Performance of Full-Scale Reinforced Concrete Columns Subjected To Extreme Earthquake Loading

Performance of Full-Scale Reinforced Concrete Columns
Subjected to Extreme Earthquake Loading
A Dissertation
SUBMITTED TO THE FACULTY OF
UNIVERSITY OF MINNESOTA
BY
Alireza Nojavan
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
Advisor:
Arturo E. Schultz
December 2015
Copyright © Alireza Nojavan 2015
Acknowledgements
This project was supported by the National Science Foundation under award No.
CMMI-10041633. I would like to acknowledge the Department of Civil, Environmental
and Geo-Engineering of the University of Minnesota for the Departmental Fellowship
Award.
I would like to also acknowledge all who made this research possible. First and
foremost, I would like to express my extreme gratitude to my advisor, Prof. Arturo E.
Schultz for his support, and valuable advice and directions during the course of my
research. He was not only a kind teacher, but also a source of motivation. I would also like
to acknowledge Prof. Catherine French, Prof. Lev Khazanovich, and Prof. Ryan Elliott for
serving on my thesis committee and for their helpful comments. Special thanks are due to
Prof. Shih-Ho Chao at University of Texas, Arlington (UTA) for directing the experimental
part of this project and for his comments.
The experimental part of this research could not be possible without help from several
others. I would like to acknowledge help from the staff at the MAST Lab of the University
of Minnesota. Special thanks are due to Paul Bergson, Rachel Gaulke, and Mary Vancura
who helped during all stages of the tests. Michael Boldischar is thanked for his help during
setting up cameras and recording and uploading test data. I am also thankful to Tanner
Swenson, Charles Vermace, Lauren Snyder, Anton Tillmann and Andrew Morgan for their
help during assembly, instrumentation, and testing of the specimens. Help of many others
at UTA in construction of the specimen is also appreciated.

i
Dedicated to
my parents and my wife
ii
ABSTRACT
Seven full-scale reinforced concrete (RC) columns were tested at the Multi-Axial
Subassemblage Testing (MAST) Laboratory of the University of Minnesota to investigate
their performance under extreme seismic events that would produce near-collapse
conditions. One of the goals of the tests was to investigate any potential differences in
performance with column size, thus, the test specimens were larger than nearly all of the
columns tested previously. In order to investigate the adequacy of current provisions, the
specimens were designed according to seismic provisions of ACI 318-11 and featured two
different cross sections (36×28 in. and 28×28 in.). Another goal of the program was to
investigate the influence of loading history, thus the column specimens were subjected to
several large displacement loading protocols, including monotonic and uniaxial and biaxial
cyclic loading protocols. The last overall goal of the program was to investigate the post-
peak behavior of the specimens at near-collapse conditions, hence loading on the
specimens continued beyond the stopping criteria in previous tests until the specimens
exhibited severe strength loss and stiffness degradation.
Results from these tests were combined with the available dataset of RC column tests
to study the effects of cross-sectional size on parameters representing seismic performance
of columns including moment capacity, effective stiffness, drift capacity, displacement
ductility, and reinforcing bar buckling. It was revealed that unlike the other parameters,
specimens featuring larger cross-sectional depths are more prone to in-plane bar buckling,
a failure mechanism that has never been reported during previous tests of RC columns.
iii
Unlike outward buckling of bars, in-plane bar buckling is not generally controlled by
confining reinforcement; rather it is the concrete surrounding the bars that restrain them
from in-plane buckling. To better understand this phenomenon, finite element (FE) models
of isolated bars as well as a three-dimensional (3D) FE model of the lower portion of the
tested specimens were analyzed. A parametric study indicates that concrete compressive
strength, bar size and overall cross-sectional size of the columns can affect bar buckling
while the effects of longitudinal bar and tie spacing are minor.
The evolution of damage during application of the various loading protocols was
quantified using several cumulative and noncumulative damage index models. In addition,
observed visual damage to the specimens was used to assess calculated damage indices
based on different models. Calculated and measured damage quantities were considered in
combination with the lateral force-deformation cyclic envelope, strength loss, stiffness
reduction, and hysteretic energy dissipation of the specimens to study the effects of applied
loading protocols on the performance of tested column specimens.
iv
Table of Contents
List of Tables ................................................................................................................... IX
List of Figures ................................................................................................................... X
CHAPTER 1 INTRODUCTION .................................................................................. 1
1.1. Background ................................................................................................... 1
1.2. Research Objectives and Summary .............................................................. 4
1.3. Thesis Organization ...................................................................................... 5
CHAPTER 2 A NEW DATASET FOR FULL-SCALE REINFORCED

CONCRETE COLUMNS UNDER COLLAPSE-CONSISTENT LOADING
PROTOCOLS ................................................................................................................... 8
2.1. Summary ....................................................................................................... 8
2.2. Introduction ................................................................................................... 9
2.3. Test Setup.................................................................................................... 13
2.4. Specimens ................................................................................................... 16
2.5. Materials ..................................................................................................... 18
2.6. Instrumentation ........................................................................................... 20
2.7. Loading Protocols ....................................................................................... 25
2.8. Test Observations........................................................................................ 28
2.9. Future Use of the Data Set .......................................................................... 36
2.10. Conclusions ................................................................................................ 38
CHAPTER 3 INFLUENCE OF CROSS-SECTIONAL SIZE ON SEISMIC

PERFORMANCE OF RC COLUMNS ........................................................................ 40
3.1. Summary ..................................................................................................... 40
v
3.2. Introduction ................................................................................................. 41
3.3. Research Significance ................................................................................. 44
3.4. Experimental Investigation and Data Selection .......................................... 45
3.5. Flexural Moment Capacity ......................................................................... 51
3.5.1. Calculation ..................................................................................... 51
3.5.2. Influence of Column Section Depth h ............................................. 52
3.6. Effective Stiffness ....................................................................................... 53
3.6.1. Calculation ..................................................................................... 53
3.6.2. Influence of Axial Loading and Aspect Ratio ................................. 55
3.7. Drift Capacity and Displacement Ductility................................................. 59
3.7.1. Calculation ..................................................................................... 59
3.7.2. Influence of Axial Loading and Aspect Ratio ................................. 59
3.8. Reinforcing Bar Buckling ........................................................................... 65
3.8.1. Characterization of Reinforcing Bar Buckling ............................... 65
3.8.2. Influence of Column Section Size ................................................... 72
3.9. Conclusions ................................................................................................. 74
CHAPTER 4 IN-PLANE BUCKLING OF LONGITUDINAL BARS IN

REINFORCED CONCRETE COLUMNS UNDER EXTREME
EARTHQUAKE LOADING.......................................................................................... 76
4.1. Summary ..................................................................................................... 76
4.2. Introduction ................................................................................................. 77
4.3. ExperimentaL Program ............................................................................... 79
vi
4.4. Test Observations........................................................................................ 80
4.5. Analytical Investigation Using Bar-Spring Model ..................................... 84
4.5.1. General Description of the Bar-Spring System .............................. 84
4.5.2. Static Analysis Using Modified Riks (Arc-Length) Method............ 87
4.5.3. Results from Bar-Spring Analysis................................................... 88
4.6. Finite Element Modeling of RC Column Specimens ................................. 90
4.6.1. General Description of the FE Model ............................................ 91
4.6.2. Estimation of Spring Stiffness ......................................................... 93
4.6.3. Modeling of Concrete Behavior ..................................................... 98
4.6.4. Explicit Dynamic Analysis of the Model....................................... 100
4.6.5. Model Validation .......................................................................... 102
4.6.6. Parametric Study .......................................................................... 105
4.7. Results and Observations .......................................................................... 106
4.7.1. Effect of Concrete Compressive Strength (f′c) .............................. 106
4.7.2. Effect of Longitudinal Bar Size and Spacing ................................ 108
4.7.3. Effect of Tie Spacing ..................................................................... 110
4.7.4. Effect of Column Cross-Sectional Size ......................................... 111
4.8. Conclusions ............................................................................................... 113
CHAPTER 5 EFFECT OF LOADING PROTOCOLS AND ESTIMATION

OF DAMAGE INDICES IN REINFORCED CONCRETE COLUMNS................ 115
5.1. Summary ................................................................................................... 115
5.2. Introduction ............................................................................................... 116
5.3. Previous Work on Damage Indices........................................................... 120
5.3.1. Noncumulative damage indices .................................................... 121
vii
5.3.2. Cumulative damage indices .......................................................... 124
5.4. Experimental Program .............................................................................. 138
5.5. Measured Response of Column Specimens .............................................. 145
5.6. Damage Analysis of Tested Specimens .................................................... 151
5.7. Correlation of Calculated and Observed Damage .................................... 160
5.8. Effect of Loading Protocols ...................................................................... 173
5.9. Conclusions ............................................................................................... 180
CHAPTER 6 SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS ... 182
6.1. Summary ................................................................................................... 182
6.2. Conclusions ............................................................................................... 184
6.3. Recommendations for Future Research .................................................... 186
REFERENCES.............................................................................................................. 188
APPENDIX A CONSTRUCTION, INSTRUMENTATION, AND

