered along the beam and column length in the
vicinity of the beam-to-column connection, and also in the column panel zone where
majority of the inelastic action occurs. The other regions of the connection will remain elastic
and hence, coarser mesh was used in those places to reduce the computational time and cost.
Initially the connection was investigated at room temperature under the simulated seismic
loading history. In the finite element simulation, the support conditions of the specimens
were chosen to represent the theoretical inflection points of the beam and the column in a
typical frame building. The specimen supports were designed to mimic the support
conditions of hinge support. The specimen was treated as a cantilever beam connected to the
column flange and the far end of the beam was subjected to the simulated earthquake loading
history. The equivalent plastic strain contour and the moment-rotation hysteresis responses

232
for the WUF-W connection under simulated seismic loading history at room temperature are
shown in Figure 17. It was observed that the connection developed plastic hinges in the beam
flanges away from the welded joint (Figure 17a) with a symmetric moment-rotation
hysteresis response. The moment-rotation plot (Figure 17b) showed that the connection
started to loose strength after 2% interstory drift angle, and there was gradual decrease in the
strength of the connection with loading cycles at higher interstory drift angles. The decrease
in the strength of the connection at 3%, 4%, 5% and 6% interstory drift angles were 13%,
24%, 31% and 40%, respectively under both positive and negative bending, which indicates
that considerable amount of energy can be dissipated by the WUF-W connection at room
temperature.

9.14 m 9.14 m 4.26 m

4.26 m
4.26 m 4.26 m

4.57 m

(a) (b)

Figure 16 (a) Reference special moment frame showing an exterior WUF-W connection
taken for the study, (b) finite element mesh of the external WUF-W connection.

To study the post-fire seismic performance of this connection, a thermal analysis was
performed on the connection to get the temperature distribution during and after the fire
event. Once the temperature history was obtained for each of the element from the thermal
analysis, the material parameters for each of the element was obtained by considering the
peak temperature of the element during the temperature history. The post-fire material

233
properties of ASTM A992 steel was obtained from the experimental tests performed by
Morrison [29] and implemented in ANSYS by using multilinear material model. Figure 18a
shows the temperature distribution in the connection after being exposed to the peak
temperature level in the prescribed time-temperature curve (Figure 4a). The distribution is
similar to the one that has been observed in the single story moment frame distortion analysis
where only one compartment was subjected to fire, and hence wont be discussed here to
avoid repetition. However, the contour plot clearly shows the variation of temperature along
the depth of the beam and columns with temperatures as high as 1000C, and as low as room
temperature.

(a)

Figure 17 Response of WUF-W connection at room temperature under simulated seismic


loading history. (a) Equivalent plastic strain contour showing formation of plastic hinges, (b)
moment-rotation hysteresis response.

Once the temperature distribution was obtained from the thermal analysis, structural
analysis was performed on the connection under simulated seismic loading history. The
equivalent plastic strain contour of the connection is shown in Figure 18b. Similar to the
response of WUF-W connection at room temperature, the WUF-W connection with fire
damage developed plastic hinges away from the welded region, but the beam bottom flange

234
exhibited more deformation compared to the top flange. This was anticipated since the
bottom flange was exposed to higher temperature (1000C) compared to the top flange which
led to material strength degradation at the bottom flange, whereas the top flange maintained
the room temperature material strength. This led to an unsymmetrical moment-rotation
hysteresis response (Figure 19a), where the degradation under negative bending was
significantly higher than that under positive bending. Under positive bending, the connection
showed increase in strength up to 2% interstory drift angle, and afterwards the strength
started to degrade gradually. The reduction in the strength at 3%, 4%, 5% and 6% interstory
drift angles were around 5%, 13%, 20% and 28% respectively.

(a) (b)

Figure 18 (a) Temperature distribution (in Kelvin) at the peak temperature level of fire
exposure in the WUF-W connection, (b) equivalent plastic strain contour of fire damaged
WUF-W connection showing formation of plastic hinges.

