You are on page 1of 279

Department of Civil and Environmental Engineering

Stanford University

LARGE-SCALE TESTING AND SIMULATION OF EARTHQUAKE INDUCED


ULTRA LOW CYCLE FATIGUE IN BRACING MEMBERS SUBJECTED TO CYCLIC
INELASTIC BUCKLING

By

Benjamin V. Fell
Amit M. Kanvinde
Gregory G. Deierlein

Report No. 172


August 2010
The John A. Blume Earthquake Engineering Center was established to promote
research and education in earthquake engineering. Through its activities our
understanding of earthquakes and their effects on mankinds facilities and structures is
improving. The Center conducts research, provides instruction, publishes reports and
articles, conducts seminar and conferences, and provides financial support for students.
Center is named for Dr. John A. Blume, a well-known consulting engineer and Stanford

Address:

The John A. Blume Earthquake Engineering Center


Department of Civil and Environmental Engineering
Stanford University
Stanford CA 94305-4020

(650) 723-4150
(650) 725-9755 (fax)
racquelh@stanford.edu
http://blume.stanford.edu

2010 The John A. Blume Earthquake Engineering Center


Copyright by Benjamin V Fell 2008

All Rights Reserved


Abstract

Special Concentrically Braced Frames (SCBFs) are popular lateral load resisting frames due to

their economy, structural efficiency and stiffness. Following the 1994 Northridge earthquake,

braced-frames became increasingly common after brittle fractures were observed at beam-column

connections in Moment Resistant Frames (MRFs). However, braced-frames are also susceptible

to fracture at the middle plastic hinge of the brace, the brace to gusset plate connection, and the

gusset plate to column or beam connection. This research primarily focuses on fracture at the

middle plastic hinge, where the combined effect of global and local buckling during cyclic

loading amplifies the plastic strain at the brace midpoint and initiates fracture. To develop a better

understanding of the localized mechanisms affecting brace fracture, this work combines a large-

scale experimental program with an intensive simulation study to investigate brace behavior

across a wide-range of material types and geometries. The simulations employ continuum-based

modeling techniques to accurately reproduce the stress and strain histories during cyclic loading

while a novel micromechanical fracture model is evaluated as a means to predict the fracture

initiation events. The fracture model operates at the continuum-level and captures the

fundamental mechanisms responsible for ductile fracture unique to Ultra Low Cycle Fatigue

(ULCF) conditions which differ from the well-defined High and Low Cycle Fatigue (HCF and

LCF) mechanisms. From the large-scale brace experiments, cross-section width-thickness and

slenderness ratios are shown to influence the brace axial deformation fracture ductility, such that

a larger width-thickness ratio and a smaller slenderness tend to reduce ductility. Furthermore, the

ii
experiments are used to evaluate the fracture model at the large-scale where small-scale

calibration tests and a multi-scale modeling procedure is used to connect the steel behavior at the

micromechanical level to the finite element simulation results. The fracture predictions are

encouraging considering the high level of complexity in modeling buckling phenomena and

imperfect constitutive model behavior. The model is used to expand the experimental test matrix

through parametric simulation of square and rectangular bracing components which, along with a

synthesis of experimental results over the last twenty years, informs a general relationship

between brace ductility and geometry.

iii
Acknowledgements

This research was supported by the National Science Foundation (NSF Grant CMS

0421492), the George E. Brown Jr. Network for Earthquake Engineering Simulation (NEES),

and the Structural Steel Educational Council (SSEC). The advice and guidance of Helmut

Krawinkler and Jack Baker (Stanford University), Stephen Mahin (University of California,

Berkeley), Charles Roeder (University of Washington), Robert Trembly (Polytechnique

Montral), Patxi Uriz (Exponent Failure Analysis Associates) Walterio Lpez and Mark

Saunders (Rutherford and Chekene) is greatly appreciated. Andy Myers (Stanford

University) has been instrumental to this research by conducting the small-scale calibration

experiments and dedicating his summer during 2005 to setup the large-scale brace tests at the

UC Berkeley Richmond Field Station. The knowledgeable support of the University of

California, Berkeley NEES lab personnel including Shakhzod Takhirov, Donald Patterson,

Donald Clyde, David MacLam, and Jose Robles is gratefully acknowledged. The authors also

acknowledge support from the John A. Blume Earthquake Engineering Center at Stanford

University and the University of California, Davis.

iv
Table of Contents

Abstract........................................................................................................................................... ii
Acknowledgements ....................................................................................................................... iv
Table of Contents ........................................................................................................................... v
List of Tables ................................................................................................................................. ix
List of Figures................................................................................................................................. x
Chapter 1 Introduction ............................................................................................................ 1
1.1 Motivation .................................................................................................................................. 1
1.2 Objectives and Scope ................................................................................................................. 3
1.3 Organization and Outline ........................................................................................................... 5
Chapter 2 Simulating Fracture in SCBF Bracing Components - Background ................ 10
2.1 Performance Evaluation of SCBF Systems ............................................................................. 12
2.1.1 SCBF System Behavior .................................................................................................... 12
2.1.2 Seismic Design Provisions for SCBF Systems................................................................. 15
2.1.3 State-of-the-art in SCBF Performance Assessment.......................................................... 17
2.1.3.1 Studies Examining Seismic Demands in SCBF Systems Through Nonlinear
Dynamic Analysis ................................................................................................................. 19
2.1.3.1.1 Sabelli (2001) Analyses ..................................................................................... 20
2.1.3.1.2 Uriz (2005) Analyses ......................................................................................... 21
2.1.3.1.3 McCormick et al (2007) Analyses ..................................................................... 23
2.1.3.1.4 Redwood et al (1991) Analyses ......................................................................... 23
2.1.3.1.5 Tremblay and Poncet (2005) Analyses .............................................................. 24
2.1.3.1.6 Izvernari et al (2007) Analyses .......................................................................... 25
2.1.3.2 Shake Table Testing .................................................................................................. 25
2.1.3.3 Summary of Demand Assessment of SCBF Systems ............................................... 26
2.1.3.4 Studies Investigating the Capacity of Bracing Members .......................................... 27
2.1.3.4.1 Square and Rectangular Hollow Steel Structural (HSS) Bracing Components . 30

v
2.1.3.4.2 Round Steel Pipe Bracing Members .................................................................. 32
2.1.3.4.3 Wide-Flanged Bracing Components .................................................................. 33
2.1.3.4.4 General Ductility Trends of HSS, Pipe, and Wide-Flanged Steel Braces.......... 34
2.1.3.4.5 Full-System Braced Frame Tests ....................................................................... 38
2.2 Modeling Techniques for SCBF Systems ................................................................................ 39
2.2.1 Demand Characterization ................................................................................................. 40
2.2.1.1 Techniques for Brace Simulation .............................................................................. 41
2.2.1.1.1 Rule-Based (or Phenomenological) Brace Models ............................................ 41
2.2.1.1.2 Concentrated Hinge Brace Models .................................................................... 42
2.2.1.1.3 Fiber-Element Based Brace Models .................................................................. 42
2.2.1.1.4 Continuum-Based Brace Models ....................................................................... 43
2.2.2 Capacity Characterization ................................................................................................ 45
2.2.2.1 Brace Geometry-Based Fracture Capacity Relationships ......................................... 45
2.2.2.2 Fiber Element-Based Critical Strain Models ............................................................. 50
2.2.2.3 Micromechanical Void Growth and Coalescence Models ........................................ 51
2.3 Summary .................................................................................................................................. 53
Chapter 3 Large-Scale Brace Component Tests ................................................................. 71
3.1 Motivation for Large-Scale Brace Tests .................................................................................. 71
3.2 Experimental Setup .................................................................................................................. 73
3.3 Test Program Scope ................................................................................................................. 73
3.3.1 Cyclic Loading Protocols ................................................................................................. 76
3.3.1.1 Standard Cyclic Loading Protocol ............................................................................ 77
3.3.1.2 Near-Field Loading Histories .................................................................................... 79
3.3.1.3 Story Drift and Brace Axial Deformation Relationships .......................................... 80
3.4 Instrumentation and Miscellaneous Testing Results ................................................................ 81
3.5 Summary .................................................................................................................................. 84
Chapter 4 Large-Scale Experimental Results and Design Implications ........................... 98
4.1 Qualitative Summary of Experimental Response .................................................................... 98
4.2 Quantitative Summary of Data For All Tests ........................................................................ 100
4.3 Effect of Test Variables on Cyclic Brace Behavior, Limit States and Design Implications.. 104
4.3.1 Effect of Width to Thickness Ratios............................................................................... 104
4.3.2 Effect of Member Slenderness ....................................................................................... 106
4.3.3 Effect of Cross Sectional Shape ..................................................................................... 107
4.3.4 Effect of Grout Filling of HSS Specimens ..................................................................... 107

vi
4.3.5 Effect of Loading Rate ................................................................................................... 108
4.3.6 Effect of Unsymmetrical Buckling ................................................................................. 110
4.4 Brace-Gusset Plate Connection Performance ........................................................................ 110
4.4.1 Experimental Results ...................................................................................................... 112
4.5 Comparison of Experimental Data to Commonly Used Formulae for Predicting Strength and
Stiffness of Bracing Members ..................................................................................................... 116
4.5.1 Elastic Stiffness .............................................................................................................. 116
4.5.2 Compressive Strengths ................................................................................................... 117
4.5.3 Maximum Tensile Strength ............................................................................................ 118
4.6 Summary ................................................................................................................................ 119
Chapter 5 Application and Evaluation of the CVGM for Large-Scale Components .... 136
5.1 Development of the Cyclic Void Growth Model (CVGM) ................................................... 139
5.1.1 Material Behavior, Constitutive Models and Calibration ............................................... 143
5.1.2 CVGM Calibration ......................................................................................................... 146
5.1.2.1 Calibration of ....................................................................................................... 147
5.1.2.2 Calibration of ....................................................................................................... 147
5.1.3 CVGM Application and Validation in Small-Scale Details ........................................... 148
5.2 Application of CVGM to Large-Scale Bracing Components ................................................ 149
5.2.1 Material Model Calibration ............................................................................................ 150
5.2.2 Large-Scale Brace Simulations ...................................................................................... 153
5.2.2.1 Modeling Global and Local Buckling ..................................................................... 153
5.2.2.1.1 Imperfections and Buckling Simulations ......................................................... 154
5.2.2.1.2 Comparison to Experimental Global Buckling ................................................ 155
5.2.2.1.3 Comparison to Experimental Local Buckling .................................................. 155
5.2.3 Brace Material CVGM Calibraiton ................................................................................ 156
5.2.3.1 Brace Material Calibration Results: ..................................................................... 156
5.2.3.2 Brace Material Calibration Results: ..................................................................... 157
5.2.4 CVGM Fracture Predictions Applied to Large-Scale Brace Components ..................... 162
5.2.4.1 Application to HSS and Pipe Cross-Sections (Standard Loading) .......................... 163
5.2.4.2 Application to Near-Field Loading Histories .......................................................... 166
5.2.4.3 Quasi-Static Versus Earthquake Loading Rates ...................................................... 167
5.2.4.4 Application to Unsymmetrical Buckling ................................................................. 168
5.2.4.5 Estimating Uncertainty in the CVGM Fracture Predictions.................................... 169
5.3 Summary ................................................................................................................................ 171

vii
Chapter 6 Parametric Simulation of HSS Bracing Members .......................................... 203
6.1 Parametric Simulation Matrix ................................................................................................ 204
6.1.1 Cyclic Loading History .................................................................................................. 205
6.2 CVGM Fracture Predictions .................................................................................................. 205
6.2.1 Effect of Width-Thickness Ratio on Brace Ductility ..................................................... 206
6.2.2 Effect of Slenderness Ratio on Brace Ductility .............................................................. 206
6.3 Synthesis of Experimental and Simulation Results ............................................................... 208
6.3.1 A Simplified Approach for Evaluating the Effect of Brace Parameters on Ductility..... 210
6.3.1.1 Plastic Hinge Length ............................................................................................... 213
6.3.1.1.1 Comparison to Fiber-based Brace Simulations ................................................ 216
6.3.2 Evaluating Brace Axial Deformation Capacities Relative to Earthquake Induced
Interstory Drift Demands ......................................................................................................... 216
6.4 Summary and Reliability of Parametric Study and Semi-Theoretical Model ........................ 222
Chapter 7 Summary, Conclusions and Future Work ....................................................... 240
7.1 Summary and Conclusions .................................................................................................... 241
7.1.1 Large Scale Brace Experiments...................................................................................... 241
7.1.2 CVGM Evaluation in Large Scale Details ..................................................................... 243
7.1.3 Parametric Simulation of Bracing Members .................................................................. 245
7.2 Future Work ........................................................................................................................... 246
References ................................................................................................................................... 249
Appendix A ................................................................................................................................. 256
Appendix B ................................................................................................................................. 263

viii
List of Tables

Table 2.1: Maximum interstory drifts (standard deviation) from nonlinear time-history analyses
on 3 and 6-story SCBF systems. ........................................................................................... 56
Table 2.2: Design level (approximately 10% in 50 years) median (standard deviation) and
maximum interstory drift from Izvernari et al (2007) ........................................................... 56
Table 2.3: Results from 6-story braced-frame shake-table test (Tang, 1987) ................................ 56
Table 2.4: Summary of HSS experimental review (63 tests) ......................................................... 57
Table 2.5: Summary of calibration constants to predict cyclic fracture life of square and
rectangular HSS bracing members ........................................................................................ 57
Table 3.1: Test parameters and loading histories ........................................................................... 85
Table 3.2: Brace material properties .............................................................................................. 85
Table 3.3: Summary of loading protocol modifications ................................................................ 86
Table 4.1: Data for standard (far-field) loading protocol tests..................................................... 123
Table 4.2: Data for near-field (NF) pulse loading protocol tests ................................................. 124
Table 4.3: Experimental results of bracing connections .............................................................. 125
Table 4.4: Comparison of brace strength and stiffness ................................................................ 125
Table 5.1: Measured material properties...................................................................................... 174
Table 5.2: Calibrated kinematic and isotropic hardening law parameters. .................................. 174
Table 5.3: Maximum measured local imperfections. ................................................................... 174
Table 5.4: Calibrated CVGM fracture parameters ....................................................................... 175
Table 5.5: CVGM fracture prediction summary .......................................................................... 176
Table 6.1: HSS simulation matrix (22 total) ................................................................................ 226
Table 6.2: Brace member lengths across various frame geometries ............................................ 226
Table 6.3: Summary of HSS experimental review (63 tests) ....................................................... 226
Table 6.4: Regression parameter values ...................................................................................... 226
Table B.1: Experimental results from square and rectangular HSS tests .................................... 263

ix
List of Figures

Figure 1.1: Outline of experimental and simulation study............................................................... 9


Figure 2.1: Large-scale Special Concentrically Braced Frame test (Uriz, 2005) and local buckling
induced brace fracture at the middle plastic hinge ................................................................ 58
Figure 2.2: (a) Chevron braced-frame story and typical connection details, (b) Out-of-plane
buckling and tension yielding and (c) Brace plastic hinge formation. .................................. 59
Figure 2.3: (a) Brace gusset-plate net section fracture and (b) Column fracture at base of shear tab
along with beam fracture from prying action of gusset plate. ............................................... 60
Figure 2.4: Plastic hinge formation in beam from bracing force imbalance. ................................. 61
Figure 2.5: Influence of brace geometry in terms of global slenderness and cross-section
compactness on fracture ductility. ...................................................................................... 62
Figure 2.6: Interstory drift time history results for the first story of a 3-story SCBF, 2% in 50
years event. ............................................................................................................................ 63
Figure 2.7: Minimum versus maximum interstory drifts for (a) 3-story and (b) 6-story SCBF
(courtesy Uriz, 2005)............................................................................................................. 64
Figure 2.8: (a) Schematic, one cycle, force-deformation response of a typical brace component
and (b-e) Progression of brace damage ................................................................................. 65
Figure 2.9: (a) Typical brace hysteretic response and definition of range, (b) Influence of Width-
thickness and (c) Global slenderness ratio on axial deformation range, range/Y. ................ 66
Figure 2.10: Brace modeling techniques........................................................................................ 67
Figure 2.11: Buckled shape, plastic strain contours, and critical fracture node from continuum
HSS4x4x1/4 brace analysis. .................................................................................................. 68
Figure 2.12: Ratio of experimental to predicted fracture deformation for three separate testing
programs. ............................................................................................................................... 68
Figure 2.13: Micromechanical process of ductile fracture in steel ................................................ 69

x
Figure 2.14: (a and b) CVGM fracture prediction and (c) Comparison to experimental fracture
time for the critical node of an HSS4x4x1/4 shown in Figure 2.11 during a standard loading
history. ................................................................................................................................... 70
Figure 3.1: (a) Plan and (b) elevation view of brace test setup. ..................................................... 87
Figure 3.2: Brace drawings for (a) HSS, (b) Pipe and (c) W12x16. .............................................. 88
Figure 3.3: Connection details for (a) HSS and Pipe and (b-c) W12x16. ...................................... 89
Figure 3.4: (a) Standard cyclic (far-field ground motions), (b) Compression and (c) Tension near-
fault pulse loading histories................................................................................................... 90
Figure 3.5: (a) Original and modified SAC loading history and (b) Cumulative plastic drift for
Chevron braced-frame and moment frame under prescribed loading histories..................... 91
Figure 3.6: One-story Chevron braced-frame. ............................................................................... 92
Figure 3.7: Axial transducer measurements for brace end plates. ................................................. 93
Figure 3.8: Intended and measured axial deformations. ................................................................ 94
Figure 3.9: Actuator force measurements versus relative axial deformation for standard cyclic
HSS4x4x1/4 test. ................................................................................................................... 94
Figure 3.10: (a) Wire-pot instrumentation plan, (b) Connection detail and (c) Pipe3STD test. .... 95
Figure 3.11: Measured out-of-plane buckling displacements. ....................................................... 96
Figure 3.12: Thermocouple recordings for (a) Quasi-static, near-field test and (b) Earthquake-rate
standard cyclic test. ............................................................................................................... 97
Figure 4.1: Typical progression of brace specimen damage. ....................................................... 126
Figure 4.2: Typical brace response for (a) Far-field loading, (b) Near-fault compression and (c)
Near-fault tension. ............................................................................................................... 127
Figure 4.3: Energy dissipated prior to (a) Fracture initiation versus local buckling and (b)
Strength loss versus fracture initiation. ............................................................................... 128
Figure 4.4: Effect of width-thickness ratio on (a) Maximum drift at fracture initiation and (b)
Normalized dissipated energy. ............................................................................................ 129
Figure 4.5: Effect of slenderness ratio on (a) Maximum drift at fracture initiation and (b)
Normalized dissipated energy. ............................................................................................ 130
Figure 4.6: Local buckling shapes. .............................................................................................. 131
Figure 4.7: Symmetric and unsymmetrical buckling. .................................................................. 132
Figure 4.8: Net section details and fracture ................................................................................. 133
Figure 4.9: (a) Experimental connection gage length and (b) Load deformation response of
Pipe3STD and Pipe5STD tests with fracture at the net section. ......................................... 134
Figure 4.10: Maximum experimental forces compared to expected capacities. .......................... 135

xi
Figure 5.1: Micromechanical process of ductile fracture in steel ................................................ 177
Figure 5.2: Scanning Electron Microscope (SEM) pictures of (a) Monotonic and (b) Cyclic
fracture surfaces (A572, Grade 50). .................................................................................... 178
Figure 5.3: (a) Uniaxial stress-strain behavior and (b) Isotropic yield surface. ........................... 179
Figure 5.4: Combined isotropic-kinematic yield surface employed in ABAQUS (2004). .......... 180
Figure 5.5: Ideal material calibration flow chart.......................................................................... 181
Figure 5.6: (a) Notched-bar geometry and (b) Typical force-deformation response. .................. 182
Figure 5.7: (a and b) Axisymmetric finite element model of notched-bar specimen and (c) Void
growth demand according to VGM. .................................................................................... 183
Figure 5.8: Exponential relationship between cyclic/ and damage............................................. 184
Figure 5.9: CVGM fracture prediction for a small-scale notched bar specimen. ........................ 184
Figure 5.10: Predicted versus experimental instance of fracture initiation for small-scale, notched-
bar cyclic coupon tests (adapted from Kanvinde and Deierlein, 2007)............................... 185
Figure 5.11: Summary of small-scale specimens for brace material calibration study................ 186
Figure 5.12: Experimental and simulation force-deformation comparisons for small-scale material
calibration tests. ................................................................................................................... 187
Figure 5.13: Experimental and simulation force-deformation comparison for large-scale
HSS4x4x1/4 brace (far-field loading). ................................................................................ 188
Figure 5.14: Brace models and meshing schemes. ...................................................................... 189
Figure 5.15: Measured brace cross-section dimensions............................................................... 190
Figure 5.16: (a) Global and (b) Local buckling mode shapes. ..................................................... 191
Figure 5.17: Cross-section wall imperfection measurements for HSS and Pipe. ........................ 191
Figure 5.18: Experimental and predicted critical buckling loads. ............................................... 192
Figure 5.19: Experimental and predicted local buckling for HSS4x4x1/4 brace (far-field loading).
............................................................................................................................................. 193
Figure 5.20: Triaxiality versus equivalent plastic strain (HSS4x4x3/8 material). ....................... 194
Figure 5.21: Calibration of NB and C......................................................................................... 195
Figure 5.22: Constant probability distribution of . ..................................................................... 196
Figure 5.23: Effect of low damage levels on variability. .......................................................... 196
Figure 5.24: CVGM fracture predictions for HSS4x4x1/4 where NB and C are used in
calculating two damage functions. ...................................................................................... 197
Figure 5.25: Comparison between experimental and predicted fracture instances (far-field
loading). .............................................................................................................................. 198

xii
Figure 5.26: Comparison between experimental and predicted fracture instances (near-field
loading). .............................................................................................................................. 199
Figure 5.27: Comparison between quasi-static and earthquake-rate loading tests. ..................... 200
Figure 5.28: Comparison between experimental and predicted fracture instances for HSS4x4x1/4
brace subjected to far-field loading with middle reinforcing plates. ................................... 201
Figure 5.29: (a) Deterministic CVGM fracture instances and (b) Cumulative probability function
for HSS4x4x1/4 subjected to far-field loading history........................................................ 202
Figure 6.1: HSS brace geometry .................................................................................................. 227
Figure 6.2: Brace loading history................................................................................................. 227
Figure 6.3: Influence of width-thickness ratio on brace ductility (fracture initiation) in terms of (a)
Axial deformation range and (b) Story drift.. ...................................................................... 228
Figure 6.4: Influence of slenderness ratio on brace ductility (fracture initiation) in terms of (a)
Axial deformation range and (b) Story drift. ....................................................................... 229
Figure 6.5: Slenderness range for minimum and maximum frame sizes (see Table 6.X) for (a) All
HSS cross-sections and (b) AISC (2005) conforming (i.e., b/t < 16) sections; (c) Influence
of brace slenderness on story drift ductility for fixed b/t (14.2) ratio and brace lengths (119
and 180 in). Also shown are past experimental (b/t=14.2 and L=119 in) results. ............... 230
Figure 6.6: Experimental brace ductility in terms of drift versus brace parameter for similar (a)
Slenderness (19 tests) and (b) Compactness (25 tests) ratios. ............................................. 231
Figure 6.7: Schematic illustration of simplified approach to evaluate effect of brace parameters on
ductility. .............................................................................................................................. 232
Figure 6.8: (a) Fiber-like fracture strain and (b) Deformation range capacity versus b/t ratio for all
experiments and simulations listed in Tables 6.1 and 6.3. .................................................. 233
Figure 6.9: Comparison between maximum experimental and predicted deformation range. .... 234
Figure 6.10: Schematic illustration of plastic hinge length calculation. ...................................... 235
Figure 6.11: Plastic hinge length as a function of increasing drift............................................... 236
Figure 6.12: Continuum finite element simulation of brace specimen showing plastic hinge length
dimension. ........................................................................................................................... 237
Figure 6.13: (a and b) Force-deformation and (c) Plastic hinge length comparison between finite
element and fiber brace models. .......................................................................................... 238
Figure 6.14: Maximum drift capacity versus width-thickness ratio. ........................................... 239
Figure 6.15: Deformation capacity (in terms of drift), divided by 1.4, versus normalized width-
thickness ratio. ..................................................................................................................... 239
Figure A.1: Experimental loading histories and brace hysteretic response. ................................ 256

xiii
1

Chapter 1
Introduction

1.1 MOTIVATION

Structural investigations following the 1994 Northridge earthquake revealed that the combination

of high fracture toughness demands caused by poor detailing of beam-column connections and

low material toughness resulted in widespread fractures in these structural details. Since

Northridge, Special Concentrically Braced Frames (SCBFs) have gained considerable popularity

as a lateral load resisting system in high seismic areas. However, braced frames are also

vulnerable to premature fracture during earthquakes due to the cyclic inelastic buckling of the

bracing elements and the resulting large force and deformation demands on the bracing

connections. In fact, recent studies (Uriz, 2005) have suggested unsatisfactory performance of

bracing systems designed with current codes, leading to fracture in the braces during design-level

earthquakes. However, research regarding SCBFs is relatively less exhaustive when compared to

that regarding moment frame systems. In addition, the mechanisms of earthquake-induced Ultra

Low Cycle Fatigue (ULCF) that could initiate fracture in steel bracing components have only

recently been explained as part of a physics-based approach to predict fracture in small-scale tests

(Kanvinde and Deierlein, 2004 and 2007). Considering the recent concern with fracture in SCBF

systems and a novel physics-based modeling approach to describe ULCF, this study intends to fill

part of the knowledge gap pertaining to brace component behavior, evaluate the ULCF fracture
2

models at a large-scale and present a simulation-based methodology that can be used to

complement experimental programs.

Currently, the tools used by structural engineering researchers to predict fracture are not as

sophisticated as other aspects of structural analysis and common fracture prediction

methodologies are often based on varying degrees of empiricism rather than fundamental

mechanics. Simplistic approaches for fracture prediction in SCBFs involve using the story drift as

an indicator of fracture, while somewhat more advanced approaches may use either critical

longitudinal strain measures, or cycle counting and fatigue life approaches for individual braces

or components (Tang and Goel, 1989). Recent studies (Uriz, 2005) have applied cycle counting

techniques through fiber-based elements to simulate fracture strain demands at a cross section,

instead of along the entire brace. While these represent important advances in the fatigue-fracture

prediction methodology, they fail to directly incorporate the effects of local buckling and the

complex interactions of stress and strain histories that trigger crack initiation in these details. In

part, this is due to the erstwhile lack of computational resources required to model phenomena

such as local buckling and to characterize the stress and strain fields at the location of fracture.

More importantly, however, there has been a lack of suitable stress/strain based fracture

prediction criteria.

In addition, simulation techniques that accurately describe the complex stress and strain states

from global and local buckling events have not been rigorously developed, especially for large-

scale components and cyclic loading histories. With recent advances in computational power,

three dimensional continuum-based modeling, such as the bracing models illustrated in Figure

1.1, has become an attractive option to model local buckling phenomenon. Continuum-based

modeling presents significant advantages over concentrated hinge or fiber-based formulations as

the buckling induced stress and strain histories that trigger fracture during earthquake loading can
3

be directly modeled. These simulations can be utilized to perform parametric studies, gain

insights into brace performance, and inform general models that characterize member ductility

through geometric properties.

1.2 OBJECTIVES AND SCOPE

Micromechanical based models which act at the continuum level have shown promise in

simulating ductile fracture initiation during cyclic loading for specimens that are relatively small

in scale (on the order of several inches) and free of complex modeling conditions (i.e., buckling)

(Kanvinde, 2004). While the mechanisms of ductile fracture mechanisms in low-carbon steel

under monotonic loading conditions are well established (Rice and Tracy, 1969), the structural

engineering community has not had suitable criteria for assessing fracture demands and capacities

during earthquake-type loading until only recently. In large part, the lack of physics-based

fatigue/fracture models for earthquake-type loading has not been extensively addressed by the

scientific community due to the unique loading conditions of the very large, yet low in number,

earthquake cycles that push a structure well into the inelastic range of material response.

Moreover, large-scale structural analysis methods with the capability of simulating complex

geometric and material nonlinearities have been impractical to employ until recent advances in

computing power. As the earthquake engineering community moves towards a more

performance-based framework of design and analysis, there is a need to develop simulation tools

which can be applied to large-scale details and general loading histories to accurately assess

ductility of steel components. Motivated by this need, the general aim of this study is to reconcile

the modeling gap that exists between current large-scale structural analysis methods that are void

of fundamental fracture prediction capabilities and the validated cyclic fracture initiation

mechanisms at the micromechanical scale. More specifically, the research addresses the

following topics
4

Experimental performance of steel bracing members subjected to earthquake-type

loading conditions. The large-scale tests include square, round and wide-flanged shapes

and investigate the effects of different brace geometries and material types on brace

ductility.

The evaluation of a general fracture prediction methodology at the large-scale which can

be applied across diverse loading situations and material types.

The application of this methodology to complement the large-scale experimental

program. In the context of this dissertation, the performance of square and rectangular

tube members is investigated through parametric simulations which employ these

advanced simulation techniques.

These objectives are investigated through various experimental, simulation and modeling

components. These components are illustrated schematically in Figure 1.1 and include

Previous experimental results on bracing members and other analytical investigations

aimed at simulating brace behavior and quantifying fracture ductility (not shown in the

figure).

Nineteen large-scale tests representative of bracing members in SCBF systems. These

experiments supply a test-bed to evaluate the micromechanical fracture initiation models

as well as provide direct performance-based comparisons for the practicing design

community.

A series of small-scale material calibration tests on coupons extracted from the large-

scale bracing members. These experiments provide the connection between the fracture

mechanisms at the micromechanical scale and the fracture predictions at the continuum

scale in the brace simulations. Accurate and consistent calibration of the multi-axial
5

plasticity model is an important aspect of linking the fracture toughness of the material at

the small-scale to large-scale behavior.

Continuum simulations of the large-scale brace tests that accurately capture the complex

stress and strain state induced by global and local buckling during cyclic loading. These

simulations provide insight into localized behavior (i.e., brace cross-section performance)

as well as global performance data (i.e., story drift at local buckling or fracture initiation).

A suite of parametric studies on bracing members performed with continuum models and

a set of complementary analyses modeling the braces with fiber elements. The

investigation shows the importance of capturing localized behavior such as cross-section

buckling and describes general models which are informed through the simulation

matrixes.

While the evaluation of the cyclic micromechanical fracture models at the large-scale is an

important scientific aspect of this study, the relationships and conclusions developed from the

parametric study present the greatest opportunity for a more immediate impact on the structural

engineering community. These general models are based on fundamental mechanics and are

informed through results from extensive parametric studies and physics-based fracture models.

By relating brace properties and deformation demands to fiber-type strain measurements, the

models present designers and code writers with a tool to make more informed decisions regarding

steel brace behavior.

1.3 ORGANIZATION AND OUTLINE

Chapter 2 provides a review of large-scale SCBF experimental testing and analytical modeling

results along with a background section looking at the state-of-the-art modeling techniques for

brace behavior and fracture prediction. First, the chapter presents results from past analytical and

experimental studies to quantify earthquake demands on braced-frame components and/or


6

systems. Next, experimental trends in brace component ductility are examined to illustrate the

dependence of ductility on brace global slenderness and cross-section compactness parameters.

After these two sections, the chapter presents the state-of-the-art in modeling techniques that have

been applied to bracing components. The models range from phenomenological models which

describe the force-deformation relationship of an axial strut to sophisticated continuum-based

models which can model global and local buckling events. Following the discussion on brace

modeling techniques, fracture prediction models are introduced that have been specifically

developed for, and applied to, inelastic buckling braces. Finally, a general, physics-based, ULCF-

fracture model that relies on stress and strain quantities at a continuum point is briefly introduced

with an example of a fracture prediction for a bracing member.

Chapter 3 is a relatively brief chapter that details the experimental setup of the nineteen large-

scale brace component tests mentioned previously. The chapter reviews the design of the

specimens, test setup, general or special instrumentation for each test and the development of the

loading histories. The primary focus of the chapter is to introduce the test matrix and objectives

of the experimental program.

Chapter 4 presents results from the nineteen large-scale tests of steel bracing members to

examine their inelastic buckling and fracture behavior as related to the seismic design of

concentrically braced frames. The brace specimens include square Hollow Structural Shapes

(HSS), Pipe and Wide-Flange sections. The effect of various parameters, including width-

thickness and slenderness ratios, cross-section shape, loading history, loading rate and grout fill

on the performance of these braces is investigated. The test data is investigated with respect to

current seismic design limits on maximum width-thickness and slenderness ratios. Also,

measurements of brace stiffness, tensile strength and compressive strength are compared with

design formulae. Future analytical studies to simulate brace buckling and fracture are outlined as
7

a way to generalize the findings of the physical tests. The experimental results are also presented

in the context of net-section limit states in brace-gusset plate connections. Finally, maximum

tensile force measurements of the steel bracing members are presented and compared to code-

based provisions that prescribe the expected yield strength (RyFyAg) of the brace for design of the

bracing connections.

Chapter 5 describes a methodology used to predict fracture in large-scale braced frame

components. The micromechanical-based ULCF fracture model (Kanvinde and Deierlein, 2007)

introduced in Chapter 2 is used to predict fracture in the brace component tests of Chapters 3

and 4. The physics-based fracture model has been shown to accurately describe the fundamental

mechanisms of void growth, collapse and coalescence in small-scale experiments. This Chapter

evaluates the performance of the physics-based fracture model across the various cross-section

shapes, material types and loading histories of the experimental program presented in Chapter 4.

The methodology demonstrates the importance of modeling the initiation, both the instance and

location, of localized cross-section buckling phenomena. Simulating local buckling is shown to

be linked to cross-section and brace properties (such as width-thickness and global slenderness

ratios), but also to material hardening parameters that are calibrated from small-scale tensile and

cyclic coupon tests. Effects of loading rate are investigated by comparing fracture predictions

from quasi-static loading tests to identical bracing members under fast, earthquake-rate tests.

Chapter 6 utilizes the fracture prediction methodology presented in Chapter 5 to conduct a

parametric study on representative bracing components of SCBF systems. By applying the

micromechanics-based fracture models, the study aims to generalize trends relating to brace

fracture performance. The focus is on Hollow Steel Structural (HSS) members that have recently

been observed to fracture prematurely during experiments. The simulations provide insights into

the failure mechanisms of inelastic buckling braces in the middle plate hinge. Moreover, the finite
8

element analyses provide a means to calculate the plastic hinge length after local buckling

initiation. Results from continuum-based modeling of the HSS braces members are compared to

those from more conventional fiber-based simulations using the OpenSEES platform. Informed

by the findings from past experimental programs and the parametric study, a semi-theoretical

model is proposed to calculate brace deformation capacity, and the associated story drift capacity,

as a function of brace geometry. The past experimental programs provide data from 63 previous

cyclic tests on square and rectangular HSS bracing members. The data is obtained from 10 testing

programs in the USA and Canada, and includes a wide range of brace parameters (slenderness

ratios ranging from 31 to 145, b/t ratios between 8.5 and 31.5).

Chapter 7 summarizes the dissertation by highlighting key observations and conclusions of the

large-scale experimental program as well as the complementary brace simulations and parametric

studies. While the research investigates braced-frame behavior from several different aspects,

from large-scale experimental results to fracture modeling to design provisions, this dissertation

does not seek to provide a final answer related to any one of these topics. Rather, the last

chapter emphasizes that the current work 1) Evaluates the use of a sophisticated modeling tool

that can fit into a more general, performance-based methodology, 2) Synthesizes experimental

and modeling results (both from this study and others) to fill parts of the knowledge gap related to

SCBF performance and 3) Highlights possible areas to be examined in future studies.


9

New insights into local


cross-section behavior
(Chapter 4 and 5; future
work)

Model development
(Kanvinde 2004) and
material calibration
(Chapter 5)

Parametric studies
(Chapter 6):

Physics-based
fracture model
evaluation
(Chapter 5)

Design considerations
(Chapter 4 and 6)

Large-scale brace experiments


(Chapters 3 and 4)

Figure 1.1: Outline of experimental and simulation study.


10

Chapter 2
Simulating Fracture in SCBF Bracing Components - Background

Several experimental and analytical investigations (Foutch et al 1987, Uriz 2005) and post-

earthquake reconnaissance reports (Tremblay et al 1995, Kelly et al 2000) have suggested that

Special Concentrically Braced Frames (SCBFs) may show unsatisfactory performance during

earthquake-type loading. Figure 2.1 illustrates a large-scale SCBF experiment by Uriz (2005) and

component tests conducted as part of the current study which both indicated local buckling-

induced fracture of the bracing members during inelastic buckling and tension yielding. In

addition to the brace fractures shown in the figure, fracture of brace-gusset plate connections

from large force and deformation demands on the connections is also a concern. Since SCBFs

rely on the buckling and yielding action of the brace components to dissipate seismic energy,

sufficient brace ductility must be provided to preclude premature fracture. Assessing the fracture

ductility of large-scale bracing components or systems is expensive, especially considering the

heavy dependence on experiment-based techniques. Moreover, testing of large-scale components

may be infeasible due to the size and strength of structural members. Considering these issues,

reliable fracture simulation and prediction techniques for SCBF systems are highly attractive and

allow for improved practicality in modeling the effects of diverse material properties and loading

conditions at the component, as well as system, level.


11

In this context, this chapter presents a broad overview of various experimental and state-of-the-art

simulation-based investigations that aim to characterize the performance (specifically the fracture

performance) of SCBF systems. Seismic performance assessment of structural systems typically

involves two distinct methodological components (1) Evaluating deformation or strain demands

at the story, component or continuum level (2) Determining the corresponding capacities through

a combination of experimental and analytical methods. A comparision between the two results in

characterization of structural performance. Although demands and capacities are abstract

concepts that interactively affect each other in reality, they form a convenient framwork for

characterizing structural performance within the limitations of current simulation techniques.

Thus, the background and literature review presented in this chapter is broadly divided into two

main parts one focusing on demand characterization, and the other focusing on capacities.

Throughout the chapter the literature is reviewed and synthesized within the context of a

Performance Based Earthquake Engineering (PBEE) methodology.

The first section of the chapter is devoted to a literature review on the performance assessment of

SCBF systems, either through experimental or analytical investigations. These provide

background for subsequent chapters. Next, various modeling techniques are examined to describe

the state-of-the art in structural analysis and fatigue-fracture prediction. With respect to the latter,

these fall into various broad categories including (1) empirical equations to predict fatigue life as

a function of the geometrical parameters of bracing elements (2) fiber based approaches that

employ a critical strain as an indicator of fracture and (3) and micromechanics-based models that

operate at the continuum level and simulate the processes of void growth and coalescense that are

responsible for ductile fracture initiation.

While the major objective of this study is to examine these continuum-based fracture models for

large scale specimens, the chapter also presents work that has advanced the state-of-the art in
12

analyis techniques which have been used on braced-frame components. These are broadly

classified as (1) phenomenological or rule-based models (2) lumped plasticity analysis techniques

(3) fiber-based elements and (4) continuum-based or finite element analyses.

The chapter concludes by summarizing the current state of research, and identifying areas where

advanced simulation and fracture models may benefit structural and earthquake research and

practice.

2.1 PERFORMANCE EVALUATION OF SCBF SYSTEMS

To provide context for the subsequent literature review, this section discusses braced-frame

systems by first describing the desired response and possible fracture events during earthquake

loading, and then reviewing modern design provisions for SCBFs. Afterwards, the literature

review is presented by first focusing on braced-frame demand evaluation and then on the ductility

capacity of these systems.

2.1.1 SCBF SYSTEM BEHAVIOR

A schematic illustration of a one story chevron (inverted-V) braced-frame is shown in Figure 2.2a

where the hollow circles indicate locations of assumed pinned connections. Shear-tabs are used in

the beam-column connections and gusset plates (flexible out of plane) connect bracing members

to beams and columns. While these are more properly classified as partially fixed connections,

they are typically considered to be pinned in design and analysis procedures, although recent

research (e.g. Uriz, 2005) has illustrated that these may be substantially more rigid than often

presumed. As illustrated in Figure 2.2b, frame deformation is accommodated by the rotation of

these connections and axial deformation of the bracing members. Thus, unlike moment frames

where the beam-column connection is fixed and inelastic behavior is concentrated in beam plastic

hinges, braced frames rely on cyclic inelastic buckling and tension yielding of the bracing
13

members to accommodate the inelastic deformations and dissipate seismic energy. The figure

shows a schematic of a chevron braced frame deformed laterally, such that one brace is buckled

in compression and the other brace is in tension. Figure 2.2c shows the frame under larger

deformations, such that the compression brace has now formed a plastic hinge in the middle of

the brace. After several loading cycles the number depending on cycle amplitude as well as

brace geometry and material type a local buckle forms at the middle plastic hinge on a

compressive excursion. This localizes the plastic strain accumulation to the region of local

buckling. Fracture initiation closely follows local buckling, with complete cross-section rupture

and strength loss occurring soon thereafter.

Besides local buckling-induced fracture at the middle plastic hinge, other possible fracture events

could occur at the slotted-end brace-gusset connections, beam-column shear tab, or the gusset to

beam-column connections. These are illustrated in Figure 2.3a-b and are described below

Slotted-end brace-gusset (net section) connection fracture is shown schematically in

Figure 2.3a where the slot extending beyond the gusset plate creates a reduced section

susceptible to fracture during severe tensile loading cycles. As the brace is loaded in

tension, inelastic strains accumulate across a short gage length at the reduced area of the

connection, initiating fracture at the net section. Fabrication flaws could promote this

type of failure by introducing surface roughness or imperfections at the slotted section.

Furthermore, the net section is adjacent to the weld between the gusset-plate and brace

such that Heat Affected Zone (HAZ) defects could decrease the ductility of the brace

base metal as a result of material phase transitions or considerable grain growth.

Beam-column shear tab fracture is illustrated in Figure 2.3b and was observed during a

large-scale test by Uriz (2005). It was assumed that fracture initiated at fabrication cracks

below the fillet weld of the shear tab to column connection. At large story drifts,
14

considerable rotation demands on the beam-column connection can pry the shear tabs

from the column flange and create localized stress intensities on the weld, and

surrounding base, material. These increased demands, combined with the variability of

base metal properties near welds could cause column fracture at large drift levels.

Gusset-plate to beam or column connection fracture shown in Figure 2.3b is also

promoted by large rotation demands at the beam-column connection and the prying

action of the gusset plate on the flanges of the beam or column. Similar to the previous

two fracture events, fracture would also initiate close to the weld in the HAZ-affected

base metal.

In addition to these fracture limit states, inverted-V chevron configurations are susceptible to

concentrated inelastic behavior, and possible plastic hinge formation, at the upper-beam to brace

connection from a large force imbalance between the compression brace (small force) and the

tension brace (large force). The formation of a hinge at the mid-point of the beam could lead to a

story mechanism as illustrated in Figure 2.4. However, braced-frame system geometries with

alternating inverted-V and V bracing members (by story) or a zipper-frame configuration with

vertical axial columns at the beam mid-points (Bruneau et al, 2005) can prevent this type of

behavior by carrying the unbalance force.

Considering these possible failure mechanisms, current American Institute of Steel Construction

(AISC, 2005) Seismic Provisions aim to reduce inelastic effects in beams and columns and ensure

high ductility of the bracing members. While significant brace buckling and yielding is expected

in SCBFs during moderate to large earthquake events, detailing requirements guard against brace

and connection fracture. The design provisions most applicable to SCBF behavior are discussed

in the following section.


15

2.1.2 SEISMIC DESIGN PROVISIONS FOR SCBF SYSTEMS

The aim of modern steel seismic codes, such as the AISC Seismic Provisions (2005), is to ensure

acceptable system behavior through ductile performance of members and connections. Thus, in

the context of SCBFs, ductile system behavior is achieved through proper connection and brace

detailing to account for large rotations and repeated inelastic buckling and tension yield

excursions. The latter (ductile brace behavior) is attained through limits on geometric features

that control buckling mechanisms. While larger brace global slenderness has been shown to

increase brace ductility (Liu and Goel, 1988 and Tremblay, 2002), current seismic provisions

ensure ductile brace performance only through limits on the cross-section width-thickness, or

compactness, ratio (i.e., b/t for HSS). Referring to Figure 2.5, a more compact cross-section will

have a smaller width-thickness ratio and is less susceptible to local buckling during cyclic or

compressive loading. For HSS cross-sections, the Seismic Provisions (AISC, 2005) limit b/t

ratios as follows

b / t 0.64 E / Fy 16 (for Fy 46ksi ) (2.1.1)

Where b/t is used generically to represent the brace width-thickness ratio as tabulated in AISC

design charts, such that b/t = B/t 3 (i.e. (B-3t)/t), where B and t are the overall length of the

buckling face and the design thickness, respectively, as shown in Figure 2.5. Similar to the b/t

limit placed on square and rectangular bracing members, the current AISC Seismic Provisions

(2005) restrict D/t for Pipe cross-sections, such that

D / t 0.044 E / Fy 36.5 (for Fy 35ksi ) (2.1.2)

Where D/t represents the brace width-thickness ratio shown in Figure 2.5 as tabulated in AISC

design charts, where D is the nominal outside diameter of the pipe and t is the design thickness.

Current AISC Seismic Provisions (2005) ensure ductile behavior by prescribing limits on Wide-

Flange width-thickness ratios

b f / 2t f 0.30 E / Fy 7.2 (for Fy 50ksi ) (2.1.3)


16

Where bf and tf are the flange width and thickness dimensions, respectively. Note that Equations

2.1.1-2.1.3 are independent of other parameters, such as the brace slenderness, which is also

presumed to have an impact on brace ductility (Tang and Goel, 1989 and Tremblay, 2002).