INSTALLATION OF SPECIMENS ........................................................................... 199
A.1. Construction .............................................................................................. 199
A.2. Instrumentation ......................................................................................... 205
A.3. Shipping and Installation .......................................................................... 211
viii
List of Tables
Table 2.1. Normal Strength Concrete Mix-Design .................................................... 19
Table 2.2. Material Properties of the Specimens ....................................................... 19
Table 2.3. Number of Different Types of Sensors for Each of the Specimens ......... 23
Table 2.4. Amount of Image and Video Data Recorded During Each Test (GB) ..... 24
Table 2.5. Applied Loading Protocols ....................................................................... 27
Table 3.1. Characteristics of the MAST Tests ........................................................... 48
Table 3.2. Properties of Rectangular Columns in the ACI 369 Database ................. 51
Table 3.3. Drift Ratios at the Onset of Bar Buckling and Fracture ........................... 71
Table 4.1. Critical Loads from FE Analysis of Bar-Spring System .......................... 90
Table 4.2. Parametric Study Matrix ......................................................................... 105
Table 5.1. Damage Classification for Park-Ang Model .......................................... 138
Table 5.2. Applied Lateral Loading to the Specimens ............................................ 140
Table 5.3. Estimated Damage Values at the End of Each Test ............................... 153
Table 5.4. Visual Damage Categories Based on Test Observations........................ 162
Table 5.5. Measured Strength Loss at the Onset of Visual Damage Categories ..... 167
Table 5.6. Estimated Damage Values for Various Visual Damage Categories....... 172
ix
List of Figures
Figure 2.1. Properties of Previous Tests on Rectangular RC Columns. .................... 10
Figure 2.2. (a) Assembly of the Specimen under the Loading Crosshead
(Vertical Actuators are not Shown for Clarity) (b) Definition of
the Coordinate System and (c) Specimen Connection to the
Crosshead. .............................................................................................. 15
Figure 2.3. Location of the Inflection Point and Definition of the Top
Moment................................................................................................... 16
Figure 2.4. (a) 3D Rendering of PF Specimens and (b) Cross-Sectional
Detailing of PF (Top) and SF Specimens (Bottom) ............................... 17
Figure 2.5. LVDTs (LV) and String Pots (SP): (a) Location along the Height
and (b) Section 1-1 ................................................................................ 21
Figure 2.6. Locations of Strain Gages: (a) on Longitudinal Bars (b) on
Transverse Hoops (c) and (d) Horizontally within the Concrete
and (e) Vertically within the Concrete ................................................... 22
Figure 2.7. Location of the Telepresence Towers and High-Resolution
Cameras at the MAST Lab ..................................................................... 23
Figure 2.8. Loading Protocols for Specimens: (a) SP1; (b) SP2, SP5, and
SP8; (c) SP3; (d) SP4; (e) SP6; (f) SP7 .................................................. 26
Figure 2.9. Progression of Damage on SP2: (a) SE Face at 2.2% Drift Ratio;
(b) SW Face at 2.2% Drift Ratio; (c) SE Face at the End of the
Test and (d) SW Face at the End of the Test .......................................... 30
x
Figure 2.10. Applied Load vs. Crosshead Displacement and Drift Ratio ................. 31
Figure 2.11. Reinforcement Strain Profiles along Column Height at 2.15%
Drift Ratio for Specimen SP2: (a) Longitudinal Bar (SL1G); (b)
Transverse Hoops (STR3G) ................................................................... 33
Figure 2.12. Concrete Core Strain Profiles along Column Height at 2.15%
Drift Ratio for Specimen SP2: (a) Vertical Strain Gages (CL1G);
(b) Horizontal Strain Gages (CTRH4G)................................................. 34
Figure 2.13. Moment vs. Curvature Relation for SP2 Measured 6 in. from the
Column Base. ......................................................................................... 35
Figure 3.1. Overview of the Test Specimens ............................................................. 46
Figure 3.2. Overview of the Test Setup at the MAST Lab and Loading
Direction ................................................................................................. 47
Figure 3.3. Behavior of Specimen SP4 under the Applied Loading Protocol:
(a) Loading Protocol, (b) Variation of Lateral Force with Drift
Ratio and Top Displacement .................................................................. 50
Figure 3.4. Normalized Moment Capacity vs. Cross-Sectional Depth for
Columns Failing in Flexure .................................................................... 53
Figure 3.5. Definition of the Effective Stiffness........................................................ 54
Figure 3.6. Normalized Effective Stiffness vs. Axial Load Ratio ............................. 56
Figure 3.7. Normalized Effective Stiffness vs. Cross-Sectional Depth ..................... 57
Figure 3.8. Normalized Effective Stiffness vs. Cross-Sectional Depth ..................... 58
Figure 3.9. Variation of Drift Capacity with Axial Load Ratio................................. 60
xi
Figure 3.10. Variation of Drift Capacity with Cross-Sectional Depth ...................... 63
Figure 3.11. Variation of Displacement Ductility with Cross-Sectional Depth ........ 64
Figure 3.12. Location of Strain Gages for Cross Sections of MAST Test
Specimens ............................................................................................... 67
Figure 3.13. Sudden Drops in Load Carrying Capacity due to Bar Fractures ........... 68
Figure 3.14. Characterization of the Onset of Bar Buckling ..................................... 69
Figure 3.15. In-Plane bar buckling and Fracture in MAST Column Tests................ 73
Figure 4.1. Specimen Geometry and Detailing ......................................................... 81
Figure 4.2. In-Plane Bar Buckling During Tests at the MAST Lab: (a)
Column Specimen under Monotonic Loading, (b) Column
Specimen Subjected to Cyclic Loading .................................................. 83
Figure 4.3. Bar-Spring System to Simulate Bar Buckling Behavior ......................... 86
Figure 4.4. Stress-Strain Behavior of Longitudinal Bars .......................................... 86
Figure 4.5. FE Analysis Results for a No. 9 Bar ....................................................... 88
Figure 4.6. General Description of FE Model: (a) Modeled Section (b) Mesh
of Concrete Material (c) Modeling of Buckling Bars within
Concrete Elements .................................................................................. 92
Figure 4.7. Preventive Effect of Surrounding Concrete from In-Plane
Buckling after Cover Spalling ................................................................ 95
Figure 4.8. Spring Stiffness Parallel to the Column Face.......................................... 98
Figure 4.9. Stress-Strain Relationship of Concrete ................................................. 100
Figure 4.10. In-Plane Buckling of Bars in the FE Model ........................................ 103
xii
Figure 4.11. Moment-Rotation Behavior of Tested and Modeled Specimens ........ 104
Figure 4.12. Effect of Concrete Compressive Strength on Buckling
Displacement in: (a) Corner Bars, (b) Middle Bars ............................. 107
Figure 4.13. Variation of Axial Loads of the Reinforcing Bars During FE
Analysis in: (a) Corner Bars, (b) Middle Bars ..................................... 108
Figure 4.14. Effect of Bar Size on Buckling Displacement in: (a) Corner
Bars, (b) Middle Bars ........................................................................... 109
Figure 4.15. Effect of Bar Spacing on Buckling Displacement in: (a) Corner
Figure 4.16. Effect of Tie Spacing on Buckling Displacement in: (a) Corner
Figure 4.17. Effect of Cross-Sectional Size on Buckling Displacement in: (a)
Corner Bars, (b) Middle Bars ............................................................... 112
Figure 5.1. Definition of Plastic Deformations in Stephens-Yao Model ................ 126
Figure 5.2. Definition of Primary and Follower Half Cycles (Kratzig &
Meskouris, 1987) .................................................................................. 135
Figure 5.3. Applied Loading Protocol to Specimen SP1 ......................................... 142
xiii
Figure 5.10. Force-Deformation Behavior of Specimen SP1 .................................. 147
Figure 5.16. Force-Deformation Behavior of Specimen SP7 along Y
Direction ............................................................................................... 150
Figure 5.17. Force-Deformation Behavior of Specimen SP7 along X
Direction ............................................................................................... 150
Figure 5.18. Estimation of the Yield Displacement and Yield Force ...................... 151
Figure 5.19. Calculated Damage Indices for Specimen SP2 Based on LUC80 ...... 157
Figure 5.26. Observed Damage at the Beginning of “Light” Damage
Category ............................................................................................... 163
xiv
Figure 5.27. Observed Damage at the Beginning of “Moderate” Damage
Category ............................................................................................... 164
Figure 5.28. Observed Damage at the Beginning of “Severe” Damage
Category ............................................................................................... 165
Figure 5.29. Observed Damage at the Beginning of “Failure” Damage
Category ............................................................................................... 166
Figure 5.30. Estimated Damage Index Values Corresponding to Different
Visual Damage Categories Based on Stephens Model......................... 168
Visual Damage Categories Based on Wang Model ............................. 168
Visual Damage Categories Based on Roufaiel Model ......................... 169
Visual Damage Categories Based on Kratzig Model ........................... 169
Visual Damage Categories Based on Park Model ................................ 170
Visual Damage Categories Based on Lybas Model ............................. 170
Visual Damage Categories Based on Banon Model............................. 171
Visual Damage Categories Based on Displacement Ductility ............. 171
xv
Figure 5.38. Generation of Cyclic Envelope for Specimen SP2 ............................. 175
Figure 5.39. Force-Deformation Cyclic Envelopes for All Tests ............................ 175
Figure 5.40. Calculated Energy Dissipation until Occurrence of 50% Strength
Loss....................................................................................................... 176
Figure 5.41. Damage Values by Lybas Model until Occurrence of 50%
Strength Loss ........................................................................................ 178
Figure 5.42. Damage Values by Stephens Model until 50% Strength Loss ............ 178
Figure 5.43. Damage Values by Roufaiel Model until 50% Strength Loss ............ 179
Figure 5.44.Damage Values by Kratzig Model until Occurrence of 50%
Strength Loss ........................................................................................ 179
Figure 5.45. Damage Values by Park Model until Occurrence of 50%
Strength Loss ........................................................................................ 180
Figure A.1. Reinforcement Details for Perimeter Frame (PF) Column
Specimens ............................................................................................. 200
Figure A.2. Reinforcement Details for Perimeter Frame (PF) Column
Specimens ............................................................................................. 201
Figure A.3. Reinforcement Details for Space Frame (SF) Column Specimens ...... 202
Figure A.4. Reinforcement Details for Space Frame (SF) Column Specimens ...... 203
Figure A.5. (a) Building Reinforcing Cage, (b) Placing the Cage in the Forms ..... 204
Figure A.6. Forming Column and Top Block.......................................................... 205
Figure A.7. Grinding and Cleaning of Bars Prior to Strain Gage Installation......... 206
Figure A.8. Attaching the Strain Gage to the Prepared Surface of the Bar ............. 207
xvi
Figure A.9. Installation of an Aluminum Anchor for External Instrumentation ..... 208
Figure A.10. Installation of Vertical LVDTs on SE Face of the Column ............... 209
Figure A.11. Installation of LVDTs and String Pots on NE Face of the
Column ................................................................................................. 210
Figure A.12. Post Tensioning of the Specimen Prior to Shipping........................... 212
Figure A.13. Lifting the Specimen Before Shipping ............................................... 213
Figure A.14. Delivery of the Specimen to the MAST Lab ...................................... 213
Figure A.15. Unloading the Specimen at the MAST Lab ....................................... 214
Figure A.16. Tilting Up the Specimen at the MAST Lab ....................................... 215
Figure A.17. Three-Piece Spacer Block .................................................................. 216
Figure A.18. Installation of the Specimen on Top of the Spacer Blocks ................ 217
xvii
CHAPTER 1
INTRODUCTION
1.1. BACKGROUND
The recently advanced performance-based design philosophy anticipates a structure to
dissipate energy during an earthquake and accepts certain levels of damage depending on
the expected risk of occurrence of earthquakes during their lifetime, and the importance of
the structure. In fact, there are different structural design objectives in the performance-
oriented seismic design philosophy, and these are based on various levels of accepted
damage (i.e., limit states) under distinct levels of earthquake loads that are expected during
lifetime of the structure. Estimated structural demands (e.g. required strength, or stiffness)
to meet the structural objectives are then compared against capacity of the structure (e.g.
available strength, stiffness, or deformation capacity). Evaluation of structural capacity, on
the other hand, relies primarily on numerical modeling tools that are calibrated against test
results. In the case of reinforced concrete (RC) columns, the test results are used to estimate
lateral load and deformation capacities under simulated seismic loadings.
1
A large number of experimental efforts have been undertaken on RC columns to
enhance knowledge on their performance under gravity and lateral loads, and to develop
and improve design code provisions. Properties of more than 300 of such tests are collected
in the ACI 369 Rectangular Column Database (Ghannoum et al., 2012) as well as in the
PEER structural performance database (Berry, Parrish, & Eberhard, 2004). Primary
variables in these tests include axial load ratio, lateral loading scheme, cross-sectional
dimensions, shear span–to-depth ratio, and amount and distribution of transverse
reinforcement.
While many tests have been conducted on RC columns, their performance, specifically
in the post-peak region, is not completely understood. This lack of knowledge in the
softening regime of response for the columns is mainly due to the fact that almost all of the
previous tests were terminated once the specimens lost 20% or less of their lateral load
capacity. While an arbitrary definition of structural failure can be tied to the
aforementioned condition, RC columns can still survive collapse when loaded beyond this
arbitrary point of failure. Therefore, to investigate the seismic performance of RC columns
and to generate and calibrate component models, especially at the near-collapse stage, RC
column specimens are required to undergo large deformation levels under which they
exhibit significant strength loss (i.e., 50% or more) and post-peak behavior (i.e., up to
residual strengths 50% or less of the peak resistance). Development of such improved
models—which is important, according to the National Institute of Standards and
Technology (NIST) and the National Earthquake Hazards Reduction Program (Advisory
Committee on Earthquake Hazards Reduction, 2008; National Institute of Standards and