On the other hand, under negative bending, the connection showed increase in strength
up to 1.5% interstory drift angle followed by a 12% decrease at 2% interstory drift angle. At
higher interstory drift angles such as 3%, 4%, 5% and 6%, the strength decreased by 28%,
43%, 55% and 64% respectively. The decrease in the strength of the fire damaged connection
at 4% interstory drift angle under negative bending was higher than the degradation observed
at 6% interstory drift angle at room temperature. As a result of significant strength

235
deterioration of the connection under negative bending, the moment-rotation hysteresis loop
exhibited hysteresis pinching which led to reduced energy dissipation capacity as observed in
Figure 19a. A comparison plot of the moment-rotation envelope of the peak negative and
positive moments at each loading cycle is shown in Figure 19b, where the moment-rotation
envelopes were plotted for the WUF-W connections with and without fire damage.

Figure 19 (a) Moment-rotation hysteresis response of WUF-W connection with fire damage
under simulated seismic loading history, (b) comparison of moment-rotation envelopes
showing peak positive and negative moments at each loading cycle.

It can be observed that for the fire damaged connection, the overall strength of the
connection reduced compared to the connection without fire damage regardless of positive
and negative bending. The reduction in the peak moments under positive and negative
bending were around 10% and 14% respectively. This was attributed to the post-fire
degraded strength of the material. However, as already mentioned, for fire damaged
connection, the strength under positive bending degraded gradually, whereas the strength
deteriorated sharply under negative bending. Moreover, under negative bending the strength
degradation initiated earlier than that under positive bending. This type of hysteresis behavior
was not observed in the overall response of the frame because of the strength contribution of
the members unaffected by fire. However, while considering the response of each individual

236
connection in the frame, this type of hysteresis behavior poses severe risk for the connections
that are directly exposed to fire.

4. Conclusions
This study demonstrated the influence of post-fire material strength degradation and
heterogeneity on the seismic response of steel structures by performing finite element
analyses on special moment resisting frames. The responses obtained from the analyses led to
the following conclusions:
Structural members of special moment frames designed for seismic application with
large beam and column sections can withstand higher temperatures (~1000C)
without significant distortion during fire, and hence, visibly they may seem to be
unaffected by the high temperature exposure. However, at such high temperature
range structural steel undergoes significant change in the mechanical properties which
essentially changes the structural behavior. Hence, it is anticipated that even though
there is no significant distortion in the structure, the post-fire degradation of structural
steel may adversely influence its seismic performance.
The seismic analysis of special moment resisting frames by considering fire damage
demonstrated that the strength degradation of fire exposed structural members is not
very influential to the overall strength of the structure, rather the lateral drift demands
were significantly influenced by the post-fire strength degradation and heterogeneity
of structural steel. It was observed that nearly all of the earthquake-induced lateral
displacement occurs at the story level where the fire exposed compartments are
located. This may lead to single story mechanism commonly known as soft story at
the fire exposed story level which can lead to catastrophic failure of the structure. In
addition, it was observed that with increase in the number of fire exposed
compartments in a story, there is decrease in the lateral strength and increase in the
drift demands at that story level.
Seismic analysis of an isolated WUF-W moment connection revealed that after fire
exposure, the temperature across the depth of a beam is unevenly distributed which

237
leads to material heterogeneity. As a result, the strength of the connection under
negative bending drops steeply compared to the strength under positive bending
which induces hysteresis pinching. Although this type of hysteretic behavior is not
seen in the overall hysteresis response of the frame, it can increase the possibility of
failure of the connection in the fire exposed compartments.
Hence based on the observations from this study, it is obvious that the fire damaged steel
structures are vulnerable to earthquake loading, if the fire exposure temperature is over
800C and the structure is rehabilitated for continued service. The notion that there is no
metallurgical change when there is no distortion might not be applicable to moment resisting
frames with large beam and column sections. Therefore, it is essential that even if no
distortion is noticeable, the fire damaged structures should be investigated thoroughly to
make decision about the reparability of the structure based on the post-fire mechanical
properties of the structural steel.