Figure 2.5 suggests that a more slender brace will also provide an increase in ductility through the

decreased curvature, and therefore strain, demand at the middle plastic hinge for a longer brace as

compared to a shorter brace for the same axial deformation. However, as the brace becomes

increasingly slender the force imbalance between the tensile yield and compressive buckling

loads will generally increase, negatively affecting the energy dissipation capabilities while

increasing the system overstrength factor and the force demand on the beam in a chevron-type

braced-frame configuration. For these reasons, the current AISC Seismic Provisions (2005) limit

the slenderness of bracing members in SCBF design by

KL / r 4 E Fy (2.1.4)

Thus, KL/r limits are prescribed as 100 (Fy = 46 ksi), 115 (Fy = 35 ksi) and 96 (Fy = 450 ksi) for

HSS A500 Grade B, Pipe A53 Grade B and A992 Wide-Flange bracing members, respectively.

AISC (2005) has recently incorporated an exception to this limit by allowing braces with KL/r <

200 and greater than Equation 2.1.4 if adequate compressive capacity is supplied by the adjoining

columns. This seeks to incorporate the positive affect of large slenderness ratios on brace

ductility.

As mentioned in the previous section, connections in SCBF systems (between the brace, beam

and column) are susceptible to several brittle modes of failure, including weld failure and net-

section fracture at the end of the slotted brace. To prevent these types of failure, the Seismic

Provisions (AISC, 2005) require the design of these adjoining connections to be based on the

maximum force that the system can transfer to the connection. Although the Seismic Provisions
17

allow calculation of this force based on pushover or nonlinear time history analyses, they

recognize that (quoting from Section C13.3) In most cases, providing the connection with a

capacity large enough to yield the member is needed because of the large inelastic demands

placed on a structure by a major earthquake. Consequently, these connections are typically

designed for forces corresponding to the expected yield force of the brace as described in

Equation 2.1.5

Py Ry Fy Ag (2.1.5)

Where Py is the expected yield force of the brace, Fy is the minimum specified yield stress of the

material and Ag is the gross cross-sectional area of the bracing member. The ratio Ry between the

expected yield stress and the minimum specified yield stress of the member material recognizes

that the expected strength will typically be larger than the minimum specified strength. Typical Ry

values are in the range of 1.1 to 1.6 (Liu et al, 2007).

2.1.3 STATE-OF-THE-ART IN SCBF PERFORMANCE ASSESSMENT

To describe the current state-of-the art in SCBF system and component evaluation, the following

sections present experimental and analytical results, as well as fracture predictive approaches

from various investigations focused on the performance of SCBF systems. In general, an

emphasis is placed on brace component behavior, where buckling induced fracture is a concern at

the center of the brace. The literature review is presented by first considering results from SCBF

demand assessment studies, followed by investigations which quantify SBCF ductility capacity.

Given this format, it is useful to first reflect on the motivation, to separate demands from

capacities for structural/earthquake engineering simulations and the resulting dependence on post-

processing techniques for performance evaluation.


18

In the overall context of the theme of demands and capacities, it is interesting to discuss that

an ideal simulation of structural response should include the direct simulation of each physical

phenomenon (local buckling, fracture initiation, fracture propagation, etc.), leading up to the

complete failure of the system. In this ideal scenario, the simulation would describe the state of

the structure after the loading event and little post-processing (i.e., such as checking the demands

against capacities) effort would be required. The necessity for a post-analysis comparison

between demands and capacities (or limit states) arises from the lack of sophistication of

modeling techniques and their inability to simulate complex phenomena. This lack of

sophistication is an important issue in earthquake engineering where structures undergo large

inelastic deformations, accompanied by various forms of buckling and fracture. While some of

these aspects are routinely incorporated in structural analysis programs, events such as fracture

are typically evaluated by comparing the analysis results (cumulative deformation, maximum

drift, maximum strain, etc.) with experiment-based empirical relationships or other models that

seek to describe the facture ductility of the material or component. This approach is advantageous

in that fracture does not need to be explicitly modeled as part of the analysis, thereby significantly

reducing the computational expense of the analysis model while providing reasonable predictions

of fracture. However, the accuracy of this approach relies on (1) Accurate simulations of

structural behavior leading up to fracture, including effects such as local buckling, etc, that drive

the fracture strains (2) General fracture models that can accurately predict fracture based on these

localized strains and (3) The accurate characterization of local material properties.

Considering the difficulty of simulating complex phenomena leading to fracture events in large-

scale systems, SCBF performance has been largely characterized through a comparison between

earthquake demands from dynamic analysis techniques and ductility capacities obtained primarily

through experimental techniques. Based on these simulations and experiments, several empirical

and semi-empirical approaches have been suggested to predict the fracture response of braces in
19

these systems. These models represent important advances in brace fracture predictions and they

are reviewed in detail in this section. However, these approaches may be difficult to generalize to

situations different than the experiments used to calibrate them. Braced-frame shake-table tests

are also reviewed in the context of assessing earthquake demands and ductility capacities of the

system, components and connections. Note that shake-table tests experimentally combine demand

and capacity analyses as an assessment of structural behavior and performance is simply the

investigation of the state of the structure following the applied ground motion. Along these same

lines, there has been recent work on incorporating fracture predictions during simulations, thereby

illustrating the on-the-fly effect of fracture on structural demands and behavior. However,

suitable analysis and fatigue-fracture prediction models which consider the mechanisms leading

to brace failure are critical elements of such simulation methods.

2.1.3.1 STUDIES EXAMINING SEISMIC DEMANDS IN SCBF SYSTEMS THROUGH NONLINEAR DYNAMIC

ANALYSIS

Several analytical studies have investigated earthquake-imposed demands on SCBF systems

through nonlinear time history dynamic analyses. In this section, three analysis studies on 3 and

6-story SCBF systems, designed according to American design codes (FEMA 1997, AISC 1997),

are reviewed along with three analysis studies on Canadian code (NRCC 1990 and 2005, CSA

1989 and 2005) compliant braced-frames. The analysis results of the US designed frames are

presented first followed by the Canadian designed frames. After the analysis results are presented,

experimental results from a braced-frame shake-table test are presented in the context of

earthquake demands. Finally, a synthesis of these results is provided to compare and contrast the

different studies.

Interestingly, the studies by Sabelli, Uriz and McCormick et al use matching frames and ground

motions with different modeling techniques for the braces. The results from these three analysis
20

programs provide some background for the later discussion on brace modeling approaches

(section 2.2).

2.1.3.1.1 SABELLI (2001) ANALYSES

The design and analysis of the braced-frames for this study is based on a site-specific design for

downtown Los Angeles following the NEHRP Recommends (FEMA, 1997) and the AISC

Seismic Provisions (AISC, 1997). The reader is directed to Sabelli (2001) for the detailed design

of the structures. A suite of 20 horizontal ground motions are used in the time history analysis of

a 3-story and 6-story inverted-V SCBF. The ground motions correspond to the design level

earthquake hazard intensity of 10% chance of exceedance in 50 years, determined according to

the elastic spectral acceleration at the first period of the structure. The reader is referred to

Somerville et al (1997) for more information on the ground motions.

The analyses were performed using the SNAP-2D structural analysis program (Rai et al, 1996),

where the braces were modeled with axial truss members. The axial struts were assigned a

phenomenological (rule-based) uniaxial hysteretic response based on experimental work by Black

et al (1980). The uniaxial force-deformation model depends primarily on the brace slenderness

and other cross-section properties. Fracture was modeled through an empirical cycle-counting

scheme. Upon fracture prediction as per this scheme, the member was removed from the

simulation model.

Referring to Table 2.1 and the 10/50 hazard level, Sabelli (2001) reported mean maximum story

drift values of 3.9 and 1.8% (standard deviations of 3.1 and 0.8) for the 3 and 6-story SCBF,

respectively.
21

2.1.3.1.2 URIZ (2005) ANALYSES

Uriz re-analyzed the previous 3-story and 6-story inverted-V SCBF structures designed by Sabelli

(2001) with the Open System for Earthquake Engineering Simulation (OpenSEES) (McKenna

and Fenves, 2004) analysis platform. The analyses of Uriz are more sophisticated as compared to

those of Sabelli. Nonlinear fiber-based elements were used to model the bracing members where

an initial camber (or member sweep) was assigned to the elements to promote global buckling. In

addition to the design level, 10% chance of exceedance in 50 years, ground motion suite, the

study employed two additional suites of ground motions corresponding to earthquake hazard

intensities of 50 and 2% chance of exceedance in 50 years. While the 10/50 hazard level is a

typical design level event, the 2/50 hazard is the Maximum Considered Earthquake (MCE) level

(IBC, 2006). Table 2.1 reports the results of the nonlinear dynamic analyses which used the sets

of scaled ground motions (10/50 and 2/50 hazard levels) developed for the SAC study (see

Somerville et al, 2007). The Table also distinguishes between analyses which incorporated a

fatigue model to track brace fracture events. The model was applied at the fiber level and once the

fracture limit state was reached, the fiber was removed, effectively diminishing the cross section.

Referring to Table 2.1 and the 10/50 ground motion analyses without the brace fatigue model,

Uriz (2005) reported median maximum story drifts of 1.5 and 1.4% (standard deviations of 0.9

and 0.8) for the 3 and 6-story SCBF, respectively. By modeling brace fracture during the analysis,

the maximum story drift recordings remain approximately the same at 1.6 and 1.1% (0.9 and 0.6)

for the 3 and 6-story SCBF, respectively. At the MCE level (2/50), the median maximum drifts

increase substantially to 5.7 and 5.1% (3.0 and 3.4) without the fatigue model and 5.7 and 4.4%

(2.4 and 2.2) with the fatigue model for the 3 and 6-story SCBF, respectively. Note that several

analyses (3 for the 3-story and 6 for the 6-story) using the 2/50 ground motions revealed brace
22

fracture, as predicted by the built-in fatigue model, and eventual structural collapse. These

analyses are not included in the calculation of the median drifts reported in Table 2.1.

In addition to the maximum story drift data from nonlinear time history analyses, it is also

interesting to note the unsymmetrical response in assessing the demands on SCBF systems and

components. Unlike moment frames, where the response is more symmetrical (under far-field

ground motions), braced frames tend to show a more unsymmetrical response, presumably

because of the unsymmetrical strength and stiffness properties once the compressive brace

buckles. To illustrate this, Figure 2.6 shows a typical story drift time history from the 3-story

SCBF system during a 2/50 ground motion. The important observation is that the response is

highly unsymmetrical, such that the max is more than two times min, i.e. min = 0.4max for this

particular story and ground motion. Figure 2.7 illustrate this point further by plotting the

minimum story drift, min, versus maximum story drift, max, for each ground motion and story of

the 3-story and 6-story SCBF, respectively (Uriz, 2005). On average, the minimum story drift is

shown on the figures as approximately 40% of the maximum story drift for both frames and the

sixty LA-based ground motions. This behavior is notable because braced frame components (such

as braces) are typically subjected to symmetric loading protocols, based on adaptations of

protocols designed to reflect demands in moment frames (Fell et al, 2006, Han et al, 2007,

Shaback and Brown, 2003 and Archambault et al, 1995). Consequently, when the maximum

equivalent drift is reported as a capacity measure, it is calculated as half the range of equivalent

drift applied to the component (e.g. Uriz, 2005). With reference to Figure 2.7, this may

underestimate the capacity of the brace which, in general, will not be subjected to symmetric

cycles under seismic excitation. This issue will be discussed in more detail in Chapter 6 which

looks at the capacity characterization of bracing components.


23

2.1.3.1.3 MCCORMICK ET AL (2007) ANALYSES

McCormick et al (2007) used the same 3 and 6-story SCBF systems as Sabelli (2001) and Uriz

(2005). Similar to Uriz (2005), the analyses were performed using the OpenSEES platform.

Nonlinear fiber-based elements are used for the beams and columns while the braces are assigned

a phenomenological model to describe the hysteretic response. This is similar to the approach by

Sabelli (2001) where the hysteretic model depends on the brace slenderness and other cross-

section and material properties. Note that a fatigue-fracture model for the bracing members was

not used in this study.

Referring to Table 2.1, results from nonlinear analyses using the 10/50 and 2/50 ground motions

by Somerville et al (1997) are reported by McCormick et al. For the design-level, 10/50, event

mean maximum story drifts are listed as 3.57 and 1.97 (standard deviation of 1.61 and 0.68) for

the 3 and 6-story, respectively. For the MCE-level, 2/50, event, the corresponding maximum

drifts are 8.13 and 4.67 (standard deviation of 3.03 and 2.70).

2.1.3.1.4 REDWOOD ET AL (1991) ANALYSES

Redwood et al analyzed several 8-story braced-frames designed as per CAN/CSA-S16.1-M89

(CSA, 1989) and the 1990 edition of the National Building Code of Canada (NRCC, 1990). At

the time of this study, three categories of braced-frame systems existed in the Canadian code

corresponding to the expected ductility; these were (i) Ductile Braced Frames (DBF) (ii) Nominal

Ductility Braced Frames (NDBF) and (iii) Braced frames with no special ductility provisions

(SBF). In the context of our discussion on SCBFs, only the DBF results are summarized here.

While the detailed design is provided in Redwood and Channagiri (1991), it should be noted that

a considerably smaller R-factor (or strength reduction factor) is used in the design of the

Canadian DBF as compared to the American designed SCBF from above.


24

The 8-story DBF was analyzed using DRAIN-2D (Kannan and Powell, 1973) where the bracing

elements were modeled with axial truss elements and a phenomenological model (Jain and Goel,

1978) to describe the nonlinear hysteretic behavior. A set of ten earthquake ground motions were

chosen, primarily from recordings in the western United States, and scaled by the Peak Ground

Velocity (PGV) to correspond to a design level event for Vancouver, British Columbia. In terms

of Peak Ground Acceleration (PGA), the scaled earthquake records varied from 0.2g to 0.6g.

Note that from UBC (1997), the design PGA for zone IV, soil profile type SD, is approximately

1.1g. Thus, considering the smaller R factor used for the design and the less intense ground

motions, as compared to the previous 3 studies, these results are not directly comparable to SCBF

behavior but are included here for completeness.

From the nonlinear dynamic analyses, the largest story drift values were recorded on the seventh

story of the 8-story DBF. At this story, the mean maximum story drift across all ten ground

motions was listed as 0.7% (standard deviation of 0.88) by Redwood et al, with a maximum drift

equal to 1.1%. The drifts of the other stories ranged from approximately 0.3 to 0.6%.

2.1.3.1.5 TREMBLAY AND PONCET (2005) ANALYSES

Tremblay and Poncet (2005) analyzed several multi-story concentrically braced frames designed

as per the 2005 edition of the National Building Code of Canada (NRCC, 2005). The primary aim

of the study was to evaluate mass and geometric irregularity affects on behavior. For comparison

with the previous investigations, only the control structure with regular mass and geometric

distributions is considered here. This structure is an 8-story braced-frame with X bracing

members in each story. Similar to the Redwood et al (1991) study, DRAIN-2D (Kannan and

Powell, 1973) is used to analyze the structures where the bracing elements are modeled with axial

truss elements and a phenomenological model (Jain and Goel, 1978).


25

A suite of ten earthquake ground motions, scaled to a design-level event for Vancouver, British

Columbia, was used in the analyses. After scaling, the PGAs ranged from 0.3 to 0.6g across the

ten records. In comparison, PGA values for the Zone IV as per the 1997 UBC design spectra

(corresponding to approximately a 10/50 hazard) disregarding near fault effects are on the order

of 1.1g. The largest drifts were consistently recorded at the 6th story of the structure. The mean

maximum story drift at this story was approximately 1.7% (standard deviation of 0.5). The

maximum drift across all ground motions was listed as 2.7% and also occurred at the 6th story.

2.1.3.1.6 IZVERNARI ET AL (2007) ANALYSES

Izvernari et al (2007) presents results from analyses on several braced-frames designed according

to the 2005 edition of the National Building Code of Canada (NRCC, 2005) as per the Limited

Ductility (LD) and Moderate Ductility (MD) categories. The LD structures are not discussed

here. A total of 5 frames with 2, 4, 6, 8 and 12-stories were analyzed with the OpenSEES

platform where, similar to Uriz (2005), the braces were modeled with fiber-based elements. A

suite of twenty ground motions (ten historical and ten simulated) were used in the nonlinear

analysis of the frames. Referring to Table 2.2, the median drift recordings range from 1.1 to 1.8%

for the five frames, where the shorter 2 and 4-story frames had smaller drifts than the 8 and 12-

story frames.

2.1.3.2 SHAKE TABLE TESTING

Shake-table experiments can provide valuable earthquake demand data for structural systems,

while at the same time, evaluating the capacity of components and connections. However, it

should be noted that shake-table experiments only provide a single performance data point due to

the single ground motion used to damage the structure. While the experiments are visually

appealing and quite exciting, the results should be viewed objectively as the performance of a

single system during subjected to a single ground motion.


26

Tang (1987), Foutch et al (1987) and Roeder (1989) report findings from a 6-story braced-frame

shake table test conducted in Tsukuba, Japan. The design of the frame was based on both US

(UBC, 1979) and Japanese (Watabe and Ishiyama, 1980) building codes and lacked the detailing

of current design codes (AISC, 2005). For example, gusset plates were not provided at the ends of

the inverted-V bracing members as the braces were welded to the flanges of the beams. Table 2.3

lists maximum story drift measurements from a shake-table test using the Miyagi-Ken-Oki

earthquake scaled to 0.51g, approximately two times that of the moderate intensity level

experiment. A corresponding hazard level of the scaled ground motion was not discussed (Tang,

1987). A maximum drift prior to any fracture event was 1.6% in the third story of the frame,

which is within the range of the analytical results of Uriz (2005) and Izvernari et al (2007)

discussed previously with 10/50 simulation-based mean/median maximum drifts between 1.1 and

1.7%. However, while evaluating this comparison it should be noted that U.S. (and Japanese)

design codes have changed considerably between the time of the tests and the nonlinear time

history analyses. After fracture in stories 2-5, the maximum drifts increase significantly due to the

softening affect of fracture. This is most noticeable in story 3 where a maximum drift of 2.5% is

recorded after complete brace fracture and strength loss.

2.1.3.3 SUMMARY OF DEMAND ASSESSMENT FOR SCBF SYSTEMS

Current SCBF design requirements state that braces could undergo post-buckling axial

deformations 10 to 20 times their yield deformation (AISC, 2005). Given a system yield level

drift of approximately 0.3% to 0.5% (corresponding to initial brace buckling), the Seismic

Provisions may be interpreted as desiring a deformation capacity of approximately 3% to 5% for

SCBF systems. In this context, the story drift demands reported by Sabelli (2001), Uriz (2005)

and McCormick et al (2007) for the 10/50 (design) event in Table 2.1 range from 1.5 to 3.9% for

the 3-story frame and 1.1 to 2.0% for the 6-story frame. At the MCE, or 2/50, level the drift

demands increased substantially to 5.7 to 8.1% and 4.4 to 5.1% for the 3 and 6-story frame,
27

respectively. Thus, for design-level events, the 10/50 analyses seem to corroborate the expected

demands of the current AISC Provisions. On the other hand, at the MCE level, these results

suggest that SCBF earthquake demands could exceed 5% drift.

The discrepancy between the results of Table 2.1 should be noted, especially considering the

same structures and ground motions were used. This highlights the significance of robust

structural analysis techniques that can accurately characterize response. The different brace

modeling techniques, discussed more in section 2.2, between the three investigations may be

responsible for these differences.

The next section of the literature review discusses experimental studies focused on the cyclic

fracture/fatigue resilience of bracing members.

2.1.3.4 STUDIES INVESTIGATING THE CAPACITY OF BRACING MEMBERS

Where the previous discussion focused on braced-frame demand assessment, this section presents

experimental results which characterize the capacity of SCBF systems and bracing members. Of

primary interest is brace component behavior, where buckling-induced fracture at the middle

hinge could lead to eventual failure. The ductility of bracing components, often expressed in

terms of a maximum, or cumulative axial deformation can be compared to story drift demands

(discussed above) through simple kinematic relationships. While one such relationship is

introduced in Chapter 3, the primary aim of this section is to review experimental work which has

characterized the ductility of inelastic buckling braces. Many of the experimental studies

presented here have sought to generalize the experimental results with fatigue-fracture models of

varying degrees of complexity. However, these will be the focus of a later section. Furthermore,

while experimental investigations on small-scale (for example, 1x1 inch cross-sections) bracing

members have been a popular means to identify important trends in ductility and strength
28

requirements (Khan and Hanson 1976, Jain and Goel 1978, Jain et al 1980), they are not covered

in depth here. Given the highly nonlinear behavior of cyclic brace response, such as local

buckling-induced fracture (where the strains are sensitive to issues such as the length of the local

buckle), similitude might not exist between large and small scale experiments. Furthermore,

material properties and residual stresses within the walls of the cross-section may vary between

small and large-scale sections. Uriz (2005) provides an excellent summary of experimental work

with smaller bracing members.

As discussed previously and illustrated in Figure 2.2, inelastic buckling and yielding of bracing

elements serve as the primary seismic energy dissipation mechanisms in SCBF systems. To

provide background for the current discussion, it is important to consider the events leading to

bracing component failure during earthquake-type cyclic loading. Referring to Figure 2.8, the

first major limit state is brace buckling at point A, which is evident by large lateral deformations

and accompanied by flaking of the whitewash paint due to large strains associated with folding at

the end gusset plates and plastic-hinging at mid-length of the brace. As seen in region A-B,

buckling is followed by a sudden drop in load. However, as the compression in the brace reduces,

the response is mainly driven by bending of the buckled brace, resulting in a more gradual drop in

load after point B. During subsequent cycles, the localized yielding in the gusset plates and mid-

point hinge becomes more severe as the amplitude of compressive loading increases. Upon

reversed loading at point C, the stiffness gradually increases (point D) as the out-of-plane

deformations decrease. As the strut straightens, the out-of-plane deformations reduce, and the

brace yields in tension at point E. The subsequent compressive excursion results in a smaller

buckling load (see point F) as compared to the first buckling event due to the Baushinger effect,

increased brace length (from tension yielding), and residual out-of-plane deformations. After

repeated compression-tension loading cycles, a local buckle forms at the middle hinge, similar to

Figure 2.8c, which amplifies the plastic strain and triggers ductile fracture soon thereafter (Figure
29

2.8d). While these events are not evident on the load-deformation curve, the photos of Figure 2.8

are fairly representative of the local buckling and fracture initiation observed in most tests. Upon

further cycling, the rupture propagates in a ductile manner across the section, i.e., for square HSS,

the buckled face ruptures first as shown in Figure 2.8e. Finally, at some point during a subsequent

tensile excursion, the entire cross-section fractures suddenly, severing the brace. This progression

is typical in large-scale braced-frames and has been detailed across numerous studies.

Experimental investigations often compare the cyclic fracture capacity of bracing members to the

brace slenderness ratio, the width-thickness ratio, or a combination of the two. Thus, a general

trend of the following sections will be to discuss the capacity of bracing members with respect to

these two geometric properties. Generally, it is difficult to observe their independent affects as

cross-section compactness is linked with the slenderness ratio through the radius of gyration. This

difficulty provides context for subsequent chapters as 1) there is a lack of general models to

explain brace ductility through geometry properties and 2) continuum-based parametric studies

can be used to expand the experimental date presented in this and future studies.

The current discussion on brace capacity begins with a review of component tests on HSS bracing

members, the most commonly used section in SCBF design. While square and rectangular HSS

have been the most wide-spread shape used in concentrically braced frames, their ability to

provide adequate performance has recently been questioned due to their perceived poor

performance during cyclic loading. Thus, results from Pipe and Wide-Flange experiments are

presented and reviewed as they have gained increased popularity over HSS shapes as bracing

members in SCBF systems. Finally, frame tests either static or dynamic are presented to gain

an understanding of system level capacity and performance.


30

2.1.3.4.1 SQUARE AND RECTANGULAR HOLLOW STEEL STRUCTURAL (HSS) BRACING

COMPONENTS

The various experimental programs on square and rectangular HSS bracing members are

summarized in Table 2.4. While a more detailed synthesis of these experiments is the topic of

Chapter 6 (combined with the full-scale tests from this investigation presented in Chapter 3 and

4), this section provides a general review of experimental research on HSS bracing members.

Given that brace ductility is a function of several parameters, it is difficult to discuss specific

conclusions across all investigations without a common basis that can be used to interpret data

from various test programs, which may have different brace lengths and cross-sectional

dimensions. Thus, the following discussion introduces experimental research on HSS bracing

members by highlighting several important trends that have been reported in the context of HSS

brace ductility.

In general, the findings from all programs listed in Table 2.4 concur that width-thickness and

slenderness ratios have the most significant effect on brace ductility, such that higher width-

thickness ratios and lower slenderness are detrimental to brace performance. For example, Han et

al (2007) found that for approximately the same slenderness ratio (KL/r 80), increasing the

width-thickness ratio by 107% (from 13.7 to 28.3) resulted in a decrease in fracture ductility by

75%. In regard to slenderness influence, Tremblay et al (2003) reported that increasing the

slenderness by 60%, while holding the width-thickness ratio constant at 25.7, increased the axial

deformation ductility by 80%.

The investigations by Gugerli, Liu, and Lee with Goel (1982, 1988 and 1988, respectively) and

later Zhao et al (2002) were focused on the relative performance of concrete-filled HSS to

unfilled HSS bracing members. The concrete acts to prevent the inward local buckling observed
31

in most hollow sections, producing a less severe local buckling shape as the cross-section walls

buckle outward. In general, it was concluded that concrete fill has a beneficial effect on ductility

by delaying cross-section local buckling. For example, Lee and Goel (1988) reported a maximum

axial deformation during cyclic loading of 1.7 inches for an unfilled HSS cross-section, whereas

the same concrete-filled brace survived a maximum deformation of 2.6 inches, fracturing

approximately 8 cycles after the unfilled brace.

Archambault et al (1995) and Tremblay et al (2003) used the effective brace length to inform

design provisions on properties such as compressive strength degradation, energy dissipation and

brace fracture resistance. A kinematic relationship was also proposed to predict out-of-plane

buckling deformations. The investigation found that using the brace effective length (KL) resulted

in an accurate prediction of the compressive buckling load and, further, that as the slenderness

increases the energy dissipation capacity of the brace decreases linearly. Lastly, an empirical

fracture resistance equation was proposed, which is proportional to the brace global slenderness

and inversely proportional to the cross-section compactness ratio. Thus, the equation is

descriptive of the behavior that one would expect from the qualitative depiction shown in Figure

2.5, and employs data from all of the studies prior to 1995 (in Table 2.4) to calibrate various

constants of the model. The relationship and its accuracy will be discussed in more detail in

subsequent sections. Shaback and Brown (2003) present similar data and conclusions on HSS

bracing members as those of Tremblay et al (2003) concerning out-of-plane deformations, energy

dissipation and compressive buckling strengths, while employing a different empirical fracture

model that had been developed previously by Lee and Goel (1988) and Tang and Goel (1987).

Similar to the approach of Tremblay et al (2003) the fracture relationship is also a function of

slenderness and cross-section compactness while using a cumulative deformation measure to

assess capacity. This model will also be reviewed in section 2.2.2. While the experimental

investigation on bracing components by Yang and Mahin (2005) provides valuable data on cyclic
32

brace performance, the series of tests were conducted primarily to study the performance of the

slotted brace, gusset-plate end connection.

2.1.3.4.2 ROUND STEEL PIPE BRACING COMPONENTS

As mentioned earlier, there has been a recent shift from using square and rectangular HSS cross-

section shapes for bracing members to Pipe and Wide-Flange shapes. In part, the restrictions on

square HSS width-thickness ratios (i.e., Equation 2.1.1) have limited the number of shapes that

are available during design, making Pipe and Wide-Flange shapes a more attractive option.

Moreover as mentioned in the previous section, the cold working of square HSS sections during

fabrication was thought to significantly reduce the fracture toughness of the corner material,

driving fracture initiation at the corners first. Thus, the more gradual bend radii of Pipes and the

rolling process of Wide-Flange sections was thought to provide tougher material as compared to

square or rectangular HSS. This section, and the next, will briefly discuss several experimental

investigations on these alternative shapes.

At the time of the writing of this dissertation, an exhaustive study on steel Pipe bracing members

was being conducted by Tremblay et al at cole Polytechnique, Montral. The study is expected

to carry significant importance as, compared to square HSS members, there are relatively few

studies that have investigated the performance of round Pipes. However, there have been several

other studies that have looked at the cyclic behavior of large-scale Pipe bracing members. One of

the earliest studies on the cyclic behavior of steel Pipe sections was by Popov and Black (1981).

Although the reported experimental results focused more on compressive strength degradation

rather than on fracture events, an important finding was that the more compact Pipe4STD (width-

thickness ratio = 19, approximately 0.52 times the current maximum ratio recommended by

AISC, 2005) and Pipe4X-Strong (width-thickness ratio of 13.4, 0.37 times the current maximum

limit) braces locally buckled at very large axial deformations.


33

More recently, Elchalakani et al (2003) conducted a series of twenty steel pipe braces under

earthquake-type cyclic loading. The testing matrix for the braces tested to fracture included a

variety of different width-thickness ratios ranging from a D/t ratio from 9 to 21 with

slenderness ratios of approximately 24 or 35. Surprisingly, the results suggested that, when

grouped by slenderness, the less slender braces (those with KL/r ratios of approximately 24)

fractured later in the loading history and at larger axial deformations than the braces with a larger

slenderness ratio. Furthermore, the investigation concluded that brace fracture ductility is more

sensitive to the cross-section width-thickness ratio as compared to the slenderness ratios. While

the results consistently confirm the less slender braces are more ductile, it should be noted that

the study queried a much larger range of D/t ratios (9 to 21) as compared to brace slenderness

ratios (24 to 35). Nonetheless, the study provides valuable fracture capacity data that can be used

to assess the performance of steel Pipe bracing members.

2.1.3.4.3 WIDE-FLANGED BRACING COMPONENTS

Similar to steel Pipe cross-sections, Wide-Flanged braces have becoming increasingly popular in

recent years due to their perceived superior ductility over HSS members during earthquake

loading. However, there are several challenges with using Wide-Flanged sections in braced-frame

construction. First, novel gusset plate end-connection details, not discussed as a part of this

chapter (see Chapter 3 for an example connection detail), must be devised to ensure out-of-plane

buckling about the member weak axis. These connection details can be quite complex as, unlike

HSS or Pipe sections, traditional slotted-end connections are not feasible for Wide-Flange shapes.

Moreover, the number of Wide-Flanged shapes that can be used as bracing members is somewhat

limited to shapes with small to moderate tensile yield to buckling load ratios. If not selected

correctly, the section properties can create a large force imbalance on the beam (in the case of

chevron-type configurations) and a reduced energy dissipation capacity. While this is accounted
34

for through Equation 2.1.4 (AISC, 2005) it does have the effect of limiting the shapes that can be

used as bracing components.

Popov and Black (1981) and Gugerli and Goel (1982) conducted a series of experiments on

Wide-Flanged bracing members. Popov and Black tested Wide-Flanged members with width-

thickness ratios ranging from 5 to 8.25 with slenderness ratios from 40 to 120 yet provide little

detail on the relative performance of the members as related to fracture ductility. Gugerli and

Goel present brace fracture capacities for very slender (KL/r from 95 to 175) Wide-Flanged

bracing members with width-thickness ratios between 7.4 and 11.5. As expected, these results

show that ductility improves with increasing slenderness and decreasing compactness.

2.1.3.4.4 GENERAL DUCTILITY TRENDS OF HSS, PIPE, AND WIDE-FLANGED STEEL BRACES

Figure 2.9 compares experimental results from three separate investigations by Shaback and

Brown (2003) on square and rectangular HSS, Elchalakani et al (2003) on round Pipe, and

Gugerli and Goel (1982) on Wide-Flanged cross-sections. The data is presented here for example

purposes to identify issues and questions with respect to assigning brace ductilities based on

slenderness and/or compactness. Furthermore, as each study applies the same loading history to

each brace, the results within each cross-section type can be discussed without considering the

influence of loading history. The shake-table and quasi-static test results from Tang (1987) and

Uriz (2005), respectively, are also shown but will be discussed in the following section. Figures

2.9b and 2.9c illustrate the influence of width-thickness and slenderness ratio, respectively, on the

maximum deformation range in tension and compression (shown in Figure 2.9a) of the brace

prior to failure, range, normalized by the yield deformation, y. The brace width-thickness and

slenderness ratios are normalized by the AISC (2005) limits described previously. It is important

to note that the brace component loading protocols varied between the three investigations where

a symmetric history in tension and compression was used for the Pipe tests, a slightly biased
35

compressive loading history for the HSS, and a compressive dominated history for the Wide-

Flanged tests. While a cumulative deformation quantity, such as energy dissipation or cumulative

plastic deformation, may be a more descriptive measure in comparing relative brace capacities,

the maximum range sustained by the brace prior to failure (Figure 2.9a) is used here as this can be

most readily transferred to a simplified maximum story drift. As illustrated in the demand section

(see Table 2.1-2.3), this quantity is most often used when discussing structural system demands

and will set the context for later chapters.

Referring to Figure 2.9b, as the width-thickness ratio (normalized by the respective AISC, 2005

limits) increases within a group of braces with similar slenderness ratios, the maximum

deformation capacity tends to decrease. For example, the Pipe braces with normalized slenderness

ratios of 0.2 (solid circles) show a definite downward trend for increasing width-thickness ratio.

This trend is also observed for HSS and Wide-Flange members when grouped according to

relative slenderness ratios. In general, the experiments selected for this example confirms the

current design methodology of placing upper-bounds on section width-thickness ratios (AISC,

2005).

Figure 2.9c illustrates the influence of slenderness ratio on maximum deformation capacity using

the same data points. The braces are grouped according to relative width-thickness ratios,

facilitating a comparison between ductility capacities as a function of slenderness. The figure

shows an apparent trend between increasing axial deformation ductility and slenderness ratio.

Interestingly, slenderness also appears to mitigate unfavorable width-thickness ratios, as the

compactness of the experiments presented here seems to decrease, on average, as slenderness

increases. However, referring back to Figure 2.9b, the influence of brace slenderness within the

Pipe and HSS sections is questionable. For example, a relative slenderness ratio decrease from

0.62 (solid squares) to 0.53 (hollow squares) for HSS members does not seem to influence
36

ductility, rather it is mostly controlled by the width-thickness ratio. Furthermore, the Pipe cross-

sections show that increasing slenderness (from 0.2 to 0.3) acts to decrease ductility. However, it

should be mentioned that the HSS and Pipe specimens sample a smaller range of slenderness

ratios as compared to the Wide-Flange braces which have a much larger range of slenderness

ratios (1.0 to 1.8). Thus, the investigations within the HSS and Pipe experiments are somewhat

limited with respect to slenderness affects. Considering the trends presented in Figure 2.9, several

key concerns and knowledge gaps can be raised pertaining to the current state-of-the-art brace

capacity assessment procedures

1. Given that brace slenderness tends to influence the axial deformation capacity of a brace

during inelastic cyclic loading, should design codes restrict the use of stocky members in

braced-frame construction? For example, in the case of the steel Pipe specimens shown in

Figure 2.9, the axial deformation capacity is severely reduced as compared to the more

slender HSS and Wide-Flanged braces. However, according to AISC Seismic Provisions,

the Pipe members have lower relative width-thickness ratios as compared to the HSS and

Wide-Flanged members. Thus, while the experiments suggest otherwise, the code deems

Pipe members as more acceptable compared to the HSS and Wide-Flanged braces

presented here.

Furthermore, should design codes restrict all braces to meet the same compactness

requirement (within cross-section type)? According to the data of Figure 2.9, doing so

may unnecessarily penalize cases where slender braces would perform well, even with a

large width-thickness ratio. For example, the large slenderness ratios (or possibly the

nature of the local buckling deformation) of the Wide-Flanged members presented in

Figure 2.9 seems to mitigate the competing influence of a large width-thickness ratio.
37

2. Does the maximum axial deformation capacity of individual bracing components, used in

Figure 2.9, directly translate to a system level ductility measure? Experimental brace

capacities are often expressed in terms of axial deformation capacity, but few have

translated this into a story drift capacity. This would facilitate a comparison between

earthquake system level demands and component ductility levels.

3. With respect to the first point, can the interactive influence of brace slenderness and

width-thickness ratio be rationally evaluated without relying on a fully empirical,

experiment-based approach? While several relationships have been developed (presented

in the following sections) to describe the ductility capacities for HSS shapes, they are

largely empirical and are not derived from a fundamental mechanics-based approach.

Thus, it could be beneficial to develop a simulation-based methodology that could

investigate the relationship between brace ductility and the governing geometric (or

material) brace properties. Although not explicitly mentioned previously, the brace

length, for example, is often constrained in most experimental due to out-of-plane testing

restraints. Furthermore, actuator force capacity and tests setup, not to mention economics,

can limit the scale and breadth of large-scale brace investigations.

4. Is brace capacity influenced by loading history? With reference to Figure 2.9, the Wide-

Flanged members survived the largest deformation capacity prior to failure, yet were

subjected to a compression dominated loading history (refer Gugerli and Goel, 1982)

with relatively small tensile excursions. Furthermore, the least ductile HSS and Pipe

braces were tested with a slightly biased compressive and standard symmetric loading

history with (i.e., equal compressive and tensile excursions as illustrated in a typical

symmetric hysteretic response in Figure 2.9a), respectively. It could be envisioned that a

symmetric history could be more severe as the tensile action during local buckle

straightening could severely increase the cumulative plastic strain in the middle plastic

hinge region.
38

These issues are specifically addressed in subsequent chapters.

2.1.3.4.5 FULL-SYSTEM BRACED-FRAME TESTS

This section presents results from two experimental investigations on large-scale CBF systems.

The first is the shake-table test on a 6-story CBF that was discussed previously in the demand

section and reported in Tang (1987), Foutch et al (1987), and Roeder (1989). The second is a

static test on a 2-story SCBF system tested as part of the work by Uriz (2005).

Listed in Table 2.3 and illustrated in Figure 2.9 are the approximate brace ductility capacities

from the 6-story shake-table test reported in Tang (1987). Interestingly, the axial deformation

capacities (range/Y) prior to brace fracture (stories 2 through 5) are similar to the capacities

discussed in the previous section from brace component tests. This is a valuable comparison as it

provides some justification to apply static, component results to inform large-scale system

behavior. The story drifts recorded prior to the onset of brace fracture range between 1.1 and

1.6%. However, as mentioned previously, this frame lacked modern detailing requirements, such

as gusset plate connections and featured braces with b/t ratios that exceeded current width-

thickness limits.

Also shown in Figure 2.9 are the brace ductility capacities (in terms of range/Y) for two braces in

the lower story of a 2-story SCBF by Uriz (2005). The second story braces did not fracture.

Similar to the shake-table results, these results are in line with the HSS component tests by

Shaback and Brown (2003), thereby supporting the use of component tests to inform system

behavior. One should also note the ability of the range/Y to accurately describe the brace

ductility. This will be important in Chapters 6 where the maximum deformation range prior to

brace fracture is used to approximate ductility. The maximum story drift for the first story prior to

any brace fracture for the quasi-static test was approximately 1%. The results illustrated in Figure
39

2.9 lend validity to the numerous component tests described previously. Furthermore, and as

discussed in the following section, if brace component behavior can be simulated correctly, the

extrapolation to large-scale behavior can be relatively straight-forward considering these results.

2.2 MODELING TECHNIQUES FOR SCBF SYSTEMS

Simulating complex physical events has long been a focus of the academic community.

Simulation models reduce the necessity of exhaustive experimentation and extend experimentally

observed phenomena to cases which may be difficult to test. Thus, through the application of

theoretical and analytical models, routine design and construction need not rely on the testing of a

replicate structure. Doing so would be an impractical, and in many cases, an impossible feat. In

this light, the current section presents structural modeling techniques in the context of the braced-

frame system. Similar to the previous section, the discussion will be separated by first considering

demand modeling, and second, capacity modeling. The former contains models which aim to

characterize quantities such as deformations, strains and stresses from an applied load, ground

motion, etc. Capacity modeling will be treated as post-processing techniques which are applied

after the analysis and aim to describe the fracture ductility of the steel bracing members.

As discussed in the introduction to this chapter, performance assessment generally takes the form

of, first, developing an analysis model that captures well-known geometric or material behavior

and, second, post-processing the results to gain insights into a variety of different phenomena

which are not included as part of the analysis model. Phenomena investigated in the second stage

are typically more difficult to incorporate into the primary model due to computational expense, a

lack of scientific understanding, or a combination of both. For example, fracture propagation

under certain conditions can be computationally prohibitive if the model is the scale of an actual

building while the elastic modulus of any structural material can be accurately described and is

included in any analysis model. Thus, while the relationship between the elastic strain and stress
40

is contained in the primary model, most fracture assessments are made following the analysis by

comparing, for example, material strains (demands) to critical toughness parameters (capacity).

In this context, current state-of-the-art modeling approaches are presented by first discussing

simulation methods that aim to characterize earthquake demands on structural systems for steel

braced-frames. Second, models that describe the capacity (usually in terms of a fracture event) of

a braced-frame system or component are presented. Generally, the capacity models are somewhat

empirical in nature and serve the purpose of expanding, or providing an explanation of, trends

from a particular experimental data set. Other models utilize data across a wide-range of

experiments, but could still be classified as empirical in nature and tend not to address the actual

physical mechanisms that control brace failures. The work by Uriz (2005) is highlighted in the

following sections as an example where performance assessment is completed during the analysis

procedure by comparing strain demands to a critical strain measure, thus eliminating a post-

processing-comparison approach for the fracture limit state. While only applied to one material

type (and one brace cross-section type), the work provides context to discuss the advantages of an

on-the-fly performance-based framework and areas where the community could benefit from

advanced simulation capabilities.

2.2.1 DEMAND CHARACTERIZATION

This section presents modeling techniques to assess demand quantities (such as deformations,

strains, etc.) for bracing components subjected to earthquake-type loading histories. For purposes

of this discussion, it is assumed that earthquake demand assessment first begins with a site-

specific earthquake hazard analysis, followed by a ground motion scaling procedure (e.g. as

outlined in Haselton, 2006). Once the ground motion(s) are determined, they serve as an input

base excitation to the structure, with the response of the superstructure being determined through

the fundamental equations of dynamic equilibrium. While copious literature is available on the
41

subjects of hazard, risk and dynamic analysis procedures, this section is specifically focused on

determining the response a bracing member once the input deformations (or histories) are known.

These could either be ground motion-induced deformations or some form of a standard loading

history.

2.2.1.1 TECHNIQUES FOR BRACE SIMULATION

Based on the previous discussion describing braced frame response, and Figure 2.8, this section

describes various techniques that have been used to simulate braced-frame response (specifically

the response of the bracing elements themselves). Some of the techniques discussed are aimed

only at providing good simulation of brace load-deformation response, whereas other, more

sophisticated techniques can simulate complex aspects of response including local buckling and

fracture.

2.2.1.1.1 RULE-BASED (OR PHENOMENOLOGICAL) BRACE MODELS

Uniaxial spring elements with rule-based phenomenological, constitutive response are often used

to model bracing elements (Zayas et al 1980, Khatib et al 1988, Sabelli 2001, and McCormick et

al 2007). As illustrated in Figure 2.10a, hysteretic rules are applied to a single degree of freedom

(axial deformation in the case of bracing elements) with the resulting behavior described by

various linear segments, transitioning between different slopes according to empirical calibration

techniques. While convenient, this approach is limited because it does not directly model

buckling events, and thus, cannot provide insight into localized brace behavior. Furthermore,

calibration is often based on empiricism and thus, these models may be more difficult to

generalize to different materials, cross-sectional shapes or loading histories. On the other hand,

phenomenological models are advantageous if the analysis is primarily aimed at assessing

global demands, for example, maximum story drift or axial deformations.


42

2.2.1.1.2 CONCENTRATED HINGE BRACE MODELS

Elastic beam-column elements with lumped plasticity end-nodes are often used in structural

analysis programs where the inelastic behavior is known to concentrate at the ends of structural

members (for example, beam elements in moment frame systems). The inelastic behavior at the

end-nodes is simulated through stress-resultant plasticity formulations (Powell and Chen, 1986;

El-Tawil and Deierlein, 1998 and Hajjar and Gourley, 1977) which operate at the cross-sectional

level, i.e., through a P-M interaction curve as shown in Figure 2.10b. For example, Higginbotham

and Hassan (1976), Ikeda and Mahin (1986), and Hassan and Goel (1991) used variations of this

approach to generalize the behavior of buckling braces. It was shown that for general cyclic

loading, the behavior of the bracing elements at the global force deformation level is captured

fairly accurately. This technique is attractive because the key aspects of response (i.e. buckling

and the post-buckling geometric nonlinearities) are explicitly modeled, and calibration is less

subjective.

While these models are generally more robust than the phenomenological models discussed

previously, they are still limited by some degree of empiricism in calibration. Furthermore,

lumped plasticity elements may not accurately capture yielding, such as may occur over the entire

length of the brace during tensile excursions. Refinements to this approach include Jin and El-

Tawil (2003), based on the prior work of El-Tawil and Deierlein (2001). This research

incorporates the distribution of plasticity along the length and across the cross-section of the

brace.