2
Technology, 2009)—requires that the shortcomings in available RC column test data be
addressed.
Additionally, many of the previous tests were conducted on columns with smaller
dimensions than those currently used in mid- or high-rise buildings mainly due to
limitations of the testing facilities. The average cross-sectional depth of the rectangular RC
columns in the more recent database (i.e., ACI 369 Rectangular Column Database, 2012)
is 12.5 in. (318 mm) with lower 5% and upper 95% bounds of 6.3 in. (160 mm) and 21.65
in. (550 mm), respectively. More specifically, an analysis of these databases indicates that
columns with flexural failure modes feature a maximum cross-sectional dimension of 24
in. (610 mm). While these tests provide a valuable dataset for RC columns, a careful study
is required to assess the validity and applicability of their results to estimate seismic
performance of actual columns with larger dimensions.
Last but not least, unlike RC bridge columns, there are limited previous experimental
efforts that have focused on the effects of applied loading protocols on RC building
columns and during which similar test specimens were subjected to monotonic and cyclic
loading protocols. Considering the uncertainty in the amplitude and number of cycles that
RC columns in a building structure may sustain during an earthquake, their performance
under distinct simulated loading protocols is needed in order to more accurately estimate
their force-deformation (i.e., “backbone”) capacity curves, as well as to accurately
characterize damage during the various stages of loading.
3
1.2. RESEARCH OBJECTIVES AND SUMMARY
To enhance understanding of the behavior of RC columns that are representative of
those commonly used in mid- and high-rise buildings and to address the shortcomings of
the aforementioned data set for rectangular RC columns, eight full-scale columns were
tested as a part of this research at the Multi-Axial Subassemblage Testing (MAST)
Laboratory of the University of Minnesota. These column specimens were subjected to a
constant axial load and various loading schemes, including uniaxial and biaxial symmetric
cyclic, near-collapse unsymmetric cyclic, and single-cycle loading protocols. The
protocols were designed to attain drift ratios exceeding 10% such that the specimens would
lose most of their lateral loading capacity and exhibit significant loss of flexural strength
and stiffness. The specimens were representative of actual columns in the ground floor of
a 20-story building located in a high seismic region and featuring two different cross-
sectional dimensions (36×28 in. and 28×28 in.), both which are larger than all flexure-
critical columns tested previously.
Results from the tests that were conducted at the MAST Lab as part of this research
were utilized along with those from previous tests that are included in the ACI 369
Rectangular Column Database (2012) to investigate the effect of cross-sectional size of RC
columns, and on the parameters that represent their seismic performance including the
moment capacity, effective stiffness, drift capacity, ductility, and drift at the onset of
buckling.
4
During the tests that were carried out at the MAST Lab, a failure mechanism was
observed that has never been observed in previous tests of RC columns. In this failure
mechanism, reinforcing bars buckled parallel to the face of the columns, which is referred
to as in-plane buckling in this study, while the common assumption is that longitudinal
bars would only buckle in the outward direction (i.e., perpendicular to the column face).
Unlike outward buckling, transverse ties were observed to have little effect in preventing
in-plane buckling of bars. This unobserved failure mechanism is investigated numerically
using 3D FE analysis of the columns to obtain a better understanding of the column
characteristics that affect it.
Last but not least, RC columns experienced severe damage during these tests resulting
in significant strength loss and stiffness degradation of the specimens. The extent of
damage in each specimen was calculated using several well-known damage index models
for RC elements. Measured damage indices were then compared against observed damage
during tests at the MAST Lab. Finally, measured and observed damage to the specimens
were considered along with cyclic force-deformation response, and energy dissipation of
the specimens to investigate the effect of the applied loading protocols.
1.3. THESIS ORGANIZATION
Chapter 2 describes characteristics of the tests that were conducted as part of this
research at the MAST Lab including details of specimens, material properties,
instrumentation, loading protocols, and brief test observations. Detailed information
regarding each test as well as measured response of specimens during loading are made
5
available via digital object identifier (DOI) links provided in the chapter. While the current
study focuses on the behavior of seven columns that were constructed with normal strength
concrete, this chapter also presents brief information regarding a specimen that was
constructed with ultra-high-performance fiber-reinforced concrete (UHP-FRC) and tested
as part of the experimental program. A detailed information regarding this specimen and
its behavior during the test can be found elsewhere (Palacios, 2015).
A study on the effect of cross-sectional size of the RC columns on their seismic
performance is investigated in Chapter 3 by analyzing moment capacity, effective stiffness,
drift capacity, displacement ductility, and longitudinal bar buckling. For the purpose of this
analysis, tested column specimens that are included in the ACI 369 Rectangular Column
Database (2012) as well as those constructed with normal strength concrete (i.e., specimens
SP1 – SP7) and tested as part of this research are considered.
In-plane buckling of longitudinal bars, an unobserved failure mechanism during
previous tests, is described in Chapter 4. The chapter also presents information regarding
the FE modeling of isolated bars, as well as that related to the 3D FE modeling of the lower
portion of tested specimens. Results from a parametric study on the effects of parameters
that control in-plane buckling of reinforcing bars is also included.
In Chapter 5, the evolution of damage during each test is quantified by cumulative and
noncumulative damage index modes. Also, the observed damage are categorized and
compared against measured damage indices. The chapter concludes by presenting the
effect of loading protocols on the performance of column specimens.