5. References
[1] Klem TJ. First Interstate Bank Building Fire, Los Angeles, CA, May 4, 1988. NFPA Fire
Investigation Report, Quincy, MA, 1988.

[2] Klem TJ. One Meridian Plaza, Philadelphia, PA, three firefighter casualties, February 23,
1991. NFPA Fire Investigation Report, Quincy, MA, 1991.

[3] Newman GM. Structural fire engineering investigation of Broadgate Phase 8 fire. Steel
Construction Institute, Ascot., 1991

[4] Routley JG, Jennings C, Chubb M. High-Rise office building fire, One Meridian Plaza,
Philadelphia, Pennsylvania. Federal Emergency Management Agency (FEMA), 1991.

[5] Moncada, J.A. Fire unchecked. NFPA Journal 2005.

238
[6] Beitel J, Nestor I. Analysis of needs and existing capabilities for full-scale fire resistance
testing. US Department of Commerce, National Institute of Standards and Technology,
2008.

[7] NIST NCSTAR 1-3D. Mechanical properties of structural steel, federal building and fire
safety investigation of the of the work trade center disaster. National Institute of Standards
and Technology, Gaithersburg, MD, 2005.

[8] Niwa H. Investigation of Windsor building fire in Spain. Report of Obayashi Corporation
Technical Research Institute (CD-ROM) 70:11-2, 2006.

[9] Isner MS. High-rise office building fire, Alexis Nihon Plaza, Montreal Canada, October
26, 1986. NFPA Fire Investigation Report, Quincy, MA, 1986.

[10] Scawthorn C, Eidinger JM, Schiff AJ, eds. Fire following earthquake. ASCE
Publications, 2005.

[11] Scawthorn C. Fires following the Northridge and Kobe earthquakes. Thirteenth Meeting
of the UJNR Panel on Fire Research and Safety 2 1996;337-345.

[12] Tide RH. Integrity of structural steel after exposure to fire. AISC Engineering Journal
1998;35:26-38.

[13] Gosain N, Drexler F, Choudhuri D. Evaluation and repair of fire-damaged


buildings. Structure Magazine, 2008.

[14] Saul FW. Structural steel bounces back. AISC Steel Construction Digest 1956;13(3),3rd
Qtr., Chicago, IL.

[15] Hapgood KB. Metallurgical analysis of bridge samples. U.S. Department of


Transportation, Federal Railroad Administration, Washington, DC, 1979.

239
[16] Smith CI, Kirby BR, Lapwood DG, Cole KJ, Cunningham AP. The reinstatement of fire
damaged steel framed structures. British Steel Corporation, Teesside Laboratories, United
Kingdom, 1980.

[17] Hineman MA. Metallurgical testing of three beams subjected to a fire test at
underwriter's laboratories. Report No. 51610, Taussig Associates, Inc., Skokie, IL, 1983.

[18] Kirby BR. Evaluation of the temperature attained by five bolts during a fire in the
Broadgate Phase 8 developmentLondon. British Steel Corporation, Swinden
Laboratories, Moorgate, Rotherham, United Kingdom, 1991.

[19] ASTM, A. 6/A 6M. Standard specification for general requirements for rolled structural
steel bars, plates, shapes, and sheet piling. American Society for Testing and Materials,
West Conshohocken, PA, 2010.

[20] Dill FH. Structural steel after a fire. Proceedings of National Steel Construction
Conference, Denver, CO, American Institute of Steel Construction, Chicago, IL, 1960.

[21] Wright W, Cayes L. Route I-78 viaduct over Waverly Yards Bridge fire, material test of
damaged girders. Federal Highway Administration, Turner-Fairbank Highway Research
Center, McLean, VA, 1990.