2.2.1.1.3 FIBER-ELEMENT BASED BRACE MODELS

Recently, the more powerful and versatile fiber-based element has been used to simulate brace

response. In general, a fiber-based model is formulated with integration points along the length of
43

the element where the member cross-section is discretized by a fiber mesh as shown in Figure

2.10c. A constitutive model is defined at the fiber level and allows for strain and stress gradients

through the cross-section.

The fiber-based element formulation can directly model the spread of plasticity along the length

of the member as well as within the depth of the cross-section, whereas a lumped plasticity model

accounts for these effects only indirectly (see Jin and El-Tawil, 2003). Unfortunately, as the fiber

discretization exists at the cross-section level (at integration points along the element), the

element can not explicitly model localized affects along the length of the brace, such as the strain

amplification during local buckling. Thus, in the context of modeling bracing members, which

can develop significant local buckles during cyclic loading, the fiber-element is somewhat

inadequate for capturing localized stress and strain demands. However, Uriz (2005) and others

(Hanbin et al, 2007; Krishnan 2008-in review) have implemented schemes to track a critical

strain measure at the cross-section level of a fiber-element to predict fracture. The approach relies

on empirical calibration to account for the affects of local buckling, but has been shown to

provide reasonably accurate fracture predictions in bracing element, as discussed later.

2.2.1.1.4 CONTINUUM-BASED BRACE MODELS

Continuum finite element models, such as shown in Figure 2.10d and 2.11 constructed through

commercial or research finite element software (e.g., ABAQUS, 2004) can be applied to model

the brace response during cyclic loading. These simulations incorporate continuum (shell or

brick) elements, large displacement and deformation formulations, and continuum cyclic

constitutive response. In addition, by simulating local as well as global imperfections, these

simulations can directly model several complex phenomena (such as local buckling) that lead to

fracture. Although computationally expensive, the resulting simulations provide a high resolution

of the stresses and strains at the local buckle (e.g. Figure 2.11), from which ductile fracture
44

initiation is assessed using micromechanics-based models, such as ones proposed by Kanvinde

and Deierlein (2004 and 2007). Furthermore, the simulations provide interesting insights into

various damage mechanisms. For example, the simulation illustrated in Figure 2.11 indicates that

while the strains in the cross-section due to global buckling and bending alone are on the order of

0.02, the strains induced through the HSS wall thickness by local buckling are more than an order

of magnitude larger ( 0.6), thereby underscoring the importance of simulating local buckling

during brace cyclic response.

Continuum-based formulations of bracing elements are advantageous as compared to integration

point fiber-based modeling as the continuum model directly evaluates inelastic brace response at

each point along the length of the member as well as through the depth of the cross-section. Thus,

contrasted with fiber-based elements, a continuum approach allows localized deformation effects,

such as local buckling, to develop along the length the cross-section of the bracing member.

While solid finite element analyses present notable advantages over lumped-plasticity and fiber-

based models, there is a heavy computational expense that accompanies these models, especially

considering the complex events of cyclic inelastic compressive buckling and tension yielding of

bracing members during earthquake-type loading. Moreover, currently, fracture initiation and

propagation events are not typically modeled as part of the analysis routine in standard finite

element software packages. In large part, this can be attributed to a lack of fundamental models

that seek to capture the complex micromechanical events that trigger fracture initiation and

propagation. Furthermore, innovative re-meshing schemes often need to be employed at an

advancing crack tip as the fracture propagates through the continuum body (Khoei et al, 2008).

While mechanics-based methodologies have been developed to address these issues (Rao et al,

2007), they are typically applied to relatively simple geometries and loading conditions and, in

general, would add a significant computational expense to large-scale inelastic cyclic brace

analyses.
45

With the advent of general, physics-based initiation and propagation models, continuum analyses

provide the framework to eliminate the separation between demand and capacity evaluation.

However, these fracture models often describe complex micromechanical processes and need to

be rigorously evaluated prior to implementation into FEM analyses. One such initiation model,

which is presented through a post-processing demand versus capacity approach, will be presented

in the following section.

2.2.2 CAPACITY CHARACTERIZATION

The previous section illustrated several modeling techniques that can be used to simulate the

various aspects of inelastic brace behavior during cyclic loading, and the corresponding

deformation or strain demands in the braces. Given these demands, this section provides a brief

review of the various approaches that have been developed to evaluate the fracture ductility of

bracing members, relative to these demands.

2.2.2.1 BRACE GEOMETRY-BASED FRACTURE CAPACITY RELATIONSHIPS

Lee and Goel (1988) described the cyclic ductility capacity of HSS bracing members with (KL/r)

> (KL/r)critical in terms of a normalized (by Y) cumulative axial deformation, f,pred, as per the

Equations below

2
F 1 b a2 B
a a a
3 KL 4
f , pred C1 Y
C2 (2.2.1)
FY , meas t H r

and,

2
F 1 b a2 B
a a3 a4
KL
f , pred C1 Y C2 (2.2.2)
FY , meas t H r critical

if (KL/r) (KL/r)critical, where C1, C2, and a1 a4 are determined by an empirical fit of the

available HSS data, and a4 = 0 (i.e., slenderness was not considered) in the original formulation
46

by Lee and Goel (1988). The measured and specified yield stresses are included as FY,meas and FY,

respectively. In Equations 2.2.1 and 2.2.2, b/t is used generically to represent the brace width-

thickness ratio, such that b/t = B/t 2, where B and t are the overall length of the buckling face

and the design thickness, respectively, as shown in Figure 2.51. The cross-section dimensions B,

H and t are the nominal dimensions of the square or rectangular brace and are also illustrated in

Figure 2.5. The axial displacement fracture prediction, f,pred = f,pred.Y, is a cumulative measure

of brace ductility and is determined by summing the compressive and tensile normalized axial

deformations (defined in Figure 2.7a) up to failure of the brace

f,exp Y 0.1 compression tension (2.2.3)

Several other experimental investigations on HSS bracing members have used the empirical

relationship proposed by Lee and Goel (1988) to calibrate the constants in Equations 2.2.1 and

2.2.2. Table 2.5 lists the results of the various calibration studies conducted on HSS bracing

members tested by Tang and Goel (1989), Archambault et al. (1995), and Shaback and Brown

(2003). It is important to note that Tang and Goel expressed ductility in terms of a standardized

number of cycles to fracture (Nf) instead of the normalized axial deformation, f,pred, used by the

others (see Tang and Goel, 1989, for more details). Also, Shaback and Brown used both

(0.1compression + tension) and (compression + tension) to calibrate Equations 2.2.1 and 2.2.2 as they

argued that the penalty of 0.1 on the normalized cumulative compressive excursions is not

representative of the full deformation demands on the brace, especially considering the largely

compressive dominant loading histories applied by the Goel et al investigations. Due to these

discrepancies, Table 2.5 does not intend to suggest average calibration values for Equations 2.2.1

and 2.2.2; rather, the constants are supplied to provide a synthesis of the fracture predictions for

HSS bracing member which take the form of the above model. The accuracy of each model

compared to three separate testing programs will be discussed later.

1
Note: Since the original formulation of Equations 2.2.1 and 2.2.2, b/t is now generally listed as B/t-3.
47

Tremblay (2002) and Tremblay et al (2003) developed similar empirical relationships for the

cyclic fracture life of HSS bracing components. The first (Tremblay, 2002) provides a rigorous

synthesis of past experimental work on the cyclic performance of HSS bracing members. Thus,

the fracture model is calibrated across six separate experimental studies with a variety of loading

histories and section properties. Interestingly, the total fracture ductility, measured in terms of the

maximum compressive and tensile ductility, compression and tension, respectively, was found to be

function only of the global brace slenderness ratio, , defined as

KL FY ,meas
(2.2.4)
r 2E

Assuming a linear relationship, the expected ductility of an HSS brace component was found to

be

f , pred compression tension 2.4 8.3 (2.2.5)

or

FY , meas
f , pred Y ,meas 2.4 8.3 L 2.4 8.3 (2.2.6)
E

Note that, f,pred in Equation 2.2.6 differs from the previous empirical models as it simply

describes the sum of the maximum compressive and tensile deformations rather than the

cumulative compressive and tensile deformations of each cycle. A later publication by Tremblay

et al (2003) suggested a relationship for the maximum rotation prior to brace failure that is more

similar in form to the original model by Lee and Goel (1988)

0.1 0.3
b d KL
f , pred 0.091 (2.2.7)
t t r

Where b/t and d/t are used generically to represent the brace width-thickness ratio, such that b/t =

B/t 4 (or d/t = D/t 4). Note these ratios differ from the B/t 2 and B/t 3 used in the previous

empirical relationships and current AISC standards (2005), respectively. Considering the
48

kinematic relationship between the rotation and axial deformation of a buckling brace with plastic

hinges at both ends and at the midpoint, Equation 2.2.7 can also be expressed as

F 0.3 2
Y ,meas b d KL
0.1
f , pred 2L 0.046 (2.2.8)
2 E t t r

Where f,pred is the sum of the maximum tensile and compressive axial deformations.

To show the accuracy of these empirical fracture predictions, Figure 2.12 illustrates the ratio of

experimental fracture measurements, f,exp, for three separate testing programs to predicted

fracture capacities, f,pred, versus the model number used to calculate the capacity according to the

following key

1. Lee and Goel (1988) model described by the general Equation 2.2.1 (see Table 2.5 for

calibration constants), where the cumulative experimental deformation, f,exp, is

calculated with Equation 2.2.3. Note this model does not consider global slenderness

affects (a4 = 0 in Equation 2.2.1).

2. Archambault et al (1995) model; similar in form to Lee and Goel (1988) model, but

considering affects of brace slenderness (i.e., a4 0).

3. Shaback and Brown (2003); Archambault et al (1995) model with different calibration

coefficients (see Table 2.5).

4. Tremblay (2002) model; linear relationship between the summed maximum compressive

and tensile deformations before fracture (f) and the global slenderness parameter

(Equation 2.2.6).

5. Tremblay et al (2003) model described by Equation 2.2.8; empirical relationship similar

in form to models 13, but with f, taken as the sum of the maximum compressive and

tensile deformations before fracture.


49

Referring to Table 2.5 and Figure 2.12, the fracture predictions of model 1 are the farthest from

the experimental fracture deformations (average f,exp to f,pred ratio of 2.6 with a COV of 0.42),

most likely from the omission of a slenderness dependent quantity. Models 2 and 3 include a

global slenderness term (KL/r), resulting in more accurate fracture capacity predictions (average

f,exp/f,pred = 1.3 and 1.2, respectively) with model 3 showing less scatter (COV = 0.30) than

model 2 (0.41). Interestingly, model 4 is one of the more accurate models (f,exp/f,pred = 0.8, COV

= 0.41) despite the absence of a b/t term in the formulation (refer Figure 2.5). The last model (5),

is the most accurate empirical-based model (f,exp/f,pred = 0.9, COV = 0.37). Interestingly,

whereas the first three models used a cumulative deformation measure to assess fracture capacity,

model 5 relies only on the summed maximum compressive and tensile deformations (or

deformation range). This is an important basis for the work presented in Chapter 6 which relies

on the assumption that brace capacity can be accurately characterized through the maximum

deformation range prior to fracture.

While empirical-based relationships are useful, they tend to be quite sensitive to variations in

loading history (Shaback and Brown, 2003) and brace material property. Moreover, the results in

Table 2.5 and Figure 2.12 are calibrated for only HSS bracing sections, under a limited set of

loading histories. To extend the methodology to other brace cross-sections, more experimental

data would be needed to calibrate accurate relationships. While the form of additional equations

would be similar to the relationships presented above, the confidence of empirical-based models

relies solely on a comprehensive data set with tests encompassing a wide range of b/t and KL/r

ratios and loading histories. Thus, one could argue that there is a need to develop general models

that implicitly reflect the relationships described above while not relying on numerous large-scale

tests data sets to calibrate brace ductility.


50

2.2.2.2 FIBER ELEMENT-BASED CRITICAL STRAIN MODELS

Uriz (2005) recently implemented an on-the-fly rain-flow counting scheme to monitor the

accumulated cyclic damage of an HSS6x6x3/8 bracing member for several different loading

histories. The model was used together with a fiber-based brace model (described previously) and

was shown to predict fracture quite accurately. The damage is expressed in terms of an equivalent

cycle, ni, for a cycle with a strain amplitude of, i, at any point and is compared to a critical strain

measure according to

crit i nim (2.2.9)

Where m and crit are calibrated according to standardized fatigue testing procedures. For the case

of monotonic loading, ni = 1 and Equation 2.2.9 simplifies to a comparison between the strain at

any point, i, and the critical monotonic fracture strain, crit. During cyclic loading (i.e., ni > 1) the

strain demand needed to induce fracture at a point is reduced by the assumption that the material

is undergoing a fatigue capacity reduction. Once the critical strain is exceeded, the material is

assumed to have fractured at that point. Referring to the previous discussion on the fiber-based

modeling approach, and considering that stress and strain is defined at a fiber level, a fiber can be

removed (strength and stiffness set equal to zero) once the strain reaches the critical strain

according to Equation 2.2.9. This allows the model to mimic the behavior of fracture initiation

through to propagation using a simple fiber-based model. In the context of performance-based

earthquake engineering, this approach is highly attractive as fiber-based models are

computationally cheap compared to more detailed continuum models, yet are more fundamental

in nature as compared to other simplified models (i.e., phenomenological models).

While the work by Uriz (2005) suggests a critical strain of crit = 0.095 with m = 0.5 for an

HSS6x6x3/8 bracing member, one may argue that these strains, being monitored at the fiber

level, are not the true strains responsible for fracture, and thus may be configuration dependent.
51

However, continuum analyses combined with novel fracture initiation models (discussed next)

could calibrate these material parameters for a wider-range of bracing member types and

geometries. A continuum-based calibration procedure combined with the computationally

inexpensive fiber models could provide an advantageous approach to simulate fracture in full-

scale steel structures.

2.2.2.3 MICROMECHANICAL VOID GROWTH AND COALESCENCE MODELS

This section presents a brief discussion of micromechanics-based models to predict fracture in

bracing elements. These models, operating at the continuum level can be used in conjunction with

continuum finite element simulations discussed earlier. Thus, rather than relying on global or

abstracted strain or deformation definitions, they rely on direct estimates of stress and strain

histories simulated through sophisticated finite element simulations. The resulting predictions

offer the advantages of being general as well as accurate, as compared to the less sophisticated

approaches, which are limited by their assumptions and oversight of critical aspects of response.

Ductile fracture and fatigue in steel is caused by the processes of void nucleation, growth, and

coalescence as illustrated in Figure 2.13 (Anderson, 1995). As the steel material experiences a

state of triaxial stress, voids tend to nucleate and grow around inclusions (mostly sulfides or

carbides in mild steels) in the material matrix and coalesce until a macroscopic crack is formed in

the material. Previous research (Rice and Tracey, 1969) has shown that void growth is highly

dependent on the equivalent plastic strain, p, and stress triaxiality, T = m/e, where m is the

mean or hydrostatic stress and e is the von Mises stress. Assuming that voids grow when the

localized triaxiality is positive and shrink when negative, Kanvinde and Deierlein (2007)

quantified cyclic void growth described by the ratio of the current void size, R, to the original

void size, R0 with a modified version of the Rice and Tracy model for monotonic loading

(Equation 2.2.10)
52

R
p p
np n np n
cyclic ln n exp 1.5T d p exp 1.5T d p 0 (2.2.10)
R0 0 np1

T 0

0 np1

T 0

For fracture to occur, the void growth demand, cyclic, should exceed the void growth capacity or

critical void size. Under cyclic loading, the monotonic ductility measure, , decays according to a

damage law, which depends on another material parameter, . Thus, the fracture criterion

according to the CVGM can be expressed as

Rcrit
cyclic ln
exp
p*
(2.2.11)
R0
The ratio of cyclic at any time in the loading history to the critical void size expresses how close

the material is to ductile crack initiation. Thus, a low ratio would suggest that the material is safe

from fracture, while a ratio closer to unity indicates a higher probability of ductile fracture at that

point. For details of the expressions in Equations 2.2.10 and 2.2.11, refer Kanvinde et al (2004

and 2007).

Unlike traditional fracture mechanics, the above fracture imitation model is not limited by the

assumption of a preexisting crack or imperfection and is valid in the presence of large-scale

yielding - often the case in regions of plastic hinge formation. Moreover, the proposed fracture

index is not directly dependent on geometry or rain-flow counting schemes to simplify an

earthquake loading history into standard cycles. Of course, if modeled correctly, the resulting

stresses and strains of the structure will be indirectly related to the geometry and loading history.

Most importantly, this fundamental approach offers important insights into localized effects that

cause fracture, which greatly improves design intuition and extends the results of this study to

situations beyond those directly tested.


53

As Chapter 5 presents this methodology in greater more detail, the use of the model will only be

highlighted in the current discussion. As an example, Figure 2.11 and 2.14 will be used to

illustrate the fracture prediction at a critical node for an HSS4x4x1/4 bracing member under a far-

field loading history (symmetric tension and compression excursions). Illustrated in Figure 2.14a

and 2.14b are Equations 2.2.10 and 2.2.11 at the critical node (continuum point which is predicted

to fracture first) shown in Figure 2.11 of the brace during the loading history shown in Figure

2.14c. Referring to the figure, it is apparent that elastic behavior is observed prior to cycle 24

after which point the bracing member buckles globally. While the brace is far from fracture

initiation, cycle 24 is the first sign of inelastic behavior both experimentally and analytically.

Local buckling is observed at approximately the same axial deformation during the first large

compressive pulse for both the experiment and ABAQUS simulation. As the plastic strain is

significantly amplified during local buckling, the critical parameter that drives fractures is the

plastic strain, and less so the triaxiality. Local buckling induced damage significantly decreases

the capacity in Figure 2.14a (or critical void size) and leads to a sharp increase in the demand to

capacity ratio (Figure 2.14b). Several cycles after local buckling initiation, the demand/capacity

ratio reaches unity and is predicted to fracture at that point in the loading history. From a

comparison with the experimental results in Figure 2.14c, it can be seen that the prediction is

quite close to the actual instance of fracture.

2.3 SUMMARY

The experimental results presented in this chapter (and later in Chapter 4) suggest that bracing

members are prone to fracture at the middle plastic hinge region from the increased plastic strain

accumulation during local buckling. Relative to current AISC (2005) limits, braces with low

slenderness (KL/r) and high width-thickness (b/t, D/t, or bf/2tf) ratios tend to fracture earlier in a

cyclic loading history compared to more slender and compact members which delay local

buckling. A large majority of research has been conducted to synthesize the combined affects of
54

slenderness, compactness and material type on brace fracture ductility; but most of the proposed

relationships have been of an empirical nature. These relationships often assign a cumulative or

maximum axial deformation capacity, which corresponds well to the scale of past analysis

techniques primarily aimed at characterizing force-deformation response. Only recently have

advances in computational power allowed researchers to simulate full-scale structures and

components directly incorporating response at the cross-section and even continuum-level.

Modeling approaches that leverage these advances in computational power may offer improved

and general fracture predictive methods. Thus, in addition to these modeling techniques (i.e.

sophisticated continuum FEM), there is a need for suitable stress/strain based fracture criteria to

accurately evaluate the complex interactions of stresses and strains at the continuum level

resulting in fracture. Recently, novel micromechanics-based model to predict Ultra Low Cycle

Fatigue (ULCF) have been developed by Kanvinde and Deierlein (2007) to address this need.

The work described in this dissertation develops a methodology that uses sophisticated FEM

simulation of analyze large-scale bracing components (incorporating complex events such as

local buckling) in conjunction with the ULCF model to predict fracture. The ULCF fracture

model operates at the continuum level and is derived from monotonic void growth and

coalescence fracture models for ductile metallic materials. The model relies on accurate

simulation of the stress and strain histories for small-scale calibration and large-scale brace

experiments. Especially important in the context of brace modeling is simulating local buckling-

which leads to significant plastic strain accumulation at the critical cross-section. Thus, the

dissertation will discuss the simulation of local buckling in detail by considering cross-section

geometry, member and section imperfections, as well as the strain hardening properties of the

material.
55

While the large-scale brace component tests and complementary continuum analyses provide a

rigorous test bed to evaluate the general, physics-based fracture prediction methodology, the

simulations also provide insights into localized brace behavior, which may be invaluable in the

context of calibrating or developing macro models such as described in preceding sections.

Finally, it is hoped that these advances in modeling will reduce the reliance on experiment-based

research and provide a useful research tool for studying design requirements of fracture-critical

structures. As discussed earlier, one of the most important advantages offered by these ULCF

models is the insight into localized effects, and their relation to global parameters that are used to

inform design and detailing considerations. For example, numerical parametric studies may be

used to develop insights into the combined influence of slenderness and width-thickness ratios on

brace ductility. The continuum-based models can extend and generalize experimentally observed

trends to untested parameter sets.


56

Table 2.1: Maximum story drifts (standard deviation) from nonlinear time-history analyses on 3
and 6-story SCBF systems.
Drift 10% in 50 years 2% in 50 years
Investigation Software/Brace Model
Value 3-story 6-story 3-story 6-story
Sabelli 3.9 1.8 SNAP-2D/Rule-
Mean NA
(2001) (3.1) (0.8) based*
1.5 1.4 5.7 5.1 OpenSEES/Fiber-
(0.9) (0.8) (3.0) (3.4) based
Uriz (2005) Median
1.6 1.1 5.7 4.4 OpenSEES/Fiber-
(0.9) (0.6) (2.4) (2.2) based*
McCormick 3.6 2.0 8.1 4.7 OpenSEES/Rule-
Mean
et al (2007) (1.6) (0.7) (3.0) (2.7) based
*Incorporated fatigue-fracture law in bracing modeling

Table 2.2: Design level (approximately 10% in 50 years) median (standard deviation) and
maximum story drift from Izvernari et al (2007)
Number Median Story Maximum
Investigation of Drift Story Software/Brace Model
Stories Drift
2 1.1 (0.4) 3.5
4 1.1 (0.6) 3.4
Izvernari et al
8 1.4 (0.8) 4.2 OpenSEES/Fiber-based
(2007)
12 1.6 (0.9) 4.6
16 1.8 (0.4) 2.7

Table 2.3: Results from 6-story braced-frame shake-table test (Tang, 1987)
HSS Story (b-3t)/t KL/r Max Fracture Drift (Def)** Damage
Shape Drift* (%) (%, range/Y)
4x4x3/16 6 20 79 0.6 NA Minor buckling
5x5x3/16 5 26 62 1.7 1.1 (7.9) Fracture initiation (N brace)
5x5x1/4 4 19 63 1.6 1.2 (9.0) Fracture initiation (N brace)
6x6x1/4 3 23 60 2.5 1.6 (9.0) Strength loss (N brace)
1.4
6x6x1/4 2 23 60 2.1 Fracture initiation (both)
(8.6N, 7.7S)
6x6x1/2 1 10 64 1.0 NA Buckling
*Maximum drift recorded during earthquake record
**Maximum drift and maximum deformation range recorded prior to fracture initiation
57

Table 2.4: Summary of HSS experimental review (63 tests)


Year Average No. of General Description of
Test Program [No.]
Published Fy,meas Tests Cyclic Loading History
Gugerli and Goel [1] 1982 60 4 Unsymmetric compressive
Liu and Goel [2] 1987 54 3 Unsymmetric compressive
Lee and Goel [3] 1988 67 6 Unsymmetric compressive
Archambault, [4] 10 Standard symmetric
1995 57
Tremblay, Filiatrault 4 Unsymmetric
Walpole [5] 1996 56 3 Standard symmetric
Shaback and Brown [6] 2003 64 8 Standard symmetric
3 Standard symmetric
Yang and Mahin [7] 2005 60
1 Unsymmetric tensile
Fell, Kanvinde, and 4 Standard symmetric
2006 69
Deierlein [8] 1 Unsymmetric compressive
Han et al [9] 2007 59 4 Standard symmetric
Yoo, Lehman, and
2008 67 12 Standard symmetric
Roeder [10]

Table 2.5: Summary of calibration constants to predict cyclic fracture life of square and
rectangular HSS bracing members
Model Investigation C1 a1 a2 a3 a4 C2 (KL/r)crit FY
1 Lee and Goel 32.68 0.6 -0.8 1.0 0 0.25 NA 46
2 Archambault et al 0.124 0.6 -0.25 0.8 2.0 0.25 70 46
3 Shaback and Brown 0.240 -1.75 -0.6 0.55 2.0 -0.125 70 51
Tang and Goel* 16.19 0 -1.0 1.0 1.0 0 60 NA
Shaback and Brown** 5.11 0.51 -1.25 0.55 2.0 -0.125 70 51
*Used number of cycles to failure, Nf, instead of (typical) cumulative normalized deformation, f,pred
**Used f,pred = (compression + tension) instead of f,pred = (0.1compression + tension)
58

Figure 2.1: Large-scale Special Concentrically Braced Frame test (Uriz, 2005) and local buckling
induced brace fracture at the middle plastic hinge.
59

(a)

Compression (out-
Tension of-plane buckling)

(b)

Compression (out-
of-plane buckling Tension
and plastic hinge
kinking)

(c)

Figure 2.2: (a) Chevron braced-frame story and typical connection details, (b) Out-of-plane
buckling and tension yielding and (c) Brace plastic hinge formation.
60

Net section
fracture

(a)

Gusset-beam
fracture

Shear tab-
column fracture

(b)

Figure 2.3: (a) Brace gusset-plate net section fracture and (b) Column fracture at base of shear tab
along with beam fracture from prying action of gusset plate.
61

Figure 2.4: Plastic hinge formation in beam from bracing force imbalance.
62

Plastic Plastic
See below
Hinge Hinge

L1 L2 > L1

Increasing global slenderness (KL/r)

Increasing brace ductility

Increasing compactness (b/t, D/t, bf/2tf)

H
b B 3t HSS

t t HSS
t1-HSS t2-HSS > t1-HSS
B

D
D
t Pipe

t1-Pipe t2-Pipe > t1-Pipe

bf
bf
2t f
t1-f t2-f > t1-f

Figure 2.5: Influence of brace geometry in terms of global slenderness and cross-section
compactness on fracture ductility.
63

0.06
Interstory Drift, (rad)

max = 0.048

0.03

min = 0.019 (0.4max)


-0.03
Figure 2.6: Story drift time history results for the first story of a 3-story SCBF, 2% in 50 years
event.
64

0.06

Symmetric
0.04 Behavior

min = 0.37max
min

0.02

0
0 0.02 0.04 0.06 0.08
max
(a)

0.06

Symmetric
0.04
Behavior
min

0.02 min = 0.42max

0
0 0.02 0.04 0.06 0.08
max
(b)

Figure 2.7: Minimum versus maximum story drifts for (a) 3-story and (b) 6-story SCBF (courtesy
Uriz, 2005).
65

P/PY (b)

1/3 D
(c)
C /Y
F
B

A
compression tension (d)

(a)

(e)

Figure 2.8: (a) Schematic, one cycle, force-deformation response of a typical brace component.
Typical progression of brace damage: (b) Global buckling (point A), (c) Local buckling, (d)
Fracture initiation and (e) Loss of tensile strength.
66

300

150

Force (k)
0

-150
range
(a)
-300
-2 -1 0 1 2
Axial Displacement (in)

25
Pipe (0.2)
Pipe (0.3)
20 HSS (0.53)
HSS (0.62)
15 WF (1.0-1.8)
range/Y

HSS (0.6)*
HSS (0.54)**
10

5 *Shake-table
**Static
(b) data
SCBF test
0
0.0 1.0 2.0
Width-thickness Ratio/AISC Limit

25

20
**Static
15 SCBF test Pipe (0.3)
range/Y

Pipe (0.4)
Pipe (0.5-0.6)
10 HSS (0.9)
HSS (1.1)
5 *Shake-table WF (0.9-1.6)
(c) data HSS (1.2-1.6)*
HSS (0.9)**
0
0.0 1.0 2.0
Slenderness Ratio/AISC Limit

Figure 2.9: (a) Typical brace hysteretic response and definition of range, (b) Influence of Width-
thickness and (c) Global slenderness ratio on normalized axial deformation range, range/Y, for
HSS (experiments from Shaback and Brown, 2003), Pipe (Elchalakani et al, 2003) and Wide-
flanged (Gugerli and Goel, 1982) members. Also shown are large-scale shake-table and quasi-
static brace ductilities from Tang (1988) and Uriz (2005), respectively. Note: legends for (b) and
(c) list the corresponding slenderness and width-thickness ratios, respectively.
67

Elastic Element

P
P

(a) (b)

Integration Points


60

40

20


-4.00E-03 -2.00E-03 0.00E+00 2.00E-03 4.00E-03 6.00E-03

-20

-40

-60

(c) (d)

Figure 2.10: Brace modeling techniques in order of increasing localized simulation abilities: (a)
Phenomenological (or rule-based) model, (c) Lumped plasticity element, (c) Fiber-based model
and (d) Continuum model.
68

Critical Node

Figure 2.11: Buckled shape, plastic strain contours and critical fracture node (as determined by
section 2.2.2.3 and Figure 2.14) from continuum HSS4x4x1/4 brace analysis.

6
Archambault et al (1995)
5 Shaback and Brown (2003)
4 Han and Kim (2007)
f,exp/f,pred

0
0 1 2 3 4 5 6
Model Number

Figure 2.12: Ratio of experimental to predicted fracture deformation for three separate testing
programs. The five empirical-based fracture models are summarized in the text.
69

Void Nucleation Necking Between Voids

Void Growth and Strain Void Coalescence and


Localization Macroscopic Crack Initiation

Figure 2.13: Micromechanical process of ductile fracture in steel


70

6
Void Size and Critical Void Size

Monotonic CVGM
Capacity () Cyclic Capacity
Prediction
(Eqn. 2.2.11)

Fracture Index
4 1
CVGM
Prediction

2
Demand, cyclic
(Eqn. 2.2.10)
(a) (b)
0 0
23 25 27 23 25 27
Cycle Number Cycle Number

4
Experimental Fracture

2 Predicted Fracture
(in)

-2

(c)
-4
20 22 24 26 28 30
Cycle Number

Figure 2.14: (a and b) CVGM fracture prediction and (c) Comparison to experimental fracture
time for the critical node of an HSS4x4x1/4 shown in Figure 2.11 during a standard loading
history.
71

Chapter 3
Large-Scale Brace Component Tests

3.1 MOTIVATION FOR LARGE-SCALE BRACE TESTS

Concentrically braced steel frames (CBFs) are attractive lateral load resisting steel-framed

systems due to their economy, structural efficiency and high stiffness. However, as indicated by

several recent studies (outlined in the previous chapter), braced frames are vulnerable to

premature fracture during earthquakes due to the interactive effects of overall flexural buckling

combined with concentrated local buckling in the plastic hinge that forms near the midpoint of

the brace. Connections between the braces and frame are also prone to fracture, prompting

proposed provisions (AISC 2005, and Yang and Mahin 2005) to mitigate this through connection

detailing that accommodates brace end rotations and provides reinforcement to avoid net section

fracture. Nevertheless, because braced systems rely on cyclic inelastic buckling of braces for

energy dissipation, buckling-induced brace fracture has a significant effect on overall system

ductility.

As discussed in Chapter 2, previous studies (e.g. Jain et al. 1978, Popov and Black 1981,

Tremblay 2000 and 2002, Shaback and Brown 2003, Lee and Bruneau 2005, Han et al 2007)

have examined the effect of various parameters, such as brace slenderness, compactness, and
72

cross-section shape, on fracture ductility and energy dissipation in braces. Qualitatively, the

studies concur that cross-sectional shapes, slenderness ratios and width-thickness ratios most

strongly affect the fracture ductility of bracing elements. Tremblay (2002) summarizes most of

these studies and found local buckling to be more severe with smaller brace slenderness, even

with compact cross-sections. Studies have also shown that more compact cross-sections tend to

have higher ductility than less compact sections. Furthermore, Tremblays review (2002) suggests

an important loading history effect, such that braces loaded with asymmetric compression cycles

are more prone to fracture as compared to those subjected to symmetric cycles of

tension/compression loading. However, data on these effects is sparse and specific quantitative

criteria to relate these parameters to brace performance is lacking, particularly in the case for Pipe

and Wide-Flange brace members.

With the goal of developing improved understanding of brace buckling and fracture, this study

involves tests of nineteen large-scale braces conducted as part of a Network for Earthquake

Engineering Simulation and Research (NEESR) project. The test specimens were subjected to

reversed-cyclic loading histories with large deformation amplitudes. The specimens are

approximately two-thirds of full-scale, as compared to braces used in typical buildings, with end

connections that represent the flexibility of commonly used gusset plate connections. The cross

sections investigated include square HSS4x4x1/4 and HSS4x4x3/8 sections, standard pipe

sections Pipe3STD and Pipe5STD, and a Wide-Flange section W12x16. These tests examine

the effects of section compactness, section geometry, loading histories, loading rates, and grout

fill in the HSS members.

The tests presented in this chapter and Chapter 4 are dual purpose tests that provide valuable data

regarding the seismic performance of braces, while serving as an evaluation test-bed for the

micromechanics-based fracture modeling methodology introduced in Chapter 2. This chapter


73

presents the test program and setup, whereas Chapter 4 focuses on the practical implications of

the tests. Based on the test results, subsequent chapters will address the fracture modeling aspects

of the study. Moreover, the results presented in this chapter (and Chapter 4) will be revisited in

Chapter 6, in the context of a numerical parametric study based on the validated fracture models.

3.2 EXPERIMENTAL SETUP

The tests were conducted at the UC Berkeley NEES facility located at the Richmond Field

Station. The facility offers state-of-the-art testing resources and versatility with respect to the

application of boundary conditions, forces, and loading rates.

As shown in Figure 3.1, the brace test rig provides a fixed-fixed boundary condition, where one

end of the brace specimen is bolted directly to a large reaction block and the other end is attached

to a moving cross-beam. The entire setup was attached to the strong floor and stood

approximately three feet high. The connection gusset plates are oriented in the vertical plane, thus

permitting buckling in the horizontal plane with an effective buckling length roughly equal to

brace length. Load is applied through two servo-hydraulic actuators, each with a 220 kip force

capacity and a +/- 10 inch stroke capacity. The tests were performed in displacement control with

the actuators set in a master-slave feedback-control to minimize end rotations and maintain a

fixed boundary condition at the translating end. The axial brace deformation, a, is measured as

the relative deformation between the two specimen end plates.

3.3 TEST PROGRAM SCOPE

Current seismic design standards (AISC 2005) distinguish between Ordinary Concentrically

Braced Frames (OCBF) and Special Concentrically Braced Frames (SCBF), where the latter have

slightly more stringent requirements. With regard to the bracing member provisions, the AISC

requirements for HSS and Pipe sections are similar for OCBFs and SCBFs. Both employ the
74

same section compactness limits for square and rectangular HSS members

( b / t 3 0.64 E / Fy ), round Pipe ( D / t 0.044 E Fy ), and W-shape braces

( b f / 2t f 0.3 E Fy ). Likewise, the overall brace slenderness limits for both system

designations are similar ( LB / r 4. E Fy ), excepting more slender braces that are permitted in

certain SCBF and OCBF systems. While the results in this dissertation are generally applicable

for both types of braced frames, the design comparisons in this paper are made in the context of

SCBF systems since these systems are prevalent in regions with high seismic activity where

design level events are expected to induce inelastic brace buckling. Moreover, design provisions

anticipate that OCBF systems will experience smaller deformation demands as compared to

SCBFs, and thus the large inelastic deformations reported in this study may not reflect expected

demands in OCBF systems.

Summarized in Table 3.1 is the test matrix for the nineteen specimens, including information on

the brace cross sections, details, and loading variables. To reflect common design practice, the

test matrix and various component details were developed in consultation with the Structural

Steel Educational Council and practicing engineers at Rutherford and Chekene Structural

Engineers. The test plan is organized to provide insights into a variety of design parameters that

affect brace buckling and fracture. The testing program includes eight square HSS specimens,

eight pipe specimens, and three wide-flange specimens. The member sizes were selected to

investigate the effect of slenderness (LB/r) and width-thickness (b/t and D/t) ratios. For example,

the two HSS sections have similar overall slenderness ratios and thus provide a comparative

assessment of the influence of width-thickness ratios. The alternative Pipe sections and W-shape

allow for an assessment of slenderness effects combined with section properties.


75

The section compactness and member slenderness ratios are summarized in the fourth and fifth

columns of Table 3.1. The numbers in parentheses are ratios of the specimen properties to the

AISC (2005) limits (listed previously) for SCBF systems. The section compactness and member

slenderness ratios for HSS and Pipe sections are all well within the AISC limits. On the other

hand, the flange compactness of the W-shape is close to the AISC limit and the weak-axis

(governing) slenderness ratio exceeds the AISC limit by about 60%. The tests were conducted

using one of three alternative loading protocols listed in Table 3.1 (distinguished by the crossed

cells for each test) and these are summarized later. Loading rates were quasi-static except for two

tests (HSS1-3 and HSS2-2), which were tested under faster earthquake loading rates.

The dimensions of the test specimens are shown in Figures 3.2 and 3.3 with the corresponding

measured material properties summarized in Table 3.2. All specimens are 10-3 long, measured

from the outside of each end plate, including a clearance of 1.5 between the end of the brace and

bolted end plate to accommodate end rotations associated with brace buckling. The gusset plates

are designed to resist buckling in compression (Astaneh-Asl 1998) and yielding in tension

(Whitmore 1950). Design forces, equal to the expected brace yield strength (RyFyAg), are used to

design all tension critical components including welds. The AISC (2005) specified material yield

and ultimate strengths (Fy, Fu ) and expected strength factors (Ry, Rt) are also listed in Table 3.2.

As shown in Figure 4.2 and Table 3.1 (with crossed cells in the Reinf. column), most of the

HSS and Pipe specimens had reinforcement to preclude net-section fractures at the connection.

Reinforcement was omitted from four pipe tests (P1-2, P1-4, P2-2 and P2-4) to investigate its

relative importance for different loading histories. Two of the HSS braces (HSS1-4 and HSS1-5;

designated with crossed cells in the Fill column) were filled with grout to assess whether the

fill would delay local buckling and subsequent fracture initiation. One of the specimens (HSS1-6)

featured 18x2x0.5 inch reinforcing plates welded at the center of its non-buckling faces (see
76

Figure 4.7). These plates were intended to encourage unsymmetric buckling of the brace (with

respect to the length), as would be produced due to dissimilar end conditions.

3.3.1 Cyclic Loading Protocols

In contrast to moment frame systems, where peak seismic story drift demands are fairly stable

with respect to design variables (Gupta and Krawinkler 1999), drift demands for braced frames

are more sensitive to variations in bracing configurations and highly nonlinear brace buckling

behavior. For example, Tremblay (2000) showed that the slenderness ratio of the bracing

elements can have a significant influence on drift demands. The loading protocols for this study

are based on an adaptation of protocols for moment frame systems that are adjusted for

earthquake loading demands in braced-frames.

As shown in Figure 3.4a, the standard cyclic protocol follows a symmetric loading history that

represents demands imposed by far-field (non near-fault) earthquake ground motions. The two

other loading protocols, shown in Figure 3.4b-3.4c, represent non-symmetric pulse type demands

as might be imposed by near-fault ground motions. These are distinguished between pulses

dominated by compression or tension loading of the brace. The loading histories are expressed in

terms of building story drift demands, where the drift ratio is assumed to be 0.2% at brace

buckling and 4% at the Maximum Considered Earthquake (MCE) demand. These displacement

demands are based on analyses of chevron configuration braced frames by the authors and others

(Izvernari et al 2007, McCormick et al 2007, Uriz and Mahin 2004, and Sabelli 2001, Uriz 2005

,McCormick et al 2007). Referring to the previous chapter, some of these investigations (Uriz

2005 and McCormick et al 2007) have shown that drift demands may exceed 4% in SCBF

systems for MCE-type events. For example, Uriz (2005) reports median drifts of 5.7 and 5.1% for

3 and 6-story SCBF frames, respectively, during ground motions with a 2% probably of

exceedance in 50 years (referred to as 2/50 ground motions). This type of MCE demand is also
77

consistent with the capacity of SCBFs as implied by the commentary to the AISC (2005) seismic

provisions, which state that braces could undergo post-buckling axial deformations 10 to 20

times their yield deformation. Assuming a system yield level drift of approximately 0.3% to

0.5%, the AISC statement may be conservatively interpreted as desiring a drift capacity of

approximately 3% to 5%.

3.3.1.1 Standard Cyclic Loading Protocol

The far-field (or general) loading history was developed by adapting one from ATC-24 (ATC,

1992) to represent SCBF behavior. This protocol is based on nonlinear time history investigations

by Gupta and Krawinkler (1999), who demonstrated that the dissipated energy demands that

result from the testing protocol are consistent (under reasonable assumptions) with realistic

seismic demands in ductile moment frames. The authors modified the moment frame loading

protocol to braced frames using concepts outlined in Krawinkler et al (2000).

Table 3.3 outlines the original ATC/SAC loading protocol. The protocol is defined in terms of

cycles of story drift angles of successively increasing magnitudes. As shown in the figure the

loading history consists of three increasing sets of six cycles ( = 0.00375, = 0.005, and =

0.0075) followed by four cycles at the approximate yield drift of a moment frame (Y = 0.01), and

four progressively increasing sets of two cycles each with the fourth set corresponding to the

Maximum Considered Earthquake MCE level ( = 0.015, = 0.02, = 0.03, and MCE = 0.04).

The modified ATC/SAC far-field protocol used in the current study for SCBF systems is also

listed in Table 3.3 and illustrated in Figure 3.4a. Referring to Table 3.3, the four cycles at the

MRF yield level (1% drift inelastic cycle group 0 in Table 3.3) are scaled to coincide with the

onset of inelasticity in an SCBF system, typically the buckling of the brace. Load steps 1-3 ( =
78

0.00375, = 0.005, and = 0.0075 from the original history) are scaled using the same factor.

The intent of this modification is to ensure a relatively consistent number of inelastic damaging

cycles between the modified and original ATC/SAC protocols. The justification for maintaining a

similar number of inelastic cycles for the modified history is based on the observations that (1)

once a structure begins to yield, the period elongates so that the demands are more ground motion

dependent rather than structure (initial stiffness) dependent and (2) recent research (Uriz and

Mahin, 2004 and McCormick et al, 2007) suggests that the MCE story drift level for SCBFs is in

the 3-5% range, which is comparable to that for MRFs. Based on this reasoning, scale factors

were developed that allowed the inelastic cycle set to increase such that (1) the number of

inelastic cycles would be preserved between the ATC/SAC and the new protocol and (2) the

largest cycles would reflect a drift level consistent with the MCE ATC/SAC protocol.

The scaling method for each group of cycles is based on a rationale similar to that used by

Krawinkler et al (2000), which assumes the cumulative drift as a measure of system damage.

Following this rationale, the deformation amplitudes and numbers of cycles in the modified

loading protocol were adjusted such that the relationship between peak drift and cumulative drift

would roughly approximate the trends (for moment frames) outlined by Krawinkler et al (2000),

and then applied to the SAC protocol. Adjusting the protocol in this way enables a convenient

interpretation of the test results, such that at any drift amplitude, the damage to the brace may be

assumed consistent with the damage induced by a ground motion that produces a similar drift

amplitude in the system.

This scaling procedure is schematically illustrated in Figure 3.5a, and involves two steps. First,

the inelastic cycle groups 0 and 4 are fixed at 0.2 and 4.0%, respectively. Next, the slope of the
79

line connecting groups 0 to 3 is modified (through trial and error) to produce the effect described

in the previous discussion.

It is important to discuss that the protocol is based on the assumption that story deformation

histories in moment frames and braced frames are similar. However, nonlinear time history data

from some studies (Uriz, 2005; refer Chapter 2) on braced frames indicate that unlike moment

frames, the unsymmetric response of the braces may result in a ratcheting response of the braced

frames. Thus, the peak drift may be significantly larger than half the drift amplitude, which

contradicts the implicit assumption of the symmetric loading protocol. A systematic consideration

of such issues will require detailed analysis of several nonlinear time history simulation histories

(in an exercise similar to that performed by Krawinkler et al, 2000 for moment frames), and is

outside the scope of this dissertation. As discussed subsequently in Chapter 6, the test results

themselves may be interpreted in a more appropriate manner, if such an analysis becomes

available at a later time.

3.3.1.2 Near-Field Loading Histories

To reflect demands imposed by near-fault ground motions, two loading protocols asymmetric

compression and asymmetric tension were used for several of the brace tests. In contrast to

moment frames where the component response is fairly symmetric, for braced frames, the pulse

like near-fault protocol must be distinguished between cases dominated by either tension or

compression response. As with the general protocol (described previously) the near-field protocol

is based on a similar one developed in the SAC project for moment frames. These loading

protocols are illustrated in Figures 3.4b and 3.4c.

As shown in Figure 3.4b, the compression dominated history is identical to the ATC/SAC near-

fault protocol. However, following completion of the near-field protocol, the far-field loading
80

protocol (of Figure 3.4a) is appended so as to extract additional information from the test in the

event that the brace survives the near-fault loading. Aside from providing data for validating the

ductile fracture models, this subsequent loading is envisioned to represent an aftershock

earthquake that follows the first large pulse of the main earthquake fault rupture.

The tension dominated history (Figure 3.4c) consists of a large monotonic pull followed by

subsequent cycles. For tension dominant loading, where one of the goals is to apply large tension

demands on the brace and connections prior to buckling, the initial negative (compression) pulse

is omitted and the positive (tension) pulse is increased to +8% drift, to impose the largest tensile

demands on the specimens given the limitations of the test setup. In anticipation of instances

where the brace may not fail during the near-fault pulse loading, the standard cyclic history is

appended to the pulse protocol.