6
Finally, a summary of this research and conclusions obtained are presented in Chapter
6. Based on these conclusions, recommendations for future study are offered.
7
CHAPTER 2
A NEW DATASET FOR FULL‐SCALE REINFORCED

CONCRETE COLUMNS UNDER COLLAPSE‐CONSISTENT
LOADING PROTOCOLS1
2.1. SUMMARY
A series of eight full-scale reinforced concrete column tests was recently carried out
at the NEES (Network for Earthquake Engineering Simulation) Multi-Axial
Subassemblage Testing (MAST) site at the University of Minnesota as part of a National
Science Foundation (NSF) NEES research program. The tests were conducted to address
the shortcomings in the available database of reinforced concrete (RC) columns tested with
large drift ratios under monotonic and cyclic loading protocols. The specimens were
designed based on ACI 318-11 and featured two different cross-sectional dimensions, both
larger than nearly all of the columns tested previously. They were subjected to several large
displacement loading protocols, including a monotonic and a cyclic biaxial loading
1
Nojavan, A., Schultz, A.E., Haselton, C., Simathathien, S., Liu, X., and Chao, S-H. (2015). “A New Dataset
for Full-Scale RC Columns under Collapse-Consistent Loading Protocols,” Earthquake Spectra, EERI, 31(2):
1211-1231.
8
protocol. Also, to investigate the effectiveness of novel materials, one specimen was
constructed with ultra-high-performance fiber-reinforced concrete (UHP-FRC). This
chapter presents a description of and potential uses for the data set that is made accessible
via a digital object identifier (DOI) (data set DOI: 10.4231/D33T9D65T).
2.2. INTRODUCTION
One of the primary goals of conducting tests on RC columns is to enhance knowledge
of their behavior and to develop component models that can predict the columns’ seismic
performance in a structural system. For the purpose of seismic performance–based design,
improved component models are required to capture the post-peak behavior of columns
more accurately. Development of such improved models—which is important, according
to the National Institute of Standards and Technology (NIST) and the National Earthquake
Hazards Reduction Program (Advisory Committee on Earthquake Hazards Reduction
(ACEHR), 2008; National Institute of Standards and Technology (NIST), 2009)—requires
that the shortcomings in available RC column test data be addressed.
A large number of experimental investigations have been carried out to understand
the behavior of RC columns subjected to various loading protocols. The properties of 326
tests on rectangular RC columns are presented by (Ghannoum et al., 2012). Primary
variables in these tests include axial load ratio, lateral loading scheme, cross- sectional
dimensions, shear span–to-depth ratio, amount and distribution of transverse
reinforcements, and bond-slip characteristics. Figure 2.1 illustrates the maximum
9
15
Flexure
12.5 Flexure-Shear
Shear
10
Maximum Drift (%)

7.5
2.5
0
0 10 20 30 40
Max. Cross-Sectional Dimension (in.)
Figure 2.1. Properties of Previous Tests on Rectangular RC Columns.
cross-sectional dimension, maximum drift ratio, and failure mode of column specimens in
previous experiments. It shows that columns with flexural failure modes feature a
maximum cross-sectional dimension of 24 in. A few of the smaller specimens were loaded
to drift ratios of 10% or more; columns with larger cross-sectional dimensions were loaded
to drift ratios smaller than 5% because they failed in more brittle modes, either in shear or
in flexure-shear.
There are no data on flexure-dominated column sections of sufficiently large
dimensions to realistically represent those used in the lower stories of multistory buildings
and loaded to sufficiently large drifts (∼10%) to elucidate their degradation characteristics
for calibration of collapse simulation models. Therefore, to investigate the seismic
performance of flexure-dominated RC columns and to generate and calibrate component
models, especially at the near-collapse stage, specimens that are representative of those
10
typical in tall buildings are required to undergo large deformation levels under which they
exhibit significant strength loss (i.e., 80%) and post-peak behavior (i.e., up to residual
strengths equal to approximately 20% of peak resistance). Several studies have explored
the effects of loading protocols on the behavior of RC members (Applied Technology
Council (ATC), 2009a; Dhakal & Fenwick, 2008; Ingham, Liddell, & Davidson, 2001;
William C. Stone, Cheok, & Stanton, 1995; Takemura & Kawashima, 1997). However, in
only a limited number of investigations were identical specimens placed under both
monotonic and cyclic loading protocols (Applied Technology Council (ATC), 2009a;
Haselton, Liel, Lange, & Deierlein, 2008). Also, in many experiments column specimens
were subjected to symmetric cyclic reversals of lateral displacements, in addition to either
a constant (Dawn E. Lehman & Moehle, 1998; Jaradat, McLean, & Marsh, 1998;
Esmaeily-Gh & Xiao, 2002) or a variable (Gilbersten & Moehle, 1980; Kreger & Linbeck,
1986) axial load. However, in an actual earthquake small loading cycles are often followed
by large unsymmetric displacement excursions that can lead to major plastic deformations
in one direction and in turn to structural collapse due to P-delta effect. Therefore, it is
necessary to investigate the effect of unsymmetric cyclic loading protocols that represent
what columns will experience during earthquakes.
To address the shortcomings of the available data set for rectangular RC columns, a
series of eight full-scale columns were subjected to a constant axial load and various
loading schemes, including uniaxial and biaxial symmetric cyclic, near-collapse
unsymmetric cyclic, and single-cycle loading protocols. The protocols were designed to
attain drift ratios exceeding 10% such that the specimens would lose most of their lateral
11
loading capacity and exhibit significant loss of flexural strength. The specimens were
representative of actual columns in the ground floor of a 20-story building located in a high
seismic region and featuring two different cross-sectional dimensions (36×28 in. and
28×28 in.) that are larger than all flexure-critical columns tested previously.
In a related matter, limited investigation was conducted on the helpful effects of ultra-
high-performance fiber-reinforced concrete (UHP-FRC) in improving damage tolerance
when applied in large column specimens. To study the effects of this emerging innovative
material in enhancing the seismic performance of columns, one of the specimens was
constructed with UHP-FRC consisting of a mixture of high-strength steel microfibers and
concrete with a compressive strength of 25 ksi at 28 days.
The experimental data recorded during the eight tests are available through DOI links
provided in this paper. The DOIs also contain all required pre- and post-test information,
including assembly of the specimens, cross-sectional details for each specimen, material
properties, location and orientation of the sensors and instrumentations (LVDTs, string
pots, tiltmeters, and strain gages), and video and still images for each specimen. The data
set is provided in complete accordance with available databases such as the ACI 369
Rectangular Column Database (Ghannoum et al., 2012), and the Pacific Earthquake
Engineering Research center (PEER) Structural Database (Berry et al., 2004). The data set
described here will provide an opportunity for researchers to explore the performance of
RC columns in various cyclic loading conditions. It will also help researchers enhance their
knowledge of the seismic performance of full-size columns under severe loading
12
conditions and to employ it to develop enhanced computational tools and seismic code
provisions for structural design.
2.3. TEST SETUP
Eight full-scale RC columns with two different cross-sectional dimensions were built
and tested under distinct axial load ratios and various lateral loading protocols. All of the
specimens were constructed and cast in an upright position in the Civil Engineering
laboratory of the University of Texas at Arlington, Texas, and were tested at the Multi-
Axial Subassemblage Testing (MAST) laboratory of the University of Minnesota. The six-
degree-of-freedom (6-DOF) loading system at the MAST lab is capable of applying up to
1,320 kips of vertical force, 880 kips of lateral force in two orthogonal horizontal
directions, 8,910 kip-ft of moments, and maximum displacements of ±20 in: and ±16 in: in
the vertical and horizontal directions, respectively (French et al., 2004). The loading
crosshead of the MAST lab is controlled by an MTS 6-DOF controller that computes the
required movement for each of the actuators (in their current configuration and along their
local coordinate system) to apply a desired boundary condition to the specimen considering
the effect of geometric nonlinearity. Each DOF defining the position of the crosshead can
be controlled either in displacement or in force mode, and the global coordinate system of
the crosshead remains constant regardless of the movement of the specimen. Users specify
the history of displacement or load for each of the six global DOFs, and the controller
computes the required forces or displacements of the local DOFs for each of the actuators.
13
Figure 2.2 illustrates the assembly of a specimen under the loading crosshead at the MAST
lab and the definition of the MAST and specimen coordinate systems.
The specimens were oriented at a 45° angle (X′ -Y′ axes) with respect to the MAST
lab primary coordinate system (X-Y axes) to provide a larger stroke capacity of ±22.6 in.
in the diagonal direction (X ′ or Y′ ) for the actuators, as opposed to ±16 in. stroke capacity
in the X (or Y) direction. The test specimen comprised the lower portion of a column in
double curvature plus a short segment above the point of inflection such that the latter was
located 10 in. below the soffit of the top block of the specimen. To keep the inflection point
at the desired location, a flexural moment was introduced by the crosshead on top of the
specimen (Figure 2.3).
14
Figure 2.2. (a) Assembly of the Specimen under the Loading Crosshead (Vertical
Actuators are not Shown for Clarity) (b) Definition of the Coordinate System and (c)
Specimen Connection to the Crosshead.
15
2.4. SPECIMENS
The specimens were representative of the lower portion of a column bent in double
curvature (Figure 2.3) at the ground floor of a 20-story prototype moment-resistant frame
building. Columns were designed according to seismic provisions in Chapter 21 of
ACI 318-11 (ACI Committee 318, 2011) and were designated as perimeter frame (PF) or
space frame (SF) depending on their location in the floor plan, representing an interior
column along the perimeter frame and an interior column in a space frame, respectively.
The PF specimens had a rectangular cross section of 36×28 in. with 16 #9 longitudinal
bars; the SF specimens featured a 28 in. square cross section with 12 #8 longitudinal bars
Figure 2.3. Location of the Inflection Point and Definition of the Top Moment
16
(Figure 2.4). In both specimen types, the longitudinal bars were tied with closed hoops
bent from #5 bars and placed with 5- or 6-in. center-to-center spacing depending on their
location relative to the column base. The column portion of all specimens were 106 in. tall
and were cast along with an 84×84×30 in. footing block and a 75×75×23 in. loading (top)
block. The footing block enabled connection to a post-tensioned, three-piece concrete base
block of 102×102×60 in. dimensions, which itself was attached to the lab strong
(b)
(a)
Figure 2.4. (a) 3D Rendering of PF Specimens and (b) Cross-Sectional Detailing of PF
(Top) and SF Specimens (Bottom)
17
floor using 1.5 in. ASTM A193 Grade-B7 threaded rods with a minimum yield strength of
125 ksi. The loading block connected to the MAST crosshead using the same threaded
rods. The entire testing assembly had a height of 219 in. under the crosshead. Figure 2.4
illustrates a typical PF specimen along with cross-sectional details of both specimen types.
2.5. MATERIALS
The specimens were built using ASTM A706 (2003) Grade-60 reinforcing steel and
normal-strength, normal-weight concrete except the last specimen (SP8), which was built
with UHP-FRC. The concrete used for the specimens was self-compacting with a
maximum aggregate size of 3/8 in. and a specified nominal 28-day compressive strength
of 5,000 psi. The concrete strength of each specimen was determined using standard
cylinder tests according to ASTM C39C (2003). For each specimen, three 4×8 in. cylinders
were tested on the day the specimen was tested. The concrete mix design for the column
specimens is presented in Table 2.1. Average measured specimen compressive strengths
on the testing day are presented in Table 2.2. The UHP-FRC specimen was constructed
with concrete with a compressive strength at 28 days of 25 ksi along with a 3% volume
fraction of high-strength steel microfibers to improve its ductility and seismic performance.
Uniaxial tensile tests were conducted on steel coupons according to ASTM A370-03a
(2003) to measure the mechanical properties of the steel bars. Samples of 18 in. lengths
were cut from each heat of reinforcing bars. The samples were instrumented with an 8-in.
gage length and pulled at a constant rate of 0.125 in/min. using a 200-kip universal testing
machine with hydraulic grips. In addition to the applied force and displacement, strains
18
were captured by an extensometer located at the middle of the gage length of each sample.
Mechanical properties of the longitudinal and transverse bars are presented in Table 2.2,
where b × h show the cross-sectional dimension of each specimen and f y, f u, and εu
represent reinforcement yield, ultimate stress, and ultimate strain, respectively.
Table 2.1. Normal Strength Concrete Mix-Design
Weight (lb) for 1 Cu. Yd. of Normal Strength Concrete
Coarse
Total
Cement Sand Aggregate Water W/C
Weight
(3/8")
748 1227 1666 404 0.54 4045
Table 2.2. Material Properties of the Specimens
Longitudinal Bars Transverse Bars