[22] Kirby BR, Lapwood DG, Thomason G. The reinstatement of fire damaged steel and iron
framed structures. British Steel Corporation, Swinden Laboratories, Moorgate,
Rotherham, United Kingdom, 1993.

[23] Kirby B.R. Large scale fire tests: the British steel European collaborative research
programme on the BRE 8-storey frame. In Proceedings of the 5th International
Symposium on Fire Safety Science (IAFSS), 1129-1140, 1997.

240
[24] Li GQ, Guo SX. Experiment on restrained steel beams subjected to heating and
cooling. Journal of Constructional Steel Research 2008;64(3):268-274.

[25] Wald FA, Simes da Silva L, Moore DB, Lennon T, Chladn M, Santiago A, Borges L.
Experimental behavior of a steel structure under natural fire. Fire Safety Journal
2006;41(7):509-522.

[26] Smith CI, Kirby BR, Lapwood DG, Cole KJ, Cunninham AP, Preston RR. The
reinstatement of fire damaged steel framed structures. Fire Safety Journal 1981;4:21-62.

[27] Outinen J, Makelainen P. Mechanical properties of structural steel at elevated


temperatures and after cooling down. Fire and Materials 2004;28(2-4):237-251.

[28] Lee J, Engelhardt MD. Mechanical properties of ASTM A992 steel after fire.
Engineering Journal 2012;49(1):33-44.

[29] Morrison M. Innovative seismic performance enhancements of welded steel building


beam-column moment resisting connection. PhD Dissertation Research, North Carolina
State University, Raleigh, NC, 2014.

[30] SAC/BD-97/02. Protocol for fabrication, inspection, testing and documentation of


beam-column connection tests and other experimental specimens. by Clark P, Frank K,
Krawinkler H, Shaw R.

[31] Adan SM. Design parameters for steel special moment frame connections. Proceedings
2011 SEAOC convention, Las Vegas, Nevada, Structural Engineers Association of
California, 2011.

[32] AISC. Prequalified connections for special and intermediate steel moment frames for
seismic applications. American Institute of Steel Construction 358, 2010a.

241
[33] AISC. Seismic provisions for structural steel buildings. American Institute of Steel
Construction 341, 2010b.

[34] ASTM E119. Standard test method for fire tests of building construction and materials.
American Society for Testing and Materials, West Conshohocken, PA, 2012.

[35] ASTM E1529. Standard test method for determining effects of large hydrocarbon pool
fires on structural members and assemblies. American Society for Testing and Materials,
West Conshohocken, PA, 2013.

[36] Quiel SE, Garlock ME. Modeling high-rise steel framed buildings under fire.
In Proceedings of the 2008 ASCE/SEI structures congress, 2008.

[37] Hu G, Morovat MA, Lee J, Eschell E, Engelhardt M. Elevated temperature properties of


ASTM A992 steel. In: Proceedings of ASCE structure congress 09, Austin, TX, USA,
2009.