3.3.1.3 Story Drift and Brace Axial Deformation Relationships

The drift demands in the loading protocols are converted to corresponding axial deformations (for

application in the experiments) through the following relationship:


cos 45 sin 45 L B 0.5L B (3.1)

Where is the story drift angle (expressed in radians), a is the corresponding axial deformation

and LB is equal to the distance between the fold lines of the gusset plates (9-9 for HSS; 9-10

for Pipe and W specimens). This relationship is based on a chevron brace configuration

shown in Figure 3.6, assuming center-line dimensions with the braces inclined at 45 degrees and

ignoring flexural deformations in the beams and columns. Given that axial deformations are

applied to the experimental brace specimens, Equation 3.1 provides an approximate measure to

compare the various brace capacities to earthquake-induced story drifts and does not take into

account all affects. For instance, Equation 3.1 is based on frame geometry that considers center-
81

line dimensions of the chevron frame, neglecting the joint size. This effect may be included in

Equation 3.1 by assuming a rigid-link distance between the gusset-plate fold lines and the

working point of the beam-column connection. Following this reasoning, a modification factor

(1+C) may be introduced in Equation 3.1, resulting in Equation 3.2, where C is the ratio of the

rigid-link length (on both ends of the brace) to the brace length LB.

a 1 C cos sin L B (3.2)

Owing to the wide-variety of brace, gusset-plate configurations in SCBF construction, it is

difficult to prescribe a consistent or precise value for the modification factor C. Recognizing the

uncertainty in other aspects of this kinematic relationship (such as the brace angle ), and

moreover the subjectivity in the characterization of the drift demands themselves (refer earlier

discussion), this chapter and the next relies on Equation 3.1 to relate the brace axial deformation

to a corresponding drift level. In the presence of these uncertainties, relationships such as the one

presented in Equation 3.2 may be used to interpret the data presented in this paper in the context

of specific frame designs, or for examining the sensitivity of the findings to other geometrical

parameters such as the connection size or brace angle.

3.4 Instrumentation and Miscellaneous Testing Results

This section describes the instrumentation used for the brace experiments. In addition to the

primary channels of load and deformation (which are discussed in detail in Chapter 4), this

section describes various ancillary measurements.

The primary measurements for each test include the total axial force resistance of the brace,

obtained from the readings of both actuators shown in Figure 3.1a, and the axial deformation

measured across the length of each brace. The latter was inferred using four displacement

transducers, one each on the top and bottom of each brace end plate. The transducers on the west-
82

end (sliding end) of the experimental setup in Figure 3.1 had a range of +/- 20 inches while the

fixed-end (east) transducers had a range of +/- 1 inch. Representative recordings for each of these

instruments are shown in Figure 3.7a-3.7b for an HSS4x4x1/4 bracing member subjected to the

standard cyclic loading history (HSS1-1). As can be seen from the figure, the deformations on the

east end are negligible compared to the sliding end. However, the figure indicates a slight

discrepancy between the top and bottom recordings at each end, suggesting a slight rotation

during large tensile excursions. Attributed to construction imperfections, these deformations are

small a maximum difference of approximately a tenth of an inch on the west end and a

hundredth of an inch on the east and are not considered to have a substantial affect on the

testing results. Lateral slip of the brace end plates (North-South movement of end plates in Figure

3.1a) is also measured with +/- 1 inch position transducers. These measurements are shown to be

small (on the order of 0.02 inch) in Figure 3.7c with the maximum lateral slippage recorded at the

inelastic tension peaks on the sliding end of the experimental test setup.

The axial brace deformation was determined from the difference between the average (of the top

and bottom) of the east and west position transducers. Thus calculated, the relative deformation

was used to trigger actuator reversal during the experiment. In contrast to direct control of the

actuator displacement (which also includes deformation of the test-setup, this procedure resulted

in a highly accurate application of the loading history as illustrated in Figure 3.8. From the figure,

it is observed that the intended and measured deformation histories are identical.

The total axial force was calculated by adding the two actuator force measurements. Figure 3.9

illustrates the axial deformation-force relationship for each actuator during the standard cyclic

loading test on the HSS4x4x1/4. As shown in the figure, the force measurements from the two

actuators are similar throughout the experiment and demonstrate the symmetry of the loading

frame (and boundary conditions) and high-quality fabrication of the experimental specimens.
83

A series of cable-extension position transducers (string pots) were used to measure the large out-

of-plane buckling displacements of the bracing members. The string pots were mounted on the

floor of the lab and attached at the brace third points as depicted in Figure 3.10. At the midpoint,

two string pots were needed to track the position of the brace point in space as the axial

deformation, from symmetry, is approximately half of the end displacement at this point. A

schematic illustrates this configuration in Figure 3.10b. At the other points, three pots were

required to track the position of the brace, where the additional device is mounted along the axis

of the bracing member. Before each test, the initial locations of the brace third points are

measured with respect to each string pot connected at that point. With these measurements, a

fairly simple program can track the position of each point given the recordings of the string pots.

Figure 3.12 shows the out-of-plane measurements at each third-point versus the brace axial force

for an HSS4x4x1/4 during a standard loading history (HSS1-1). As expected, the mid-point out-

of-plane deformations are larger than the third-points.

Thermocouples were used on several of the experiments to ascertain the influence of rate affects

on steel fracture behavior. It was found that rate affects did not significantly alter the behavior or

performance of the bracing members as compared to the slower quasi-static experiments. The

results of the faster rate experiments and a discussion on rate-induced temperature increases in the

context of ductility and fracture predictions are presented in more depth in Chapters 4 and 5,

respectively. Figure 3.12a illustrates the temperature increase due to inelastic buckling and

tension yielding for an HSS4x4x1/4 brace during the quasi-static near-field compression loading

history. The maximum recorded temperature occurs on the compression side of the brace during

the first inelastic excursion to -6.0% drift. Note the cooling period during the elastic cycles

between the end of the near-field history and the appended far-field history. Figure 3.12b shows a

significantly larger temperature increase, reaching a maximum of nearly 200F, recorded for an

HSS4x4x3/8 brace during an earthquake-rate.


84

3.5 Summary

To provide background for the upcoming chapters, this chapter introduced the test matrix, setup

and loading histories applied to nineteen large-scale bracing component members, representative

of members in Special Concentrically Braced Frame (SCBF) systems. The tests, designed to

represent typical conditions in steel braced frames, complement previous studies by investigating

three cross section types (HSS, Pipe and Wide flanged sections) of varying section compactness

and brace slenderness. The test program also investigates the effects of loading histories, loading

rates, connection reinforcement, and grout filling of HSS sections. Results presented in the

subsequent chapters will heavily reference the background information presented in this chapter.
85

Table 3.1: Test parameters and loading histories


Test Specimen Parameters Loading Protocol # #
Test Bracing b/t or D/t KLB/r Reinf. Fill Std. Pulse Pulse
LB
No. Member (AISC**) (AISC**) FF NF-C NF-T
HSS1-1
HSS1-2
14.2 77
HSS1-3*
HSS4x4x1/4 (0.89) (0.77)
HSS1-4 9-
HSS1-5 9
HSS1-6#
HSS2-1 8.5 80
HSS4x4x3/8
HSS2-2* (0.53) (0.80)
P1-1
P1-2 21.6 63
Pipe5STD
P1-3 (0.59) (0.55)
P1-4
P2-1
9-
P2-2 16.2 102
Pipe3STD 10
P2-3 (0.44) (0.89)
P2-4
W1
7.5 153
W2 W12x16
(1.04) (1.59)
W3
*fast (earthquake) loading rate

**the numbers in parentheses are the section width-thickness or brace slenderness ratio
normalized by the criteria specified by the AISC Seismic Provisions (2005)

#HSS1-6 was reinforced at the midpoint (top and bottom) with 18 long plates (see Figure 4.7)

# #FF: Standard far-field loading (Figure 3.4a); NF-C: Near-field compression loading (Figure
3.4b); NF-T Near-field tension history (Figure 3.4c)

Table 3.2: Brace material properties


Measured Specified
Brace Fy ,meas Fu , meas
Properties Properties
Cross-
Fy,meas Fu,meas Fy Fu Fy Fu
Section Ry Rt
(ksi) (ksi) (ksi) (ksi)
Corner: Corner: Corner: Corner:
HSS1* 74 80 1.6 1.4
Center: Center: Center: Center:
67 71 1.5 1.2
46 58 1.4 1.3
Corner: Corner: Corner: Corner:
HSS2* 73 78 1.6 1.4
Center: Center: Center: Center:
72 79 1.6 1.4
P1 47 61 1.4 1.0
35 60 1.6 1.2
P2 54 67 1.5 1.1
W 60 69 50 65 1.1 1.1 1.2 1.2
*center material properties are from ASTM specified rectangular tensile coupons. All other data
are from non-ASTM cylindrical coupons
86

Table 3.3: Summary of loading protocol modifications


Inelastic Number Original SAC Modified SAC
Group of Loading History Loading History
Number Cycles Drift (%) Drift (%) a (in)
6 0.375 0.08 0.04
elastic
6 0.50 0.10 0.06
loading
6 0.75 0.15 0.09
0 4 1.00 Y 0.20 B 0.12
1 2 1.50 1.03 0.61
2 2 2.00 1.85 1.10
3 2 3.00 2.68 1.59
4 2 4.00 - MCE 4.00 - MCE 2.38
5 2+n 5.00 5.00 2.99
Y: Approximate Moment Frame yield; B: Approximate braced-frame inelastic buckling; MCE:
Maximum Considered Earthquake
87

Sliding Beam
Constraint Frame
N
Actuator 2 (220 kip, +/- 10 in)

10-3
Cyclic Reaction
5-2
Specimen Wall
Loading Brace buckling

Actuator 1 (220 kip, +/- 10 in)

(a)

Constraint
Sliding Frame
Beam
Cyclic
Loading
Actuator Reaction
3-3 Wall

(b)

Figure 3.1: (a) Plan and (b) elevation view of brace test setup.
88

Reinf. Plates: 2
HSS1: 10 HSS1: 8x2x1/4
HSS2: 1-3 HSS2: 8x2x3/8

HSS-section

Gusset Plate (typ)


1 (typ)
HSS1: 1-1
HSS2: 1-6
(a)

1 (typ) Reinf. Plates:


P1: 1-0 1
P1: 11x3x
P2: 6
P2: 6x2x

1-8
(typ) Pipe-section

P1: 1-3
P2: 9
10-3 (typ)
(b)

8x5x3/8
6 1
Slotted Plate

W12x16 CJP

See Figure 3.3


7.5
(c)

Figure 3.2: Brace drawings for (a) HSS, (b) Pipe and (c) W12x16.
89


Gusset (typ.)

9/16


Reinforcing PL
(a)

CJP

8x5x3/8 45
6
Plate

6 R=3/8
6 2 13/16
W12x16
Gusset Web

CJP


45
(b)
6
8x5x3/8
6 Plate

9/16 slot
W12x16
5 Flange

6 2
Gusset
(typ.)
(c)

Figure 3.3: Connection details for (a) HSS and Pipe and (b-c) W12x16.
90

6 3.54
Maximum Considered (4)
4 2.36

Axial Deformation (in)


2 1.18

Drift (%)
Expected Buckling (0.2)
0 0

-2 -1.18

-4 -2.36

-6 -3.54
0 10 20 30
Cycle Number

(a)

4 2.36
Compression dominated
2 1.18

Axial Deformation (in)


near-field loading history

0 0
Drift (%)

-2 -1.18

-4 -2.36
Far-field loading history
-6 appended to near-field -3.54

-8 -4.72
0 10 20 30
Cycle Number
(b)

8 4.72
Far-field loading history
6 appended to near-field 3.54
Axial Deformation (in)

4 2.36
(%)
Drift (%)

2 1.18
Drift

Tension dominated
0 near-field loading history 0

-2 -1.18

-4 -2.36
0 10 20 30
Cycle Number
(c)

Figure 3.4: (a) Standard cyclic (far-field ground motions), (b) Compression and (c) Tension near-
fault pulse loading histories.
91

6
Original SAC
Modified SAC

Peak Drift (%)


4 Fixed at 4%
Fixed at 0.2% --MCE
--Buckling (B)
2
Variable Slope

0
0 2 4 6
Inelastic Cycle Group Number
(a)

1.5
HSS4x4x1/4
HSS4x4x3/8
Pipe3STD
1
P (rads)

Pipe5STD
W12x16
Moment Frame
0.5

0
0 2 4 6
Inelastic Cycle Group Number
(b)

Figure 3.5: (a) Original and modified SAC loading history and (b) Cumulative plastic drift for
Chevron braced-frames during modified history and a moment frame under original SAC history.
92

hinge a

LB

Figure 3.6: One-story Chevron braced-frame.


93

3
Top
Bottom
2

Displacement (in)
1

-1
(a)
-2

3
Top
Bottom
2
Displacement x 10 (in)
-2

-1
(b)
-2

2
East
West
Displacement x 10 (in)
-2

(c)
-1

Figure 3.7: Axial transducer measurements for (a) West and (b) East top and bottom brace end
plates and (c) Lateral (North-South) displacements of end plates. For (a) and (b) positive
displacement corresponds to tension of the bracing member.
94

Imposed Axial Deformation (in)


1

0
Elastic
cycles
-1

-2
-2 -1 0 1 2
Measured Axial Deformation (in)

Figure 3.8: Intended and measured axial deformations.

150
Actuator 1
100 Actuator 2
Axial Force (kip)

50

-50

-100
-2 -1 0 1 2
Axial Deformation (in)

Figure 3.9: Actuator force measurements versus relative axial deformation (between east and
west end plates) for standard cyclic HSS4x4x1/4 test.
95

(a)

Brace

Washer
Cable

String pot device


(b) Unistrut connection
to lab floor

(c)
Figure 3.10: (a) Wire-pot instrumentation plan, (b) Connection detail and (c) Pipe3STD test.
96

300
West
200 East

Force (kip) 100

-100
(a)
-200
0 4 8
Lateral Displacement (in)

300

200
Force (kip)

100

-100
(b)

-200
0 4 8 12 16
Lateral Displacement (in)

Figure 3.11: Measured out-of-plane buckling displacements at (a) East and west (refer Figure
3.1a) brace third-points and (b) Brace mid point.
97

95
Local Buckling Face
Opposite Face

Temperature ( F)
85
o

75

65
(a)

225
Local Buckling Face-MidPoint
Opposite Face-MidPoint
185 Opposite Face-Offset
Temperature ( F)
o

145

105

65
(b)

Figure 3.12: Thermocouple recordings for (a) Quasi-static, near-field test and (b) Earthquake-rate
standard cyclic test.
98

Chapter 4
Large-Scale Experimental Results and Design Implications

This chapter summarizes results from the tests described previously in Chapter 3, in the context

of the performance of SCBF systems and current design provisions. The chapter begins with a

qualitative description of brace response and the observed damage states. This is followed by a

discussion of brace performance for all nineteen tests with the aim to examine the influence of

various test parameters. Two distinct types of failure are observed in the tests; one involves

fracture at the center of the brace due to local buckling at the middle plastic hinge region, whereas

the second involves failure of the brace-gusset plate connections. To address these issues

individually, the discussion of test results includes two separate sections. Based on the

experimental data, the last section of the chapter examines the applicability of strength equations

for bracing members commonly used for the capacity design of connections.

4.1 QUALITATIVE SUMMARY OF EXPERIMENTAL RESPONSE

The typical sequence of events leading up to fracture of an HSS brace is illustrated in Figures

4.1a-d for test HSS1-1, with the corresponding load versus deformation response shown in Figure

4.2a. The initial elastic cycles do not induce any visually observable deformation in the specimen.

The first major limit state is global brace buckling (Figure 4.1a) at a drift ratios of approximately
99

0.3%1, accompanied by large lateral deformations and flaking of the whitewash paint at the end

gusset plates and near the mid-point of the brace. Upon further loading, a plastic hinge develops

at the mid-point of the brace, which experiences local buckling (Figure 4.1b) at a drift ratio of

about 2%. Subsequently, cyclic loading triggers ductile fracture initiation (Figure 4.1c), which for

HSS1-1 occurred after the first reversed cycle to 2.7%. Soon after initiation, the fracture

propagates by ductile tearing through the section (Figure 4.1d) leading to a noticeable loss of

force capacity in the hysteretic response. In the square HSS, the buckled face ruptures first at the

corners and then propagates up the sides, leading to complete severance of the brace and loss of

strength. As the imposed story drifts increase up to 4%, the lateral deformations of the brace

become quite large - on the order of LB/8 (15-18 inches). It should be noted that unlike global

buckling and strength loss, which can be observed accurately through sudden drops in the load-

deformation plot, the precise instants of local buckling and fracture initiation are somewhat more

subjective to ascertain, since they are inferred through visual and photographic observations.

However, this has a relatively minor impact on overall performance assessment, since the

catastrophic event of final fracture and strength loss of the brace almost immediately follows

local buckling and fracture initiation.

Shown in Figure 4.2b is the load-deformation response of HSS1-2 subjected to the near-fault

compression dominated loading history. In this case, global buckling occurs at a smaller

compressive load (119 kips for HSS1-2 as opposed to 157 kips for HSS1-1), due to the tensile

elongation during the initial excursion to 2.0% drift. Local buckling occurred during the large

compressive pulse at a drift of 2.6%. Subsequent cycling of the already buckled brace about its

residual drift (refer to the near-field loading histories in Chapter 3) did not produce appreciable

1 This and other similar drift ratios reported in this section are specific to the tests described in this section,
which are used only for a qualitative illustration of brace response. These drift ratios may vary significantly
for other tests depending on test parameters. Table 4.1 and 4.2 summarizes these for all the specimens.
100

straining in the plastic hinge region. The brace ultimately survived twelve cycles of the appended

standard cyclic loading history before fracture initiation at a drift of -1.1%.

As compared to the other loading protocols, the tension dominated near-fault pulse presents a

more critical test of net section fracture at the brace end. Three of the five tension dominated tests

were reinforced at the net-section (P1-1, P2-3 and W3) and survived the large tension pull,

whereas the two pipe sections that were not reinforced (P1-4 and P2-4) fractured at the net-

section during the initial tension pull. The results of test P1-1, shown in Figure 4.2c, is typical of

the tension dominated response where net-section fracture does not occur. The loading history

begins with an initial tensile excursion to 8% drift, followed by a few cycles of compression and

tension loading. The appended standard cyclic loading ultimately leads to local buckling and

fracture similar to that described previously in Figures 4.1a-d. Being of a different section type,

the pipe section data in Figure 4.2c are not directly comparable to the HSS results in Figures 4.2a

and 4.2b. However, compared to other pipe section tests (P1-1 and P1-2), the initial tension

excursion causes a reduction of about 30% in the compression buckling strength (summarized in

Tables 4.1 and 4.2). This reduction is larger than the 25% reduction observed between the two

HSS tests (HSS1-1 and HSS1-2), shown in Figures 4.2a and 4.2b. Strain hardening during the

large tension pulse increases the maximum tensile strength of test P1-3 by about 20%, relative to

other pipe tests (P1-1 and P1-2).

4.2 QUANTITATIVE SUMMARY OF DATA FOR ALL TESTS

Data for all nineteen tests are summarized in Tables 4.1 and 4.2 (for standard cyclic and near-

fault pulse loading, respectively), including the maximum measured forces, deformation and drift

levels corresponding to key damage states, and total dissipated energy up to failure. Referring to

Table 4.1, the following data are reported for the limit states of standard cyclic loading tests: (1)

a GB - axial deformation at global buckling, (2) the maximum deformations sustained prior to
101

the occurrence of a LB - local buckling, a FI - fracture initiation (both observed visually), and

a SL - strength loss (3) the drift levels as per Equation (3.1) for each of these events

( GB , LB , FI , SL ), (4) aP SL - the cumulative plastic deformation and the corresponding

P
cumulative plastic drift ( SL ) sustained prior to strength loss and (5) ESL / E y - the normalized

dissipated energy, i.e., the summation of energy dissipated up to the point of strength loss

normalized by the tension yield energy (calculated as the product of the measured yield strength

and the yield displacement). For example, referring to Figure 4.2a and Table 4.1, HSS1-1

experiences global buckling at GB = 0.3%, local buckling at the peak drift of 1.9% during the

second compression cycle to 1.9%, fracture initiation at a drift of 1.7% during the second tension

cycle to 2.7%, and strength loss at a drift of 2.5% during the second cycle to 2.7%. Note that in

Table 4.1 the maximum drifts for FI and SL are both reported as 2.7%, since this is the

maximum drift that is sustained prior to fracture initiation and strength loss. Local buckling

occurred on the first compressive cycle to 1.9%, thus in this case, the drift at the instant of

buckling is the maximum drift sustained by the brace to that point. In cases where the specimen

survives the standard protocol (Figure 3.4a), the cycles at 5% drift were repeated. For example,

the W-shape specimen (W1) sustained 14 cycles at 5% drift prior to fracture initiation. These

maximum values by themselves may not provide a complete description of the brace

performance, since they do not incorporate information regarding the loading history. In this

context, an examination of the cumulative plastic deformation or energy dissipation provides a

better relative performance assessment, although it is difficult to compare these quantities directly

to imposed seismic demands that are typically expressed as peak drifts. Thus, the maximum

values of sustained drift (and axial deformation) are used for discussion, mainly due to their

simplicity and convenience of interpretation with respect to system level drift demands. Refer to
102

Appendix A for detailed documentation regarding the instants (during the loading histories) of

local buckling, fracture initiation and strength loss limit states.

Referring to the near-fault loading data of Table 4.2, all of the specimens except for P1-4 and P2-

4 survived the near-fault loading without fracture, and only HSS1-2 and W2 experienced local

buckling during the large compression pulse. The drift at global buckling is not reported for the

near fault cases, since the buckling drift is not unique and depends on the prior loading history

(e.g., see Figures 4.2b and 4.2c). Drifts for damage states reached during the appended standard

cyclic loading protocol are reported as relative values, with the datum being the residual drift at

the end of the near-fault loading. Referring to Figure 3.4b-c, the near-fault residual drift was 3%

in either the positive (tension) or negative (compression) sense. Referring to HSS1-2 (Figure 4.2b

and Table 4.2), local buckling occurred during the large near-fault compression pulse at -2.6%

drift. Fracture initiation and strength loss did not occur until well into the subsequent standard

cyclic loading, where fracture initiation occurred at 1.9% drift during the first tension cycle to

1.9% drift and strength loss occurred at 2.7% drift during the first tension cycle to 4% drift.

Having already survived the first compression cycle to 4% drift, the maximum value for SL is

listed as 4%.

Referring to Table 4.1, lateral brace buckling (global buckling) occurs at 0.3% for most of the

specimens, with slenderness values ranging from KLB/r = 63 to 80. The buckling drift is lower

(0.2%) for the more slender W-shape brace with KLB/r = 153 and larger (0.4%) for the grout

filled HSS brace. These values are generally consistent with previous studies (e.g. Tremblay

2000).
103

Compared to global buckling, larger variations were observed among drifts at local buckling,

which indicates the sensitivity of the local buckling limit state to the cross sectional shape, width-

thickness ratio, and grout fill. For the standard loading (Table 4.1), the maximum drifts at local

buckling ranged from 1.9% to 5%, where the larger resistance was observed in the more stocky

HSS and Pipe sections and the W-section. Local buckling occurred at larger drifts for the near-

fault compressive dominated cases (HSS1-2, HSS1-5, and W2 in Table 2). For example,

comparing HSS1-1 and HSS1-2, the initiation of local buckling occurred at 1.9% drift under

standard loading and 2.6% under the near-fault pulse. Similarly, comparing HSS1-4 and HSS1-5,

local buckling for the grout filled HSS initiated at 2.7% drift under standard loading and survived

a 6% drift pulse during the near-fault loading.

In general, fracture initiation and strength loss closely follow local buckling, where fracture

initiation and loss of strength occurred between 2.7% and 5.0% drifts. In this context, Figure 4.3

further illustrates the dependence of brace fracture on local buckling (tests with connection

failure, i.e., P1-4 and P2-4, are not shown). Figure 4.3a plots the normalized the energy dissipated

prior to local buckling versus a similar quantity for fracture initiation. From the figure (and the

associated linear regression fit between the local buckling and fracture energies), it is interesting

to observe that when all brace specimens and loading histories are considered, fracture initiation

succeeds local buckling after a relatively constant interval = 10.5 (indicated as the y-intercept on

Fig 4.3a), measured in terms of normalized dissipated energy. A similar trend is observed

between fracture initiation and strength loss (see intercept indicated on Figure 4.3b) such that

strength loss succeeds fracture by a constant interval = 3.8 (again in terms of normalized

dissipated energy). Thus the key observation is that once local buckling occurs, fracture initiation

and strength loss occur soon after. While the lag between any two of these limit states is

relatively constant, strength loss appears to follow fracture initiation quite quickly, in contrast to a

slightly larger separation between local buckling and fracture initiation.


104

As observed for local buckling, the near-fault pulse loading was not as damaging as the standard

cyclic loading, where all of the tests sustained the pulse loading without fracture initiation. The

fracture endurance (especially dissipated energy) for members subjected to the standard cyclic

loading after the compression pulse was similar to that observed under the standard cyclic loading

alone. The endurance of specimens that were subjected to the tension dominated pulse was

improved compared to their respective standard cyclic tests, largely due to cycling the specimen

about a residual tension elongation, thereby delaying local buckling.

4.3 EFFECT OF TEST VARIABLES ON CYCLIC BRACE BEHAVIOR, LIMIT STATES AND

DESIGN IMPLICATIONS

While the previous section summarized general trends and observations with respect to all the

experiments, this section investigates the effects of cross section geometry, width-thickness ratio,

slenderness ratio, loading rates and histories on brace performance. To provide a more

meaningful discussion of brace capacity in the context of expected demands, the experimental

results are compared to an assumed 2% design drift and a 4% Maximum Considered Earthquake

(MCE) event level drift. Recall the prior discussion (Chapter 3) that presented the rationale for

using 4% as an approximate measure of MCE drift (2/50 ground motions). On similar lines (and

based on data from Sabelli, 2001; Uriz, 2005; and McCormick, 2007) a value of 2% drift is

considered indicative of the mean demands expected in 10/50 ground motions. However, it

should be recognized that these values are largely subjective, and a rigorous system performance

assessment is needed to accurately establish acceptability criteria for braces. Such an assessment

is outside the scope of this dissertation.

4.3.1 Effect of Width to Thickness Ratios

Referring again to Figure 4.1b, brace fracture is driven by the amplified local strains induced by

the interactive effects of global and buckling during reversed cyclic loading. The drifts at fracture
105

initiation and the normalized energy dissipation capacity are plotted versus the normalized width-

thickness ratios in Figure 4.4 and versus global slenderness in Figure 4.5. The horizontal lines

drawn at 4% and 2% drift (Figures 4.4a and 4.5a) are considered to represent the minimum

required and design drift capacities of SCBF systems, respectively. As the fracture ductility is

known to be controlled by a combination of slenderness and width-thickness ratios (Tang and

Goel 1989), one should be mindful that the trends in the plots of Figures 4.4 and 4.5 are inter-

related.

Within each cross section type, the tendency for local buckling and fracture initiation increases

with increasing width-thickness ratios, resulting in reduced drift capacity at fracture. The two

HSS sections with similar slenderness ratios and varying width-thickness ratios (HSS1-1, HSS1-

3, HSS2-1, and HSS2-2, shown by the hollow squares in Figure 4.4) provide the most direct

evidence of the relationship between section compactness and fracture ductility. In this case, the

reduction in width-thickness ratio by 40% resulted in about a 65% increase in fracture drift

capacity and a 90% increase in energy dissipation capacity. Similar trends are observed when

comparing the pipe sections (P2-1, P2-2, P1-2, and P1-3, shown by circles in Figure 4.4), where a

25% reduction in the diameter to thickness ratio increased the fracture drift capacity by 50% and

energy dissipation by about 60%.

While all of the HSS and Pipe braces are well below the compactness limits of the AISC SCBF

provisions, the drift capacities of the less compact cross-sections do not meet the acceptance

criteria of 4% drift. The average drift at fracture initiation is equal to 2.9% for the HSS1 section

and 3.4% for the P1 (pipe) section. These results suggest that the current AISC SCBF

compactness limits for HSS and pipe sections may be unconservative and, in the least, warrant

further review. The data in Figures 4.4a and 4.4b suggest that a reduction to about three-fourths

of the current compactness limits for HSS and Pipe sections would achieve a drift capacity of 4%.
106

On the other hand, if one considers the 2% story drift (corresponding to design events), all but

one (HSS1 during the near-fault compression history) of the nineteen braces survive this drift

without fracture.

The W-shape braces exhibit high ductility despite having a flange width-thickness ratio that

exceeds the AISC compactness limit by about 5%. This can be attributed in part to the high

slenderness of the specimen which limits plastic strains in the central plastic hinge. Perhaps

equally significant is that the local buckling shape in the W-section induces less severe material

strains as compared to the HSS or Pipe sections.

4.3.2 Effect of Member Slenderness

As the LB/r slenderness increases, the brace buckling is more elastic and the smaller cross section

dimensions (relative to the brace length) lead to smaller strain demands at the central plastic

hinge. Previous studies (e.g. Jain et al. 1978, Tang and Goel 1989) have documented the

beneficial effects of increased slenderness on brace performance. In fact, some studies have

determined slenderness ratio to be the most important parameter controlling brace response (e.g.

Lee and Bruneau 2005).

Referring to the plot of slenderness ratio versus fracture drift and dissipated energy in Figures

4.5a and 4.5b, the data (particularly of the pipe braces) suggest that fracture ductility increases

slightly with member slenderness. However, since the member slenderness between tests of a

given cross section is not varied as much as the section compactness, it is difficult to draw clear

conclusions about LB/r slenderness from the data. The large capacities obtained for the W-shape

underscore the effect of overly large slenderness on ductility. Despite having a large width-

thickness ratio, the high member slenderness, as well as the local buckling shape, of the W12x16

appears to contribute to its large fracture resistance.


107

4.3.3 Effect of Cross Sectional Shape

Representative inelastic local buckling mode shapes are shown in Figure 4.6 for the three types of

cross sections, plus the filled section. As the local buckling shapes are quite different in form, the

resulting strain concentrations that trigger fracture are different as well. The HSS shapes exhibit

severe local buckling and crimping at the corners that greatly amplifies the local strain (Figure

4.6a) leading to fracture at that location. On the other hand, the local buckling deformations in

Pipes and W-shapes are more gradual, leading to a less severe strain gradient at the critical

location (Figures 4.6b and 4.6c). Thus, the Pipe and W-sections are inherently more resilient to

local buckling induced fracture. However, a suitable combination of slenderness and width-

thickness ratios can mitigate the effects of cross section geometry. As suggested by the data in

Figure 4.4a and 4.4b, the HSS section performance improves with reduced width-thickness ratios.

Conversely, in spite of their favorable shape, Pipes with large diameter-thickness ratios and lower

slenderness may not provide the required ductility.

4.3.4 Effect of Grout Filling of HSS Specimens

Previous experimental investigations (Liu and Goel 1988) indicate that concrete filled sections

may exhibit higher ductility and withstand more cycles of reversed loading as compared to

hollow sections. The concrete fill delays local buckling and minimizes the severity of strain that

drives fracture initiation. When local buckling occurs in concrete-filled tubes, the tubes tend to

buckle outward (Figure 4.6d), such that the cyclic strain demands are reduced compared to the

unfilled section (Figure 4.6a). This increased ductility is account for by AISC (2005) by

specifying relaxed b/t limits on filled-HSS braces

b / t 1.4 E / Fy 35.5 (for Fy 46ksi ) (4.3.1)

Note that the hollow b/t limit of 16 is significantly less than the value obtained by Equation 4.3.1.
108

The effect of grout fill can be seen by comparing HSS1-1 to HSS1-4 and HSS1-2 to HSS1-5,

where HSS1-4 and 5 have high strength grout fill ( f c' = 6-8 ksi). Referring to Table 4.1, the drift

at fracture initiation for the filled HSS1-4, FI = 4.1%, is about 50% larger than for the unfilled

HSS1-1; the dissipated energy for the HSS1-4 is about 20% larger. For the near-fault

compression loading (Table 4.2), the filled HSS1-5 exhibits a large (160%) increase in drift

capacity and (170%) dissipated energy compared to the unfilled HSS1-2. While these tests

confirm the beneficial effects of the fill, the degree of improvement is highly variable and

warrants further study, particularly considering constructability costs and larger cross sections

where size effects associated with cracking of the grout or concrete fill may play a role.

Furthermore, after normalizing the b/t ratio of the HSS4x4x1/4 (14.2) by the AISC (2005) limits

for composite HSS members (35.5) results in a ratio of approximately 0.4. However, comparing

the ductilities of HSS1-4 and HSS2-1 (with FI = 5.0%; normalized b/t limit = 0.53) suggests that

Equation 4.3.1 may be too relaxed and warrants further investigation in the context of SCBF

bracing components.

4.3.5 Effect of Loading Rate

Two experiments (HSS1-3 and HSS2-2) were conducted at higher loading rates, comparable to

rates that would occur during earthquakes, as in contrast to the quasi-static rates used in the other

tests. The earthquake loading rate was determined using the approximate secant stiffness at the

design drift and corresponding elongated period (0.8 seconds) for a typical SCBF frame. The

resulting peak loading excursion rate of 6 in/sec is about 360 times faster than the slow rate of

0.017 inch/sec (1 inch/min) used in the other tests.

High loading rates and the associated high strain rates can induce elevated stresses, due to rate-

dependent yielding and strain hardening behavior (Anderson 1995), which may increase the
109

tendency for brittle cleavage fracture. Conversely, the higher loading rates may cause a

temperature rise in the regions of high localized strain, which will tend to improve fracture

resistance as is evident in the Charpy energy curve (e.g. Barsom and Rolfe 1987). Thus, the

higher loading rates can have competing adverse and beneficial impacts on fracture ductility,

depending on the structural component geometry, stress constraint, presence of cracks, ambient

temperature, and material properties.

Referring to Table 1, the differences in fracture ductility between the high and low rate test

(HSS1-1 versus HSS1-3 and HSS2-1 and HSS2-2) are not significant. It should be noted that a

direct comparison between the tests is difficult because the loading histories for the high and low

rate tests are not identical. This variation results from the reduced ability to accurately control the

actuators at the high-rate tests, such that the imposed displacements are somewhat larger than the

specified target limits. The apparent insensitivity of fracture resistance to loading rate is

consistent with the fact that the brace fractures occur due to ductile tearing in regions of relatively

low-constraint, such that the modest effects of loading rate and temperature change (for the

ranges considered) do not significantly alter the fracture mechanisms. In addition to the observed

fracture response, thermo-couples installed on the surface of the braces provide a comparison of

temperature rise in the high-rate tests. The maximum recorded brace surface temperature for the

high-rate test, HSS2-2, was nearly 200F, compared to the peak 90F reading during a quasi-

static test, HSS1-2. Since the typical brittle-ductile transition temperatures for mild structural

steels range between -100F to 70F (Koteski et al 2005), the temperature increase from 90F to

200F does little to affect the ductility of fracture in the brace.


110

4.3.6 Effect of Unsymmetrical Buckling

One of the experiments HSS1-6 had two 18x2x0.5 reinforcing plates welded at the center to

the top and bottom (non-buckling faces) of the brace to deliberately induce unsymmetric buckling

(see Figure 4.7). The objective of this detail was to encourage unsymmetric buckling, such as

may be caused in field details due to dissimilar end conditions or imperfections. As indicated by

kinematics-based calculations, this type of unsymmetric response may increase the brace plastic

hinge rotations for a given axial deformation, relative to the idealized situation in which the brace

plastic hinge forms at the center of the brace. In all other respects, the specimen was similar to

HSS1-1. As expected, the ductility decreased by 43% from a maximum sustainable drift before

fracture of 1.57% in HSS1-1 (symmetric) to 1.1% in HSS1-6. The welded attachment itself is not

believed to influence fracture substantially (other than by causing the non-symmetric buckling

pattern) since fracture initiation occurred at a distance of approximately two inches from the end

of the reinforcing plate (see Figure 4.7).

These observations highlight the extent to which non-ideal conditions may affect the response,

especially when interpreting test results from specimens that are highly idealized representations

of field construction. The degree to which these non-ideal conditions, or unloaded attachments

(such as the plates in HSS1-6), will influence the response is uncertain. However, the above

discussion supports the requirement of a protected zone for design of SCBF systems which

guards the lateral load resisting elements against nonstructural attachments that negatively impact

the desired response (AISC, 2005).

4.4 BRACE-GUSSET PLATE CONNECTION PERFORMANCE

Fracture often controls the strength and ductility of critical structural members and connections.

The 1994 Northridge earthquake revealed the vulnerability of steel beam-column connections to
111

brittle fracture. Research since Northridge (e.g. FEMA, 2000) has resulted in significant

improvements in design practices regarding beam-column details in moment resisting frames.

However, earthquake-critical details commonly used in other lateral load resisting systems such

as Special Concentrically Braced Frames (SCBFs) have received relatively lesser attention in

research. A consequence of this is the limited understanding of the fracture resistance of these

details.

Brace-gusset plate connections in SCBF systems are an example of practical relevance where

earthquake-induced net section fracture may be a governing limit state. SCBF systems rely on

cyclic yielding and buckling of bracing members, subjecting the brace-gusset connection to large

tensile loads. Figure 4.8 illustrates commonly used brace connections for tubular HSS, Pipe and

Wide-Flanged sections. Figures 4.8a and 4.8b show the brace-gusset connection for these sections

where the tubular brace is slotted to accommodate the gusset plate. As shown in the figure, the

slot, extending beyond the gusset plate is typical of these connections and creates a reduced

section susceptible to fracture during severe tensile loading cycles. Fractures at these sections,

illustrated in Figure 4.8d, have been observed in earthquakes (Tremblay et al, 1995) and simple

calculations illustrate that the nominal strength of the net section may often be smaller than the

demands imposed by the yielding brace indicating that fracture in these details may well be a

likely mode of failure. Moreover, experiments by Tremblay et al (2003), Uriz (2005) and this

investigation have confirmed the likelihood of such fractures. Figure 4.8c shows a similar detail

for a Wide-Flanged brace where weld access holes create a reduced section.

In view of these developments, an objective of this chapter is to present experimental data from

the nineteen large-scale cyclic tests of brace specimens to examine the fracture resistance and

strength of brace-gusset connections, considering the effect of net-section reinforcement, loading

history and cross-section type. Recent tests (Liu et al, 2006, Yang and Mahin, 2005, Korol (1996)
112

and Cheng et al, 1998) have confirmed the likelihood of such failures in realistically sized

members under earthquake type loading. Moreover, the Seismic Provisions (AISC, 2005) require

that the design strength of CBF or SCBF braces based on net-section rupture (RtFuUAn) should

be at least equal to the expected tensile strength of the brace (RyFyAg) to avoid net-section

fracture. Typically, Ry is 1.4 for HSS tubes, 1.6 for Pipe sections and 1.1 for wide-flanged shapes

(Liu et al, 2007). Assuming U = 0.85, and given that An must be smaller than Ag, a simple

calculation shows that the net section strength of the member will be smaller than the tensile axial

strength, unless the net section is reinforced. The concerns are even more important as thicker

gusset-plates result in smaller net areas. However, the net section is often not reinforced in

practice, mainly because of the lack of guiding experimental data. This investigation

complements previous research by examining additional cross-sections and shapes (e.g. Pipes and

Wide-Flanged sections).

4.4.1 Experimental Results

Qualitatively, the specimens exhibited two main types of response. As discussed in the previous

section, a large majority (17 of 19) of the braces showed global buckling (Figures 4.1a) during the

compressive cycles, followed by local buckling (Figures 4.1b) of the cross section at the central

plastic hinge. The local buckles resulted in severe strain amplifications, eventually leading to

fracture at the center of the brace (Figures 4.1c and 4.1d). However, in the context of this section,

these specimens serve as examples of brace-gusset connections that survived earthquake-type

loading histories where both the strength and ductility are limited by brace, rather than

connection, response. For typical earthquake loading histories, once buckling occurs, damage

localizes at the center of the brace, essentially protecting the connection region from further

damage or fracture. Of the nineteen tests, only two (Pipe sections without reinforcement)

exhibited net-section fracture at the brace-gusset connection. Figure 4.8d shows a photo of the

fracture at the connection (for Pipe3STD), whereas Figure 4.9b shows the load deformation curve
113

of these experiments where the connection deformation is measured over each end region (one-

third length refer to Figure 4.9a) of the brace to distinguish between response observed at each

end connection. In both of these experiments, the specimens were subjected to a near fault tension

dominated history (refer Chapter 3, Figure 3.4c) where a large tensile pulse was applied before

any cyclic loading could concentrate buckling-induced damage at the center. In both cases, the

braces fractured on the first tensile pulse before buckling-induced damage could localize at the

center.

Deformation data for the brace specimens was recovered in two formats. The first measure is the

overall deformation of the brace, measured over the entire brace length (see Figure 3.1a), which

can be converted to a corresponding drift ratio as discussed earlier. While this enables a

convenient comparison with expected story drift demands, it does not capture local effects, e.g. if

deformations localize at one end before the other. Thus, additional deformation measurements

were made over each end region (one-third length refer Chapter 3 and Figure 3.1a) of the brace

to distinguish between response observed at each end connection.

Table 4.3 summarizes deformation capacities (in terms of drift) corresponding to fracture

initiation in the specimens. Also indicated in the Table are connection deformations for the braces

which failed at the net section (tests P1-4 and P2-4) prior to any damage localization at the center

of the brace from global and local buckling. The drift capacities correspond to the maximum

drifts sustained by the member before fracture was observed. For example, if a brace fractures

during a tensile cycle at a specific drift, but encountered a larger drift during a prior compressive

excursion, the larger drift is reported. The Table also includes a comparison of the maximum

experimental tensile strength capacities with nominal capacities (based on a minimum of gross

yield and net-section strengths). Several observations regarding brace and connection behavior

can be made based on the data presented in Table 4.3.


114

When HSS or Pipe braces are reinforced at the net-section with reinforcing plates none of the

specimens (total 12) exhibited net section fracture. Irrespective of cross section type or

loading history, the damage localized at the center due to buckling and low cycle fatigue. In

fact, several of the specimens sustained drifts as large as 5-8% prior to failure, well in

exceedance of the expected seismic drift demands in SCBF systems. This indicates that

reinforcement effectively prevents net section fracture in these details.

Four Pipe specimens (P1-2, P1-4, P2-2, and P2-4) were not provided with reinforcement

plates at the net-section. Of these, two (P1-4 and P2-4) fractured during the first tensile pull

of the near-fault tension dominated history. This loading history was applied as a worst-case

scenario for tensile fracture at the connection, because it applied large deformation demands

before buckling damage could localize at the center of the brace. Despite the severity of the

loading history, the details fractured at deformations corresponding to drift levels of 6.4 and

5%, still fairly large as compared to expected demands. The two other unreinforced pipe

specimens (P1-2 and P2-2), subjected to far-field loading histories, did not fracture and

survived drifts as large as 2.68 and 5.0%, before fracturing at the center.

Three tests (W1, W2 and W3) featured W-section braces, subjected to compression and

tension dominated near-fault, as well as a regular far-field loading history. All these

specimens survived drifts between 5 and 8%, fracturing at the center due to buckling induced

damage. For the W-section braces, even the tension dominated near-fault history failed to

produce net section fracture, despite imposing tensile drift demands as large as 8%. This large

ductility can be attributed to the connection detail (see Figure 4.8c), where the weld access

hole produces a long reduced section resulting in smaller strains at the connection.
115

Shown in bold-face in Table 4.3, the specimens P2-3 and P2-4 (Pipe3STD) are controlled by

net section failure as opposed to brace yielding, i.e., Pn Rt FuUAn Py Ry Fy Ag . Note, the

-factor is removed in the comparison of the two capacity quantities. Of these, P2-3 failed at

the center (under a far field loading history) whereas P2-4 failed at the connection due to net

section fracture. Also interesting to note is that in test P1-4, where

Pn Rt FuUAn Py Ry Fy Ag (suggesting that gross yielding would govern), the specimen

failed by net section fracture. This can be attributed to strain hardening in the brace that

increases the brace yielding force resulting in connection failure.

In context of the previous point, refer the last column of Table 4.3 that summarizes the ratio

between the experimental and predicted brace force Pmax/min(Py, Pn). Recall that Pn governed

only for P2-3 and P2-4. For all the other tests, is apparent that the measured maximum tensile

strengths (Pmax) are up to 25% larger than the expected yield strength ( Pu Ry Fy Ag ). This

effect is most significant for the HSS specimens where the maximum tensile forces were 13

to 21% larger than the expected strength, Py. This suggests that the current practice of

designing bracing connections based on the expected yield force may be underestimate force

demands in connections, mainly because it ignores the effect of strain hardening that

amplifies the brace forces. Others have noted this too, e.g. Tremblay et al (2002) who

suggested amplifying brace forces by appropriate factors (), such that Pbrace Ry Fy Ag

incorporates the effects of strain hardening. This is discussed in the following section.

Thus, the experimental data suggests that (1) net-section reinforcement in brace-gusset

connections effectively prevents fracture, (2) in far-field type earthquake loading histories, braces

typically exhibit damage localization at the center thereby protecting the net-section from failure

and (3) fracture at the net-section occurs only during the most severe loading histories where a
116

large tension pulse precedes cyclic loading. Even in these cases, the deformation capacity of the

member exceeds the expected seismic drift demands in SCBF systems (4) maximum tensile

forces observed in the large-scale brace tests may be somewhat larger than the expected yield

force (RyFyAg) typically used to design the bracing connections. This final point will be discussed

in the following section.