Age of
bh fc
Spec. testing
(in.) (psi) Bar fy fu Bar fy fu
(days) u u
Size (ksi) (ksi) Size (ksi) (ksi)
SP1 36×28 4,860 28 9 74 112 0.2 5 69 104 0.12
SP2 36×28 5,400 28 9 74 112 0.2 5 66 108 0.12
SP3 36×28 5,300 28 9 74 112 0.2 5 65 105 0.12
SP4 36×28 5,370 20 9 74 112 0.2 5 65 105 0.12
SP5 28×28 5,270 21 8 59 102 0.15 5 64 103 0.12
SP6 36×28 4,610 19 9 74 112 0.2 5 65 105 0.12
SP7 36×28 4860 14 9 74 112 0.2 5 62 103 0.11
SP8 28×28 23000 24 8 59 102 0.15 5 63 104 0.12
19
2.6. INSTRUMENTATION
Linear variable differential transformers (LVDTs) were installed in a vertical
configuration on each specimen to measure column deflections from which curvature could
be calculated. Also, horizontal LVDTs and diagonal string potentiometers (pots) were
installed to measure specimen shear deformation at 36 and 96 in. from the column base, as
shown in Figure 2.5. Additional string pots were connected to the specimen at 96 in. from
the top of the base block to track displacement at the inflection point. A tiltmeter was
installed at this point to measure specimen rotations. Strain gages were used to record
strains on longitudinal bars and transverse hoops and in the concrete core in each specimen;
their location is shown on Figure 2.6. In the figure, “XX” after the each strain gage name
refers to its specific location along the specimen height. The total number of different
sensor types for each specimen is given in Table 2.3.
Besides the sensors, high- and extra-high-resolution still cameras as well as
high-resolution video cameras were applied during each test. Still images were captured
by eight Canon SX110IS 9.0-megapixel still-image cameras located at two different
elevations on four telepresence towers at the four corners of the MAST strong floor, as
shown in Figure 2.7. Also, high-definition (HD) videos were recorded by eight Sony
EVI-HD1 video cameras located next to the image cameras on telepresence towers. Each
still camera and video recorder was adjusted to focus on a single face of the column
specimens. Still images were taken either at every 1-in. displacement of the crosshead or
at the peak displacement of each drift cycle, whereas videos were recorded continuously
20
Figure 2.5. LVDTs (LV) and String Pots (SP): (a) Location along the Height and
(b) Section 1-1
21
Figure 2.6. Locations of Strain Gages: (a) on Longitudinal Bars (b) on Transverse
Hoops (c) and (d) Horizontally within the Concrete and (e) Vertically within the
Concrete
22
Table 2.3. Number of Different Types of Sensors for Each of the Specimens
LVDT String Pots Strain Gages

Spec.
Long. Horiz. (in Vert. (in Tiltmeter Total
Vert. Horiz. Diag. Horiz. Hoops
Bars concrete) concrete)
SP1 20 8 4 2 52 36 34 7 - 163
SP2 20 8 4 2 53 36 34 7 1 165
SP3 20 8 4 2 48 36 42 7 1 168
SP4 20 8 4 2 52 42 34 11 1 174
SP5 20 8 4 2 68 36 34 9 1 182
SP6 20 8 4 2 68 36 32 10 1 181
SP7 38 10 8 4 100 32 36 12 - 240
SP8 20 8 4 2 64 34 38 11 1 182
Figure 2.7. Location of the Telepresence Towers and High-Resolution Cameras at the
MAST Lab
23
during the tests. Additionally, three high-resolution cameras, including two Nikon D600
24.3-megapixel digital SLRs and a Nikon D3200 24.2-megapixel digital SLR captured
high-resolution still images continuously during each test at a rate of one image per 10 s.
Moreover, time-lapse images were captured by a 6-megapixel Canon Powershot Pro Series
S3 every 30 s during the test and every 5 min during specimen preparation.
Table 2.4 summarizes the amount of visual data (still images and videos) recorded
during each test. Contact sensors (mini-accelerometers) and air-coupled ultrasonic
transmitters were used on specimens SP6, SP7, and SP8 for nondestructive evaluation
(NDE) based on ultrasonic testing. A nondestructive testing (NDT) technique was
conducted on SP6 using ultrasonic shear-wave tomography, known as MIRA, to detect
internal defects in the cross section. Discussion of the applied ultrasonic methodology and
the results obtained are discussed elsewhere.
Table 2.4. Amount of Image and Video Data Recorded During Each Test (GB)
Still and Extra High

Total Total image and
Sp. timelapse Resolution Videos
images video
images images
SP1 7.7 75.4 83.1 33.97 117.0
SP2 6.7 286.1 292.8 24.1 316.9

SP3 8.1 112 120.1 16.9 137.0
SP4 7.8 128.7 136.5 17.3 153.8
SP5 6.2 106.2 112.4 15.8 128.2
SP6 4.4 41.3 45.7 10.0 55.7
SP7 10.2 200.9 211.1 41.5 252.6
SP8 7.7 153.8 161.5 30.7 192.2
24
2.7. LOADING PROTOCOLS
The RC specimens were subjected to six different loading protocols to investigate the
effect of various seismic loading events. At the beginning of each test, an axial load was
applied and its magnitude and vertical orientation were kept constant during the tests using
the eight actuators that drive the MAST crosshead. The movement of each actuator is
computed internally by the MTS controller to ensure the commands defined for the global
DOFs at the crosshead. Therefore, there is no need for the user to modify the movement to
account for geometric nonlinearity. The magnitude of the axial load for the specimens
constructed with normal-weight concrete was selected according to the estimated structural
demands of the PF and SF columns in the prototype structure and with consideration
of their cross-sectional dimensions and concrete strengths such that the axial load ratio
(P⁄f′cAg) was 0.15 and 0.3 for PF and SF specimens (except the UHP-FRC specimen),
respectively, where Ag is the gross cross-sectional area of concrete for each specimen.
While the axial load was kept constant, the specimens were subjected to lateral
displacement cycles, as shown on Figure 2.8 and described in Table 2.5. A single-cycle
loading protocol (i.e., monotonic cycle) was applied to the first specimen (SP1), whereas
all other specimens were subjected to progressively increasing displacement cycles until
they exhibited severe damage and strength loss. The loading stopped when one of the
following conditions was satisfied: (1) the specimen showed significant strength loss such
that the residual strengths were less than or equal to 20% of the peak strengths in both
25
12 12
8 8
Drift (%) 4 4
Drift (%)
0 0
0 1 2 3 0 15 30 45 60
-4 -4
-8 -8
-12 -12
Cycle Cycle
(a) (b)
12 12
8 8
4 4
Drift (%)
0 Drift (%) 0
0 15 30 45 60 0 15 30 45 60
-4 -4
-8 -8
-12 -12
Cycle Cycle
(c) (d)
6
12
4.5
8
3
4
DriftY' (%)
Drift (%)
1.5
0 0
0 2 4 6 8 10 -6 -4.5 -3 -1.5 0 1.5 3 4.5 6
-4 -1.5
-8 -3
-12 -4.5
Cycle -6
DriftX' (%)
(e) (f)
Figure 2.8. Loading Protocols for Specimens: (a) SP1; (b) SP2, SP5, and SP8; (c) SP3;
(d) SP4; (e) SP6; (f) SP7
26
Table 2.5. Applied Loading Protocols
Loading
bh Protocol Description of the
Specimen TYP. P/ fc Ag
(in.) in Figure Loading Protocol
2.8
SP1 PF 36×28 0.154 a Monotonic
SP2 PF 36×28 0.139 b Symmetric Cyclic (ACI 374)