242
CHAPTER 8: CONCLUSIONS AND RECOMMENDATIONS

1. Conclusions
This dissertation dealt with experimental and finite element investigation of the localized
failure mechanism of welded steel moment connections as well as moment resisiting frames.
The objective of the study was to study the influence of several important parameters on the
seismic respose of the connections, address the extent of influence and wherever possible
propose modifications to improve the performance of the connections. The parameters that
were investigated in the course of the study are influence of end plate stiffeners and bolt
arrangement on the seismic performance of extended end plate connection, influence of
initial and weld induced residual stresses and weld sequence on the localized failure
mechanism of the WUF-W and WUF-B connections, and finally the effect of fire damage
(material strength degrdadation and heterogeneity) on the seismic performance of the
moment resisiting frames.
In Chapters 2 and 3, a modified design of extended end plate connection was porposed
and validated. In order to eliminate the stress concentration caused by the end plate stiffener
in AISC proqualifed eight bolt stiffened extended end plate connection, different
arrangements of bolts were studied by using finite element analysis and an optimized bolt
arranegement was obtained which led to equal distribution of bolt forces. However, this
introduced a different failure mechanism through higher strain demands on the weld. Hence,
the beam flanges were heat treated to predefine the plastic hinge location and if failure occurs
it will occur in the heat treated location of the beam flanges and the required ductility
capacity of the connection will be achieved. This again introduced another failure mechanism
through lateral or lateral torsional buckling of the beam due to reduced strength of the
material in the heat treated region. As a result, a stiffener was welded to the beam web near
the connection region. The modified connection showed significant improvement in ductility
and energy dissipation capacity, and the connection showed no strength degradation upto 5%
interstory drift angle. The modified connection design was validated by an experiment on full

243
scale beam-column connection and showed excellent performance under simulated seismic
loading history.
Chapters 4, 5 and 6 dealt with experimental and finite element study of the localized
failure mechanism of welded steel moment resisting connections by considering the initial
and weld induced residual stresses. In the experimental study, two WUF-B connections were
tetsted under fatigue loading in order to investigate the effect of weld sequence on the
performane of the connection. The results from the experimental investigation demonstrated
accumulation of axial strain near the weld toe and weld access hole, and different welding
procedure induced different amount of axial strain accumulation, which consequently led to
different fatigue life of the two specimens. In order to simulate the influnece of weld
sequence on the low-cycle fatigue failure mechanism of the connections, advanced numerical
techniques were developed for simulating the initial and weld induced residual stresses and
the simulation resposes were validated against responses from the experimental study and
literature.
The simulation responses showed that significant amount of residual stresses develop
close to the weld toe and weld access hole of the connection because of the welding
procedure, which gradually diminishes with increase in the distance from the weld toe. The
peak value of residual stresses can be higher than the yield stress of the base material and it
can be either tensile of compressive or a combination of both. Initial residual stress values are
negligible around the edges of the beam, and it increases with an increase in the distance
from the edge with a peak value at the centerline along the beam length. As a result, the
residual stresses close to the weld region are predominantly welding induced residual
stresses. The presence of residual stresses may lead to axial strain ratcheting near the weld
toe and weld access hole, and subsequently may lead to low-cycle fatigue failure of the
connection. Hence, the low-cycle fatigue failures that have been observed in the cyclic
testing of the post-Northridge connections can be attributed to the development of damage
accumulation near the weld toe and weld access hole as a result of the axial strain ratcheting.
Different weld sequences induce different distribution and magnitude of residual stresses
around the welded joint, and hence, the strain ratcheting would be different for different weld

244
sequences. Post-weld heat treatment can be used to reduce the magnitude of residual stresses
which is expected to increase the fatigue life of moment resisting connections.
Chapter 6 dealt with the seismic response of fire damaged steel buildings where the
influence of post-fire material properties and heterogeneity of structural steel on the seismic
performance of moment resisting frames was investigated. The results of the analysis
revealed that the strength degradation of fire exposed structural members is not very
influential to the overall strength of the structure, rather the lateral drift demands were
significantly influenced by the post-fire strength degradation and heterogeneity of structural
steel. It was observed that nearly all of the earthquake-induced lateral displacement occurs at
the story level where the fire exposed compartments are located. This may lead to single
story mechanism commonly known as soft story at the fire exposed story level which can
lead to catastrophic failure of the structure. In addition, it was observed that with increase in
the number of fire exposed compartments in a story, there is decrease in the lateral strength
and increase in the drift demands at that story level. Seismic analysis of an isolated WUF-W
moment connection revealed that after fire exposure, the temperature across the depth of a
beam is unevenly distributed which leads to material heterogeneity. As a result, the strength
of the connection under negative bending drops steeply compared to the strength under
positive bending which induces hysteresis pinching. Although this type of hysteretic behavior
is not seen in the overall hysteresis response of the frame, it can increase the possibility of
failure of the connection in the fire exposed compartments.