4.5 COMPARISON OF EXPERIMENTAL DATA TO COMMONLY USED FORMULAE FOR

PREDICTING STRENGTH AND STIFFNESS OF BRACING MEMBERS

This section compares the experimental brace stiffness and maximum strength (compressive and

tensile yield) data to commonly used analytical values. Table 4.4 lists the calculated and

measured strengths and elastic stiffness for each of the tests along with statistics for each

comparison. For example, where the first column lists the expected critical buckling loads of each

brace, Pc,exp, the second column, RPc, is the ratio of the experimental to predicted buckling load.

Referring to the table, the elastic stiffness, compressive buckling load, and expected tensile yield

and ultimate forces will be discussed in this section.

4.5.1 Elastic Stiffness

The initial elastic stiffness is calculated as Kel = EAg/L, where E = 29,000 ksi, Ag is the nominal

brace cross-sectional area, and L is the brace length (measured from end-to-end of the gusset

plates). For the grout filled braces, the stiffness of the fill is calculated similarly and is assumed to

act in parallel with the brace. The average test-prediction ratio for the stiffnesses is 1.04, with a

coefficient of variation of 0.11. This may be attributed to errors in estimating the grout stiffness

and other simplifications, such as errors in stiffness estimation of the gusset plates regions.
117

4.5.2 Compressive Strengths

The expected compressive strengths, Pc,exp, are calculated based on the AISC Specification

(2005), except that the expected yield strengths are used and the effective buckling length is

assumed equal to the distance between the gusset plate fold lines (see Figure 3.2; taken as the LB

listed in Table 3.1). For example, HSS1-1 had a measured compressive resistance of 157 kips,

whereas the corresponding values calculated as per AISC are:

Pc,exp = Fcr A g = (30.7 ksi)(3.37 in 2 ) = 103kips


2E 2 (29,000 ksi)
Fe = = = 47.5 ksi
1.0 (118")
2 2
KL

r 1.52"
(4.3.2)
KL E
Since 4.71 77.6<118
r Fy
Fy
46

Fcr 0.658 Fe Fy 0.658 47.5 46 ksi 30.7 ksi

The estimated critical buckling loads presented in Table 4.4 use Ry to account for the increase in

yield stress from the minimum specified value to the expected value:

Pc ,exp = Fcr Ry Ag = (36.5 ksi)(3.37 in 2 ) = 123kips


2E 2 (29,000 ksi)
Fe = = = 47.5 ksi
1.0 (118")
2 2
KL

r 1.52"
(4.3.3)
KL E
Since 4.71 77.6<100
r R y Fy
Ry Fy
1.4 46
Fcr Ry
0.658 Fe Ry Fy 0.658 47.5 1.4 46 ksi 36.5 ksi

For the grout filled braces, HSS1-4 and HSS1-5, the fill strength was assumed equal to its

nominal specified value of fc = 7 ksi. The strengths for these specimens are calculated as per

AISC (2005), which involves adding the contributions from the strengths of the grout and steel,

as described by Equations 4.3 through 4.6 below


118

Po Ag Ry Fy 0.85 Ac f c ' (4.3.4)


Pe 2 EI eff KL B
2
(4.3.5)

Where EI eff Es I s C3 Ec I c and C3 0.9 for this investigation.

(a) When Pe 0.44 Po

Po

Po 0.658 e
P
Pc ,exp (4.3.6)

(b) When Pe 0.44 Po

Pc ,exp 0.877 Pe (4.3.7)

Where Ag, Fy, Is and Ac, fc, Ic are the properties of steel and grout, respectively.

Overall, for all tests, the ratio of measured to calculated expected compressive strengths is 1.24

with a standard deviation of 0.30, where the expected compressive strengths are, in general,

smaller than the experimental buckling loads. The smaller expected compressive strengths may

be a result of assigning an effective length factor, K = 1.0, in the above calculations. Assuming

the gusset plates do provide some fixity and assigning K = 0.9 decreases the measured to

calculated ratio to 1.08 with a standard deviation of 0.22. The remaining error may be attributed

to the larger measured yield strengths, Fy,meas, above RyFy and the varying effective length factors

from the different brace and gusset-plate geometries.

4.5.3 Maximum Tensile Strength

The expected tensile yield, Py,exp = RyFyAg, and ultimate, Pu,exp = RtFuAg, brace strengths are

calculated as the product of the expected material strengths and the gross cross-sectional area

(this was introduced previously in the context of brace-gusset connections). Figure 4.10a

compares the maximum measured experimental tensile forces with the expected yield and
119

ultimate strengths. The average ratio of the maximum measured brace strength to the expected

brace yield strength is 1.14 with a standard deviation of 0.07. In some cases (e.g. test W3), the

measured strength is as large as 1.25 times the expected yield strength, which while conservative

from the perspective of member design, can be unconservative for determining capacity design

requirements for connections. This is especially true for tension near-field loading for the Pipe

and Wide-Flanged sections. The average ratio of the measured strength to the expected ultimate

strength is 0.92 with a standard deviation of 0.08. On average, an improved estimate of the

expected brace strength would be to use an average of the expected yield and ultimate strengths,

i.e., Pmax = 0.5(RyFy+ RyFy)Ag. This is illustrated in Figure 4.10b. Using this measure, the ratio of

the maximum experimental brace forces to this capacity measure is 1.01 with a standard deviation

of 0.08).

4.6 SUMMARY

This chapter presents findings and design implications based on nineteen large scale tests of

concentrically-loaded steel braces subjected to earthquake type cyclic loading. The tests are part

of a larger project that has the dual aims of validating and applying new micromechanics based

models to simulate ductile fracture under cyclic loading and to develop practical behavioral

information and design guidance of steel braced frames.

The tests, designed to represent typical conditions encountered in steel braced frames,

complement previous studies by investigating three cross section types (HSS, Pipe and Wide

flanged sections) of varying section compactness and brace slenderness. The test program also

investigates the effects of loading histories, loading rates, connection reinforcement, and grout

filling of HSS sections.


120

Qualitatively, the tests all followed a similar sequence of events leading to failure. Global

buckling of the brace (at displacements corresponding to 0.2-0.4% story drift) leads to the

formation of a plastic hinge at the midpoint of the brace. Subsequently, local buckles form in the

hinge region (at 2% to 5% story drift) that amplify the strains and trigger fracture initiation (at

2%-8% story drift). Soon after this, the fracture propagates through the entire cross section,

severing the brace. Brace buckling is accompanied by large out of plane displacements that pose

threats to surrounding architectural enclosures.

One of the main conclusions of this study is that brace fracture ductility is primarily a function of

section compactness and to a lesser extent member slenderness and loading history. Specifically,

fracture ductility increases with more compact cross sections and more slender members. Further,

the standard loading protocols (modeled to represent general or far-field ground motions) are

more damaging than loading protocols developed to represent pulse-like near-field ground

motions. The tests further demonstrate that the local buckles in HSS sections result in more

severe straining of the steel material, leading to fracture initiation near the corners of the brace.

This is in contrast to Pipe and Wide-Flange sections that exhibit more gradual local buckling

modes. This is not to imply that Pipes and Wide-Flange braces are naturally more ductile than

HSS. In fact, the Pipe5STD (large D/t) brace fractured at a smaller drift ratio than the

HSS4x4x3/8 (small b/t).

The tests suggest that the section width-thickness ratios in the AISC Seismic Provisions (2005)

for HSS and Pipe sections may not result in adequate deformation capacities for seismic design.

HSS members with width-thickness ratios equal to about 90% of the limiting compactness

criteria, and subjected to the general loading protocol, fractured at drift ratios in the range of 2.7%

to 3.0%. Pipe members with diameter to thickness ratios equal to 60% of the limit fracture at drift

ratios of 2.7%. While the drifts achieved by these members are larger than the approximate
121

design level drift of 2%, they are smaller relative to the 4% drift demand criteria implied by

several previous investigations and current design requirements. On the other hand, W-shape

braces, which slightly violated the compactness criteria, sustained drift ratios of up to 5%. These

results are sensitive to loading history, since the endurance for all of the braces increased

considerably (up to two or three times) when subjected to the near-fault loading protocol that

subjected the braces to fewer reverse loading cycles. Tests to investigate the effect of loading rate

on fracture performance demonstrated essentially no difference in response between quasi-static

and earthquake loading rates.

Tests of braces filled with high-strength grout fill did not increase the fracture capacity as much

as expected. As expected, the grout fill postponed local buckling and increased the fracture

resistance by about 160% in the near-fault tests while only a modest increase in drift ratio (50%)

was observed in the far-field tests. Thus, the effectiveness of grout fill to improve braced frame

performance warrants further study. One of the braces, fitted with reinforcing plates to induce

unsymmetric buckling showed significantly reduced ductility. This indicates that experimental

data from idealized specimens must be interpreted with caution, since imperfections in field

details may negatively impact response.

Comparison of measured and calculated strengths for brace strength and stiffness generally

confirm expectations and the legitimacy of standard assumptions. In particular, ratios of measured

compressive buckling strengths to calculated strengths (using the standard AISC column curve

equation and expected yield strengths with Ry factors specified by AISC) have a mean value of

1.24 and a standard deviation of 0.30. Ratios of measured tensile strengths are estimated fairly

well by the average of the expected yield and ultimate brace strengths (calculated using RyFy and

RtFu values specified by AISC) with a mean value of 1.01 and a standard deviation of 0.08.
122

While the brace test data provides valuable insights into the brace buckling and fracture behavior,

it is difficult to generalize the experimental findings since (a) the tests cover only a limited range

of parameters and configurations, and (b) in and of themselves, the tests provide limited data to

quantify the localized stress and strain combinations that trigger fracture and fatigue. Thus,

analytical simulations are necessary to generalize the data; for example, to develop quantitative

relationships between various parameters (such as slenderness and width-thickness ratios) and

fracture ductility. These topics are addressed in the following chapters.


123

Table 4.1: Data for standard (far-field) loading protocol tests


Test Pt,max Pc,max a GB , in a LB , in a FI , in a SL , in aP SL , in ESL
No. (kip) (kip) ( GB , %) ( LB , %)** ( FI , %)** ( SL , %)** ( aP SL , rad) Ey
0.18 1.10 1.57 1.57 22.3
HSS1-1 247 157 29.6
(0.3) (1.9) (2.7) (2.7) (0.4)
0.20 1.26 1.73 1.73 23.1
HSS1-3* 255 161 41.1
(0.3) (2.1) (3.0) (3.0) (0.4)
HSS1-4 0.21 1.57 2.40 2.40 31.0
257 194 39.3
(filled) (0.4) (2.7) (4.1) (4.1) (0.5)
0.21 1.10 1.10 1.57 17.7
HSS1-6# 249 163 25.0
(0.35) (1.9) (1.9) (2.7) (0.3)
0.17 2.99 2.99 2.99 77.9
HSS2-1 348 186 62.5
(0.3) (5.0) (5.0) (5.0) (1.3)
0.20 2.64 2.64 2.64 78.2
HSS2-2* 362 184 73.5
(0.3) (4.5) (4.5) (4.5) (1.3)
0.18 1.57 1.57 2.40 30.5
P1-1 241 181 35.3
(0.3) (2.7) (2.7) (4.0) (0.5)
0.18 1.57 2.36 2.36 39.8
P1-2 243 177 41.8
(0.3) (2.7) (4.0) (4.0) (0.7)
0.16 2.99 2.99 2.99 101.5
P2-1 132 80 65.3
(0.3) (5.0) (5.0) (5.0) (1.7)
0.16 2.99 2.99 2.99 91.3
P2-2 130 84 60.9
(0.3) (5.0) (5.0) (5.0) (1.5)
0.09 2.99 2.99 2.99 222.5
W1 286 93 81.7
(0.2) (5.0) (5.0) (5.0) (3.8)
*fast (earthquake) loading rate

**values for the limit states of local buckling (LB), fracture initiation (FI) and strength loss (SL) are reported as the maximum value of axial
deformation (a, in) or drift (, %) that was sustained by the specimen prior to the limit state

#HSS1-6 was reinforced at the midpoint (top and bottom) with 18 long plates (see Figure 4.7)
124

Table 4.2: Data for near-field (NF) pulse loading protocol tests
Appended Standard
NF Loading #
Loading #
Test Pt,max Pc,max
No. (kip) (kip) a , in a , in aP SL , in ESL
Event ( , %)** Event
( , %)** ( aP SL , rad) Ey
Near Fault Compression Pulse Loading
-1.50 FI 1.10 (1.9) 62.2
HSS1-2 249 119 LB 19.3
(-2.6) SL 2.36 (4.0) (1.1)
None, LB 2.99 (5.0)
HSS1-5 -3.54 119.7
263 136 Survived FI 2.99 (5.0) 52.2
(filled) (-6.0) (2.0)
Pulse of: SL 2.99 (5.0)
-2.68 FI 2.99 (5.0) 140.6
W2 287 82 LB 36.4
(-4.5) SL 3.66 (6.2) (2.4)
Near Fault Tension Pulse Loading
1.61 (2.7)
92.2
P1-3 292 127 2.40 (4.0) 52.7
(1.6)
2.40 (4.0)
None, LB 2.87 (4.9)
+4.72 136.6
P2-3 149 57 Survived FI 2.87 (4.9) 67.1
(+8.0) (2.3)
Pulse of: SL 2.87 (4.9)
2.32 (3.9)
475.5
W3 323 75 2.32 (3.9) 95.0
(8.0)
3.07 (5.2)
+3.86 3.6
P1-4 279 -- FI/SL at -- -- 14.7
(+6.5) (0.1)
net
+2.99 2.8
P2-4 144 -- section -- -- 14.5
(+5.1) (0.05)
#limit states observed during the standard, far-field loading protocol that is appended to the near-field loading are reported in terms of drift ratios
measured relative to the residual drift of +/- 3% that existed at the end of the near-field loading (see Figure 3.4b and 3.4c). Otherwise, the values
are described in the same way as for those reported in Table 4.1 for the standard (far-field) loading.
125

Table 4.3: Experimental results of bracing connections


Fracture/ Pt, max
Cross **
Test Detail Type Failure Type Maximum Drift
Section min(Py , Pn )
(%)
P1-3 Pipe5STD Reinforced Fracture in middle of brace 8.0# 1.21
P1-4 Pipe5STD* Unreinforced Net section Fracture at end 6.5 (3.9 in) 1.16
P2-3 Pipe3STD Reinforced Fracture in middle of brace 8.0# 1.05
P2-4 Pipe3STD* Unreinforced Net section Fracture at end 5.1 (3.0 in) 1.17
W2 W12x16 NA Fracture in middle of brace 8.0# 1.25
*Failure at net section. Connection deformations are listed in parentheses.

** Py R y Fy Ag ; Pn Rt FuUAn 0.85 Rt Fu Ag Rt Fu An brace . Number in bold are controlled


plate
by net section failure (i.e. Pn Rt FuUAn Py Ry Fy Ag )

#Denotes maximum drift sustained without fracture at net section. Failure occurred during the
subsequent cyclic loading (refer Table 4.2 for details).

Table 4.4: Comparison of brace strength and stiffness


Test Pc,exp RPc RPc Py,exp Pu,exp Kel
RPy RPu RKel
No. (kip)* (K = 1.0) (K = 0.9) (kip) (kip) (kip/in)
HSS1-1 1.27 1.14 1.14 0.97 1.13
HSS1-2 0.96 0.86 1.15 0.98 1.13
123 217 254 825
HSS1-3 1.30 1.17 1.18 1.00 1.11
HSS1-6 1.32 1.10 1.15 0.98 1.14
HSS1-4 1.24 0.77 1.18 1.01 0.77
156 217 254 1216
HSS1-5 0.87 1.18 1.21 1.04 0.78
HSS2-1 1.10 0.98 1.13 0.97 1.06
169 308 360 1165
HSS2-2 1.09 0.97 1.18 1.01 0.9
P1-1 1.04 0.98 1.00 0.78 1.07
P1-2 1.02 0.96 1.01 0.78 1.04
174 241 310 1039
P1-3 0.73 0.69 1.21 0.94 1.12
P1-4 N/A N/A 1.16 0.90 1.08
P2-1 1.51 1.28 1.06 0.82 1.09
P2-2 1.58 1.34 1.04 0.81 1.07
53 125 161 537
P2-3 1.07 0.91 1.19 0.93 1.12
P2-4 N/A N/A 1.15 0.89 1.12
W1 1.85 1.50 1.10 0.85 1.00
W2 50 1.63 1.32 259 1.11 337 0.85 1136 1.08
W3 1.49 1.21 1.25 0.96 1.04
average 1.24 1.08 1.14 0.92 1.04
std.dev. 0.30 0.22 0.07 0.08 0.11
*Assumes K = 1.0
126

(a) (b)

(c) (d)

Figure 4.1: Typical progression of brace specimen damage (a) Global buckling, (b) Local
buckling, (c) Fracture initiation and (d) Loss of tensile strength.
127

1200 Max. Max.


(a) Force Drift
1099 kN 2.7%
800
Fracture

Force (kN)
400 Initiation
1.7%

0
Local Loss of
-400 Buckling Tensile
1.9% Global Buckling Strength
GB =0.3%, 698 kN 2.5%
-800
-4 -2 0 2 4
Drift (%)

1200
Max.
(b) Force
800 Fracture 1108 kN
Initiation
-1.1% (1.9%) Loss of
Force (kN)

400 Tensile
Strength
-0.3%
0 (2.7%)
Max.Drift
-7.0% Local
-400 (-4.0%) Global Buckling
Buckling
-2.6% GB =1.0%, 529 kN
-800
-8 -6 -4 -2 0 2 4
Drift (%)

1600
(c) Max.
1200 Force
1299 kN
Fracture
800 Initiation
Force (kN)

5.9% (2.9%) Loss of


Max. Drift Tensile
400 -1.0% Strength
(-4.0%) 6.4%
0 (3.4%)

Local
-400 Buckling Global Buckling
0.3% (-2.7%) GB =6.8%, 565 kN
-800
-2 0 2 4 6 8 10
Drift (%)

Figure 4.2: Typical brace response for (a) Far-field loading (HSS1-1 shown), (b) Near-fault
compression (HSS1-2) and (c) Near-fault tension (P1-3). Drifts in parenthesis are relative to
residual drift after near fault loading. Drifts underlined are reported in Tables 4.1 and 4.2.
128

100
(a)

EFI/EY = 0.99ELB/EY + 10.5


EFI/EY

50 HSS1 (Grout Fill)


HSS1
HSS2
P1
P2
10.5 W
0
0 50 100
ELB/EY

100
(b)

ESL/EY = 1.01EFI/EY + 3.8


ESL/EY

50 HSS1 (Grout Fill)


HSS1
HSS2
P1
P2
3.8 W
0
0 50 100
EFI/EY

Figure 4.3: Energy dissipated prior to (a) Fracture initiation (FI) versus local buckling (LB) and
(b) Strength loss (SL) versus fracture initiation.
129

6
HSS1 (Grout Fill)

HSS1
Fracture Drift (%)

4 HSS2

P1

P2
2
W

(a)

0
0 1 2
Width-Thickness Ratio/AISC Limit

100
HSS1 (Grout Fill)

HSS1

HSS2
E/EY

P1
50
P2

Connection
(b)
Failure
0
0 1 2
Width-Thickness Ratio/AISC Limit

Figure 4.4: Effect of width-thickness ratio on (a) Maximum drift at fracture initiation and (b)
Normalized dissipated energy.
130

6
HSS1 (Grout Fill)

HSS1
Fracture Drift (%)

4 HSS2

P1

P2
2
W

(a)

0
0 1 2 3
Slenderness Ratio/AISC Limit

100
HSS1 (Grout Fill)

HSS1

HSS2
E/EY

P1
50
P2

W
Connection
Failure
(b)

0
0 1 2 3
Slenderness Ratio/AISC Limit

Figure 4.5: Effect of slenderness ratio on (a) Maximum drift at fracture initiation and (b)
Normalized dissipated energy.
131

(a) (b)

(c) (d)

Figure 4.6: Local buckling shapes for (a) HSS, (b) Pipe, (c) Wide Flanged and (d) Grout-filled
HSS cross sections.
132

CL

(a)

Reinforcing Plates

CL

(b)

Figure 4.7: (a) Symmetric local buckling at brace midpoint and (b) Middle reinforcement plate
(top and bottom)-induced unsymmetrical buckling.
133

Net
Section
Reinforcement plate

Net
Section

(a) (b)

Weld access hole

(c) (d)

Figure 4.8: Net section details for (a) Reinforced HSS4x4x1/4, (b) Pipe3STD, (c) W12x16; (d)
Net section failure of Pipe3STD.
134

Sliding Beam
Constraint Frame
N
Actuator (220 kip, +/- 10 in)

10-3
Monotonic Specimen Reaction
Wall
Loading

Connection Gage Length (3-3)

(a)

300

Pipe5STD
200
Force (kip)

Pipe3STD

100

(b)
0
0 1 2
Connection Deformation (in)

Figure 4.9: (a) Experimental connection gage length (refer Figure 3.10a for locations of wire-
pots) and (b) Load deformation response of Pipe3STD and Pipe5STD tests with fracture at the net
section.
135

400
RyFyAg
RyFyAg

Design Capacity (kips)


RtFuAg
RtFuAg
Overestimated
200

Underestimated

(a)
0
0 200 400
Pmax (kips)

400
1/2[RyFyAg + RtFuAg] (kips)

200

(b)
0
0 200 400
Pmax (kips)

Figure 4.10: Maximum experimental forces compared to (a) Py,exp=RyFyAg and Pu,exp=RtFuAg
capacities and (b) Average of Py,exp and Pu,exp.
136

Chapter 5
Application and Evaluation of the CVGM for Large-Scale

Components

Prevailing approaches to characterize the fatigue and fracture performance of braced frame and

other structural components are based mostly on empirical or semi-empirical methods. For

braces, previous research has relied on critical longitudinal strain measures at the material or

cross sectional level, or cycle counting and fatigue-life approaches at the component level (Tang

and Goel, 1989). In fact, recent studies (e.g. Uriz and Mahin, 2004) have combined these

approaches by applying fatigue-life approaches to a fiber strain at a brace cross-section. While

these approaches represent important advances in the fatigue-fracture prediction methodology for

structures, they do not directly incorporate the effects of local buckling or the complex

interactions of stress and strain histories that trigger crack initiation in bracing components.

Consequently, large-scale testing is still required to characterize the fracture performance of these

details (Lehman et al., 2008).

In part, the dependence on experiment-based approaches can be attributed to the lack of

computational resources required to simulate phenomena such as local buckling that create

localized stress and strain gradients that cause fracture. However, where fracture is of concern,
137

the reliance on simplistic models is primarily due to the lack of suitable stress/strain based

fracture criteria to accurately evaluate the complex interactions of stresses and strains. This is

particularly the case when fracture occurs in structural components subjected to large-scale

yielding and cyclic loading where traditional fracture mechanics approaches are not accurate.

Moreover, many of these situations (especially those found in braced-frames) do not contain a

sharp crack or flaw, which is another necessary assumption for the use of traditional fracture

mechanics. Finally, earthquakes produce Ultra Low Cycle Fatigue (ULCF) in structures where

very few (typically less than 10) cycles of extremely large magnitude (several times yield) are

typical during the dynamic response of a building. This ULCF behavior is quite different from

low or high cycle fatigue, which occurs in bridges and mechanical components. Consequently,

continuum-based models that capture the fundamental physics of the ULCF/fracture phenomena

are required to capture the complex stress-strain interactions leading to fracture. The models

presented in this chapter simulate the micromechanical processes of ULCF to predict fracture

from a fundamental physics-based perspective. They are fairly general, can be applied to a wide

variety of situations as they work at the continuum level, and are relatively free from assumptions

regarding geometry and other factors. Finally, these models require inexpensive tension coupon

type tests for calibration.

Kanvinde and Deierlein (2007) suggested an explanation for the processes which may govern

micromechanical ULCF behavior during earthquake-type cyclic loading. The resulting

continuum-based, Cyclic Void Growth Model (CVGM) has shown promise in simulating the

physical void growth, shrinkage and damage events that lead to ductile fracture initiation during

cyclic loading in small-scale experiments. The model has been rigorously examined by Kanvinde

and Deierlein (2007) at the small-scale across a variety of steel types, geometries and loading

histories. In this chapter, the CVGM is applied to the brace specimens described in Chapters 3

and 4 to extend the work of Kanvinde and Deierlein and assess the accuracy and limitations of the
138

model at a larger scale. Referring to the previous discussion, the brace fracture at the middle

plastic hinge presents a case where large-scale yielding and smooth geometries limit the

applicability of traditional fracture mechanics. Furthermore, the testing matrix presented in

Chapter 3 provides a rich collection of different steel types, cross-sectional shapes and loading

histories to assess the capabilities of the CVGM at the large-scale. However, the global and local

buckling-induced behavior of the braces in the presence of cyclic inelastic response under

multiaxial stresses presents challenges in simulating these components which must be resolved to

accurately predict the fracture.

A successful large-scale evaluation of the physics-based fracture model provides a powerful

framework which can be used to investigate structural component or system behavior without

experimentation. For example, while the results from the nineteen brace tests, presented in

Chapter 4, suggested trends between brace ductility and geometric and material properties, the

experimental results alone do not form a basis for providing conclusive recommendations in the

context of brace performance and detailing guidelines. In fact, large-scale testing results are

rarely exhaustive due to insufficient lab capabilities and high material and laboratory costs that

accompany large-scale testing. Thus, the development of simulation methodologies provides

researchers with powerful tools to generalize experimental data with the ultimate goal of an

improved assessment of fracture susceptibility in steel systems.

To provide support for such a methodology for large-scale structural components, this chapter

reviews a finite-element simulation study and the performance of micromechanical-based fracture

models applied to the brace experiments of Chapters 3 and 4. The chapter begins by reviewing

the formulation of the Cyclic Void Growth Model (CVGM) and the calibration procedure for the

CVGM parameters. Next, the small-scale CVGM validation results from a prior study (Kanvinde

and Deierlein, 2007) are briefly introduced to illustrate the accuracy of the CVGM fracture
139

predictions across multiple small-scale experimental specimens, seven different steels, and

diverse loading histories. The fracture prediction methodology is then developed for the large-

scale brace component tests presented as part of this study. Here, the results of finite element

simulations, and corresponding CVGM fracture predictions, of each bracing member are

presented to examine the performance of the model across various material types and loading

histories. The fracture predictions are discussed in a deterministic as well as a probabilistic

framework, where the latter incorporates material uncertainty. Finally, the implications of this

study are discussed in the broader context of the state of the art in fracture modeling in

earthquake engineering.

Since the micromechanical-based fracture model is based on continuum stress and strain

quantities, the model is sensitive to the accurate description of the material constitutive response

as well physical phenomena which interactively affect the continuum stresses, especially for post-

buckling behavior. Thus, issues associated with finite-element modeling, such as proper

calibration of material constitutive models, and modeling of initial brace imperfections and cross-

section local buckling are given special attention when developing the fracture prediction

methodology.

5.1 DEVELOPMENT OF THE CYCLIC VOID GROWTH MODEL (CVGM)

This section presents the fracture criterion for the CVGM model. The model is expressed in terms

of continuum stresses and strains which intend to reflect void growth and coalescence during

inelastic cyclic loading. Next, material constitutive model calibration is discussed as the accuracy

of the CVGM depends on correctly simulating phenomena such as necking and buckling which

are highly sensitive to material constitutive response. Then, the calibration procedure for the

CVGM fracture parameters is reviewed. Finally, the accuracy of the CVGM fracture predictions

for small-scale (on the order of several inches) details is discussed.


140

The CVGM is derived from the findings of McClintock (1968) and Rice and Tracey (1969) that

the monotonic growth rate of a single spherical void, dr/r, in an elastic-perfectly plastic

continuum is exponentially related to the hydrostatic stress state, m

dr 1.5 m p
0.283exp d (5.1.1)
r Y

Where Y is the yield stress, and dp is the incremental equivalent plastic strain defined by

2 p
d p d ij .d ijp (5.1.2)
3

For an incremental plastic loading excursion, from pn-1 to pn, Equation 5.1.1 can be integrated to

determine the new void size in relation to the void size at the previous step

np p
dr
n
1.5 m p
r p
C exp
Y
d

(5.1.3)
np1 n 1

np
np 1.5 m p
ln r p C exp d (5.1.4)
n1
np1
Y

If the new void size at pn is Rn, and the original size at pn-1 is Rn-1, the above expression can be

used to describe the ratio of the new void size to the previous void size

p
R n 1.5 m p
ln n C exp
Rn 1 np1 Y
d (5.1.5)

Subsequently, DEscata and Devaux (1979) suggested replacing the yield stress by the effective,

or von Mises, stress, e, to provide a better description of void growth in the presence of strain

hardening. Following this refinement, the ratio m/e referred to as the stress triaxiality, T, is a

scalar quantity which affects the rate of void growth in ductile metals during inelastic loading.

Equation 5.1.5 is based on analytical derivations of the growth of a single void and does not

explicitly account for the interaction effects of neighboring voids. However, several researchers
141

including Hancock and Mackenzie (1976), Panontin and Sheppard (1995), Chi and Deierlein

(2000) and recently Kanvinde (2004) have demonstrated the effectiveness of this type of criterion

to predict fracture during monotonic tensile loading. To facilitate calibration, and thus the

prediction methodology itself, Equation 5.1.5 is best expressed in terms of the void ratio at any

plastic loading increment, Rn, with respect to the original void size, R0

np
R
ln n C exp 1.5T d (5.1.6)
p

R0 0

With this Void Growth Model (VGM), as it will be referred to henceforth, simple notched-bar (to

generate the necessary triaxial stress state) tensile tests and complementary finite-element

simulations are used to calibrate the material dependent, critical void size, Rcrit. From the

equivalent plastic strain and triaxiality evolution at the critical fracture point, it is assumed that at

fracture, Equation 5.1.6 describes a fracture parameter,

ln
Rcrit
R0
crit
p

exp 1.5T d
p
(5.1.7)
C 0

A monotonic fracture criterion can then be proposed as

np

exp 1.5T d (5.1.8)


p

While Equation 5.1.8 is developed by considering the growth of a single void, the calibration of

implicitly accounts for void interaction. By evaluating at the first occurrence of a macroscopic

crack, the parameter describes void-cluster coalescence rather than a critical size of a single void.

Thus, the criterion tracks the critical void size that will lead to necking instabilities between voids

as illustrated in Figure 5.1. In general, when the stress and strain gradients are relatively shallow,

the fracture criterion is satisfied over a large volume of material. If the criterion is evaluated in

the presence of sharp stress and strain gradients (for example, at a crack tip), then it must be

satisfied over a characteristic volume of the material. This volume is considered a material
142

property and is usually presented as the characteristic length, l*. In the context of the shallow

stress and strain gradients from the notched-bar and brace geometries discussed in this chapter, it

will not be necessary to consider the characteristic length effect.

While the intent of the VGM is to model void growth under positive triaxiality (T > 0), i.e. tensile

hydrostatic stress, it cannot simulate void collapse under a negative (or compressive) triaxial

stress state during cyclic loading such as observed during earthquakes. To account for void

collapse during negative triaxialities, Kanvinde (2004) proposed the following modification to the

VGM

R np n np n
p p

cyclic ln n exp 1.5T d


p
exp 1.5T d p 0 (5.1.9)
R0 0 np1

T 0

0 p
n1

T 0

This expression will be referred to as the CVGM and introduces a second term to the original

VGM to account for loading excursions with negative triaxialities (T < 0). Subtracting this term

simulates the effect of void collapse mechanisms. Following Kanvinde (2004), different loading

rates for growth and collapse are not considered. Therefore, is assumed to be equal to 1 for this

study. In addition, cyclic, or ln(Rn/R0), is restricted to values greater than or equal to zero. This

may be interpreted to imply that Rn/R0 remains greater than or equal to one, or Rn R0, i.e., the

void size is assumed to remain larger than or equal to the unity. However, this cannot be

independently verified, but as indicated by Kanvinde and Deierlein (2007) and discussed

subsequently in this chapter, this assumption provides results that are in good agreement with

experimental observations.

Similar to the VGM, a fracture criterion is also specified for cyclic loading. However, unlike

monotonic loading, the fracture toughness is assumed to reduce from the effects of cyclic loading.
143

The formulation by Kanvinde and Deierlein (2007) account for this damage by specifying the

CVGM fracture criterion as

cyclic f D (5.1.10)

Where f(D) is a damage function which reduces (i.e., f(D) 1) the monotonic fracture parameter,

. The damage is assumed to occur primarily during compressive loading when voids collapse

into a more oblate shape, introducing damage at the corners of the void. This damage is verified

in Figure 5.2 by examining the fractured surfaces of a steel material after monotonic and cyclic-

induced fracture. Referring to the figure, Scanning Electron Microscope (SEM) images of

fractured surfaces from Kanvinde and Deierlein (2007) are shown for monotonic and cyclic

loading (Figure 5.2a and 5.2b, respectively). A contrast can be seen between the smaller,

somewhat squished dimples formed prior to cyclic fracture initiation as compared to the larger,

more spherical dimples from monotonic fracture. To reflect the damage to the fracture toughness,

an exponential damage function is specified by Kanvinde and Deierlein (2007) as


f D exp p* (5.1.11)

Where is a model parameter and the damage is determined as the equivalent plastic strain at any

load reversal point (p*) determined by a switch from a negative to positive triaxiality. Thus, if

plotted versus cycle number, the damage takes the form of a decreasing step function. This will

be illustrated in the following sections.

5.1.1 MATERIAL BEHAVIOR, CONSTITUTIVE MODELS AND CALIBRATION

Considering Equations 5.1.9 and 5.1.10 above, the accuracy of the fracture predictions depends

on an accurate finite element simulation of the material constitutive (stress and strain) response at

the critical fracture locations. Since the simulations are conducted at a continuum, rather than a

micromechanical scale, the material constitutive models cannot directly simulate the

micromechanical processes responsible for material plasticity (such as dislocation motion). This
144

results in two issues in the context of this study (1) The constitutive models themselves may not

be sophisticated enough to capture complex hardening behavior, especially under cyclic non-

proportional loading, resulting in situations where it is impossible to develop calibrated parameter

sets that simulate response accurately for all experiments and (2) The calibration is based on

phenomenological, rather than physical procedures, resulting in a high degree of subjectivity and

tedium, in the calibration process.

When developing constitutive relationships for steel and other common alloys, several

simplifying assumptions are typically made to describe the average, macroscopic, material

behavior. First, it is common to assume that steel is both homogeneous and isotropic. Material

homogeneity is best described through the statement that, for any loading condition, every

material point in the body will show an identical response. Isotropy is the assumption that the

material has identical properties in all directions and is not orientation dependent. For small

deformations, the material response can be classified as linear elastic in that the stress can be

uniquely related to the strain, and vice versa, through a material dependent constant referred to as

the elastic (or Youngs) modulus. However, at larger deformations, ductile materials yield and

deform inelastically as illustrated in the uniaxial stress versus strain response of Figure 5.3a.

Here, the elasticity assumption no longer holds as the work done on the body can not be fully

recovered during load reversal. This is primarily due to permanent deformations that accompany

dislocation (i.e., line imperfections in the material matrix) motion. For implementation into an

analysis program such as ABAQUS (2004), uniaxial, inelastic stress/strain data (along with the

elastic properties) can be used with an isotropy assumption to describe the material response

along any loading plane. Following this assumption, this type of material model is typically

referred to as an isotropic hardening model and is usually represented by the growth of a (yield)

surface in multi-dimensional stress space such as Figure 5.3b instead of the uniaxial stress

increase of Figure 5.3a.


145

During cyclic loading, cold-working anisotropic effects are introduced through dislocation

multiplication and pile-up. The movement of dislocations in one direction generates local back

stresses which tend to assist movement in the opposite direction upon load reversal. This

produces the well-known Bauschinger effect, decreasing the yield stress in compression, for

example, if the material was first loaded in tension. Since loading direction determines the

evolution of the stress state, the isotropic assumption is no longer valid. To account for this

anisotropy, the yield surface (i.e., consider the surface shown in Figure 5.3b) is allowed to move

as well as grow, thus introducing a combined kinematic-isotropic hardening rule. The kinematic

hardening rule is often described in terms of a back stress (typically designated as ), named from

the effect of dislocation pile-up behavior described previously. For example, the Armstrong-

Frederick (1966) model is often used to decribe the incremental back stress vector, dij

d ij
C
0
ij ij d p ij p (5.1.12)

Where C and are material parameters, dp and p are the incremental and current equivalent

plastic strain quantities, respectively, defined by equation 5.1.2. The growth of the yield surface,

or the isotropic behavior, is described by 0


0 Y Q 1 e b
p

(5.1.13)

Where Q and b characterize the maximum size of the yield surface and the saturation rate,

respectively, and Y is the value of the yield stress as before. With the back stress tensor fully

defined, the yield surface, f, can be written as

3
f
2
Sij ij Sij ij Y 2 0 (5.1.14)

Where Sij is the deviatoric stress tensor. Figure 5.4 illustrates this combined hardening model in

principal stress space (ABAQUS, 2004). As illustrated in Figure 5.4, the ratio of C/ is

proportional to the radius of the yield surface and describes the saturation limit of the back stress
146

component, . Considering the saturation value of Equation 5.1.14 at large plastic strains, the

model is bounded by a limit surface of the size proportional to Y + Q + C/. Armstrong and

Frederick (1966), and later, Lemaitr and Chaboche (1990), along with the ABAQUS Theory

Manual, version 5.5 (ABAQUS, 2004) can provide further details.

From Equations 5.1.13 and 5.1.14, the isotropic-kinematic hardening law necessitates the

calibration of five material parameters, Y, Q, b, C and , to fully describe the model. The yield

stress, Y, is relatively easy to obtain from a standard tension coupon experiment. The isotropic

hardening parameters, Q and b, are calculated from a uniaxial cyclic test with multiple

increasing symmetric plastic excursions. From this data, a first estimate can be made on the

maximum size and the rate of growth of the elastic envelope. Once the yield stress and isotropic

hardening parameters are specified, the kinematic hardening components can be calibrated to

provide the best fit across cyclic and monotonic tests. Inevitably, this process (outlined in Figure

5.5 and discussed more in subsequent sections) requires some iteration to properly tune the model

parameters.

5.1.2 CVGM CALIBRATION

This section describes the calibration procedure for the two CVGM parameters, and . Unlike

the material constitutive model which is calibrated with smooth uniaxial coupon tests, the CVGM

parameters are calibrated with circumferentially notched-bar specimens. The notch, such as

shown in Figure 5.6a, produces a high-triaxiality with a very shallow gradient over the central

region of the notched cross section. Since the VGM reflects the processes of void growth and

coalescence that typically occur under high triaxiality values (T > 0.4; Bao and Wierzbicki,

2005), the notched bar geometries are fabricated to generate relatively high triaxiality values.

Referring to Figure 5.6a, the magnitude of the triaxiality is controlled by the ratio of the neck
147

diameter, D0, to the notch radius, r*. A smaller r*/D0 suggests a larger constraint, and therefore a

higher triaxiality in the notch region, as compared to a larger r*/D0 which provides less

constraint.

5.1.2.1 CALIBRATION OF

To calibrate , the notched bar tests are complemented by finite element simulations, such as

described in detail by Kanvinde (2004) and shown in Figure 5.7. These feature an axisymmetric

finite element model, incorporating the material constitutive relationships described in the

previous section. Figure 5.6b shows good agreement between the simulated and the experimental

load-deformation response, including the point of experimental fracture. The stress-strain

histories leading up to this point are integrated according to Equation 5.1.8, described earlier, to

generate an estimate of for each experiment. The notch geometry ensures that the highest

triaxiality is observed at the center of the notched cross section, requiring this integration only at

one node, i.e. the central node of the notched cross section (Refer Figure 5.7). Repeating this

procedure for each experiment provides statistical information regarding the parameter .

The parameter, may also be indirectly inferred if the upper shelf Charpy V Notch Energy is

known or specified. For this purpose, an empirical relationship (Kanvinde and Deierlein, 2006)

between the Charpy V-notch (CVN) upper-shelf energy, ECVN (ft-lbs) may be used.

0.018 ECVN 1.30 (5.1.15)

5.1.2.2 CALIBRATION OF

Once is calibrated, cyclic notched-bar experiments and complementary finite element

simulations are used to calibrate . Referring to Equation 5.1.10 and Figure 5.8, is obtained

from an exponential regression analysis through the average monotonic fracture parameter (0, )
148

and multiple cyclic data points (p*, cyclic) such that the x-coordinate indicates the accumulated

damage at the beginning of the failure cycle, whereas the y-axis coordinate plots the cyclic/ at

the instant of fracture observed in the experiment. Referring to a previous discussion, the damage

is assumed to be the equivalent plastic strain at the beginning of the tensile excursion of

experimental fracture. Also indicated on Figure 5.8, an exponential function of the form

f(p*)=exp(-.p*) is fit to this data through a regression fit, resulting in a calibrated value of the

CVGM parameter, . The regression fit is constrained to pass through the point (0, ) to

appropriately reflect monotonic fracture.

Multiple cyclic tests are used to query a range of damage values prior to fracture. The damage

prior to fracture can be controlled by varying the loading history and/or the specimen triaxiality.

In general, specimens with lower triaxiality will accumulate greater damage prior to fracture as

compared to specimens with high triaxiality. The calibration range can be determined from the

expected magnitude of damage where fracture is predicted. This may necessitate an approximate,

a priori simulation of the component of interest. The notched-bar tests used to predict large-scale

brace fracture in this study highlight the possible shortcomings of using cyclic material

calibration experiments which do not sample appropriate damage levels. This will be addressed in

a subsequent section. In addition, the cyclic loading histories for the notched-bar calibration tests

are designed according to the applicable range of ULCF where a relatively small number (< 5-10

cycles) of very large inelastic cycles are applied prior to fracture.

5.1.3 CVGM APPLICATION AND VALIDATION IN SMALL-SCALE DETAILS

Once the model parameters ( and ) are calibrated, along with the material constitutive model,

the CVGM can be used to predict fracture. Figure 5.9 illustrates the fracture prediction for a

cyclically loaded notched bar. A detailed finite element analysis is used to determine the stress
149

and strain histories at the center of the notched-bar the location of experimental fracture

initiation. The increasing solid line is calculated according to Equation 5.1.9 and represents the

demand, or cyclic, at the critical material point, while the dashed line is the fracture capacity

described by Equation 5.1.11. Note that the capacity, following Equation 5.1.11 and prior

discussions, degrades in a stepwise fashion. Fracture is predicted to occur when the demand

exceeds the capacity, as indicated on Figure 5.9. The instant in the simulation corresponding to

this event can be compared to the experimental fracture instance to evaluate accuracy of the

CVGM. Note that the prediction shown in Figure 5.9 is based on mean and values.

Probabilistic aspects of these predictions will be discussed in a subsequent section.

Kanvinde and Deierlein (2007), demonstrated the effectiveness of the CVGM in predicting

fracture in various material types and notch geometries, exposed to different loading histories.

Referring to Figure 5.10 (adapted from Kanvinde and Deierlein, 2007), the accuracy of the

CVGM is demonstrated by comparing the equivalent plastic strain at the predicted fracture

instance, panalytical, to the equivalent plastic strain at experimental fracture initiation, pexperimental.

The monotonically increasing equivalent plastic strain is used in the comparison for the

convenience it allows in identifying a specific instance within a cyclic loading history. Thus, the

solid line corresponds to a perfect prediction while the dashed lines indicate a 25% error margin

between the prediction and the experiment. In general, the study concluded that the CVGM

performs well for the material types, loading histories and scale of the 46 notched-bar

experiments.

5.2 APPLICATION OF CVGM TO LARGE-SCALE BRACING COMPONENTS

Considering the accuracy of the CVGM predictions applied to small-scale notched bar

experiments, this section evaluates the methodology for large-scale details. The focus of the

discussion will be validating the CVGM using the representative SCBF brace component tests of
150

Chapters 3 and 4. Similar to the previous discussion, the methodology will consist of material

model and CVGM calibration, as well as large-scale finite element brace simulations. Analogous

to the notched-bar analyses, the accuracy of the brace fracture predictions depend on accurately

simulating the stress and strain histories at the critical point of fracture. In the context of the

buckling-induced fracture processes in brace component behavior, the accurate simulation of

complex phenomena such as local buckling is shown to control the fracture prediction accuracy.

5.2.1 MATERIAL MODEL CALIBRATION

To reintroduce the large-scale testing matrix (review Chapter 3 and 4 for more detail), Figure

5.11 illustrates the three cross-section types used in the experimental program where two HSS

(HSS4x4x1/4 and HSS4x4x3/8), two Pipe (Pipe3STD and Pipe5STD) and one Wide-Flanged

shape (W12x16) comprise the nineteen experiments. Also shown in the figure are locations along

the cross-section where material coupons were extracted for constitutive model calibration as well

as CVGM parameter calibration. The dimensions for the circumferentially notched and smooth

tensile coupons are shown along with the number of specimens extracted from each location. The

large-notched (LN) bar is subjected to a cyclic loading history (illustrated in Figure 5.12e) while

the other specimens are monotonic tension experiments. Also shown in Figure 5.11, are typical

flat tension coupons from both HSS cross-sections and fabricated according to ASTM Standard

E8 (2005): Standard Test Method for Tension Testing of Metallic Materials.