Symmetric Cyclic (ACI 374)
SP3 PF 36×28 0.141 c
Plus Monotonic Push
Symmetric Cyclic (ACI 374)
SP4 PF 36×28 0.140 d
Plus Monotonic Push
SP5 SF 28×28 0.285 b Symmetric Cyclic (ACI 374)
SP6 PF 36×28 0.163 e Near-Collapse
SP7 PF 36×28 0.154 f Biaxial
SP8 SF 28×28 0.065 b Symmetric Cyclic (ACI 374)
directions exhibited during the test; and (2) the hydraulic actuators reached their maximum
stroke or rotational capacity.
Except for SP1 and SP6, all other loading protocols were designed in accordance with
ACI 374-05 (ACI Committee 374, 2005) guidelines. These protocols comprised three
full-reversed displacement cycles at each drift level followed by a small cycle at 1/3 of the
preceding drift level. This loading pattern was applied to SP2, SP5, and SP8 with
progressively increasing drift cycles until one of the stopping criteria was satisfied; SP3
and SP4 experienced a monotonic push after a certain number of cycles. Displacement
cycles were applied along the primary axis of the specimens (Y′ axis in Figure 2.2(b)),
27
except for SP7, which was subjected to a biaxial loading protocol as shown on Figure
2.8(f).
SP6 was subjected to a near-collapse loading protocol that was developed using results
from time-history analysis of a series of low- to high-rise buildings subjected to a set of
far-field earthquake records. The near-collapse loading protocol consisted of symmetric
cycles followed by large unsymmetric displacement cycles under which typical columns
in the time history analysis showed significant yielding and strength loss.
2.8. TEST OBSERVATIONS
Figure 2.9 shows specimen damage initiated after application of early cycles when
flexural cracks were observed on the SE and NW faces close to the column base. Also,
shear-flexure cracks were observed on the SW and NE faces and distributed through the
lower portion of the specimens. Cracking and spalling of the cover concrete was followed
by yielding of the longitudinal bars. However, the specimens reached their peak strengths
at larger displacement levels because of the resistance mobilized by the confining pressure
of the transverse hoops as well as strain hardening of the longitudinal bars. Cracking and
crushing of the concrete along the column perimeters and, more importantly, buckling and
fracture of the longitudinal bars resulted in the loss of flexural resistance. In some
specimens, spalling of the cover concrete and lateral pressure of the longitudinal bars
caused some hoops to open. Buckled bars fractured at following large drift ratios, which
resulted in sudden decreases in lateral load capacity. At the end of the tests, severe damage
28
to the perimeter concrete and fracture of the longitudinal bars in the extreme faces were
observed.
Figure 2.10 shows applied lateral loads for SP1, SP2, and SP3 plotted against their
lateral drifts (measured at the idealized inflection point) and lateral displacements. The
specimens subjected to these loading protocols [(a), (b), and (c)] were selected as
representative of all loading protocols because they included a single cycle, a series of
gradually increasing cycle groups, and a series of gradually increasing cycle groups
followed by a collapse-level drift cycle. The loading protocols addressed one of the primary
goals of this project that was testing of RC columns under large drift ratios and capturing
their post-peak performance at the near-collapse state. In the first specimen (SP1), the
loading was terminated at an approximate 12% drift ratio when the actuators reached their
rotational capacity. However, the progression of damage and strength loss in SP2 and SP3
occurred at a greater rate such that the capacity of these specimens deteriorated to 20% or
less of their peak strength at smaller drift ratios than that of SP1.
Strains in longitudinal bars, transverse hoops, and the concrete core were recorded
during each test using embedded strain gages that are shown in Figure 2.6. On longitudinal
bars, the strain gages were installed at closely spaced points along the lower portion of the
column and in the footing block, by which the onset of yielding, strain hardening, and
fracture of longitudinal bars could be identified during application of the loading protocols.
As shown in Figure 2.11(a), at a 2.15% drift ratio for specimen SP2, yielding occurred
over a large portion of one of the longitudinal bars. Nominal yielding in Figure 2.11 can
29
(a) (b)
(c) (d)
Figure 2.9. Progression of Damage on SP2: (a) SE Face at 2.2% Drift Ratio; (b) SW
Face at 2.2% Drift Ratio; (c) SE Face at the End of the Test and (d) SW Face at the
End of the Test
30
Lateral Displacement (in.)
-16 -12 -8 -4 0 4 8 12 16
300
SP1
200
Lateral Force (kip)

100
0
-100
-200
-300
-12 -8 -4 0 4 8 12
Drift (%)

-12 -8 -4 0 4 8 12
300
SP2
200
Lateral Force (kip)
100
0
-100
-200
-300
-8 -4 0 4 8
Drift (%)

-16 -12 -8 -4 0 4 8 12 16
300
SP3
200
Lateral Force (kip)
100
0
-100
-200
-300
-12 -8 -4 0 4 8 12
Drift (%)
Figure 2.10. Applied Load vs. Crosshead Displacement and Drift Ratio
31
be recognized as a strain equal to 0.0025 in./in., as indicated by the dashed line. Also, at
the same drift level transverse hoops started to yield at a location approximately 12 in.
above the column base, as shown in Figure 2.11(b). On transverse hoops, strain gages were
installed along the column height and at different locations in the cross section (Figure 2.6).
Besides longitudinal bars and transverse hoops, strains at different locations in the
concrete core were measured using vertical and horizontal strain gages. Figures 2.12(a) and
2.12(b) show strain profiles measured along the height of SP2 and at a 2.15% drift ratio. In
these figures, CL1G corresponds to the vertical strain gage located at the midpoint of the
specimen’s SE face; CTRH4G measures the horizontal strains (as shown on Figure 2.6).
Horizontal strains measured by CRTH4G fell beyond the cracking strains for the concrete,
which was confirmed by the formation of many horizontal flexural cracks as observed in
Figure 2.9.
32
60
40
Height (in.)
20
-20
0 5 10 15 20
Strain ( 10-3 in./in.)
(a)
30
20
Height (in.)
10
-10
-20
0 0.5 1 1.5 2 2.5 3
(b)
Figure 2.11. Reinforcement Strain Profiles along Column Height at 2.15% Drift Ratio
for Specimen SP2: (a) Longitudinal Bar (SL1G); (b) Transverse Hoops (STR3G)
33
15
10
Height (in.)
5
-5
-0.6 -0.4 -0.2 0 0.2
(a)
15
10
Height (in.)
-5
0 50 100 150 200
(b)
Figure 2.12. Concrete Core Strain Profiles along Column Height at 2.15% Drift Ratio
for Specimen SP2: (a) Vertical Strain Gages (CL1G); (b) Horizontal Strain Gages
(CTRH4G)
34
Along with strain gages, vertical LVDTs were installed on the specimens’ SE and NW
faces (Figure 2.5) to estimate flexural curvature along the column height. Such estimation
was carried out for the lower portion of specimen SP2, as shown in Figure 2.13, which
illustrates base moment for specimen SP2 against curvature measured 6 in. above the
column base. Curvature at this point was calculated using measurements for the vertical
LVDTs on the column’s SE and NW faces. Also, the base moment included the first-order
moment caused by application of the lateral load over the column height as well as the
second-order moment resulting from the P-delta effect. The LVDT cores were connected
to aluminum plates located on the footing block. These plates were anchored at a distance
from the column so that they would not get loose after cracking of the concrete at the
column base.
3000
2000
1000
MBase (kip.ft)
-1000
-2000
-3000
-6000 -4000 -2000 0 2000 4000 6000
-6
 (10 /in.)
Figure 2.13. Moment vs. Curvature Relation for SP2 Measured 6 in. from the Column
Base.
35
2.9. FUTURE USE OF THE DATA SET
Recorded data from eight full-scale column tests are available at NEEShub (NEES
2009) for researchers and structural engineers to investigate the observed behavior of the
column specimens during the tests. These data include measurements from the MAST lab
6-DOF control system and the instrumentation (LVDTs, string pots, tiltmeters, and strain
gages). Additionally, the high- and extra-high-resolution still images and high-resolution
videos provide a visual record of damage for each face of the column specimens during
testing. The data set will enhance understanding of the behavior of RC and UHP-FRC
columns subjected to near-collapse loading protocols and can be used to calibrate finite-
element models intended to simulate their behavior. Also, it can be utilized to improve
collapse-predictive component models for performance-based design, to evaluate ACI
code methodology for estimation of shear and moment capacity, to quantify the evolution
of energy dissipation and secant (or tangent) stiffness during cyclic loading, and to assess
the effects of loading protocols on damage state.
First, the test data from this experimental program can enhance knowledge of the
behavior of RC and UHP-FRC columns subjected to large seismic events, and they can be
used to assess and improve finite-element modeling. Previous experiments on RC columns
were performed either on small- or medium-scale specimens or under small drift ratios.
Accordingly, finite-element models were typically calibrated with test results that did not
capture the complete range of column post-peak behavior. The loading protocols in this
experimental program were extended to large drift ratios (∼12%) until the specimens
36
exhibited significant strength loss (i.e., 80% or more of their maximum strength in each
loading direction) and stiffness deterioration. The data from this research program will
enable finite element model calibration and validation against a wide range of structural
performance. Developed models can then be utilized to assess the seismic performance of
structures under various axial load ratios, cross-sectional dimensions, and loading
protocols. Also, the tests performed on columns constructed with UHP-FRC material will
enhance knowledge of the seismic performance and damage tolerance capabilities of
columns constructed with this innovative material, and they will provide a tool to improve
structural models that capture their behavior.
An important application of the recorded test data is the assessment and improvement
of existing collapse-predictive component models in performance-based seismic design. A
set of comprehensive test data is required for the development of representative nonlinear
models and implementation of a reliable analysis for performance-based design. Improved
component models represent an important step in predicting the collapse safety of existing
buildings that are designed in accordance with modern codes as well as in designing
reliable structures. They have been recognized as a requirement for the advancement of
performance-based seismic engineering procedures (Applied Technology Council (ATC),
2009b). The tests conducted in this research were designed considering the critical need
for collapse prediction in performance-based design procedures.
The proposed methodology in the ACI code for estimation of column shear and
moment capacity can be evaluated using flexural moments, axial loads, and internal strains
37
measured during each test. Nominal shear and moment capacities calculated following the
ACI code approach can be compared to the specimen’s actual strength exhibited during
each test to assess the validity of the current design methodology for RC columns,
including the capacity design procedures for evaluating seismic demands in RC moment
frames. Also, the effects of the applied axial load level and the lateral loading protocol on
shear and moment capacity can be estimated.
Finally, recorded test data during each test provide a measure to quantify the evolution
of damage as represented by energy dissipation (or absorption) and the change in secant
(or tangent) stiffness under cyclic loading protocols. Energy dissipation and stiffness
change are both affected by characteristics of the loading protocol (i.e., target
displacement, axial load level, and number of cycles). These characteristics can affect the
extent of specimen damage produced. A larger target displacement, a higher axial load
level, and a greater number of deformation cycles will typically result in more damage to
the concrete as well as yielding, buckling, or ultimately fracturing of the reinforcing bars.
The progression of damage in a RC column, on the other hand, is accompanied by energy
dissipation and a reduction in secant (or tangent) stiffness. Therefore, recorded test data
can be used to calculate energy dissipation and secant (tangent) stiffness at different test
stages to estimate the extent of damage and identify the parameters that accelerate it.
2.10. CONCLUSIONS
A set of eight full-scale RC and UHP-FRC column tests was conducted at the MAST
laboratory at the University of Minnesota to address the shortcomings in available test data
38
for rectangular columns in high-rise buildings. The tests aimed to enhance understanding
of the behavior of RC columns representative of those typically used in the ground floor of
tall moment-resisting frame buildings at a near-collapse state, and to utilize this enhanced
understanding to develop improved collapse-predictive component models for
performance-based design. The data set can be accessed through a DOI and includes details
of specimens, material properties, sensor and instrumentation layouts and their recorded
data, and high- and extra-high-resolution images and high-resolution videos recorded
during each test. It provides an opportunity for researchers and engineers in the field to
enhance their knowledge of the seismic performance of RC columns under severe loading
conditions. The recorded test data can be utilized to assess and improve finite element
modeling tools as well as existing collapse simulation models to more accurately capture
post-peak column behavior. Additionally, the data set can be employed to evaluate the
current ACI code for estimating nominal RC column moment and shear capacity and for
assessing the evolution of energy dissipation and stiffness deterioration under various
loading protocols.
39
CHAPTER 3
INFLUENCE OF CROSS‐SECTIONAL SIZE ON SEISMIC