2. Recommendation for Future Works


Although the present research showed its effectiveness in studying the local failure
mechanism of welded steel moment connections, but it was not possible to address all the
factors that influence the localized failure. There are new ideas of research arising from this
dissertation research. Based on the the responses obtained from this study, following are
some of the recommendations for future research works.

245
2.1 Modified Extended End Plate Connection with Extended Shear Tab
In the seismic performance enhancement of AISC prequalified eight bolt stiffened
extended end plate connection, heat treatment was implemented at the flanges of the beam to
ensure formation of plastic hinge away from the connection region. This was based on the
weakening mechanism of the beam similar to what is used for reduced beam section moment
connection. Alternatively, the formation of plastic hinge in the beam flanges can be ensured
by strengthening the beam in the connection region using an extended shear tab. The
extended shear tab will delay the strength degradation by strengthening the connection
region, and also push the plastic hinge away from the welded joint. This type of the
configuration of the connection was not investigated in this study and there may be a
possibility that with the modified bolt arrangement and extended shear tab, the connection
can show similar response as observed for the modified connection proposed in this study.

2.2 Further Study on Seismic Analysis of WSMCs with Residual Stresses


In the seismic analysis of moment resisting connection with residual stresses, there are
several issues that need further investigation. In this study only two weld sequences were
considered for the analysis of WUF-W connection. Using the same procedure used in this
study, several weld sequences can be studied to get an optimized weld sequence that will
produce the least amount of residual stresses in the connection region. Moreover, in this
study the effect of the weld between the beam web and column flange was not considered.
The same analysis used in this study can be performed by including a fillet weld between the
beam web and column flange to investigate its influence on the overall residual stress
distribution. In addition, the presence of void or inclusion in the weld may affect the peak
value of the residual stress and its distribution. The weld residual stress analysis can be
performed with the inclusion or void in the weld in order to study its influence on the overall
residual stress distribution as well as on the low cycle fatigue response of the moment
resisting connections.
Although in the residual stress calculation, the effect of material heterogeneity based on
the peak temperature was included, however, in the fatigue analysis of moment resisting

246
connections, the heterogeneity of the material was not accounted for because of the
unavailability of cyclic response of ASTM A992 material at high temperatures. Further
investigation is required on the seismic analysis of moment resisting connections by
considering the heterogeneity in the material to study the influence of material softening on
the local strain responses of the connections. There were no experimental measurements of
residual stresses induced by welding at the weld region of the connection for which the
simulated residual stress response could not be validated. Similarly, no initial residual stress
measurements were available for wide flange members consisting of ASTM A992 steel.
These residual stress data is required from experimental measurements to validate the
simulated responses. Finally, the implemented residual stress simulation technique can be
further used to study the influence of residual stresses as well as welding procedure on the
local failure mechanism of other types of moment resisting connections such as reduced
beam section (RBS) connection, extended end plate connection.

2.3 Fire Damage Analysis with Water Quenched Structural Steel


In the seismic analysis of fire damaged moment resisting frames, it was assumed that the
fire exposed structural steel is cooled down slowly to room temperature as the fire slowly
extinguished over a long time. Further study is required to investigate the performance of the
fire damaged steel buildings under seismic loading history by considering water quenched
material properties of structural steel to consider quick extinguishing of fire with firefighter
sprayed water.

2.4 Simulation with Advanced Constitutive Models


Although all the simulation results presented in this study showed good conformity with
the experimental responses, for more complex analysis where creep, stress relaxation and
time dependent behavior is important, it is highly recommended that more advanced
constitutive model is used to get a better prediction of the responses. This type of advanced
constitutive model will require additional terms such as time sensitivity, influence of dwell
time, temperature dependence and finally, multiaxial parameters.

247

You might also like