Using the experimental data from these specimens, Figure 5.5 illustrates the calibration procedure

for the combined isotropic-kinematic hardening model described previously. First, an average

yield stress (Y) is calculated from the smooth and ASTM tension coupons using a 2%-strain

offset rule. These measurements are listed in Table 5.1 along with other material properties. Next,

the circumferentially large-notched cyclic experiments are used to approximate the isotropic

hardening law (i.e., Q and b). The specimen is modeled in ABAQUS using the axisymmetric
151

element formulation and loaded with the cyclic history shown in Figure 5.12e (cycles of +0.15/-

0.05 inches across a 1 inch gage length). From experience, the initial kinematic hardening

coefficients are taken as C = 300 and = 50 with Q, and b set to 15 and 5, respectively. The

isotropic parameters are adjusted using 2-4 analyses to obtain a first approximation. An iterative

approach is used as it is difficult to rigorously characterize the growth of the yield surface with

the relatively large, and few numbers of, plastic excursions during the cyclic loading of the

notched-bar test. Ideally, a uniaxial specimen with a symmetric loading history comprised of a

large cycle count would be used to measure the growth of the yield surface. However, as

mentioned previously, a large cycle count combined with a smooth geometry does not facilitate

the CVGM calibration. Therefore, to reduce fabrication and testing costs, the approximate cyclic

material behavior is obtained from the notched-bar experiments with the final check being the

large-scale brace force-deformation behavior (see Figure 5.5).

After the isotropic hardening parameters, Q and b, are approximated from the large-notch

experiments, the smooth and small-notched monotonic tests, along with the first pull of the large-

notched cyclic test, are used to calibrate the kinematic hardening rule. The isotropic parameters

can be slightly adjusted during this stage, keeping in mind the approximate values obtained from

the large-notch experiments. The new kinematic and isotropic parameters are then applied to the

large-notch analysis to ensure the cyclic behavior is acceptable. Figure 5.12 compares the

simulation results after this calibration procedure with the experimental data from the corner

material specimens of the HSS4x4x1/4 cross-section. A single set of parameters (Y = 73, Q =

850, b = 160, C = 14.5 and = 5.27) for the combined hardening law is able to describe the

behavior of the various specimens fairly well. This list of parameters is recorded in Table 5.2

along with the other material types. The same procedure is used to calibrate the material for the

walls of the HSS brace. After both material types are defined, a large-scale brace continuum
152

model (discussed more in a later section) is analyzed with the small-scale parameters to verify the

constitutive response at the large-scale. Figure 5.13 compares the simulated and experimental

force-deformation response for an HSS4x4x1/4 bracing member subjected to a far-field loading

history. Also of importance is the instance of local buckling in the simulation as compared to the

experiment. Buckling is addressed more in a later section.

While the good agreement between the experiments and simulations presented in Figures 5.12

and 5.13 is representative of all material types and bracing members, only the HSS4x4x1/4 and

Pipe5STD brace/material types are calibrated with a single set of hardening parameters (i.e., Y,

Q, b, C and ). The HSS4x4x3/8 and Pipe3STD brace/material types required different

parameter sets to describe the behavior of the small and large-scale specimens. Ideally, a unique

constitutive model parameter set could predict material behavior across the variety of scales,

geometries and loading histories. However, this is somewhat of an unrealistic aim as the

constitutive model should be considered imperfect, and whose parameters may not be applicable

across a wide range of inelastic strain levels (ABAQUS, 2004). In fact, in the case of the

HSS4x4x3/8 and Pipe3STD braces which, in general, survived several more inelastic cycles as

compared to the less compact HSS4x4x1/4 and Pipe5STD (see Chapter 4) the repeated loadings

could be highlighting an inherent cumulative error effect of the combined hardening model.

Considering that the objective of the current study is to apply and evaluate the CVGM at the

large-scale, the finite element simulations, and therefore the constitutive model, are regarded as a

tool which can reproduce the stress and strain histories of the experimental specimens. In the case

where the calibration procedure outlined in Figure 5.5 is not applicable across all experiments, the

parameters are adjusted accordingly so as to capture the observed experimental behavior. For

example, for the small-scale specimens, the load-deformation behavior was used to determine

accurate material modeling, while for the large-scale brace experiments, it was considered

paramount to predict the instance of local buckling. While local buckling is dependent on brace
153

geometry through brace slenderness and cross-section compactness, material hardening also

influences behavior. The modeling of the large-scale brace simulations and buckling phenomena

is discussed in the following section.

5.2.2 LARGE-SCALE BRACE SIMULATIONS

Figure 5.14 illustrates a representative finite element model for a brace specimen constructed

using the ABAQUS simulation software (ABAQUS, 2004). The simulations featured three-

dimensional, twenty-node brick elements with reduced integration (type C2D20R in the

ABAQUS element library), which were determined to simulate through-thickness bending of the

specimen walls with a high degree of accuracy. The simulations were based on the measured,

rather than the nominal dimensions of the braces (see Figure 5.15). To reflect experimental

conditions accurately, a 1.5 inch clear distance between the end of the brace and the gusset plate

was incorporated in all brace models (Figure 5.14). Symmetry in the boundary conditions,

specimen and the deformation mode are leveraged to construct the quarter model as shown in the

figure.

Typical meshes for various cross sections are shown in Figure 5.14. Referring to the figure, the

mesh is finer at the center of the brace where severe plastic deformations are expected. The mesh

gradually transitions to a coarse mesh at the third point of the brace, such that the element lengths

increase in a geometric progression. A total of 25 elements are used along the middle-third of the

half-brace, with the elements in the hinge region one-fourth the size of the elements at the ends of

this region. A detailed mesh-refinement study verified the accuracy of this mesh.

5.2.2.1 MODELING GLOBAL AND LOCAL BUCKLING

Since the strains responsible for fracture are controlled by global and local buckling of the brace,

it is important to accurately simulate these events in the brace simulations. To achieve this, global
154

and local imperfections are introduced to the brace model by scaling eigenvectors from elastic

buckling analyses. The following subsections discuss these aspects of modeling, and examine the

accuracy of the adopted approach.

5.2.2.1.1 IMPERFECTIONS AND BUCKLING SIMULATIONS

Elastic buckling analyses are performed with ABAQUS to determine the buckling mode shapes

of each brace specimen. The global buckling analysis involves the application of a compressive

axial load at the gusset plate while a separate local buckling involves the application of a

compressive load at the partition between the fine and coarse mesh region, 20 inches from the

brace mid-point. Figure 5.16 shows the buckling mode shapes obtained from the elastic buckling

analyses for the HSS4x4x1/4. These buckling modes are scaled appropriately and incorporated as

deformation imperfections into the inelastic cyclic loading simulations discussed subsequently.

The scaling procedure is discussed next.

The global imperfections were introduced by scaling the global buckling mode shape, such that

the peak deflection at the center was L/1000, which is equal to the maximum sweep of

compression members assumed by AISC (2005). Thus, the elastic global buckling eigenvector is

scaled such that the maximum out-of-plane imperfection is 0.12 inches. Following the

measurement procedures by Schafer and Pekz (1998), the local imperfection magnitudes are

determined from digital end-mill measurements along the walls of four-foot sections of each

brace cross-section type (with the exception of the Wide-Flange section). These measurements

are illustrated in Figure 5.17 and listed in Table 5.3 for the HSS and Pipe cross-sections. The

maximum values from each brace cross-section listed in Table 5.3 were chosen as the local

imperfection magnitudes in the brace analyses. Interestingly, these magnitudes do not influence

the final solution. In fact, virtually identical stress and strain histories are obtained without
155

incorporating local imperfections. Considering the relatively small imperfections, the solution

tends to be dominated more by the plastic strains accumulated during repeated cyclic loading.

5.2.2.1.2 COMPARISON TO EXPERIMENTAL GLOBAL BUCKLING

The first inelastic event during the far-field cyclic loading experiments is brace buckling, while

the near-field loading histories yield the brace in tension prior to buckling (refer to Chapters 3 and

4 for loading history descriptions). Buckling, often described by the maximum, or critical,

compressive load is dependent on member geometry, material properties, and imperfection

magnitudes. Figure 5.18 compares the experimental buckling loads (presented earlier in Chapter

4) to the simulated buckling loads from the finite element models. On average, the ratios of the

predicted to experimental buckling load are 1.06 and 1.35 for the far-field and near-field

simulations, respectively. The HSS4x4x1/4 simulations are used as an example to compare the

larger predicted buckling loads for the near-field loading history (Figure 5.18b) to the far-field

loading (Figure 5.18a). These observations from the near-field buckling analyses may be

explained by the lack of axial residual stresses in the model as well as the tension loading prior to

buckling (see Chapter 4) which has the effect of removing the initial global imperfections.

However, in the context of predicting fracture, it is not critical to accurately predict the buckling

load as the plastic strain accumulation from local buckling overwhelms global buckling strains.

Local buckling simulation is discussed in the next section.

5.2.2.1.3 COMPARISON TO EXPERIMENTAL LOCAL BUCKLING

Referring to the qualitative description of brace failure from Chapter 4, the amplified inelastic

strain demand from local buckling is primarily responsible for fracture initiation in the middle

plastic hinge of the brace. For accurate application of the CVGM, the instant of local buckling

initiation must be correctly modeled to accurately reproduce the continuum stress and strain

histories at the critical fracture point. Figure 5.19 illustrates the progression of global and local
156

buckling for an HSS4x4x1/4 bracing member at the compressive peaks of a far-field loading

history. The figure compares the recorded experimental events to the continuum-based simulation

events. The solid dot on the second compressive excursion to 1.1 inches indicates the first visible

local buckling event recorded during the experiment. This point is accurately predicted by the

simulation as well. This is representative of all brace analyses presented in this chapter.

5.2.3 BRACE MATERIAL CVGM CALIBRATION

This section presents the results of the small-scale material calibration study for the various brace

material types. While a previous section discussed the calibration of the material constitutive law,

here the focus is on calibrating the CVGM parameters, and .

5.2.3.1 BRACE MATERIAL CALIBRATION RESULTS:

Monotonic tension notched bar specimens, extracted from various locations within each brace

cross-section, were used to calibrate the fracture parameter . Figure 5.11 shows the locations

where material coupons were sampled from each brace type. Referring to the HSS experimental

results in Chapter 4, fracture initiates at the corner of the tubes rather than on the face. Therefore,

more calibration tests are performed on the HSS corner material as compared to the wall material.

Referring to the Pipe cross-section material, the welded seam material is different as compared to

the base metal. For the Wide-Flange, only the flange material is investigated as local buckling

initiates in and remains restricted to the flanges.

Table 5.4 presents the calibration results for each material type investigated as part of the small-

scale study. The materials are listed by cross-section type and the location within the cross-

section. For each monotonic notched-bar experiment, the fracture deformation, f, is presented

along with the corresponding fracture parameter, (the calibration results are presented in the
157

next section). Axisymmetric finite element simulations are used to determine the stress and strain

histories at the fracture location of the notched specimen and Equation 5.1.7 is used to estimate .

In this way, each experiment/simulation combination provides a single estimate of , while

multiple experiments provide statistical data on . Table 5.4 provides average values, along with

the Coefficient of Variation (COV), for each material type.

Referring to Table 5.4, the HSS4x4x1/4 corner material is the toughest material tested with an

average of 5.98. It is interesting to note the low toughness of the same (nominal) material type

(A500 Grade B specified by AISC, 2005) at the corner of the HSS4x4x3/8 where the average is

found to be 3.62. This may be due to different plate materials and thicknesses used to form the

two cross-sections. A comparison between the mean toughness values of the HSS wall and corner

materials illustrates the effect of cold working on the steel ductility. The HSS4x4x1/4 corner

material was found to be 12.4% less tough as compared to the wall material, while a 3.7%

decrease was observed in the HSS4x4x3/8. However, it is relevant to note that despite the

apparent trends observed in the mean toughness values, the variations in the wall toughness

measurement are large (COV = 22 and 35% for the HSS4x4x1/4 and HSS4x4x3/8, respectively).

Referring to the toughness values of the two Pipe materials, the 3 inch Pipe toughness of 4.40 is

considerably larger than the less ductile 5 inch Pipe with an average equal to 2.46.

5.2.3.2 BRACE MATERIAL CALIBRATION RESULTS:

Referring to the discussion in the first section of this chapter is calibrated through an

exponential regression fit. The regression analysis uses experimental data and finite element

simulations of cyclic notched-bar tests. The notched-bar tests measure the toughness of the

material as a function of inelastic damage during cyclic loading. Since the small-scale calibration

informs the large-scale fracture prediction, the calibration tests should ideally sample material
158

behavior over the range of damage expected in the large-scale specimen. This recommendation is

offered after studying the CVGM fracture predictions of the large-scale braces. While the

predictions are discussed in the next section, the different damage levels between the calibration

and brace experiments lead to inconsistent fracture predictions in some cases. As illustrated in

Figure 5.20, the lower triaxiality in the braces (0.4-0.5) also results in the higher damage levels as

compared to those observed in the notched bars, where the triaxiality is significantly higher (in

the range of 1.2-1.3). Furthermore, only three cyclic tests were conducted for each material type

(four for the HSS4x4x1/4 corner material) and the loading history was not varied between

specimens. Thus, the cyclic data is somewhat limited from the small-scale study. The values

calibrated for each material type are presented in Table 5.4. At the conclusion of this section a

more detailed discussion is provided on the effect of insufficient small-scale cyclic data on

fracture prediction in the large-scale components.

Figure 5.21 illustrates the degradation of fracture capacity in the small-scale cyclic calibration

tests. From section 5.1, the x-coordinate indicates the accumulated damage at the beginning of the

failure cycle and the y-axis coordinate plots the cyclic/ at the instant of fracture observed in the

experiment. Following the arguments of Kanvinde and Deierlein (2007), the damage is

considered equal to the equivalent plastic strain (p*) at the beginning of the final tensile pull to

fracture. Also shown in Figure 5.21 is an exponential function which is fit to this scatter data

through a regression fit, resulting in a calibrated value of the second CVGM parameter . The

regression fit is constrained to pass through the point (0, ) to appropriately reflect monotonic

fracture.

The HSS wall and Pipe seam material are not included in Figure 5.21 since fracture did not

initiate in these materials during the large-scale tests. The Wide-Flange braces will not be
159

included in the large-scale fracture predictions, so this material is also excluded. Following the

procedure outlined in the preceding paragraph, -values are generated for each material type, and

are summarized in Figure 5.21, and listed in Table 5.4. The subscript NB in the figure and the

Table indicates that the trend fit is done using only the notched-bar (NB) experiments (in contrast

to a more expanded data set, discussed next).

It is also useful to compare the damage incurred to the large-scale braces with the damage to the

notched-bar experiments. Introduced previously, Figure 5.21 also includes scatter points for the

braces in a manner similar to the scatter points for the notched bars described previously. It is

interesting to note that in general, these points are below the trend fit based only on the notched

bar scatter points. The dashed line on the figure shows a similar trend fit; however, this fit is

based on the entire data set (including the brace scatter points). Correspondingly, the subscript C

on the calibrated -values indicates that it is based on a combined data set. The C values, for

each of the HSS and Pipe steels are higher than the corresponding NB values. This indicates that

calibrating the -value based only on a data set that queries a limited range of damage may be

inaccurate. Moreover, for the specific situation discussed here, the larger damage observed in the

brace specimens implies that using only the notched bar specimens, will result in a lower bound

estimate of the -value. Since lower -values shifts the trend fit upwards, brace fracture

predictions based on the NB values, in general, correspond to a later instance in the loading

history as compared to the experimental observations.

While the next section describes the prediction results in more detail, here the likelihood of

obtaining the results from the notched-bar experiments is examined assuming that the true

behavior of the material is the exponential function described by the C-value. In other words, if

the small-scale experiments were repeated, what is the probability that the characteristics of the
160

notched-bar data points (i.e., mean and standard deviation), relative to the exponential function

described be the C-value, will be repeated? This is interesting to consider because, for each

material type, it quantifies the likelihood that the notched-bar experiments follow the material

behavior described by the C exponential function. Therefore, the combined damage function

represents the population while the notched-bar experiments represent the sample. To investigate

this, an error quantity, , is first introduced to describe the vertical difference from any

experimental point to the damage law. From the notched-bar and brace experiments, a normal

probability distribution is generated where the mean at = 0 is, by definition, located on the

exponential damage law defined by C. This is illustrated in Figure 5.22. Second, a null

hypothesis is proposed which states the mean epsilon of the population is equal to zero (H0: =

0) while an alternative hypothesis states that the mean is not equal to zero (HA: 0). Finally,

utilizing the Students t-distribution and assuming the population error () is normally distributed,

a P-value can be computed to assess the probability that the null hypothesis is true. A small P-

value suggests the null hypothesis may not be true and leads to the conclusion that the notched-

bar results are unlikely assuming the behavior of the population is governed by the C exponential

function from the notched-bar and brace experiments. For the purposes of this discussion, a small

P-value is defined as less than 5% (Montgomery et al, 2001).

Following the hypothesis stated above, the P-values for the HSS4x4x1/4, HSS4x4x3/8,

Pipe3STD and Pipe5STD materials are calculated as 30.0, 3.4, 0.1 and 87.6%, respectively.

Considering the large P-values for the HSS4x4x1/4 and Pipe5STD materials, the null hypothesis

can not be rejected. From this, it can not be stated that with 95% certainty the mean error of the

notched-bar data is inconsistent with the exponential damage function defined by C. While the

HSS4x4x3/8 fails the hypothesis test, the P-value is relatively close to the predefined degree of

confidence (0.95). On the other hand, the null hypothesis is certainly rejected for the Pipe3STD
161

material due to the low sample standard deviation in . A small standard deviation increases the

value of the test statistic since it is unlikely (with respect to the null hypothesis) that repeated

experimentation will consistently produce data points which are equal distance from the

exponential function. From this hypothesis testing we can conclude that for two of the brace

materials, the null hypothesis can not be rejected at a high degree of confidence, while the null

hypothesis for the Pipe3STD and HSS4x4x3/8 notched-bar samples is rejected with 95%

confidence. In summary, two of the four brace material types fail the t-test, concluding that for

these materials we can say with 95% confidence that the notched-bar data is inconsistent with the

true behavior of the material.

While hypothesis testing describes the likelihood that the notched-bar observations are not

consistent with the assumed true behavior, it does not reconcile the difference between the NB

and C calibration values for the two exponential functions. However, referring to Figure 5.21, it

is interesting to note that for the material types which failed the hypothesis testing, the brace

points are consistently below the dashed line representing the combined fit while the notched-bars

are above. Discussed in the next section, these notched-bar tests also provide the worst fracture

predictions. This may highlight a fault in CVGM such that different mechanisms are active at

lower triaxialities.

Furthermore, calibrating at small or large damage levels can influence the variation in . As

discussed previously and explained through the low triaxiality (relative to notched-bars) of the

brace experiments shown in Figure 5.20, the damage ranges for the two data sets are quite

different, such that the notched-bars incur less damage prior to fracture initiation. Assuming the

constant distribution (Myers, 2009) shown in Figure 5.22, the variation in the different values

can be explained by the behavior illustrated in Figure 5.23. Referring to the figure, the bounds
162

through plus and minus one standard deviation of the error distribution are tighter if the

calibration includes large damage values as compared to calibrating the function with samples

which have smaller damage values. As expected, this behavior is verified through Monte-Carlo

simulations. Thus, in lieu of matching the expected level of damage at the fracture location with

the calibration tests, the small-scale experiments could be designed to produce large damage

values to reduce the variation in .

5.2.4 CVGM FRACTURE PREDICTIONS APPLIED TO LARGE-SCALE BRACE COMPONENTS

This section presents the fracture predictions for the large-scale bracing members where the stress

and strain histories from the brace simulations and the fracture parameters ( and ) are used to

determine the instant when the CVGM fracture criterion (Equation 5.1.10) is satisfied. The

predictions are compared to the experimental instances of fracture presented in Appendix A and

discussed in Chapter 4. The accuracy and limitations of the CVGM are investigated through the

wide-range of brace cross-section and material types as well as the various loading histories and

rates of the large-scale testing matrix.

Deterministic fracture predictions based on the mean and values are first presented for the

far-field, general loading history experiments of the HSS4x4x1/4, HSS4x4x3/8, Pipe3STD and

Pipe5STD. Next, the CVGM is applied to the bracing members subjected to asymmetric, near-

field loading histories, to investigate the influence of loading history on the accuracy of the

predictions. Then, the fracture predictions are presented for the two HSS bracing members

subjected to earthquake loading rates. Finally, an unsymmetrical deformation shape, produced by

reinforcing plates at the middle of an HSS4x4x1/4 brace (HSS1-6), further examines the

performance of the CVGM with different stress and strain histories.


163

Following the deterministic results, a probabilistic approach is developed to account for the

uncertainty introduced into the predictions by the regression analysis used to fit the exponential

function. The error quantity (i.e., ) distributions, discussed previously, are used to generate

probabilities of failure at each instance during the loading history.

5.2.4.1 APPLICATION TO HSS AND PIPE CROSS-SECTIONS (STANDARD LOADING)

Figure 5.24 illustrates the CVGM prediction for an HSS4x4x1/4 bracing member (HSS1-1)

subjected to the standard loading history. Using the finite element analysis results and the

calibrated CVGM material parameters listed in Table 5.4, the critical node is determined to be at

the corner of the square cross-section. This point is identified on the finite element mesh in Figure

5.24 and coincides with the experimental location of fracture initiation. Also shown in the figure

is the increasing demand, or void growth (cyclic), and the decreasing capacity, or critical void

size, at the critical point of the brace cross-section. At the intersection of cyclic and the decreasing

capacity function, the fracture criterion (Equation 5.1.10) is satisfied and fracture initiation is

predicted to occur. The figure compares the expected fracture instances for two different capacity

functions according to the NB and C parameters calculated previously. Referring to the figure,

cyclic exceeds the fracture capacity calculated with the larger -value (C) one cycle before the

prediction with the smaller NB. This is expected considering that a larger -value decreases the

fracture capacity at a faster rate as compared to a smaller .

Figure 5.25a shows the CVGM fracture instances from Figure 5.24 as deterministic points on the

standard loading history of the HSS4x4x1/4 test where the predicted (i.e., using NB) and

experimental instances of fracture are approximately the same. Also shown in Figure 5.25a is the

expected fracture instance according to the larger C-value used to define the capacity of the

brace material. Referring to the figure, using the C-value, calibrated with both the notched-bar
164

and brace experiments, does not result in a good comparison to the experimental instance of

fracture. However, on average (across all HSS4x4x1/4 specimens), it is expected that the CVGM

damage function (Equation 5.1.11) will be more accurately characterized by the C-value because

it is based on the experimental points from the large-scale brace tests. The probabilistic

methodology presented in the following section will provide a better framework to assess the

performance of the CVGM with the two different -values.

Figure 5.25 also illustrates the fracture predictions for the HSS4x4x3/8, Pipe3STD and Pipe5STD

cross-section types subjected to the standard loading history. Referring to Figure 5.25d, the

CVGM estimates the instance of fracture initiation accurately for the Pipe5STD bracing member,

while the predictions for the HSS4x4x3/8 and Pipe3STD (Figures 5.25b and 5.25c, respectively)

are several cycles after the experimental fracture instances. These predictions are the result of the

small NB-values calibrated with the notched-bar tests. The smaller -values incorrectly estimate

the experimental damage of the Pipe3STD and HSS4x4x3/8 bracing members. However, the

larger C-values, calibrated with both the notched-bar and brace experiments, provide better

estimates of the damage to the Pipe3STD and HSS4x4x3/8 brace materials.

These results indicate that either 1) the CVGM is not an appropriate criterion to predict fracture in

large-scale bracing members or 2) calibrating the -value based only on a data set that queries a

limited range of damage may be inaccurate. Referring to the first point, the model may be

inappropriate to model fracture under the brace loading conditions due to several factors

including, but not limited to,

The relatively low triaxiality range at the critical location of fracture (0.4-0.5) in the brace

experiments. Considering the small-scale validation study by Kanvinde and Deierlein


165

(2007) included specimens with relatively large constraint (producing triaxialities in the

range of 1.0-2.0), the CVGM has not been tested in situations with low-triaxiality. This

could trigger somewhat different micromechanical failure mechanisms other than those

assumed by the CVGM.

The exponential damage function employed by the CVGM may be fundamentally flawed,

suggesting there is another function which better characterizes damage. While the

exponential degradation was shown to work well at the small-scale, the stress and strain

histories of the brace experiments are unlike the histories from the notched-bar tests. This

may alter the rate of damage accumulation in the braces relative to the previously

observed behavior in the small scale validation tests by Kanvinde and Deierlein (2007).

While investigating alternative functions is outside the scope of the current study, it is a

recommended topic for future work.

However, the expected fracture instances using the C-values illustrate that even if the functional

form of the damage law is inaccurate, and/or the low triaxiality alters the assumed void growth

mechanisms, the CVGM is still able to describe the behavior of the brace specimens.

Referring to the second point, the inaccurate -values could be a product of inappropriate small-

scale calibration tests such that the damage levels are inconsistent with the brace experiments.

Furthermore, referring to Figure 5.23, calibrating with the notched-bar data set may produce

larger variations in as compared to a data set with larger damage levels. Thus, if small-scale

experiments were designed to sample larger damage levels, the calibration may be more

consistent with the brace points. Nonetheless, the predictions from the HSS4x4x3/8 and

Pipe3STD brace types expose some faults of the CVGM if is not calibrated properly.
166

5.2.4.2 APPLICATION TO NEAR-FIELD LOADING HISTORIES

To examine the capabilities and limitations of the CVGM further, two near-field loading histories

(see Chapter 3) are used to generate stress and strain histories in the bracing members which are

unlike those observed in the specimens tested with the standard loading history. Considering the

CVGM operates at the continuum level, it is interesting to investigate the accuracy of the model

in the presence of different loading cases. For example, while the predictions from standard

loading may be accurate, a more compressive (or tension) dominated stress and strain history may

influence the accuracy of the model. Investigating various loading histories is also of practical

importance considering the unsymmetrical behavior of braced-frames during earthquake loading

(see Chapter 2).

In this section, the near-field loading histories presented in Chapters 3 and 4 are used to examine

the accuracy of the CVGM during asymmetric loading histories. Referring to the discussion in

Chapter 3, the asymmetric compression history is marked by a large pulse in compression

followed by deformation cycles about a residual compressive offset. This is applied to an

HSS4x4x1/4 bracing member. An asymmetric tension history is applied to both the Pipe3STD

and Pipe5STD bracing members and contains a large initial tensile pulse followed by deformation

cycles applied at a residual tensile offset.

Referring to Figure 5.26a, using the NB-value to describe the decreasing fracture capacity for the

HSS4x4x1/4 bracing member subjected to the compressive near-field history results in a

deterministic fracture prediction which is approximately four cycles after the experimentally

observed fracture instance. However, the C-value shifts the expected fracture instance closer to

the experimental. Figure 5.26b illustrates the prediction (i.e., using the NB-value) for the

Pipe3STD bracing member subjected to the tension near-field history. The figure shows the same
167

effect which was observed during the standard loading history of the Pipe3STD brace where the

low NB-value pushes the prediction several cycles past the instance of experimental fracture. On

the other hand, the fracture prediction instance for the Pipe5STD bracing member during the

asymmetric tension loading history is only one cycle after the experimental fracture instance.

Similar to the HSS4x4x1/4, the C-value provides a more accurate fracture initiation instance for

both Pipe specimens.

5.2.4.3 QUASI-STATIC VERSUS EARTHQUAKE LOADING RATE

Chapters 3 and 4 describe two earthquake-rate loading experiments, where the standard loading

history is applied to each of the HSS brace cross-sections (HSS4x4x1/4 and HSS4x4x3/8) at 360

times the loading rate of the quasi-static experiments. The justification for this rate is discussed

more in the previous chapters. Referring to the experimental observations discussed in Chapter 4,

the earthquake loading rates did not significantly alter the behavior of the bracing members as

compared to the quasi-static experiments. Thus, the accuracy of the CVGM fracture predictions

for these fast-rate loading experiments is not expected to be notably changed from the quasi-static

tests.

Inconsistent displacement limits (from actuator over-shooting described in Chapter 4) between

the quasi-static and fast-rate experiments make it somewhat difficult to judge if the increased

strain rate or temperature in the region of fracture had a substantial effect on the performance of

the brace. However, after both experimental histories are input into the finite element model of

the brace and fracture initiation is predicted according to the CVGM, the resulting deviations

between the prediction and the experiment are essentially the same for both tests. Therefore,

within the precision of the models, rate effects do not seem to affect fracture significantly. This is

illustrated in Figure 5.27 where (a) and (b) are replicates of the quasi-static results from Figure

5.25 and Figures 5.27c-d correspond to the earthquake-rate tests.


168

5.2.4.4 APPLICATION TO UNSYMMETRICAL BUCKLING

To further examine the capabilities of the CVGM, two eighteen inch reinforcing plates were

welded at the center to the top and bottom (non-buckling faces) of an HSS4x4x1/4 bracing

member to deliberately induce unsymmetric buckling. In all other respects, the specimen was

similar to the HSS4x4x1/4 experiment during far-field loading discussed previously. The welded

attachment is not believed to influence fracture substantially (other than by causing the non-

symmetric buckling pattern) since fracture initiation occurred approximately two inches from the

end of the reinforcing plate (see Figure 5.28a).

As expected, and shown in Figure 5.28c, the experimental ductility is less than that recorded for

the symmetric experiment shown in Figure 5.25a. This is due to the larger strains that develop

when the plastic hinge is not at the center of the brace and can be explained by considering the

kinematics of a buckling brace with varying plastic hinge locations. This behavior is modeled by

incorporating reinforcing plates into the brace finite-element model where a tie constraint is used

to replicate the welded connection between the plates and the brace. To simulate the

unsymmetrical buckling effect, the plates are offset by 0.1 inches from the brace mid-point.

Figure 5.28a illustrates the results of the continuum-based analysis which predicts local buckling

on one side of the reinforcing plates. The predicted fracture location is determined to be at the

same location as the experimental specimen, i.e. one inch from the edge of the reinforcing plate.

Referring to Figure 5.28c, the CVGM fracture prediction (using NB) is shown to be one cycle

after the experimental instance of fracture. Again, the prediction using the C-value shows an

improved accuracy over the damage function calibrated with just the notched-bar experiments.
169

5.2.4.5 ESTIMATING UNCERTAINTY IN THE CVGM FRACTURE PREDICTIONS

While the deterministic fracture predictions in the previous section describe the mean behavior of

CVGM, they do not provide the likelihood of fracture at any point in the loading history. This

may be useful for determining conservative lower bounds on predictions, investigating the

influence of material uncertainty, or providing a tool to examine the influence of loading history

on the CVGM predictions in a probabilistic sense. While Myers (2009) provides a more detailed

explanation of the probabilistic approach presented here, a simplified version of his method is

applied to the large-scale brace simulations.

Referring to Figure 5.22, an error quantity () with a normal distribution is assigned with a mean

of zero along the exponential trend-line. Furthermore, the distribution is assumed constant for

each damage point along the exponential function. Thus, the CVGM fracture criterion presented

in the first section of this chapter can be expressed as

cyclic
f D exp p* (5.2.1)

Where is a random variable with a normal probability distribution. Solving for at any time (ti)

in the loading history, the difference between the CVGM demand (cyclic/) and the fracture

toughness can be calculated with

cyclic
i exp p* (5.2.2)

The expression for i provides a bound on the integral of the probability density function of ,

such that on one tensile excursion

f | B d
c


Pr[T f ti | B c ] 0
0
(5.2.3)
1 f | B d
c


170

Where Tf is a random variable corresponding to the instance of fracture, ti is any instance during

the loading history, f() is the probability density function and Bc is the complement of the event

that the brace fractured at the peak of the previous tensile excursion. Thus, Equation 5.2.3 is

conditioned on the event that the brace did not fail on the previous cycle. Note that the expression

is also normalized to account for 0 > - . The probability of fracture at any analysis step is the

joint probability, determined by scaling Equation 5.2.3 by the maximum likelihood that fracture

did not occur on the previous cycle

Pr[T f ti , B c ] Pr[ B] Pr[T f ti | B c ]Pr[ B c ] (5.2.4)

Where Pr[Bc] is 1-Pr[B]. Furthermore, during compressive loading (T < 0) the probability is

assumed to be equal to the maximum probability from the previous tensile excursion. This

prevents the cumulative probability function from decreasing. As an example, the probability of

failure during the far-field loading history of an HSS4x4x1/4 bracing member is shown in Figure

5.29b. The figure shows two cumulative distribution functions (CDFs) for each of the -values

calibrated previously as well as a vertical line at the observed experimental fracture location. As

expected from the deterministic fracture prediction in Figure 5.25a, the prediction using the NB-

value from the notched-bar calibration tests intersects the experimental fracture location at

approximately 36%. Note that an exact deterministic prediction translates to a 50% likelihood of

fracture at the experimental fracture instance. If the C-value is used to describe the damage

function, the expected (i.e., at the 50th percentile) instance of fracture initiation is on the previous

tensile cycle while the probability at the experimental fracture instance is approximately 85%.

Also interesting to note is the sharp increase of the CDFs across these two tensile excursions of

the far-field loading history. This can be mostly attributed to the influence of local buckling on

the stress and strain state at the critical cross-section and infers that a probabilistic analysis is not

necessary for the specific case of predicting large-scale brace fracture. Moreover, the sharp nature

of the CDF further underscores the importance of modeling local buckling phenomena accurately.
171

The probabilistic analysis is applied to each brace simulation from the previous section. Table 5.5

lists the probability of fracture initiation at each experimental fracture instance using both -

values. Referring to the Table, an approximate 0% probability of fracture initiation is calculated

at the experimental fracture instance for the HSS4x4x3/8 and Pipe3STD bracing members if NB

is used to describe the fracture capacity. This is consistent with the deterministic fracture

predictions of these two cross-sections and can be attributed to the calibration issues as well as

the sharp nature of the CDF discussed previously. The analysis of the HSS4x4x1/4 and Pipe5STD

bracing members calculates non-zero probabilities where the CVGM predictions are shown to be

the most accurate for the Pipe5STD brace type. For each cross-section type, using the C-values

provides probabilities of fracture which are closer to 50%. Similar to the deterministic

predictions, the probabilistic analyses demonstrate that if calibration specimens are designed

according to an expected damage range then the CVGM can accurately simulate cyclic fracture

behavior.

5.3 SUMMARY

This chapter introduced a micromechanical-based fatigue/fracture initiation model developed and

validated at the small-scale by Kanvinde (2004). The monotonic Void Growth Model (VGM) and

the corresponding Cyclic-VGM (CVGM) operate at the continuum level and utilize stress and

strain histories from finite element simulations to predict fracture during monotonic and cyclic

loading. Therefore, the accuracy of the models relies on accurate material constitutive simulation

and, related to this, correct simulation of localized phenomena which lead to fracture. Calibration

procedures for two CVGM fracture parameters, and , are outlined in the chapter. While the

monotonic fracture parameter, , is calibrated using notched-bar tensile tests, the degradation

parameter, , requires more judicious cyclic calibration experiments. It is found to be important

for the damage range of the calibration experiments to approximately match the expected damage
172

of the component (large-scale or otherwise) where the fracture prediction is being made. In lieu of

this, Figure 5.23 illustrates that calibrating the exponential function with large damage levels will

reduce the variation in . The chapter highlights the shortcomings of the CVGM fracture

predictions applied to large-scale bracing members if this condition is not observed. However,

using the brace experiments and simulations as calibration points, it is shown that the form of the

damage model (exponential) accurately tracks the damage processes leading to eventual fracture

in the bracing elements. To illustrate this, a combined damage law is calibrated based on the

notched-bar experiments as well as the large-scale brace experiments. Statistical P-values are

used to characterize the likelihood of observing the notched-bar experimental results if the

combined damage law is the true behavior of the material. Finally, a probabilistic framework is

introduced utilizing the inherent error, or uncertainty, in fitting a trend-line to the cyclic damage

points. While the predictions are shown to be somewhat inaccurate for the HSS4x4x3/8 and

Pipe3STD specimens, the HSS4x4x1/4 and Pipe5STD fracture predictions are quite accurate

across a variety of loading histories. Moreover, the unsatisfactory performance of the CVGM

when applied to the HSS4x4x3/8 and Pipe3STD brace types is suspected to be a result of

insufficient small-scale calibration tests and not the formulation of the model itself.

In light of the evaluation of the CVGM performance discussed in this chapter, the methodology

can be applied to several practical situations. First, the large-scale brace simulations provide a

tool to inform inelastic buckling behavior during earthquake-type loading. For example, the

ability to accurately simulate the localized inelastic strains at the middle plastic hinge region

during local buckling provides valuable insights into the mechanisms which eventually produce

brace fracture. Furthermore, as design codes incorporate more performance-based techniques, the

sole reliance on large-scale testing results to evaluate ductility limits for steel components may

not provide a comprehensive description of all factors controlling fracture limit states. Through

the CVGM-based fracture methodology presented in this chapter, parametric simulation studies
173

can expand testing programs to provide improved insights into the behavior of large-scale brace

components. This is the topic of the next chapter.


174

Table 5.1: Measured material properties (refer to Figure 5.12 for locations of each specimen).
Material (Specimen) Type No. Tests Y, ksi U, ksi F, in/in = F/Y
A500 Gr. B (ASTM)
HSS4x4x1/4 (Wall) 3 67 (1.8) 71 (1.5) 0.98 (2.2) 249 (3.9)
HSS4x4x3/8 (Wall) 3 72 (5.2) 79 (4.1) 0.90 (1.8) 182 (4.3)
A500 Gr. B (UN)
HSS4x4x1/4 (Corner) 2 74 (2.2) 80 (2.1) 1.14 (6.6) 242 (8.4)
HSS4x4x1/4 (Wall) 2 70 (3.4) 74 (1.9) 1.02 (0.0*) 216 (0.05*)
HSS4x4x3/8 (Corner) 2 73 (1.8) 78 (1.6) 1.04 (2.3) 224 (2.9)
HSS4x4x3/8 (Wall) 2 80 (6.0) 84 (7.7) 0.89 (5.0) 184 (10.8)
A53 Gr. B (UN)
Pipe3STD (Wall) 2 54 (1.0) 67 (1.3) 0.96 (0.0*) 259 (0.6*)
Pipe5STD (Wall) 3 47 (9.2) 61 (2.2) 0.86 (3.6) 251 (12.3)
A992
W12x16 (Flange) 2 60 (0.1) 79 (0.3) 0.89 (4.9) 192 (5.0)
*Nearly identical measured fracture diameters

Table 5.2: Calibrated kinematic and isotropic hardening law parameters.


Brace Material Type Y, ksi C, ksi Q, ksi b
HSS4x4x1/4 (Corner) 73 850 160 14.5 5.25
HSS4x4x1/4 (Wall) 68 300 25 10 6
HSS4x4x3/8 (Corner) 64 1450 70 18.5 6
HSS4x4x3/8 (Wall) 64 1450 70 18.5 6
Pipe3STD (Wall) 55 500 35 52 2
Pipe3STD (Seam) 55 500 35 52 2
Pipe5STD (Wall) 51 489 26 13 7
Pipe5STD (Seam) 51 489 26 13 7
All Gusset Plates 50 500 38 17 5

Table 5.3: Maximum measured local imperfections (x 10-3 inches) along 3-foot section of each
brace cross-section.
HSS4x4x1/4 HSS4x4x3/8 Pipe3STD Pipe5STD
Seam (HSS-location 0.8(Left) 1.5 (Left)
within wall; Pipe-along 1.2(Right) 1.0 (Right) 1.9 4.1
seam) 1.1 (Middle) 0.6 (Middle)
Adjacent to Seam (HSS- 2.3(Left) 1.4(Left)
location within wall; 2.7 (Right) 0.5 (Right) 3.6 6.4
Pipe-180 from seam) 2.8 (Middle) 1.3 (Middle)
*Global camber assumed as L/1000
175

Table 5.4: Calibrated CVGM fracture parameters.


Sample
Brace Material Type fexp (in) NB C
No.
1 0.0150 5.83
2 0.0153 5.98
HSS4x4x1/4 (Corner) 3 0.0156 6.13 0.09 0.17
Avg. 0.0153 5.98
COV 1.96% 2.51%
1 0.0156 5.78
2 0.0192 7.88
HSS4x4x1/4 (Wall) 0.04 --
Avg. 0.0174 6.83
COV 14.63% 21.74%
1 0.0102 3.29
2 0.0112 3.70
HSS4x4x3/8 (Corner) 3 0.0116 3.87 0.05 0.13
Avg. 0.0110 3.62
COV 6.56% 8.33%
1 0.0093 2.82
2 0.0137 4.69
HSS4x4x3/8 (Wall) 0.09 --
Avg. 0.0115 3.76
COV 27.05% 35.21%
1 0.0196 4.03
2 0.0213 4.44
Pipe3STD (Wall) 3 0.0224 4.72 0.15 0.25
Avg. 0.0211 4.40
COV 6.69% 7.89%
1 0.0197 4.05
2 0.0203 4.19
Pipe3STD (Seam) 3 0.0204 4.22 --
Avg. 0.0201 4.15
COV 1.88% 2.18%
1 0.0099 1.98
2 0.0106 2.24
Pipe5STD (Wall) 3 0.0129 3.17 0.16 0.21
Avg. 0.0111 2.46
COV 14.10% 25.40%
1 0.0048 0.59
2 0.0078 1.30
Pipe5STD (Seam) 3 0.0092 1.74 --
Avg. 0.0073 1.21
COV 30.94% 47.96%
1 0.0186 2.98
2 0.0198 3.30
W12x16 (Flange) 3 0.0198 3.30 0.34 --
Avg. 0.0194 3.19
COV 3.57% 5.79%
176

Table 5.5: CVGM fracture prediction summary


Probability of Fracture**
Loading History
Bracing Deterministic (%)
Test No.
Member Std. Pulse Pulse Figure No. NB Damage C Damage
FF NF-C NF-T Law Law
HSS1-1 HSS4x4x1/4 5.25a 35 85
HSS1-2 HSS4x4x1/4 5.26a 13 42
HSS1-3* HSS4x4x1/4 5.27c 3 26
HSS1-6# HSS4x4x1/4 5.28 1 13
HSS2-1 HSS4x4x3/8 5.25b Approx. 0 47
HSS2-2* HSS4x4x3/8 5.27d Approx. 0 43
P1-1 Pipe5STD 37 43
5.25d
P1-2 Pipe5STD 23 26
P1-3 Pipe5STD 5.26c 57 74
P2-1 Pipe3STD Approx. 0 45
5.25c
P2-2 Pipe3STD Approx. 0 26
P2-3 Pipe3STD 5.26b Approx. 0 17
*fast (earthquake) loading rate

#HSS1-6 was reinforced at the midpoint (top and bottom) with 18 long plates (see Figure 5.28)

**Calculated at experimental fracture instance


177

Void Nucleation Necking Between Voids

Void Growth and Strain Void Coalescence and


Localization Macroscopic Crack Initiation

Figure 5.1: Micromechanical process of ductile fracture in steel


178

(a)

(b)

Figure 5.2: Scanning Electron Microscope (SEM) pictures of (a) Monotonic and (b) Cyclic
fracture surfaces (A572, Grade 50). Note the larger dimples of the monotonic as compared to the
cyclic surface.
179

100
n
= Y + K

True

50

Model
Experiment
0
0 0.025 0.05
True
(a)

1
1 = 2 = 3

2
(b)

Figure 5.3: (a) Uniaxial stress-strain behavior and (b) Isotropic yield surface.
180

S33
2
3 Y Q C
limit surface
2C
F
3
ij
Limiting
location
of ijdev 2
Y
3

S11
S22

yield surface

Figure 5.4: Combined isotropic-kinematic yield surface employed in ABAQUS (2004).


181

Yield stress, Y
(uniaxial
tension coupon)

Isotropic hardening
parameters, Q and b
(notched-bar cyclic test;
initial values: C=300,
=30, Q=15, and b=5)

Kinematic hardening
parameters, C and
(notched-bar
monotonic test)

Verify that parameters work


well for uniaxial tension and
notched-bar cyclic tests

Yes; use parameters No; repeat small-scale


in large-scale model calibration loop

Verify large-scale model


properly simulates localized
phenomena that lead to
fracture (i.e., local buckling)

No; adjust parameters


to track local buckling
Yes: done and check against
small-scale tests

Figure 5.5: Ideal material calibration flow chart.


182

Experimental
fracture, fexp
Force

D0
r*

Model
Experiment
Deformation
(b)

(a)
Figure 5.6: (a) Notched-bar geometry and (b) Typical force-deformation response. See next figure
for notched-bar continuum model.
183

Node 1 Node 2

(a) (b)

Node 1
R Node 2
ln
R0

exp
f

Deformation (in)
(c)

Figure 5.7: (a and b) Axisymmetric finite element model of notched-bar specimen. The contours
show (a) Equivalent plastic strain and (b) Triaxiality in the notched region. The large triaxiality at
Node 1 is primarily responsible for the discrepancy in (c) the Void growth demand between the
two nodes.
184

f (D) = exp(- p* )
cyclic/

1 cyclic test

p*
Figure 5.8: Exponential relationship between cyclic/ and damage (equivalent plastic strain at the
beginning of the tensile excursion to fracture, p*).