PERFORMANCE OF RC COLUMNS2
3.1. SUMMARY
Current ACI code provisions ACI 318-14 (ACI Committee 318, 2014) for the design
of reinforced concrete (RC) columns are based on previous experiments on column
specimens that featured smaller cross-sectional dimensions compared to those currently
used in mid- or high-rise buildings. A study on a database of RC columns including those
recently tested at the MAST Laboratory of the University of Minnesota as part of this
research, and ACI 369 Rectangular Column Database (Ghannoum et al., 2012) suggests
that flexural moment capacity, drift capacity, and displacement ductility of columns are
not generally affected by column dimension (i.e. cross-sectional depth). However, the bar
buckling can start earlier in larger columns. Additionally, observations from the MAST
tests reveals a bar buckling mode that is seldom observed in tests on smaller columns, in
2
Under review by ACI Structural Journal.
40
which the bars buckle parallel to the face of the column. A study that is presented herein
suggests that larger columns are more prone to this in-plane mode of bar buckling.
Keywords: RC column, cross-sectional size, bar buckling
3.2. INTRODUCTION
Many experiments have been carried out on rectangular RC columns. Properties of
more than 300 of such tests are collected in the ACI 369 Rectangular Column Database
(2012) as well as in the PEER structural performance database (Berry et al., 2004). The
primary goal of most of these tests was to investigate the behavior of RC columns subjected
to gravity and seismic loadings and to develop and improve design code provisions for
building columns. However, many of the tests were conducted on columns with smaller
dimensions than those currently used in mid- or high-rise buildings mainly due to
limitations of the testing facilities. The average cross-sectional depth of the rectangular RC
columns in the more recent database (i.e., ACI 369 Rectangular Column Database (2012))
is 12.5 in. (318 mm) with the lower 5% and upper 95% bounds of 6.3 in. (160 mm) and
21.65 in. (550 mm), respectively. The smallest and largest columns in these databases that
failed either in flexure or flexure-shear mode featured a cross-sectional depth of 3.15 in.
(80 mm) (Zhou, Higashi, Jiang, & Shimizu, 1985) and 27.56 in. (700 mm) (Yarandi, 2007).
While these tests provide a valuable dataset for RC columns, a careful study is required to
assess the validity and applicability of their results to estimate seismic performance of
actual columns with larger dimensions.
41
A clear understanding of the seismic performance of RC columns is crucial in
designing earthquake resistant structures. In performance-based design, columns are
designed to sustain a certain level of damage under the effect of applied loads, but to resist
collapse. The state of damage in RC columns is usually linked to moment capacity, drift
capacity, drift ratio at yield, and the drift at the onset of buckling.
RC columns are typically designed with sufficient flexural capacity to resist external
loads. The flexural moment capacity of RC columns can be affected by many parameters
including axial load magnitude, concrete compressive strength, and amount of transverse
reinforcement. A well-designed RC column should be able to resist flexural moments due
to lateral loading and associated P-Delta effects, and experience large deformations before
failure. Therefore, the maximum flexural moment of a RC column during testing can be
used to evaluate its seismic performance as well as the accuracy of the ACI approach to
estimate flexural moment capacity.
Additionally, accurate estimation of the effective stiffness of RC columns is important
in design because it affects load carrying capacity, deformation demands, and dynamic
response characteristics. Most existing procedures specify the effective stiffness as a
fraction of the gross stiffness depending on the axial load ratio (ACI Committee 318, 2014;
American Society of Civil Engineers (ASCE), 2000, 2007; Paulay & Priestley, 1992).
Further research (Elwood & Eberhard, 2009) showed that the effective stiffness of concrete
columns is also strongly proportional to aspect ratio (the ratio of shear span to column
depth in the direction of loading, L/h).
42
Moreover, drift capacity and displacement ductility are important parameters in
performance-based design as they indicate the expected lateral deformation when reaching
a certain level of strength loss, typically 20% of the peak lateral load capacity. The effect
of column aspect ratios on the drift capacity depends on the level of axial load ratio.
Increasing aspect ratio (L/h) will result in an increase in the drift capacity provided that the
axial load ratio is small. In columns with large axial load ratios, on the other hand,
increasing L/h can pronounce the P-Delta effect and result in a lower drift capacity (Bae &
Bayrak, 2008).
Last but not least, seismic performance of RC columns can be significantly affected
by reinforcing bar buckling as it is often followed by significant strength degradation and
stiffness deterioration that lead to extensive and costly damage. Reinforcing bar buckling
in RC columns can initiate right after spalling of the cover concrete or yielding of
transverse hoops (Bae & Bayrak, 2008) and well before global collapse of the specimens.
Early bar buckling can drastically affect ductility of RC columns and limit their drift
capacity. Therefore, understanding the drift ratio at which the bar buckling initiates, is
significantly important for practicing engineers. To achieve this understanding, analytical
and experimental studies have been carried out to investigate the buckling of longitudinal
bars. Early investigations employed Euler buckling theory to explain the observed buckling
behavior of bars (Bresler & Gilbert, 1961; Scribner, 1986; Papia, Russo, & Zingone, 1988).
Other researchers conducted analytical (Gomes & Appleton, 1997; Dhakal & Maekawa,
2002) and experimental studies on isolated bars (Mander, Priestley, & Park, 1984; Monti
& Nuti, 1992; Rodriguez, Botero, & Villa, 1999; Bayrak & Sheikh, 2001) to investigate
43
the constitutive relationship of steel reinforcing bars including the effect of buckling, and
to study the parameters that affect bar buckling. However, experimental investigations on
RC columns with specific focus on bar buckling are limited (Suda, Murayama, Ichinomiya,
& Shimbo, 1996; Moyer & Kowalsky, 2003; Brown, Lehman, & Stanton, 2008). These
experimental investigations have sought to define drift or displacement demands
corresponding to the onset of bar buckling as a critical damage state in seismic
performance-based design.
In this study, the selected database of RC columns is used to study the influence of
cross-sectional size on seismic performance characteristics of columns, namely the
moment capacity, drift capacity, ductility, and drift at the onset of buckling. For each
seismic characteristic, the calculation scheme used with the information available in the
database, as well as the key controlling parameters are discussed and employed to
categorize the columns. The influence of cross-sectional size is then studied for columns
in each category.
3.3. RESEARCH SIGNIFICANCE
RC columns can incorporate cross-sectional dimensions significantly larger than those
tested previously and which were utilized to verify ACI code design provisions. A series
of tests on rectangular RC columns with larger cross-sectional dimensions was carried out
as part of this research to fill the gap in the existing database of RC columns. Results from
these tests were employed along with an available database of RC columns (Ghannoum et
al., 2012) to investigate the effects of cross-sectional size on seismic performance of RC