1.5

Critical Void Size
Void Growth and

Predicted
1 Fracture Time

0.5

0
Analysis Time

Figure 5.9: CVGM fracture prediction for a small-scale notched bar specimen.s
185

4
AP50
AP110
3 AP70HP
JP50
AW50
Analytical

JP50HP
2
JW50
p

0
0 1 2 3 4
p
Experimental

Figure 5.10: Predicted versus experimental instance of fracture initiation for small-scale, notched-
bar cyclic coupon tests (adapted from Kanvinde and Deierlein, 2007).
186

Corner:
3.0 1.0
2S
3 SN
3 LN
Wall:
2 UN 0.5 3.0 0.1 1.5
3 Flats
2 SN r = 0.5
2 LN
HSS HSS4x4x1/4, 3.0 1.0
HSS4x4x3/8 0.75

HSS Flats Smooth (S)


Seam: (thickness: 0.25) monotonic tension
2S
3 SN
1 LN
Wall: D0 = 0.1 D0 = 0.1
2 UN
3 SN
3 LN
Pipe Pipe3STD,
Pipe5STD r* = 0.05
r* = 0.025
3.5
(typ)

Flange:
2S
3 SN
3 LN

Small-notched (SN)
0.2 monotonic tension
Wide-Flanged (typ)
W12x16 Large-notched (LN)
cyclic

Figure 5.11: Summary of small-scale specimens for brace material calibration study.
187

700 700

Load (lbs)
Load (lbs)

350 350

Model
1 Model
(a) 2 (b) Experiments
0 0
0.00 0.04 0.08 0.12 0.05 0.075 0.1
Deformation (in) Neck Diameter (in)

1000 1000
Load (lbs)

Load (lbs)

500 500

Model Model
1 1
2 2
(c) 3 (d) 3
0 0
0 0.01 0.02 0 0.01 0.02
Deformation (in) Deformation (in)

0.02 1500

1000
Fracture
Deformation (in)

0.01 500
Load (lbs)

0 -500 Model
1
-1000 2
(e) (f) 3
-0.01 -1500
-0.01 0 0.01 0.02
Deformation (in)

Figure 5.12: Experimental and simulation force-deformation comparisons for small-scale material
calibration study on HSS4x4x1/4 corner material. Shown above is (a) Smooth bar force-
deformation behavior and (b) Fracture load versus fracture diameter; (c) SN force-deformation
behavior and (d) First tension excursion for the cyclic LN experiments; (e) Cycle loading history
and (f) LN cyclic force-deformation relationship.
188

300
Experiment
200 Model
Force (k)

100

-100

-200
-2 -1 0 1 2
Axial Displacement (in)

Figure 5.13: Experimental and simulation force-deformation comparison for large-scale


HSS4x4x1/4 brace (far-field loading).
189

1=0
2=0
3 loading

n=10, p=2
8 15 7

16 14
20
n=25, p=3 20 19
5 13
6
4 11
3
17 12 18
2=0 2 10
3=0
3 1
1 9 2

HSS4x4x1/4 HSS4x4x3/8

Pipe3STD
Pipe5STD

W12x16

Figure 5.14: Brace models and meshing schemes.


190

Outside corner Outside corner


radius = 0.625 radius = 0.800
4 4

4 4 0.355
0.254

(a) (b)

5.56

3.5

0.216 0.254

(c)

(d)
4.08

0.270

12.0 0.20

(e)

Figure 5.15: Measured brace cross-section dimensions for (a) HSS4x4x1/4, (b) HSS4x4x3/8, (c)
Pipe3STD, (d) Pipe5STD and (e) W12x16.
191

(a) (b)

Figure 5.16: (a) Global and (b) Local buckling mode shapes.

8
Imperfection Magnitude, in (10-3)

Pipe3
Pipe5
HSS4x4x1/4
HSS4x4x3/8
4
Along HSS
Along Pipe Seam Seam

Figure 5.17: Cross-section wall imperfection measurements for HSS and Pipe.
192

300 300
Experiment Experiment
200 Model 200 Model
Force (k)

Force (k)
100 100

0 0

-100 -100
(a) (b)
-200 -200
-2 -1 0 1 2 -6 -4 -2 0 2
Axial Displacement (in) Axial Displacement (in)

250
Experimental Buckling Load (k)

HSS4x4x1/4 10% margin


200 HSS4x4x3/8 lines
Pipe3STD
150 Pipe5STD

100

50 Near-field
(c) loading
0
0 50 100 150 200 250
Predicted Buckling Load (k)

Figure 5.18: Predicted buckling loads for (a) Far-field loading as compared to (b) Near-field
loading for HSS4x4x1/4 brace. (c) Experimental and predicted critical buckling loads.
193

Axial Deformation (in)


1

0.61" 0.61"
-1 (1) (2) 1.1" 1.1"
(1) (2)
-2
20 22 24 26 28
Cycle Number

0.61 (1)

0.61 (2)

1.1 (1)

1.1 (2)

Figure 5.19: Experimental and predicted local buckling for HSS4x4x1/4 brace (far-field loading).
.
194

2
Brace
Notched-Bar
1
Triaxiality

-1

-2
0 2 4 6
eP
Figure 5.20: Triaxiality versus equivalent plastic strain (HSS4x4x3/8 material).
195

1 1
= 0.09 = 0.05

C = 0.13
cyclic/

cyclic/
0.5 C = 0.17 0.5

Notched-Bar Notched Bar


Brace (a) Brace (b)
0 0
0 2 4 6 8 0 2 4 6 8
P*
P*
1 1
= 0.15
= 0.16
cyclic/

cyclic/

0.5 C = 0.25 0.5


C = 0.21
Notched Bar Notched-Bar
Brace (c) Brace (d)
0 0
0 2 4 6 0 2 4 6
P* P*

Figure 5.21: Calibration of NB and C using the notched-bar (NB) and combined (C) data points,
respectively, for (a) HSS4x4x1/4, (b) HSS4x4x3/8 corner material and (c) Pipe3STD and (d)
Pipe5STD base metal.
196

Probability
Distribution of

f (D) 0.5
C = 0.17

Notched-Bar
Brace
0
0 2 4 6 8
P*

Figure 5.22: Constant probability distribution of .

cyclic/

Damage

Figure 5.23: Effect of low damage levels on variability.


197

6 6
Experimental
Void Size and Critical Void Size
Void Size and Critical Void Size

Fracture

4 CVGM 4
Prediction: NB CVGM
Prediction: C

2 2

0 0
23 24 25 26 27 28 23 24 25 26 27 28
Cycle Number Cycle Number

Figure 5.24: CVGM fracture predictions for HSS4x4x1/4 where NB and C are used in
calculating two damage functions (circle and triangle, respectively).
198

4 4
Experimental Fracture
Notched-Bar Fit
2 Combined Fit 2
(in)

(in)
0 0

-2 -2
(a) (b)
-4 -4
20 22 24 26 28 30 20 24 28 32 36
Cycle Number Cycle Number

4 4

2 2
(in)
(in)

0 0

-2 -2
(c) (d)
-4 -4
20 24 28 32 36 20 22 24 26 28 30 32
Cycle Number Cycle Number

Figure 5.25: Comparison between experimental and predicted fracture instances for (a)
HSS4x4x1/4, (b) HSS4x4x3/8, (c) Pipe3STD and (d) Pipe5STD. Note both Pipe sections have
two experimental points.
199

2 6

0 4
(in)

(in)
-2 2

-4 Experimental Fracture 0
Notched-Bar Fit
Combined Fit (a) (b)
-6 -2
0 6 12 18 24 30 36 0 6 12 18 24 30 36 42 48
Cycle Number Cycle Number

4
(in)

0
(c)
-2
0 6 12 18 24 30 36
Cycle Number

Figure 5.26: Comparison between experimental and predicted fracture instances for (a)
HSS4x4x1/4 brace subjected to the asymmetric compressive near-field loading history, (b)
Pipe3STD and (c) Pipe5STD subjected to the asymmetric tension near-field loading history.
200

4 4
Experimental Fracture
Notched-Bar Fit
2 Combined Fit 2
(in)

(in)
0 0

-2 -2
(a) (b)
-4 -4
20 22 24 26 28 30 20 24 28 32 36
Cycle Number Cycle Number

4 4

2 2
(in)

(in)

0 0

-2 -2
(c) (d)
-4 -4
20 22 24 26 28 20 24 28 32 36
Cycle Number Cycle Number

Figure 5.27: Comparison between HSS4x4x1/4 and HSS4x4x3/8 (a and b) Quasi-static (shown
previously in Figure 5.26) and (c and d) Earthquake-rate tests.
201

(a)

(b)

2
Experimental Fracture
Notched-Bar Fit
1 Combined Fit
(in)

-1
(c)
-2
20 22 24 26 28
Cycle Number

Figure 5.28: Comparison between experimental and predicted fracture instances for HSS4x4x1/4
brace subjected to far-field loading with middle reinforcing plates.
202

4
Experimental Fracture
Notched-Bar Fit
Combined Fit
2
(in)
0

-2
(a)
-4
26 27 28
Cycle Number

1 Experimental Fracture
Probability of Fracture

Notched-Bar Fit
Combined Fit

0.5

(b)
0
26 27 28
Cycle Number

Figure 5.29: (a) Deterministic CVGM fracture instances and (b) Cumulative probability function
for HSS4x4x1/4 subjected to far-field loading history.
203

Chapter 6
Parametric Simulation of HSS Bracing Members

To generalize experimental findings using the CVGM methodology, this chapter presents a

parametric simulation study on square and rectangular HSS bracing members subjected to

earthquake-type cyclic loading. The braces are modeled with continuum and fiber elements to

compare the advantages and disadvantages of the modeling techniques in the context of

simulating brace fracture. The continuum-based Cyclic Void Growth Model (CVGM), evaluated

in Chapter 5, is used to predict the fracture initiation in the brace simulations. These predictions

expand the experimental testing matrix presented in Chapter 4 by examining a wider range of

geometric properties which may influence brace ductility.

The parametric study results are combined with rectangular and square HSS experimental data

from the last twenty years to examine the factors which control brace ductility. Based on this

combined experimental and simulation data set, a semi-theoretical model is proposed, which

relates the brace axial deformation capacity, and associated story-drift capacity to the bracing

member geometry (i.e. slenderness and cross sectional geometry). An illustrative example is

provided wherein the semi-theoretical model is used to suggest a limiting cross-section width-

thickness, or b/t, ratio for square and rectangular bracing members, to provide acceptable
204

performance. Referring to Chapter 2, limits on cross-section b/t ratios are used by the AISC

Seismic Provisions (2005) to ensure acceptable ductility with respect to brace local buckling and

fracture. Thus, the results presented in this chapter provide an example where the advanced

simulation tools presented in Chapter 5 can be applied to develop insights into localized behavior

affecting fracture in large-scale steel structures, and subsequently improve design provisions.

6.1 PARAMETRIC SIMULATION MATRIX

Table 6.1 summarizes the simulation matrix, which includes 22 brace simulations with varying

cross-section sizes and width-thickness, aspect and slenderness ratios. The cross-section sizes

range from 4 to 8 inches square with width-thickness ratios between 8.5 and 14.2 for the square

cross-sections. Three slenderness ratios of 40, 80 and 120 are selected to complement the testing

matrix presented in Chapter 3 where the experimental HSS4x4x1/4 and HSS4x4x3/8 both had

slenderness ratios of approximately 80. Four rectangular brace simulations are conducted to

investigate the influence of aspect ratio, or B/H, on brace ductility where B is the overall length of

the buckling face and H is the perpendicular depth of the cross-section as shown in Figure 6.1.

Corner bend radii are assumed to be approximately 2t where t is the thickness of the cross-

section, according to the minimum specified fabrication limits for HSS bend radii from AISC

(2005). The brace cross-sections are attached to end gusset plates to replicate realistic boundary

conditions. The gusset plates are designed according to the procedure discussed in Chapter 3.

The HSS4x4x1/4 material type, presented in Chapter 5, is used for the parametric study discussed

in this chapter, given that it resulted in excellent predictions of fracture. Similar to the finite

element models presented in Chapter 5, a fine mesh is used in the middle third of the half-brace,

transitioning to a coarse mesh at the ends such that the element lengths increase in a geometric

progression. This allows for a fine mesh at the middle plastic hinge region where local buckling is

expected and a coarse mesh at the end of the brace where minimal inelastic strains develop. For
205

the larger sections, the HSS4x4 meshes developed in Chapter 5 are used as guidelines to

determine the number of elements along the dimensions of the cross-section.

6.1.1 CYCLIC LOADING HISTORY

Figure 6.2 illustrates the cyclic loading history for the parametric brace simulations expressed in

terms of story drift. The peak drifts for each loading cycle are identical to the experimental far-

field history presented in Chapter 3. The story drift is converted to axial deformations according

to the relationship


a cos 45 sin 45 LB 0.5LB (6.1.1)

Where is the story drift angle (expressed in radians), a is the corresponding axial deformation

and LB is the length of the bracing member between gusset plate fold lines. Refer to Chapter 3 for

more details.

6.2 CVGM FRACTURE PREDICTIONS

This section presents the results of the parametric study where the CVGM is used to predict the

instance of fracture in each of the brace simulations of Table 6.1. The CVGM fracture parameters

are assumed to be equal to the calibrated values from the HSS4x4x1/4 brace material where is

equal to 5.98 and is 0.17. Note that is calibrated from the combined data set of the notched

bar and brace experiments because it characterizes the cyclic damage of the experiments more

accurately. The fracture predictions expand the experimental testing results from Chapter 4 and

inform relationships which govern brace ductility. From the discussion in Chapters 2 and 4,

width-thickness (b/t) and brace slenderness ratio (KL/r) were shown to affect brace ductility.

Thus, the fracture predictions are discussed in the context of these parameters.
206

6.2.1 EFFECT OF WIDTH-THICKNESS RATIO ON BRACE DUCTILITY

In this section, the results of the simulation study are used to determine the effect of width-

thickness ratio on brace ductility. The maximum axial deformation range, range=max+|min| (See

Figure 2.9 or 6.7), prior to fracture initiation is used to describe the capacity of the bracing

components. While cumulative ductility capacities (i.e., cumulative plastic deformation) are, in

general, more comprehensive, the consistent loading history diminishes the usefulness of a

cumulative measure. Furthermore, the semi-theoretical model in Section 6.3 will use the

deformation range because it may be evaluated relative to maximum story drift demands.

Referring to Figure 6.4a, the influence of brace width-thickness ratio on axial deformation

ductility is difficult to ascertain due to the effect of the brace length on the ductility. If the brace

simulations had been the same length, the ductility trend would be similar in appearance to Figure

2.9. Thus, converting the axial deformation range to the story drift range, range, allows the effect

of the width-thickness ratio to be more easily identified. Figure 6.4b illustrates the expected

decreasing ductility for increasing width-thickness ratio after the deformation range is normalized

by the brace length. Referring to the figure, for a slenderness ratio of KL/r = 80, a 40% reduction

in width-thickness ratio resulted in about a 100% increase in fracture drift ductility. This is

consistent with the other slenderness ratios as well as the experimental observations presented in

Chapters 2 and 4.

6.2.2 EFFECT OF SLENDERNESS RATIO ON BRACE DUCTILITY

This section uses the results of the parametric study to investigate the influence of brace

slenderness ratio on ductility. Referring to Figure 6.4a, brace slenderness has a dramatic effect on

the axial deformation ductility, such that increasing the slenderness ratio of an HSS6x6x3/8

bracing member from 40 to 120 increases the ductility by nearly 330%. This trend was also
207

shown in Chapter 2 across a variety of cross-section types. However, as illustrated in Figure 6.4b,

expressing ductility in terms of the maximum story drift range considerably lessens the effect of

brace slenderness ratio. In terms of drift, increasing the HSS6x6x3/8 slenderness ratio from 40 to

120 increases the ductility by 43%. Note that for the group of simulations with b/t < 14, the

ductility increases for a given slenderness because of the decreasing width-thickness ratios

included in that group (i.e., 10.8, 9.9 and 8.5, refer Table 6.1). As expected, this is not observed

for the group of simulations where b/t = 14.

Referring to Figure 6.4, the vertical line represents the AISC (2002 and 2005) upper-limit on

brace slenderness (100). However, it should be noted that AISC Seismic Provisions (2005) has

adopted an exception which allows slenderness ratios greater than 100 (for A500 Gr. B HSS

braces), but less than 200, if the expected yield force of the brace is less than the compressive

capacity of the adjoining columns. While the intent of the new provision is to improve brace

ductility through slenderness ratio, slenderness ratios above 100 may be somewhat unlikely

considering the larger HSS shapes used in high seismic regions (i.e., HSS8x8 and above). For

example, Table 6.2 lists the corresponding brace lengths (assuming a 45 Chevron orientation) for

story heights equal to 13 and 20 feet and frame bay widths equal to 20, 30 and 40 feet. Referring

to the table, assuming typical braced-frame dimensions, the range of brace lengths is between

16-5 and 28-3 (197 and 339 in). Using these lengths, and an effective length factor (K) equal to

0.9, Figure 6.4a illustrates the resulting slenderness ratios for all available square HSS shapes.

Horizontal lines are also drawn at the AISC (2005) slenderness limits of 100 and 200, where the

200 limit can only be used if the adjoining columns have adequate compressive capacity. Figure

6.4b illustrates the available square HSS shapes with b/t ratios less than the maximum b/t limit of

16 prescribed by the AISC Seismic Provisions (2005). Referring to these Figures, for shapes

which are relatively common in SCBF design (larger than 5x5), the largest slenderness ratio is

approximately 120. Thus, the parametric study (and the following synthesis of experimental data)
208

appropriately queries slenderness ratios which can be expected in typical braced-frame

construction.

In the context of the previous discussion, it may be more appropriate to investigate the influence

of slenderness ratio on brace ductility through the radius of gyration, r, by fixing the overall brace

length and b/t ratio. This eliminates the compactness (discussed previously) and length effects on

brace ductility. Using the results from the parametric study, Figure 6.4c illustrates story drift

capacities for fixed brace lengths equal to 119 and 180 inches (b/t =14.2) with varying r.

Referring to the figure, the ductility does not increase for the 119 inch length (slenderness range

from 40 to 80) while a negligible increase is observed for the 180 inch length (80 to 120). Also

shown in Figure 6.4 are results from past experimental testing programs (discussed next) where,

similar to the parametric study results, a fixed brace length of 119 inches and a constant b/t limit

equal to 14.2 examines the influence of increasing r on story drift ductility. The experimental

trends seem to confirm the results of the simulations such that an increasing slenderness ratio (50

to 145) has negligible effects on ductility.

6.3 SYNTHESIS OF EXPERIMENTAL AND SIMULATION RESULTS

This section uses the results from the parametric study simulations along with HSS brace tests

conducted over the last twenty years to develop a semi-theoretical model to predict ductility as a

function of brace geometry. The relationship can inform design procedures of HSS braces in

SCBF systems and evaluate their performance relative to expected seismic demands. Specifically,

the methodology generalizes the experimental and simulation results and proposes a relationship

that may be used to determine limiting b/t ratios considering various aspects of brace response.

A review of ten experimental studies (a total of 63 experiments) on HSS bracing members is

summarized in Table 6.3. Important results from all the experiments within each program are
209

summarized in Appendix B. Referring to Appendix B, these programs have examined the effect

of various parameters, such as brace slenderness, compactness, and cross-section shape, on

fracture ductility and energy dissipation in braces. The yield stress for all the HSS steel materials

listed in Table 6.3 and Appendix B is fairly consistent (Mean = 62 ksi, COV = 0.09).

In general, the findings from all programs concur that the width-thickness ratios and slenderness

ratios control brace ductility, such that higher width-thickness ratios and lower slenderness are

detrimental to brace performance. However, synthesizing the data from diverse experimental

programs presents several challenges. Since the different programs have dissimilar test setups,

brace geometries (length and cross-sections) and loading protocols, it is difficult to directly

interpret results from these various programs. To illustrate this point, Figures 6.6a and 6.6b

include example plots of the observed maximum drift (expressed as half the maximum drift

range, max) for selected experimental data points from various test programs. Figure 6.6a plots

the drift against the b/t ratio for 19 braces of approximately similar slenderness ( 80). On the

other hand, Figure 6.6b plots the drift capacity versus brace slenderness for 25 braces of

approximately equal b/t ratios ( 14.2). Referring to the figures, two observations may be made

1. Although the b/t ratios appear to negatively influence the brace ductility, the experiments

exhibit maximum drift ductilities (expressed as half of the total range, as is often done for

symmetric cyclic loading protocols) that are smaller (average 2.6%) than the expected

4% during MCE events, suggesting that almost all HSS braces are deficient, many even

under design level events (that correspond to 2% drift). Referring to Figures 6.3 and 6.4,

this is also true for the simulation results where the maximum drift ductilities (again,

calculated as half range) are, on average, 3.4%.


210

2. Similar to the parametric study results presented in Figure 6.4b and 6.5c, Figure 6.6b

does not indicate a strong positive effect of the brace slenderness on ductility (for

constant b/t).

Both these observations are somewhat surprising and raise two key issues

1. Given that brace ductility is a function of several parameters, a common basis is required

for interpreting data from various test programs, which may often have different brace

lengths and cross-sectional dimensions. This common basis can then be used to

generalize results of the various programs. This is the topic of the next section.

2. A consistent approach to evaluate the brace drift capacities relative to drift demands is

desirable, especially given the variability in loading protocols. This is the topic of Section

6.3.2.

6.3.1 A SIMPLIFIED APPROACH FOR EVALUATING THE EFFECT OF BRACE PARAMETERS ON

DUCTILITY

A simplified approach is proposed to examine the effect of various parameters such as the brace

buckling length and cross sectional dimensions on brace ductility. The main assumptions of this

approach are summarized schematically in Figure 6.7 and listed below

1. Neglecting elastic deformations in the brace, the range of axial brace deformations

range=max+|min| can be kinematically related to a gross strain quantity corresponding to

fracture (similar to a fiber strain) at the center of the brace. Referring to Figure 6.7 and

Equation 6.3.1, LB is the brace buckling length, and is distinct from the effective length,

KL
211

H 2 range 2H 2 max min


F (6.3.1)
Lh ,est LB H B LB

2. The plastic hinge length, Lh,est, at the center of the brace is equal to the average of the

width, B, of the cross-section over which the local buckle forms and the depth of the

cross-section, H. Refer to the following discussion for an explanation of Lh,est.

Thus, the relationship presented in Equation 6.3.1 can be used to convert the brace axial

deformation range to the local (fiber-level) strains corresponding to fracture. This relationship is

advantageous because it provides a common variable that can be examined across various

simulations, experiments and testing programs and eliminates the effect of brace length,

dimensions, and loading protocol. However, these gross strains do not incorporate the

amplification due local buckling of the cross-section wall. To include the local buckling effect,

Figure 6.8a plots the fiber-level strain, F, determined according to Equation 6.3.1 versus the b/t

ratio for all the 63 tests. Also shown are the 22 simulation points from the parametric study.

Referring to the figure, a strong relationship is observed between the b/t ratio and F. This may be

expressed through a regression fit (shown as dashed lines on Figure 6.8a) as shown below

b
b
F a (6.3.2)
t

Where a and b are constants which are calibrated with a least squares fit. Referring to Table 6.4

and Figure 6.8a, the above regression fit is determined for the experiments (63 points),

simulations (22) and a combined data set of the experiments and simulations (85). The relatively

close values of the regression parameters, a and b, across the three data sets is encouraging in that

the simulation results are corroborated by the experimental observations. For comparison, Figure

6.8b directly plots range against b/t for all the experiments. Referring to the figure, F results in a

clearer dependence on b/t.


212

The regressed relationship from Equation 6.3.2 may be used in conjunction with Equation 6.3.1 to

develop a predictive relationship for the brace axial deformation capacity such that

2
L a H B b b
p r edicted
range B (6.3.3)
2 2 H t

Figure 6.9 plots the ratio of the measured brace axial deformation capacity to the predicted

deformation capacity (range,exp/range,pred) for all the experiments. Given the variability between

various test programs, the results presented in Figure 6.9 are encouraging (mean

range,exp/range,pred=1.03, and COV of 26%).

It is important to note that the relationships presented in this section are based on the brace

buckling length, LB, instead of the brace effective length (with respect to elastic buckling). This

can be explained by considering that once the brace buckles and forms a central plastic hinge, the

strains are driven by a kinematic relationship which is a function of the buckled brace geometry

(rather than the slenderness ratio, which corresponds to the elastic buckling curve). Thus, if the

brace buckles elastically, i.e. KL/r > 118 (for Fy = 46 ksi), the mechanisms discussed in this

section may not be active. Out of the 63 experiments listed in Table 6.3, only 4 had KL/r values

greater than 118 while 6 simulations are approximately equal to the elastic buckling limit (120).

Moreover, the relationships assume loading history independence and quantify the capacity in

terms of a single parameter, range. While this is a simplified estimate, it provides good agreement

with test data and can be conveniently interpreted. This may be refined through a more detailed

analysis of test data.

In summary, this section presents an approach to determine brace axial deformation capacities

given brace parameters. However, these predictions (as well as data from other experiments) must

be evaluated relative to expected story drift demands. This is discussed in the Section 6.3.2.
213

6.3.1.1 PLASTIC HINGE LENGTH

The spread of plasticity along the length of structural components during inelastic loading may be

approximately characterized with the plastic hinge length. Although it is subjectively defined,

(given the irregular shape of the plastic zone) it often serves as a convenient measure of the size

of the zone of concentrated plasticity. Previous research has expressed plastic length as a function

cross-section geometry. For example, in circular bridge piers, the plastic hinge length is often

considered equal to half the diameter of the reinforced concrete column (Priestley and Park,

1987). Similarly, the plastic hinge length of rolled steel members is often assumed to be equal to

the depth of the cross-section across the neutral axis. Although the plastic hinge is commonly

visualized as a zone of concentrated plastic deformation within a member, the plastic hinge length

itself is somewhat difficult to characterize in a precise manner for a variety of reasons. For

example, weak moment gradients within a member may extend the zone of plasticity well outside

the commonly visualized hinge region. Thus, in the context of this discussion, the plastic hinge

length is interpreted as a parameter which enables the convenient determination of fiber-like

strains based on global deformations, rather than a precise physical quantity that may be

objectively characterized.

In this section, the plastic hinge lengths of the brace simulations are examined. The hinge length

is shown to be dependent on local buckling such that the buckling wavelength of the HSS wall

controls the distribution of inelastic strains. To calculate the plastic hinge length in the brace

simulations, the curvature along the length of the bracing member is integrated to determine the

total rotation between the gusset end plates. Assuming this rotation is uniformly concentrated in a

concentrated plastic hinge at the brace mid-point, an approximate hinge length can be interpreted

such that
214

LB

Lhmax ( x)dx (6.3.4)


0

Where Lh is the plastic hinge length, max is the maximum curvature at the center of the brace,

and (x) is the curvature along the length of the brace, LB. The curvature along the length of the

brae is determined through differentiation of the brace profile on the face which does not locally

buckle. The plastic hinge calculation is illustrated schematically in Figure 6.10.

Referring to Figures 6.11a and 6.11b, the plastic hinge lengths at the compressive peaks of the

standard loading history are calculated according to Equation 6.3.4 for the HSS4x4x1/4 and

HSS6x6x3/8 bracing member simulations, respectively. Figures 6.11c and 6.11d illustrate the

plastic hinge lengths for the HSS4x2x1/4 and HSS4x2x3/16 with B/H ratios of 2 and 0.5,

respectively. The vertical dashed lines show the location of local buckling while the horizontal

dashed lines correspond to an estimated plastic hinge length once adequate deformation occurs to

clearly form the plastic hinge. This length will be discussed below. Two vertical lines are used in

Figures 6.11a and 6.11b because local buckling occurred at a later cycle for the braces with large

slenderness ratios (L/r=120).

To associate the plastic hinge length results from Equation 6.3.4 to the brace geometry (and

explain the horizontal dashed lines in Figure 6.11), Figure 6.12 illustrates the deformed shape of

an HSS4x4x1/4 during local buckling. Connecting the plastic hinge length with the brace

dimensions is also convenient in light of the semi-theoretical model described above. Thus, an

approximate plastic hinge length is developed considering the spread of plasticity with respect to

the post-buckling amplified strains. Referring to Figure 6.12, the hinge length is mostly influence

by local buckling, such that the buckling wavelength is B and the effective buckling wavelength

is approximately B/2. Furthermore, considering the plastic hinge length is used in the semi-

theoretical model as a parameter to estimate fiber-like strains, the hinge length is determined
215

along the center-line of the brace at a distance of H/2 from the buckling face. From this, the

plastic hinge length is estimated as

B
Lh ,est mH (6.3.5)
2

Where m is the slope of the plane section which extends from the local buckling face to the

center-line of the brace. Equation 6.3.5 can be explained physically by the fact that while the

plastic hinge length is dependent on the local buckling wavelength of the compression face, the

depth provides rotational constraint such that a larger depth develops a larger hinge length (for

equal B). The slope m can be considered proportional to the out-of-plane buckling deformation of

the brace divided by half the brace length. However, realizing the complexity of the inelastic

behavior during global and local buckling, Equation 6.3.5 is simplified by assuming m = 0.5

Thus, while the expression captures the physical behavior of the plastic hinge length during local

buckling, it does not necessitate a dependence on the global deformation shape. In light of the

semi-theoretical model proposed in the next section, as well as the complex nature of inelastic

buckling phenomena, Equation 6.3.5 with m = 0.5 serves as a reasonable estimate for the plastic

hinge length.

Referring back to Figures 6.11a-b, the hinge lengths from the square cross-section simulations

tend to converge to Lh,est, i.e., the wall dimension of the square tubes, regardless of slenderness

ratio and cross-section size. Moreover, referring to Figures 6.11c and 6.11d, the brace simulations

with aspect ratios not equal to 1 (HSS4x2x1/4 and HSS4x2x3/16, respectively) illustrate that the

finite-element hinge lengths also approach Lh,est after local buckling initiates. Therefore, square

and rectangular simulations suggest that the hinge length after local buckling can be

approximated as the average of the cross-section dimensions, B and H. In addition, these

simulation results further emphasize the influence of local buckling on the strain amplification at

the middle plastic hinge of bracing members.


216

6.3.1.1.1 COMPARISON TO FIBER-BASED BRACE SIMULATIONS

Modeling bracing components with line models, such as the fiber element, is frequently used in

analyses of braced-frame behavior (Uriz et al, 2008 and Izvernari and Tremblay, 2007). The

popularity of fiber-based modeling can be attributed to the accuracy of the model in predicting

system and component level behavior such as story drifts and deformation demands in the

presence of material and geometric nonlinearities. Furthermore, the computational expense of

continuum models for large-scale structural modeling renders fiber-models a more efficient

choice in the context of modeling global response. However, fiber-models cannot simulate

phenomena such as local buckling. Thus, while the fiber-model can accurately predict global

behavior and provides computational advantages over continuum models, the model may not be

suitable in assessing localized behavior which can lead to fracture events. Figure 6.13 compares

the results from a continuum and fiber model of an HSS4x4x1/4 bracing member during the far-

field loading history shown in Figure 6.2. The fiber-based model uses approximately 20 elements

along the length of the brace where, similar to the continuum model, more elements are used

along the middle-third of the half-brace. Referring to the figure, while the load-deformation

behavior from the fiber model matches the experimental results well, the plastic hinge length

remains relatively constant during the loading. Compare this to the continuum-based predictions,

which show the sudden decrease in plastic hinge length after local buckling occurs. Thus, the

amplified strain demands from local buckling are not predicted accurately using the fiber model.

6.3.2 EVALUATING BRACE AXIAL DEFORMATION CAPACITIES RELATIVE TO EARTHQUAKE

INDUCED STORY DRIFT DEMANDS

To provide limits on brace geometries, the brace axial deformation capacities, range, determined

either directly from experiments or through an approach such as the one suggested in the previous

section, must be evaluated relative to expected story drift demands in SCBF buildings. For this
217

discussion, the drift demands will be discussed in terms of the maximum story drift, max. Several

components, already presented in this chapter or previous chapters, are combined to develop a

relationship between the axial deformation capacity and the expected maximum drift. These are

1. Estimating the maximum story drifts, max, that may be encountered in SCBF buildings

during earthquakes. This was discussed in Chapter 2.

2. Examining the relationship between the maximum story drift demands max and the range

of drift, range = max +|min|, expected within a story. This is also discussed in Chapter 2.

3. Developing a relationship between range and the brace axial deformation, range. This was

discussed in Chapter 3.

Referring to the first point, current design requirements for SCBFs (AISC, 2005) state that

braces could undergo post-buckling axial deformations 10 to 20 times their yield deformation.

Given a system yield level drift of approximately 0.3 to 0.5%, the Seismic Provisions could be

interpreted as desiring a deformation capacity of approximately 3 to 5% for SCBF systems.

Referring to the nonlinear time history simulations in Chapter 2, mean story drifts were

determined to be between 4.4 (6-story from by Uriz) and 8.1% (3-story frame by McCormick et

al) at the Maximum Considered Earthquake (MCE) demand level (2% chance of exceedance in

50 years). However, for the ground motions and buildings used as part of these studies, large

coefficients of variations (COV) are found to accompany these results (approximately 50 and

37%, respectively). Considering the discrepancy between the analysis results of Chapter 2 and the

interpretation of the AISC Seismic Provisions (2005), it is difficult to select an expected

maximum story drift demand. While it is outsides the scope of the current work, for the purposes

of this chapter, max is selected as 4%. Additional simulations and further analysis of existing

simulation data are suggested to refine this value.


218

With reference to the second point, Chapter 2 (Figure 2.6) illustrates the highly unsymmetrical

drift time history of SCBF systems under a 2/50 ground motion (Uriz, 2005) such that the

maximum drift is approximately two and a half times the maximum drift in the other direction of

loading, i.e., min=0.4max. Unlike moment frames, where the response is more symmetrical

(under far-field ground motions), braced frames tend to show a more unsymmetrical response,

presumably because of the unsymmetrical strength and stiffness properties once the compressive

brace buckles. However, braced frame components (such as braces) are typically subjected to

symmetric loading protocols, based on adaptations of protocols designed to reflect demands in

moment frames (Fell et al, 2006, Han et al, 2007, Shaback and Brown, 2001 and Archambault et

al, 1995). Consequently, when the maximum equivalent drift is reported as a capacity measure, it

is calculated as half the range of equivalent drift applied to the component (e.g., Chapter 4). This

may underestimate the capacity of the brace which, in general, will not be subjected to symmetric

cycles under seismic excitation. In fact, one may argue that this is one of the main reasons why an

examination of all the experimental data (as illustrated previously in Figure 6.6) indicates an

unusually low capacity for HSS braces. For a better interpretation of test data with respect to

realistic seismic demands, Equation 6.3.6 below incorporates the unsymmetrical nature of the

brace deformation discussed in Chapter 2

range max min max 1


1 (6.3.6)
max range
1

Where = 0.4 is chosen from Figure 2.7. While more investigations are required to characterize

this parameter accurately, these will likely reveal that is smaller than one which is implicitly

assumed while interpreting results of symmetric loading protocols. Thus, based on Equation

6.3.6, using the estimate of = 0.4 < 1 to interpret the test data increases the calculated drift

capacity, such that for the test data discussed earlier and shown in Figure 6.6, the average drift
219

capacities are 3.6% for the experiments (0.71/0.5 = 1.4 times 2.6% from above) and 4.8% for the

simulations described in the beginning of this section.

Finally, referring to the final point introduced at the beginning of this section, the story drift

range, range, can be related to the brace axial deformation, range, with the kinematic relationship

presented in Chapter 3. Assuming that the story drift is accommodated by the deformation of the

braces (a conservative assumption from the standpoint of brace capacity), the kinematic

relationship was presented as

range 1 C LB range cos sin (6.3.7)

Where range is the story drift angle (expressed in radians), range is the corresponding axial

deformation, is the brace angle, and C is the ratio of the rigid-link length (on both ends of the

brace) to the brace length, LB.

Thus, based on the ideas presented in this section, the brace axial deformation capacity, range,

may be converted to a corresponding drift capacity by combining Equations 6.3.6 and 6.3.7 such

that

1 1 1 range
max (6.3.8)
1 1 C cos sin LB

As discussed earlier, this relationship may be used to directly characterize brace capacities based

on experimental estimates of range, or alternatively, it might be used in conjunction with

Equation 6.3.3 to develop a relationship between the critical width-thickness ratio and the brace

capacity, expressed in terms of maximum drift demand

1
b 2H b
21 C 1 cos sin max (6.3.9)
t crit aB H
220

By substituting = 0.4, a = 0.98 and b = -0.56 from Table 6.4, and C = 0 and = 45 from the

assumptions discussed in Chapter 3, Equation 6.3.9 may be simplified to

1.78
b 2H
0.71 max
0.89
(6.3.10)
t crit BH

By substituting B = H in Equation 6.3.10, i.e. for square cross-sections

b
0.71 max
0.89
(6.3.11)
t crit

Figure 6.14 plots Equation 6.3.11 in the (b/t)-max space, indicating that lower (b/t) ratios will

result in higher max capacities. If an expected drift demand of max = 4% is substituted into

Equation 6.3.11, the limiting (b/t) ratio is calculated as 12.5. Shown by the dashed line on Figure

6.13, this is approximately 22% lower than the AISC limiting (b/t) = 16. On the other hand,

substituting the AISC value (b/t) = 16 into Equation 6.3.11 results in max = 3.0%, indicating that

although the current design provisions may not guarantee the 4% drift capacity implied by the

code, they may not be as deficient as have been recently suggested (especially when interpreting

results of experiments subjected to symmetric loading protocols).

Also illustrated in Figure 6.14 is an interesting dependence on a cross-section aspect ratio term,

2H/(B+H), for rectangular sections in Equation 6.3.10, where B is the overall length of the

buckling face and H is the distance between the compression (buckling) face and the tension face

(see Figure 6.12). If the cross-section is oriented to buckle about the weak-axis, 2H/(B+H) is less

than one and Equation 6.3.10 implies a more ductile configuration through an amplified (b/t)critical.

For the uncommon condition when the brace buckles about the strong axis, the 2H/(B+H) ratio is

greater than unity and the critical with-thickness ratio is reduced. Referring to Figure 6.12, this

effect can be explained by considering the strain gradient through the depth of the cross section,

H. For equal curvature, a larger H acts to increase the strain demand at the extreme fiber of the

cross-section and promotes a less ductile configuration.


221

Another interesting observation is that, according to Equations 6.3.9-6.3.11, the width-thickness

limit is not dependent on the length or the slenderness of the brace. Since this is somewhat

counterintuitive, it merits additional discussion. Referring to Equation 6.3.3, the brace length LB

affects the axial deformation capacity range in a positive manner. However, when the capacity is

expressed in terms of the story drift (i.e., combining Equation 6.3.3 with Equation 6.3.8), the LB

term cancels out. Physically, this may be explained by considering brace elements of identical

cross-sections included in a large and small braced-frame. The brace in the larger frame will have

a larger LB, and consequently a proportionally larger axial deformation capacity range. If both

these frames are subjected to an equal story drift, the larger brace will be subjected to

proportionally larger deformations, such that both will encounter similar strains in the plastic

hinge region. Furthermore, referring to Figure 6.5c, for identical cross-section b/t ratios and frame

geometries (i.e., equal bracing member lengths), the drift capacity was shown to remain constant

for varying radii of gyration. This can be explained through Equation 6.3.1 such that with an

increase in r for square cross-sections, H and B will be proportionally larger, thereby increasing

the length of the plastic hinge. Thus, this acts to offset the increase in strain which accompanies a

larger cross-section depth H.

While the model presented in this chapter is primarily focused on inelastic buckling behaviors,

future work is recommended to develop expressions for brace lengths which may produce large

slenderness ratios. In the context of the AISC Seismic Provisions (2005), this could allow for

larger (b/t)critical ratios for unusually slender braces. Also, the estimates of the (b/t) limits

presented in Figure 6.14 should be used with caution when applied to slenderness ratios that are

significantly outside the range (less than 31 and greater than 120) of the experimental and

simulation data presented here. For example, as discussed earlier, for KL/r > 118, i.e. elastic

buckling, the mechanisms discussed in this chapter might not be active. Similarly, for very low

KL/r, other mechanisms, such as local buckling under pure axial compression without global
222

buckling, may be active. Additionally, the relationships may not be valid for braces with different

material properties or cross-section types such as Pipe or Wide-Flanged sections. This

experiments synthesized in this investigation were comprised of square and rectangular brace

experiments with average yield strengths of 62 ksi and a coefficient of variation equal to 9%.

Moreover, the simulation study assumes the material yield strength to be approximately 60 ksi

and a fracture toughness of =5.98 (HSS4x4x1/4 material type from Chapter 5). Using Equation

5.1.15, this fracture toughness corresponds to a very large CVN value of approximately 404.

Thus, in the absence of further study, the results of this chapter are applicable only within the

experimental and simulation parameters investigated. The next section discusses the reliability of

the presented approach.

6.4 SUMMARY AND RELIABILITY OF PARAMETRIC STUDY AND SEMI-THEORETICAL

MODEL

This chapter synthesizes the results of a parametric study and experimental data from ten

independent testing programs to develop improved insights into the fracture capacity of HSS

bracing members. The parametric study suggests trends in ductility as a function of width-

thickness and slenderness ratio while expanding the results of the experimental program in

Chapter 4 to investigate a wider-range of brace parameters. Applying the same loading history

across all simulations allows for clear relationships to be obtained between ductility and brace

geometry. The simulations also establish the length of the plastic hinge after cross-section local

buckling as the average of the width, B , of the cross-section over which the local buckle forms of

the buckling face and the depth of the brace cross-section, H. This is a critical component in the

development of a semi-theoretical model used to synthesize the results of the experimental and

simulation studies. Perhaps of more interest, however, is the validation of using advanced

simulation techniques to expand the set of experimental data. Referring to Table 6.4 and Figure

6.8, separate least-squares fits on the experimental (63 tests) and simulation (22) data sets
223

provides very similar regression parameters. Even considering the experimental programs of

Table 6.3 with varying loading histories and geometries, the simulation results (obtained with

CVGM fracture predictions) seem quite similar plotted in the F versus (b/t) space in Figure 6.8.

While the simulations and experimental results suggest certain trends, it is somewhat challenging

to reconcile experimental data from diverse test programs. Therefore, to eliminate the effects of

geometric variables such as brace length and cross-sectional dimensions, a kinematic relationship

is proposed to relate the brace axial deformation capacity, range, to a fiber strain-like quantity,

F, in the central plastic hinge region of the buckling brace. A relatively strong trend is observed

between this strain and the (b/t) ratio and a regressed power-law model is used to express this

relationship. This relationship is then used to determine the brace axial deformation capacity,

range, as a function of the width-thickness ratio, (b/t).

Once the brace axial deformation capacity is determined, either directly from experimental data

or based on the proposed relationship, it is useful to express it in terms of equivalent drift capacity

for meaningful comparison with seismic demands. This includes developing a kinematic

relationship between the brace axial deformation capacity, range, and the equivalent drift range.

An important issue associated with this is the unsymmetrical nature of story drift time histories.

Therefore, determining the maximum drift capacity using half of the total range results in

conservative estimates, but is typically done when reporting testing results. Based on observations

from various nonlinear dynamic simulations, a preliminary estimate of max = 0.71range is

proposed. This is combined with the relationship between range and (b/t), and results in a limiting

function for the maximum (b/t) ratio given a desired level of drift capacity. To examine the

efficacy of this approach, Figure 6.15 plots all the drift capacities from the 63 experiments and 22

simulations against their (b/t) ratios normalized by the critical limit calculated using Equation
224

6.3.10. Following the reasoning outlined previously, the drift capacities are calculated as max =

0.71range, where range = max +|min| is the range of equivalent drift experienced by the test

specimen.

Plotted on the figure is the horizontal line corresponding to the 4% drift capacity expected from

SCBF systems. Also drawn on the figure is the vertical line corresponding to (b/t)/(b/t)critical = 1,

where (b/t)critical is calculated according to Equation 6.3.10 and is the critical (b/t) ratio required to

meet the drift demand of 4%. While this value is constant for square braces ((b/t)critical = 12.5),

results are plotted in this way because the data points also include those corresponding to non-

square cross-sections. These two limits divide the space into four quadrants, where the Roman

numerals are followed by the count of data points in that particular quadrant

Quadrant I (22%): Experiments which are safe (i.e., do not fail before 4%) and

predicted to be safe (i.e., less than the b/t limit of 1.0)

Quadrant II (16%): Experiments which are safe, but predicted to fail

Quadrant III (12%): Experiments which fail, but are predicted to be safe

Quadrant IV(50%): Experiments which fail, and are predicted to fail

A perfect model would result in 100% of the data points in either Quadrant I or IV where the

experimental data matches model predictions. As shown in Figure 6.15 for the proposed

relationship, approximately 72% of the data points lie in these two Quadrants. In addition, only a

small percentage of data points (12%) lie in Quadrant III where the relationship

(unconservatively) predicts safety. Thus, given the variability amongst the test programs and the

subjectivity in characterizing demands in SCBF systems, the proposed approach is fairly reliable

when evaluated against a large data set.


225

The results presented in this chapter are based on calibrated empirical relationships, and therefore

should be used with caution when applied to brace parameters (especially slenderness ratios

smaller than 60 or greater than 120 and materials dissimilar to those listed in Table 6.3). In

addition, the approach makes several simplifying assumptions, for example, it neglects the effect

of loading history on brace capacity. While these assumptions may not affect the efficacy of the

approach (in an average sense) with respect to a large sample of test data, they may result in

significant errors when applied to individual components. Finally, further study is needed to

accurately quantify both the maximum drift demands, the nature of the deformation histories in

SCBF systems and brace behavior at larger slenderness ratios.


226

Table 6.1: HSS simulation matrix (22 total)


Shape B H B/H b/t LB Approx. LB/r
HSS4x4x1/4 4 4 1 14.2 60,120,179 40,80,120
HSS4x4x3/8 4 4 1 8.5 60,120,179 40,80,120
HSS4x2x1/4 4 2 2 14.2 62 80
HSS4x2x3/16 2 4 0.5 8.5 111 80
HSS6x6x3/8 6 6 1 14.2 90,180,271 40,80,120
HSS6x6x1/2 6 6 1 9.9 90,180,271 40,80,120
HSS6x2x3/8 6 2 3 14.2 61 80
HSS6x2x3/16 2 6 0.3 8.5 161 80
HSS8x8x1/2 8 8 1 14.2 120,241,362 40,80,120
HSS8x8x5/8 8 8 1 10.8 120,241,362 40,80,120

Table 6.2: Brace member lengths across various frame geometries


Frame Bay Width (feet)
Story Height (feet)
20 30 40
13 16-5 19-10 23-10
20 22-4 25-0 28-3

Table 6.3: Summary of HSS experimental review (63 tests)


Year Average No. of General Description of
Test Program [No.]
Published Fy,meas Tests Cyclic Loading History
Gugerli and Goel [1] 1982 60 4 Unsymmetric compressive
Liu and Goel [2] 1987 54 3 Unsymmetric compressive
Lee and Goel [3] 1988 67 6 Unsymmetric compressive
Archambault, [4] 10 Standard symmetric
1995 57
Tremblay, Filiatrault 4 Unsymmetric
Walpole [5] 1996 56 3 Standard symmetric
Shaback and Brown [6] 2003 64 8 Standard symmetric
3 Standard symmetric
Yang and Mahin [7] 2005 60
1 Unsymmetric tensile
Fell, Kanvinde, and 4 Standard symmetric
2006 69
Deierlein [8] 1 Unsymmetric compressive
Han et al [9] 2007 59 4 Standard symmetric
Lehman et al [10] 2008 67 12 Standard symmetric

Table 6.4: Regression parameter values


Data Set a b
Experimental 0.90 -0.53
Simulation 1.13 -0.62
Combined 0.98 -0.56
227

Outside corner
radius = 2t
H

Buckling
face B t

Figure 6.1: HSS brace geometry

6
Maximum Considered (4) #cycles Drift (%)
4
6 0.08
2 6 0.10
6 0.15
Drift (%)

Expected Buckling (0.2)


0 4 0.20 B
2 1.03
-2 2 1.85
2 2.68
-4 2 4.00 - MCE
2+n 5.00
-6
0 10 20 30
Cycle Number

Figure 6.2: Brace loading history


228

20
40
80
range (in) 15 120

10

(a)
0
5 10 15 20
Width-thickness (b/t)

0.12
40
80
0.08 120
range (rad)

0.04

(b)
0
5 10 15 20
Width-thickness (b/t)

Figure 6.3: Influence of width-thickness ratio on brace ductility (fracture initiation) in terms of (a)
Axial deformation range and (b) Story drift.
229

20
Small b/t (<14)
Large b/t (14)
range (in) 15

10

(a)
0
0 50 100 150
Slenderness (L/r)

0.12

0.08
range (rad)

0.04
Small b/t (<14)
Large b/t (14) (b)
0
0 50 100 150
Slenderness (L/r)

Figure 6.4: Influence of slenderness ratio on brace ductility (fracture initiation) in terms of (a)
Axial deformation range and (b) Story drift.
230

500 500
L=204in L=204in
400 L=281in 400 L=281in
Slenderness

Slenderness
300 300
(KL/r)max (KL/r)max
200 200

100 100
(a) (b)
0 0
0 5 10 15 0 5 10 15
Cross-section size (in) Cross-section size (in)

0.08
range (rad)

0.04

L=180 in
L=119 in
(c) L=119 (Experimental)
0
0 50 100 150 200
Slenderness (L/r)

Figure 6.5: Slenderness range for minimum and maximum frame sizes (see Table 6.2) for (a) All
HSS cross-sections and (b) AISC (2005) conforming (i.e., b/t < 16) sections; (c) Influence of
brace slenderness on story drift ductility for fixed b/t (14.2) ratio and brace lengths (119 and 180
in). Also shown are past experimental (b/t=14.2 and L=119 in) results.
231

0.06
KL/r ~ 80

0.5 x range (rad)

0.03

(a)
0.00
0 10 20 30 40
Width-thickness, b/t

0.06
b/t ~ 14.2
0.5 x range (rad)

0.03

(b)
0.00
0 50 100 150 200
Slenderness, KL/r

Figure 6.6: Experimental brace ductility in terms of drift versus brace parameter for similar (a)
Slenderness (19 tests) and (b) Compactness (25 tests) ratios.
232

300

200
Force (k)

100

-100
range
-200
-1.5 -1 -0.5 0 0.5 1
Axial Displacement (in)

Strong relationship

From experiments H
hinge
fracture
Lh ,pred . . F f b / t
2
range From regression
/2
(= 0.5(H+B), based on
FEM simulations)
LB

Figure 6.7: Schematic illustration of simplified approach to evaluate effect of brace parameters on
ductility.
233

0.4
Experiments
Simulations
Experiment Fit
Simulation Fit
Combined Fit

0.2
F

(a)
0
0 10 20 30 40
b/t

Experiments
0.06 Simulations
0.5 x range (rad)

0.03

(b)
0.00
0 10 20 30 40
b/t

Figure 6.8: (a) Fiber-like fracture strain and (b) Deformation range capacity versus b/t ratio for all
experiments and simulations listed in Tables 6.1 and 6.3.
234

2
Experiments
Simulations
range,exp/ range,pred

0
0 2 4 6 8 10 S
12
Test Program

Figure 6.9: Comparison between maximum experimental and predicted deformation range.
235

(x)

LB

max ( x)dx
0

x
LB
Lh

(a)

(x)

max
LB

( x)dx
0

x
LB

Lh

(b)

Figure 6.10: Schematic illustration of plastic hinge length calculation where the total rotation is
equated to an equivalent area defined by max and Lh. Shown above is the curvature profile (a)
Before and (b) After local buckling.
236

6 4
40 40
80 80
120 120
4
Lh/Lh,est

Lh/Lh,est
2

(a) (b)
0 0
0 0.02 0.04 0.06 0 0.02 0.04 0.06
(rad) (rad)

4 4
Lh/Lh,est

Lh/Lh,est

2 2

(c) (d)
0 0
0 0.02 0.04 0.06 0 0.02 0.04 0.06
(rad) (rad)

Figure 6.11: Plastic hinge length as a function of increasing drift for (a) HSS4x4x1/4, (b)
HSS6x6x3/8, (c) HSS4x2x1/4 and (d) HSS4x2x3/16 analyses.
237

Lh

(a)

m
B B/2 Lh

H/2

(b)

Figure 6.12: Continuum finite element simulation of brace specimen showing plastic hinge length
dimension in (a) Isometric and (b) Top view.
238

300
Experiment
200 Continuum Model

100

Force (k)
0

-100
(a)
-200
-2 -1 0 1 2
Axial Deformation (in)

300
Experiment
200 Fiber Model

100
Force (k)

-100
(b)
-200
-2 -1 0 1 2
Axial Deformation (in)

8
Continuum Model
Fiber Model
6
Lh/Lh,est

(c)
0
0 0.02 0.04 0.06
rad)

Figure 6.13: (a and b) Force-deformation and (c) Plastic hinge length comparison between finite
element and fiber brace models.
239

0.10
B=H
B/H=2
B/H=0.5

max (rad) 0.05

0.03

12.5
0.00
0 5 10 15 20 25
b/t
Figure 6.14: Maximum drift capacity versus width-thickness ratio.

0.09
I: 19 II: 14
safe/predicted safe/predicted
safe to fail
0.71range (rad)

0.06

III: 10 IV: 43
0.03 fail/predicted fail/predicted
safe to fail

0
0 0.5 1 1.5 2 2.5
(b/t)/(b/t)critical

Figure 6.15: Deformation capacity (in terms of drift), divided by 1.4, versus normalized width-
thickness ratio.
240

Chapter 7
Summary, Conclusions and Future Work

This chapter summarizes various findings and conclusions from the study, while discussing areas

that need further examination. Within the broad context of simulating structural fracture using

advanced modeling methods, this dissertation had three specific goals (1) Investigation of the

cyclic inelastic buckling and fracture of steel braces in concentrically braced frames (2)

Development of a methodology wherein novel physics-based models are evaluated for full-scale

structural components (3) Application of this methodology for generalizing experimental data

sets to inform design considerations, while demonstrating the effectiveness of micromechanics-

based fracture modeling.

The bracing members investigated in this study are representative of those in Special

Concentrically Braced Frames (SCBFs) and are subjected to loading histories which impose

realistic seismic demands across a wide-range of ground motion intensities. Square Hollow Steel

Sections (HSS), steel Pipe and Wide-Flanged cross-sections are investigated in a rigorous testing

matrix which investigates a range of width-thickness and slenderness ratios along with unique

conditions such as grout-filled HSS, earthquake-rate loading and unsymmetrical buckling. In

addition to the practical data generated by the large-scale tests, the experimental specimens also
241

serve as a test-bed to validate a micromechanics-based modeling approach in large-scale details.

The Cyclic Void Growth Model (or CVGM), previously developed by Kanvinde and Deierlein

(2007), is evaluated using the experimental observations from the brace tests during earthquake-

type cyclic loading. Since a continuum-based model is used, the accuracy of the modeling

approach is contingent on accurate characterization of stress and strain histories in the presence of

phenomena such as local and global buckling.

This chapter reviews the conclusions of the large-scale experimental program along with a

summary of the model validation and parametric simulation studies. Limitations of the results and

method are discussed, leading to recommendations for future work.

7.1 SUMMARY AND CONCLUSIONS

The following three sections present a summary of the previous chapters. First, general

conclusions from the large-scale experimental study are presented. This is followed by a section

focusing on the fracture prediction methodology for the large-scale brace specimens. Finally, the

parametric simulation study on square and rectangular HSS sections, as well as the semi-

theoretical model used to generalize the experimental and simulation results, is reviewed.

7.1.1 Large-Scale Brace Experiments

Results of the large-scale testing program on nineteen bracing elements subjected to earthquake-

type cyclic loading are reported in Chapters 3 and 4. The experiments featured brace specimens

detailed according to current codes, and were subjected to various types of cyclic loading

histories designed to reflect realistic seismic demands. The testing matrix included a diverse

blend of parameters including cross section width-thickness, slenderness, type of cross section,

loading history, loading rate and special details such as grout filled braces. Various limit states,
242

such as local buckling, fracture initiation and loss of strength were monitored, and related to

system drift levels.

The braces subjected to cyclic loading failed due to fracture at the center, which was triggered by

strains highly amplified due to local buckling. Consequently, cross section width-thickness ratios

were found to strongly influence brace ductility for all cross sections, and higher width-thickness

ratios resulted in a severe decrease in ductility. Importantly, in some experiments, current AISC

limits for width-thickness ratios could not ensure acceptable performance, resulting in fracture at

unacceptably low story drift levels.

In addition to width-thickness, slenderness was determined to be another important factor

affecting brace fracture, in that more slender braces suffered relatively lower levels of inelasticity,

delaying fracture. In fact, the axial deformation at fracture was determined to be governed by a

combination of slenderness and width-thickness. For example, the wide-flange section with an

undesirable width-thickness ratio exhibited excellent ductility, due to its very high slenderness

(above the elastic limit) and the less severe nature of the local buckling shape as compared to

HSS sections. Since brace slenderness is a system level design variable, it might not be feasible to

provide large slenderness with the sole intent to prevent fracture. Moreover, large slenderness can

reduce energy dissipation in the brace, and place excessive tensile demands on connections (due

to overstrength). In addition, when the axial deformation capacity is expressed in terms of

equivalent story drift, the positive effects of slenderness tend to be diminished for slenderness

ratios below 120 (see Chapter 6 discussion of HSS simulation study).

In addition to slenderness and width-thickness, various other factors were considered. Of these,

the nature of the local buckling shape was found to differ in severity across the various cross-

section types (HSS, Pipe and Wide-Flanged). The square HSS local buckling shapes were found
243

to be particularly severe while the Pipe and Wide-Flange shapes developed more gradual

buckling shapes. However, this does not imply that Pipe and Wide-Flange behavior are naturally

superior to HSS (as illustrated with the Pipe5STD tests described in Chapter 4). Filling the braces

with concrete resulted in a somewhat larger ductility in one of two tests, but given the logistical

challenges to this, it may be more economical to achieve similar levels of ductility by using either

a more compact shape or an alternate cross section. Rate effects were examined and determined

to be relatively unimportant.

Connection performance regarding net section fracture at slotted brace-ends was investigated by

subjecting these to tension dominated near-fault loading histories with a large initial tensile pulse.

These tests, conducted for pipe sections and one wide-flange section, confirmed previous findings

that net section reinforcement increases ductility substantially and prevents fracture at the

connection. In fact, for the pipe specimens, the large difference between yield and ultimate

strengths resulted in large ductilities even for unreinforced connections. Overall, the variations in

the expected versus nominally specified material properties demonstrate the degree to which the

net section fracture response may differ between different structures. The test data did confirm

that the expected yield strength (RyFyAg) and the expected ultimate strength (RtFuAg) tend to

bracket the maximum measured strength fairly well. Furthermore, an accurate prediction of the

maximum brace tensile force was found to be the average of the expected and ultimate strengths.

7.1.2 CVGM Evaluation in Large-Scale Details

Referring to Chapter 5, the Cyclic Void Growth Model (CVGM) developed by Kanvinde and

Deierlein (2007) is used to predict the instant of fracture initiation in the large-scale experimental

brace specimens. Smooth (and ASTM-specified flat for HSS wall material) tensile coupons along

with circumferentially notched bars are tested to calibrate the material constitutive model

parameters as well as the CVGM fracture parameters ( and ). Large-scale finite-element brace
244

simulations are employed to investigate the stress and strain histories at the critical fracture

location during global and local buckling mechanisms. The simulations incorporate global and

local imperfections, as well as multiaxial von-Mises plasticity with isotropic and kinematic

hardening.

The CVGM fracture criterion is expressed as cyclic exp(.D), where cyclic is a void growth

function and D is the damage, assumed to be the equivalent plastic strain at a load reversal point

marked my a switch from negative to positive triaxiality. While the fracture parameter can be

calibrated conveniently and accurately through monotonic notched bar experiments, calibrating

the -parameter is more challenging. In fact, for two out of the four steels examined in the current

study, fracture predictions of the braces based on a -value calibrated from notched bar tests were

highly inaccurate. A closer inspection revealed that the notched bar specimens failed at low

damage levels (owing to the high triaxiality present in the notched bars), relative to the braces,

which had lower triaxialities. While the low triaxialities in the braces may trigger entirely distinct

fracture mechanisms (such as void shearing), an analysis of the data indicates that that the source

of the error is most likely in the selection of calibration tests themselves, rather than the damage

model. Thus, for future use of the CVGM it is recommended that the cyclic tests sample large

damage levels, or appropriately sample the magnitude of damage in the specimen where the

CVGM will be applied.

The CVGM fracture predictions are also presented in a probabilistic framework incorporating the

effect of material uncertainty. The analysis indicates that, for all specimens, the probability of

failure increases sharply 2-3 cycles after local buckling is observed in the simulations (for the

standard loading histories). This illustrates the strong influence of local buckling on the CVGM

predictions and the importance of simulating it accurately. In fact, the quick succession of
245

fracture following local buckling indicates that it may not be as important to simulate fracture, if

local buckling is accurately modeled. However, the sensitivity of local buckling initiation to

material model parameters is not explicitly investigated in this study.

7.1.3 Parametric Simulation of Bracing Members

A parametric simulation study on square and rectangular bracing members (total of 22 analyses)

is performed to expand the experimental data set presented in Chapter 4 and demonstrate the

applicability of advanced simulations methods. The simulations, which are similar to the ones

used for the CVGM evaluation, incorporate large deformations, multi-axial plasticity and brick

elements. The Cyclic Void Growth Model (CVGM) is used to predict the fracture initiation

instances of the 22 simulations and investigate a wide-range of parameters which affects brace

failure. The finite-element simulations are also used to determine the length of the plastic hinge

after cross-section local buckling. Referring to Chapter 6, the plastic hinge length is found to be

approximately equal to the average of the width of the cross-section, B, over which the local

buckle forms of the buckling face and the depth of the brace cross-section, H.

Experimental and simulation data from this study is synthesized with experimental data from 10

previous testing programs to present a semi-theoretical relationship that may be used to predict

the fracture deformation of HSS braces directly from brace properties such as cross-sectional

dimensions and slenderness ratios. The relationship is shown to reconcile fracture data from

several diverse experimental programs and simulation studies, which encompass a wide range of

test variables. The use of this relationship to develop design considerations is presented as an

illustrative example. Specifically, for the range of b/t and KL/r ratios investigated in this study,

the semi-theoretical relationship reveals 1) Overall slenderness has a minimal effect on ductility

when expressed in terms of a story drift and 2) Weak axis bending (H < B) leads to a more ductile

configuration.
246

7.2 FUTURE WORK

This dissertation presents a large and small-scale experimental testing program, nonlinear finite

element modeling, the evaluation of a physics-based fracture criterion at the large-scale, and the

development of a semi-theoretical model to estimate brace ductility. The results raise some

interesting scientific questions regarding the feasibility of using micromechanics-based models to

predict fracture in full-scale components. In addition to the modeling aspects, the study also

motivates further examination of several practical issues. All of these issues are now summarized-

1. Development of appropriate loading protocols for braced frames: While this study

presented a loading history based on a modified SAC loading protocol, it was not

developed through rigorous nonlinear dynamic analyses of braced-frame systems

(comparable to the SAC loading history for moment frames). Furthermore, analyses by

Uriz (2005) and others (McCormick et al 2007) suggest that a symmetric loading history

for braced-frame components may not be appropriate due the unsymmetrical response of

braced-frames during cyclic loading. Thus, the experimental results presented in Chapter

4 with respect to brace fracture ductility may be over-conservative in light of the

symmetric loading history used in this (and previous) studies.

2. Characterizing braced-frame demands: Considering the previous discussion, with

advances in structural modeling techniques and increasing computational power, there is

an opportunity to better characterize braced-frame demands across varying building

geometries, such as frame and story height, as well as brace properties (i.e., slenderness

ratios) and system configurations (Chevron, cross-bracing, etc.). This will assist in

performance comparisons between brace ductilites and expected seismic demands.

3. Experimental investigation of alternate cross-sections: Considering the recent popularity

of Pipe and Wide-Flanged sections over HSS, a comprehensive experimental or


247

simulation study (similar to Chapter 6) is warranted for these shapes. While the

experiments conducted as part of this investigation expanded the database of Pipe and

Wide-Flanged tests, there is a need to better characterize the ductility of bracing members

with these cross-sections.

4. Using the CVGM to inform fiber-based fracture criterion: Referring to Chapter 2, the

efficiency of fiber-based elements for earthquake engineering analyses has contributed to

their popularity in modeling bracing, and other structural, components. Following the

approach developed by Uriz (2005), an empirical-based fracture criterion could be

calibrated with the CVGM and then used in the analyses of full-scale braced-frames. This

would allow for on-the-fly analysis procedures by combining demand and capacity

characterizations which are typically separate in the context of fracture events.

Referring to the previous section and Chapter 5, several limitations of the CVGM have been

highlighted by applying the model in the context of brace fracture

1. Low triaxiality: The low triaxiality ranges observed in the brace tests may invalidate the

underlying assumptions of the CVGM. Considering the development of the CVGM (at

the small-scale) did not investigate these low triaxiality ranges (Kanvinde and Deierlein,

2007), the micromechanical fracture behavior of the braces may not be governed by the

void growth and cyclic damage mechanisms assumed by the CVGM. Thus, it is

important to investigate the fatigue and fracture mechanisms which are triggered in the

presence of a low triaxiality stress state. This may motivate the need for alternative

damage models other than the exponential function presented in Chapter 5.

2. Crack Propagation: Although not explicitly mentioned in Chapter 4, the brace tests with

near-field loading histories demonstrated that cross-section tensile strength loss may not

occur immediately after fracture initiation (refer Appendix A). This is mostly attributed
248

to the loading history offset at 3% drift for these histories such that the stress and strain

state needed to propagate the crack does not develop as rapidly during near-field loading

as compared to the more severe tension histories in the far-field tests. Thus, the

simulation of ductile propagation events may provide a more reliable measure of brace

ductility. Considering the unsymmetrical behavior of braced-frames, this may prove

useful in characterizing the various performance limit states of bracing members.


249

References

ABAQUS (2004). ABAQUS Users Manual, Version 5.8, Hibbitt, Karlsson, and Sorensen, Inc.,
Providence, RI.
American Institute of Steel Construction (AISC). (1997). Seismic provisions for structural steel
buildings, Chicago, IL.
American Institute of Steel Construction (AISC). (2002). Seismic provisions for structural steel
buildings, Chicago, IL.
American Institute of Steel Construction (AISC). (2005). Seismic provisions for structural steel
buildings, Chicago, IL.
American Institute of Steel Construction (AISC). (2005). Steel construction manual, 13th Ed.,
Chicago, IL.
Anderson, T.L. (1995). Fracture mechanics, 2nd Ed., CRC Press, Boca Raton, FL.
Archambault, M-H, Tremblay R, and Filiatrault A. (1995). E tude du comportement seismique
des contreventements ductiles en X avec profiles tubulaires en acier. Rapport no.
EPM/GCS-1995-09. In: Montreal, Canada: Departement de genie civil, E cole
Polytechnique.
Armstrong, P.J. and Frederick, C.O. (1966). A mathematical representation of the multiaxial
Baushinger effect. CEGP Report RD/B/N731. Berkeley Nuclear Laboratories, Berkeley,
UK.
ASTM Standard E8. (2003) Standard Test Methods for Tension Testing of Metallic Materials.
American Society for Testing and Materials.
Astaneh-Asl, A. (1998). Seismic behavior and design of gusset plates. SteelTIPS, Technical
Information and Product Service, Structural Steel Educational Council. Moraga, CA.
Bao, Y. and Wierzbicki, T. (2004). On fracture locus in the equivalent strain and stress
triaxiality space. Int. J. Mech. Sci., 46(1), 81-98.
Barsom, J.M., and Rolfe, S.T. (1987). Fracture and fatigue control in structures Applications of
fracture mechanics, Prentice Hall, Englewood Cliffs, NJ.
Black, G. R., Wenger, B. A., and Popov, E. P. (1980). "Inelastic Buckling of Steel Struts Under
Cyclic Load Reversals." UCB/EERC-80/40, Earthquake Engineering Research Center,
Berkeley, CA.
250

Bruneau, M., Engelhardt, M., Filiatrault, A., Goel, S.C., Itani, A., Hajjar, J., Leon, R., Ricles, J.,
Stojadinovic, B., and Uang, C.-M. (2005). Review of selected recent research on US
seismic design and retrofit strategies for steel structures. Progress in Structural
Engineering and Materials, 7(3), 103-114.
Cheng, J.J., Kulak, G.L. and Khoo, H.-A. (1998). Strength of slotted tubular tension members.
Can. J. Civ. Eng., 25(6), 982-991.
Chi, W-M., and Deierlein, G.G. (2000). Prediction of steel connection failure using
computational fracture mechanics. Blume Center TR136, Stanford University, Stanford,
CA.
CSA. (1989). Limit States Design of Steel Structures. CANlCSA-Sl6.1-M89, Canadian Standard
Association, Rexdale, ON.
CSA. (2001). Limit States Design of Steel Structures. CAN/CSA-S16-01, Canadian Standard
Association, Toronto, ON.
CSA. (2005). Limit States Design of Steel Structures. CSA-S16S1-05 Supplement No. 1 to
CAN/CSA-S16-01. Canadian Standard Association, Toronto, ON.
DEscata, Y. and Devaux, J.C. (1979). Numerical study of initiation, stable crack growth and
maximum load with a ductile fracture criterion based on the growth of holes. ASTM STP
668, American Society of Testing and Materials, Philadelphia, PA, 229-248.
El-Tawil, S. and Deierlein, G.G. (1998). Stress-resultant plasticity for frame structures. J. Eng.
Mech., ASCE, 124(12), 1360-1370.
Fell, B.V., Kanvinde, A.M., Deierlein, G.G., Myers, A.M., and Fu, X. (2006). Buckling and
fracture of concentric braces under inelastic cyclic loading. SteelTIPS, Technical
Information and Product Service, Structural Steel Educational Council. Moraga, CA.
FEMA. (1997). 1997 NEHRP Recommended Provisions for seismic regulations for nee
buildings and other structures. Building Seismic Safety Council, Washington D.C.
FEMA. (2000). FEMA-350: Recommended design criteria for new steel moment-frame
buildings. Federal Emergency Management Agency, Washington, D.C.
Foutch, D.A., Goel, S.C. and Roeder, C.W. (1987). Seismic testing of full-scale steel building-
Part I. J. Struct. Eng., ASCE, 113(11), 2111-2129.
Gupta, A., and Krawinkler, H. (1999). Prediction of seismic demands for SMRFs with ductile
connections and elements. SAC Background Document, Report No. SAC/BD-99/06.
Gugerli H. (1982). Inelastic cyclic behavior of steel members. Ph.D. Thesis, Department of
Civil Engineering, University of Michigan, Ann Arbor, MI, 1982.
251

Hajjar, J.F. and Gourley, B.C. (1977). A cyclic nonlinear model for concrete-filled tubes. I:
Formulation. J. Struct. Eng., ASCE, 123(6), 736-744.
Han, S.-W., Kim, W. T., and Foutch, D. A. (2007). Seismic behavior of HSS bracing members
according to width-thickness ratio under symmetric cyclic loading. J. Struct. Eng.,
ASCE, 133 (2), 264-273.
Hanbin, G., Kawahito, M. and Masatoshi, O. (2007). Ultimate strains of structural steels against
ductile crack initiation. Structural Engineering/Earthquake Engineering, 24(1), 13s-22s.
Hancock, J. W., and Mackenzie, A. C. (1976). On the mechanics of ductile failure in high-
strength steel subjected to multi-axial stress states. J. Mech. Phys. Solids, 24(3), 147
169.
Haselton, C.B. and Deierlein, G.G. (2007). Assessing seismic collapse safety of modern
reinforced concrete moment frame buildings. Blume Center TR156, Stanford
University, Stanford, CA.
Hassan, O.F. and Goel, S.C. (1991). Modeling of Bracing Members and Seismic Concentrically
Braced Frames. UMCE 91-1, University of Michigan, College of Engineering, Ann
Arbor, MI.
Higginbotham, A.B. and Hanson, R.D. Axial hysteretic behavior of steel members. J. Struct.
Div., ASCE, 102(7), 1365-1381.
Ikeda, K. and Mahin. S.A. (1986). Cyclic Response of Steel Braces. J. Struct. Eng., ASCE, 112
(2), 342-361.
Izvernari, C., Lacerte, M., and Tremblay, R. (2007). Seismic performance of multi-storey
concentrically braced steel frames designed according to the 2005 Canadian Seismic
Provisions. Proceedings of the 9th Canadian Conference on Earthquake Engineering,
Ontario, Canada, June 2007.
Liu, J., Sabelli, R., Brockenbrough, R.L., and Fraser, T.P. (2007). Expected yield stress and
tensile strength ratios for determination of expected member capacity in the 2005 AISC
Seismic Provisions. Engineering Journal, AISC, 44(1) First Quarter, 15-25.
Jain, A.K., Goel, S.C., and Hanson, R.D. (1980). Hysteretic cycles of axially loaded steel
members. J. Struct. Div. ASCE, 106(8), 17771795.
Jain, A.K., Goel, S.C., and Hanson, R.D. (1978). Inelastic response of restrained steel tubes. J.
Struct. Div. ASCE, 104(6), 897-910.
Jin, J. and El-Tawil, S. (2003). Inelastic cyclic model for steel braces. J. Eng. Mech., ASCE,
129(5), 548-557.
252

Kahn, L.F., and Hanson, R.D. (1976). Inelastic cycles of axially loaded steel members. J.
Struct. Div. ASCE, 102(5), 947959.
Kannana, A.E. and Powell, G.H. (1973). DRAIN-2D - a general purpose computer program for
dynamic analysis of inelastic plane structures. Reports No. EERC 73-6 and EERC 73-
22, University of California, Berkeley, CA.
Kanvinde, A.M and Deierlein, G.G. (2004). Micromechanical simulation of earthquake induced
fractures in steel structures. Blume Center TR145, Stanford University, Stanford, CA.
Kanvinde A.M. and Deierlein, G.G. (2007). A cyclic void growth model to assess ductile
fracture in structural steels due to ultra low cycle fatigue. J. Eng. Mech., ASCE, 133(6),
701-712.
Kelly, D. J., Bonneville, D. R., and Bartoletti, S. J. (2000). 1994 Northridge earthquake: damage
to a four-story steel braced frame building and its subsequent upgrade. 12th World
Conference on Earthquake Engineering, Upper Hutt, New Zealand, 2000.
Khatib, F., Mahin, S.A., and Pister, K.S. (1988). "Seismic behavior of concentrically braced steel
frames." UCB/EERC-88/01, Earthquake Engineering Research Center, University of
California, Berkeley, CA.
Khoei, A.R., Azadi, H. and Moslemi, H. (2008). Modeling of crack propagation via an
automatic adaptive mesh refinement based on modified superconvergent patch recovery
technique. Eng. Fract. Mech., 75(10), 2921-2945.
Korol, R.M. (1996). Shear lag in slotted HSS tension members. Can J. Civ. Eng., 23(6), 1350-
1354.
Koteski, N., Packer, J.A., and Puthli, R.S. (2005). Notch Toughness of Internationally Produced
Hollow Structural Sections. J. Struct. Eng., ASCE 131 (2), 279-286.
Krishnan, S. (in review). 3-D Modeling of Steel Braced Structures. Submitted for consideration
to the Nonlinear Analysis Special Issue, Earthquake Engineering and Structural
Dynamics.
Lee, K., and Bruneau, M. (2005). Energy dissipation of compression members in concentrically
braced frames: Review of experimental data. J. Struct. Eng., ASCE, 131(4), 552-559.
Lee, S., and Goel, S.C. (1987). Seismic behavior of hollow and concrete-filled square tubular
bracing members. Research Rep. No.UMCE 87-11, University of Michigan, Ann Arbor,
MI.
Lehman, D.E., Roeder, C.W., Herman, D., Johnson, S. and Kotulka, B. (2008). Improved
seismic performance of gusset plate connections. J. Struct. Eng., ASCE, 134(6), 890-
901.
253

Lemaitr, J., and Chaboche J.-L. (1990). Mechanics of Solid Materials, Cambridge University
Press.
Liu, Z., and Goel, S.C. (1988). Cyclic load behavior of cement-filled tubular braces. J. Struct.
Eng., ASCE, 114(7), 1488-1506.
McClintock, F.A. (1968).A criterion for ductile fracture by the growth of holes. Proc. ASME
Meeting APM-14, American Society of Mechanical Engineers, June 1214, 1968.
McCormick, J., DesRoches, R., Fugazza, D., and Auricchio, F. (2007). Seismic assessment of
concentrically braced steel frames with shape memory alloy braces. J. Struct. Eng.,
ASCE, 133 (6), 862-870.
McKenna, F. and Fenves, G.L. (2004). Open System for Earthquake Engineering Simulation
(OpenSees). Pacific Earthquake Engineering Research Center (PEER), University of
California, Berkeley, CA. (http://opensees.berkeley.edu/index.html)
Montgomery, D.C., Runger, G.C. and Hubele, N.F. (2001). Engineering Statistics, 2nd Ed., John
Wiley & Sons, Inc., New York, NY.
Myers, A.T. and Deierlein, G.G. (2009). In preparation. Blume Center Report, Stanford
University, Stanford, CA.
NRCC. (1990). National building code of Canada 1990. National Research Council of Canada,
Ottawa, Ont.
NRCC. (2005). National building code of Canada 2005. National Research Council of Canada,
Ottawa, ON.
Panontin, T. L. and Sheppard, S. D. (1995). The relationship between constraint and ductile
fracture initiation as defined by micromechanical analyses. Fracture mechanics: 26th
Volume, ASTM STP 1256, ASTM, West Conshohoken, Pa., 5485.
Priestley, M.J.N and Park, R. (1987). Strength of ductility of concrete bridge columns under
seismic loading. ACI Struct. J., 84(1), 61-76.
Popov, E. P., and Black, R. G. (1981). Steel struts under severe cyclic loading. J. Struct. Eng.,
ASCE, 107(9), 1857-1881.
Powell, G.H. and Chen, P. (1986). 3D beam-column element with generalize plastic hinges. J.
Eng. Mech., ASCE, 112(7),627-641.
Rice, J. R., and Tracey, D.M. (1969). On the ductile enlargement of voids in triaxial stress
fields. J. Mech. Phys. Solids, 17, 201-217.
Rai, D.C., Goel, S.C., and Firmansjah, J. (1996). SNAP-2DX (Structural Nonlinear Analysis
Program). Research Report UMCEE96-21, Department of Civil Engineering, University
of Michigan, Ann Arbor, MI.
254

Redwood, R.G., Lu, F., Bouchard, G. and Paultre, P. (1991). Seismic response of concentrically
braced steel frames. Can. J. Civ. Eng., 18(6), 1062-1077.
Redwood, R.G. and Channagiri, V.S. (1991). Earthquake resistant design of concentrically
braced steel frames. Can. J. Civ. Eng., 18(5), 839-850.
Roeder, C.W. (1989). Seismic behavior of concentrically braced frame. J. Struct. Eng., ASCE,
115(8) 1837-1856.
Rao, B.N. and Rahman, S. (2003). An enriched meshless method for non-linear fracture
mechanics. Int. J. Numer. Meth. Eng., 59(2), 197223.
Sabelli, R. (2001). "Research on improving the design and analysis of earthquake resistant steel
braced frames." FEMA / EERI.
Krawinkler, H.K., Gupta, A., Median, R.A. and Luco, N. (2000). Loading histories for seismic
performance testing of SMRF components and assemblies. SAC/BD-00/10, SAC Joint
Venture, Sacramento, CA.
Schafer, B.W., and Pekoz, T. (1998). Computational modeling of cold formed steel:
Characterizing geometric imperfections and residual strains. J. Constr. Steel Research,
Elsevier, 47(3), 193-210.
Shaback, B., and Brown, T. (2003). Behavior of square hollow structural steel braces with end
connections under reversed cyclic axial loading. Can. J. Civ. Eng., 30(4), 745-753.
Somerville, P.G. Smith, N., Punyamurthula, S., and J. Sun. (1997). Development of ground
motion time histories for phase 2 of the FEMA/SAC Steel Project. SAC BD/97-04, SAC
Joint Venture, Sacramento, CA.
Tang, X. (1987). Seismic analysis and design considerations of braced frame steel structures.
Ph.D. Thesis, University of Michigan, Ann Arbor, MI, 1987.
Tang, X., and Goel, S. C. (1989). Brace fractures and analysis of phase I structures. J. Struct.
Eng., ASCE, 115(8), 1960-1976.
Tremblay, R. (2000). Influence of brace slenderness on the seismic response of concentrically
braced steel frames. Behavior of Steel Structures in Seismic Areas: Proceedings of the
Third International Conference STESSA. 527-534.
Tremblay, R. (2002). Inelastic seismic response of steel bracing members. J. Constr. Steel
Research, Elsevier, 58(5), 665-701.
Tremblay, R., Archambault, M-H., and Filiatrault, A. (2003). Seismic Response of
Concentrically Brace Steel Frames Made with Rectangular Hollow Bracing Members. J.
Struct. Eng., ASCE, 129(12), 1626-1636.
255

Tremblay, R. and Poncet, L. (2005). Seismic performance of concentrically braced steel frames
in multi-storey buildings with mass irregularity. J. Struct. Eng., ASCE, 131(9) 1363-
1375.
Tremblay, R., Timler, P., Bruneau, M., and Filiatrault, A. (1995). Performance of steel structures
during the 1994 Northridge earthquake. Can. J. Civ. Eng., 22(3), 338-360.
UBC. (1979). Uniform Building Code. International Conference of Building Officials, Pasadena,
CA.
UBC. (1997). Uniform Building Code. International Conference of Building Officials, Pasadena,
CA.
Uriz, P. (2005). Towards earthquake resistant design of concentrically braced steel structures.
Ph.D. Thesis, University of California, Berkeley, 2005.
Uriz, P., Filippou, F.C., and Mahin, S.A. (2008). Model for cyclic inelastic buckling of steel
braces. J. Struct. Eng., ASCE, 134(4), 619-628.
Uriz, P., and Mahin, S.A. (2004). Seismic performance assessment of concentrically braced steel
frames. Proceedings of the 13th World Conference on Earthquake Engineering,
Vancouver, Canada, August 2004.
Walpole WR. (1996). Behaviour of cold-formed steel RHS members under cyclic loading.
Research Report 96-4. Christchurch, New Zealand: Department of Civil Engineering,
University of Canterbury.
Watabe, M. and Ishiyama, K. (1980). Earthquake resistant regulations for building structures in
Japan. Earthquake Resistant RegulationsA World List, Building Research Institute,
Tsukuba, Japan.
Whitmore, R.E. (1950). Experimental investigation of stresses in gusset plates. Masters Thesis,
University of Tennessee Engineering Experiment Station Bulletin No. 16.
Yang, F. and Mahin, S. (2005). Limiting net section fracture in slotted tube braces. SteelTIPS,
Technical Information and Product Service, Structural Steel Educational Council.
Moraga, CA. (in preparation).
Zayas, V. A., Popop, E.P., and Mahin, S.A. (1980). Cyclic inelastic buckling of tubular steel
braces. Rep. No. UCB/EERC-80/16, Earthquake Engineering Research Center,
University of California, Berkeley, CA.
Zhao, X.-L., Grzebieta, R. and Lee, C. Void-filled cold-formed rectangular hollow section
braces subjected to large deformation cyclic axial loading. J. Struct. Eng., ASCE,
128(6), 746-753.
256

Appendix A

Figure A.1 illustrates the loading histories and load-deformation plots for the nineteen brace

experiments. The limit states of global and local buckling, fracture initiation, and strength loss are

reported on each figure in terms of story drift, while the stiffness and maximum tensile and

compressive forces are shown on the hysteretic plots. The test numbers correspond to Table 3.1

and titles are also provided to distinguish the specimens, loading histories, and other attributes.

Figure A.1: Experimental loading histories and brace hysteretic response (continued on next
page)
257

Figure A.1: Experimental loading histories and brace hysteretic response (continued on next
page)
258

Figure A.1: Experimental loading histories and brace hysteretic response (continued on next
page)
259

Figure A.1: Experimental loading histories and brace hysteretic response (continued on next
page)
260

Figure A.1: Experimental loading histories and brace hysteretic response (continued on next
page)
261

Figure A.1: Experimental loading histories and brace hysteretic response (concluded on next
page)
262

Figure A.1: Experimental loading histories and brace hysteretic response


263

Appendix B

This appendix presents experimental results from past tests on large-scale bracing members. The

table summarizes the brace geometries and deformation capacities for each experiment used in

the development of the semi-theoretical model presented in Chapter 6. It should be noted that the

deformation capacities may differ slightly from the true experimental results owing to the

difficulty of assimilating data from diverse testing programs.

Table B.1: Experimental results from square and rectangular HSS tests (continued on next page)
Deformation
Test Program (Year) Test Brace Properties
Capacity
[Material] I.D.
Shape B H LB tgusset max min
TW2 HSS5x3x1/4 5 3 136 0.57 -1.7
Gugerli and Goel (1982) TW3 HSS4x2x1/4 4 2 138 0.78 -2.55
Not
[ASTM-A500, Gr. B] TW4 HSS7x5x1/4 7 5 132 0.57 -1.7
Used
TW6 HSS6x3x3/16 6 3 136 0.57 -1.7
Liu and Goel T633H HSS6x3x3/16 6 3 113 0.63 0.43 -1.70
(1987) T424H HSS4x2x1/4 4 2 120 0.63 0.85 -2.55
[ASTM-A500, Gr. B] T422H HSS4x2x1/8 4 2 120 0.50 0.68 -1.70
1 HSS5x5x3/16 5 5 116 0.63 0.33 -0.85
2 HSS5x5x3/16 5 5 126 0.63 0.32 -0.85
Lee and Goel
4 HSS4x4x1/8 5 5 126 0.50 0.21 -2.13
(1988)
5 HSS4x4x1/4 4 4 122 0.63 0.24 -3.40
[ASTM-A500, Gr. B]
6 HSS4x4x1/4 4 4 130 0.63 0.21 -2.13
7 HSS4x4x1/4 4 4 130 0.63 0.32 -1.70
S1A HSS5x3x3/16 3 5 181 0.44 1.70 -1.70
S1B HSS5x3x3/16 3 5 181 0.44 1.70 -1.70
S2A HSS4x3x3/16 3 4 182 0.44 2.52 -2.52
S2B HSS4x3x3/16 3 4 182 0.44 2.39 -2.39
S3A HSS3x3x3/16 3 3 182 0.38 3.31 -3.31
Archambault, Tremblay, S3B HSS3x3x3/16 3 3 182 0.38 3.32 -3.32
Filiatrault S4A HSS5x2.5x3/16 2.5 5 182 0.44 1.52 -1.52
(1995) S4B HSS5x2.5x3/16 2.5 5 182 0.44 1.80 -1.80
[G40.21-350W] S5A HSS4x3x1/4 3 4 182 0.50 3.28 -3.28
S5B HSS4x3x1/4 3 4 182 0.50 2.70 -2.70
S1QA HSS5x3x3/16 3 5 181 0.44 1.69 -1.69
S1QB HSS5x3x3/16 3 5 181 0.44 1.81 -1.81
S4QA HSS5x2.5x3/16 2.5 5 182 0.43 1.52 -1.52
S4QB HSS5x2.5x3/16 2.5 5 182 0.43 1.80 -1.80
264

Table B.1: Experimental results from square and rectangular HSS tests
Walpole RHS1 HSS6x4x1/4 6 4 106 1.39 -1.39
RHS2 HSS6x4x1/4 6 4 80 Not 0.97 -0.97
(1996)
Used
[AS1163-C350] RHS3 HSS6x4x1/4 6 4 99 0.49 -0.49
1B HSS5x5x5/16 5 5 134 1.00 1.36 -2.67
2A HSS6x6x5/16 6 6 157 1.00 1.25 -2.03
2B HSS6x6x3/8 6 6 157 1.00 1.31 -2.58
Shaback and Brown
3A HSS5x5x1/4 5 5 173 1.00 1.02 -2.91
(2001)
3B HSS5x5x5/16 5 5 173 1.00 1.47 -2.49
[G40.21-350W]
3C HSS5x5x3/8 5 5 173 1.00 1.34 -3.78
4A HSS6x6x5/16 6 6 193 1.00 1.16 -2.95
4B HSS6x6x3/8 6 6 192 1.00 1.18 -3.56
1 HSS 6x6x3/8 6 6 115 1.00 1.55 -1.90
Yang and Mahin
3 HSS 6x6x3/8 6 6 115 1.00 1.25 -1.01
(2006)
4 HSS 6x6x3/8 6 6 115 1.00 2.90 -0.70
[ASTM-A500, Gr. B]
5 HSS 6x6x3/8 6 6 115 1.00 1.93 -1.51
85-14A HSS4x4x1/4 4 4 133 0.63 1.33 -1.33
Han and Foutch
70-18 HSS5x5x1/4 5 5 133 0.63 1.00 -1.00
(2007)
82-19 HSS4x4x3/16 4 4 135 0.63 0.67 -0.67
[ASTM-A500, Gr. B]
77-28 HSS4x4x1/8 4 4 132 0.63 0.33 -0.33
HSS1-1 HSS4x4x1/4 4 4 118 0.50 1.59 -1.59
Fell, Kanvinde, Deierlein HSS1-2 HSS4x4x1/4 4 4 118 0.50 1.18 -3.54
(2007) HSS1-3 HSS4x4x1/4 4 4 118 0.50 1.24 -1.77
[ASTM-A500, Gr. B] HSS2-1 HSS4x4x3/8 4 4 118 0.50 3.00 -3.00
HSS2-2 HSS4x4x3/8 4 4 118 0.50 2.50 -2.50
HSS-2 HSS5x5x3/8 5 5 162 0.50 1.53 -2.14
HSS-3 HSS5x5x3/8 5 5 162 0.50 1.53 -3.06
HSS-4 HSS5x5x3/8 5 5 158 0.50 1.53 -2.96
HSS-5 HSS5x5x3/8 5 5 162 0.38 1.73 -3.16
Lehman, Roeder, Herman, HSS-6 HSS5x5x3/8 5 5 162 0.38 1.73 -3.06
Johnson, Kotulka HSS-7 HSS5x5x3/8 5 5 153 0.88 1.33 -2.86
(2008) HSS-8 HSS5x5x3/8 5 5 166 0.38 2.35 -2.65
[ASTM-A500, Gr. B] HSS-9 HSS5x5x3/8 5 5 162 0.50 1.33 -2.45
HSS-10 HSS5x5x3/8 5 5 161 0.50 1.94 -2.55
HSS-11 HSS5x5x3/8 5 5 153 0.88 1.12 -1.53
HSS-12 HSS5x5x3/8 5 5 139 0.50 1.43 -2.14
HSS-13 HSS5x5x3/8 5 5 153 0.50 1.94 -2.14

You might also like