44
columns including moment capacity, effective stiffness, drift capacity, displacement
ductility, and bar buckling.
3.4. EXPERIMENTAL INVESTIGATION AND DATA SELECTION
Seven full-scale RC column tests, representative of modern mid- and high-rise RC
moment frame buildings, were conducted at the Multi-Axial Subassemblage Testing
(MAST) Laboratory of the University of Minnesota. These columns featured
cross-sectional dimensions (36×28 in. and 28×28 in.) (914×711 mm. and 711×711 mm.)
larger than most of the previous tests on RC columns subjected to cyclic loading protocols
(Berry et al., 2004; Ghannoum et al., 2012). Each specimen was constructed horizontally
along with a 84×84×30 in. (2134×2134×762 mm.) footing block and a 75×75×23 in.
(1905×1905×584 mm.) loading block (Figure 3.1) and was placed on top of a spacer block
under the MAST loading crosshead (Figure 3.2). Each test was started with application a
constant axial load along Z′ axis followed by a lateral displacement loading protocol along
X′ axis. Additionally, to focus this study on the effects of loading protocols and to eliminate
changes in the response of columns due to dislocation of the inflection point a flexural
moment was applied on top of the specimen to maintain the inflection point at 96 in. (2438
mm.) above the column base corresponding to one half of the height of the columns in the
ground floor of the prototype building. However, during an actual earthquake, the
inflection point typically shifts upward along the length of RC columns and hence a lower
lateral load as compared to the case with fixed inflection point would be required to
generate the plastic moment at their base. Therefore, RC columns in practice may exhibit
45
lower drift capacity and drift ratio at the onset of buckling as those that were recorded
during tests at the MAST Lab. Information regarding cross-sectional dimensions as well
as concrete compressive strength (f′c), axial load ratio (P/f′cAg), and aspect ratio (L/h) of
these columns is presented in Table 3.1. The table also shows distinct lateral loading
protocols applied to the specimens during each test.
Figure 3.1. Overview of the Test Specimens
46
Figure 3.2. Overview of the Test Setup at the MAST Lab and Loading Direction
47
Table 3.1. Characteristics of the MAST Tests
Specimen
b×h
f'c, ksi P/ Mtes 80 Keff
Loading Protocol in. ×in. L/h μΔ
(MPa) f'cAg t/Mn (%) /Kg
(mm×
mm)
36×28 4.86
SP1 Monotonic (914×711) (33.51)
2.67 0.15 1.30 10.74 0.19 10.15
Symmetric Cyclic 36×28 5.40

SP2 (ACI 374-05 (2005)) 2.67 0.14 1.26 7.13 0.26 5.31
(914×711) (37.23)
SP3 Plus Monotonic Push 2.67 0.14 1.22 7.83 0.19 7.35
(914×711) (36.54)
SP4 Plus Monotonic Push 2.67 0.14 1.35 10.19 0.21 8.78
(914×711) (37.02)
SP5 (ACI 374-05 (2005) ) 3.43 0.28 1.33 4.09 0.38 4.53
(711×711) (36.34)
36×28 4.61
SP6 Near-Collapse (914×711) (31.78)
2.67 0.16 1.28 12.47 0.17 9.60
36×28 4.90
SP7 Biaxial (914×711) (33.78)
2.67 0.15 1.19 5.09 0.14 4.79
A single cycle loading protocol was applied to the first specimen (SP1) to capture the
monotonic force-displacement envelope of the test specimens in a single cycle. The second
specimen (SP2) was subjected to progressively increasing displacement reversals which
were defined according to the guidelines in the ACI 374-05 (ACI Committee 374, 2005).
The applied lateral loading protocols were continued until the specimens lost most of their
load carrying capacity (more than 80% of peak lateral load capacity). The lateral loading
on SP3 and SP4 started with displacement cycles like those used in SP2. However, these
two specimens were subjected to a final monotonic displacement segment after a specified
number of cycles. The applied loading protocol and hysteretic force-displacement (or drift
48
ratio) of specimen SP4 is represented in Figure 3.3. The last two specimens (i.e., SP6 and
SP7) were subjected to a near-collapse cyclic loading protocol which was designed as part
of this research, and a biaxial cyclic loading protocol, respectively. Further details
regarding the applied loading protocols can be found elsewhere (Nojavan et al., 2014,
2015) and are not shown here for brevity. Results from these tests were employed along
with ACI 369 Rectangular column database (Ghannoum et al., 2012) to investigate the
influence of cross-sectional depth on the seismic performance of columns. The column
dataset was screened to include only columns with concrete compressive strength between
2.5 ksi (17.24 MPa) and 10 ksi (68.94 MPa) in order to focus the study on structural
columns constructed with normal and moderately high strength concrete. Additionally, it
is assumed that columns in a typical building design based on standard code practice will
not suffer shear failure and reach flexural capacity when subjected to extreme loadings.
Thus, column specimens failing in shear were excluded from the dataset. A complete list
of the test specimens utilized for the purpose of this study is not shown here for brevity.
Instead, the range and statistics of important parameters (i.e. minimum, maximum, mean,
and coefficient of variation (CV) of concrete compressive stress (f′c), reinforcing yield
stress (fy), cross-sectional depth (h), and reinforcing bar diameter (db)) for the selected
specimens are presented in Table 3.2.
49
12
Drift (%)
0
-4
-8
-12
0 15 30 45
Cycle
(a)
(b)
Figure 3.3. Behavior of Specimen SP4 under the Applied Loading Protocol: (a)
Loading Protocol, (b) Variation of Lateral Force with Drift Ratio and Top
Displacement
50
Table 3.2. Properties of Rectangular Columns in the ACI 369 Database
Parameter Min. Max. Mean COV
2.55 9.83 4.58

f′c, ksi (MPa) 0.276
(17.6) (67.8) (31.6)
46.8 85.1 62.1
fy, ksi (MPa) 0.154
(322.9) (586.9) (428.2)
3.15 27.6 13.6
h, in. (mm) 0.351
(80) (700) (345)
0.24 1.25 0.71
db, in. (mm) 0.281
(6) (32) (18)
3.5. FLEXURAL MOMENT CAPACITY
3.5.1. Calculation
The normalized flexural moment capacity of columns is defined here as the ratio of
the maximum measured flexural moment during the test (Mtest), including the P-Delta
contribution (i.e. moment due to axial load and column lateral deflection), to its nominal
moment capacity (Mn) calculated according to the provisions in ACI 318-14 (2014). To
compute the nominal flexural moment capacity (Mn) for each column in the database, the
nominal axial load-flexural moment (P-M) interaction diagram was computed using strain
compatibility, and assuming an ultimate concrete compressive strain of 0.003 and an
elastic-perfectly plastic reinforcing steel material model. Measured concrete and steel
material properties were used. The location of the neutral axis was computed based on a
linear strain distribution and was used to determine the depth of the compression region.
For each level of strain, the compressive force provided by concrete was calculated based
51
on an equivalent rectangular concrete stress block, following provisions in the ACI 318-14
(2014), and neglecting the confinement effect. The compressive force resultant in the
concrete was considered along with the forces provided by reinforcing steel bars to
compute the net axial force and the moment about the geometric centroid of the section.
The moment associated with the applied axial load on the P-M interaction diagram was
then taken as the flexural moment capacity of the column specimen (Mn).
3.5.2. Influence of Column Section Depth h
The normalized moment capacity (Mtest / Mn) for all the columns in the database is
plotted against their cross-sectional depths (Figure 3.4). It is noticed that columns failing
in flexure exhibit a normalized moment capacity of at least one while this ratio is generally
smaller for the columns that failed in shear (not shown for brevity). In other words, all of
the column specimens developed their nominal moment capacity, computed according to
ACI 318-14 (2014) regardless of their cross-sectional depth, unless they failed in shear.
This observation confirms the ACI 318-14 (2014) flexural design provisions for RC
columns to be appropriately conservative. The data in Figure 3.4 also indicate that the
normalized moment capacity is not generally dependent on the cross-sectional depth of
columns (i.e. the linear least squares trend line including the tests conducted at the MAST
lab is nearly flat) for flexure dominated columns and the MAST Lab tests do not alter this
trend.
52
3
Prev. Tests
2.5 MAST Tests

Trend Line
Mtest/Mn
1.5
1
Mtest=Mn
0.5
0
0 10 20 30 40
h (in)
Figure 3.4. Normalized Moment Capacity vs. Cross-Sectional Depth for Columns
Failing in Flexure
3.6. EFFECTIVE STIFFNESS
3.6.1. Calculation
A simple approach was employed to compute the effective stiffness (Keff) of the
columns (Figure 3.5). A secant passing through a point on the Lateral Force-Displacement
(F-Δ) cyclic envelope corresponding to 70% of the maximum applied lateral load on the
specimen (0.7Fmax) is extended to intersect with the horizontal line at Fmax as shown in
Figure 3.5. The yield displacement is defined as intercept of a vertical line from the
intersection point on the horizontal axis. This approach is similar to that used by Elwood
and Eberhard (2009) to estimate EI eff calc except that they used the lateral load
53
F
Keff
Fmax
0.7Fmax
F-∆ Cyclic
Envelope
∆y ∆
Figure 3.5. Definition of the Effective Stiffness
corresponding to a specified concrete compressive strain and yielding of reinforcing steel
instead of Fmax and 0.7Fmax. Reported values for Fmax and drift ratio at yielding (δy) by
Ghannoum et al. (2012) and Sivaramakrishnan (2010) that were calculated according to
the aforementioned approach, were utilized to compute the effective stiffness of the
columns in the dataset. Considering that concrete cracking typically happens before
yielding of the longitudinal bars in RC columns, it would appear to be more realistic for
the effective stiffness to be normalized by the stiffness computed based on cracked and
transformed section properties. Mehanny et al. (2001) applied this idea to normalize the
effective stiffness, but their proposed approach was shown to overestimate the normalized
stiffness of both rectangular and circular columns (Elwood & Eberhard, 2009).
Consequently, the gross section stiffness (Kg) was used to normalize the calculated
effective stiffness, as was done by Elwood and Eberhard (2009).
54
3.6.2. Influence of Axial Loading and Aspect Ratio
The normalized effective stiffness of RC columns (Keff / Kg) which is equal to EIeff /
EIg can vary significantly depending on the applied axial load level and aspect ratio
(Elwood & Eberhard, 2009). To investigate the effect of cross-sectional depth on effective
stiffness, the columns in the database were subdivided into distinct categories depending
on their axial load ratio (P/f′cAg) and aspect ratio (L/h) in order to separate the effects of
these two parameters and identify the influence of column depth more easily. Four
categories of axial load ratio (P/f′cAg<0.2, 0.2≤P/f′cAg<0.3, 0.3≤P/f′cAg<0.4, and
0.4≤P/f′cAg) and four categories of aspect ratio (1≤L/h<2, 2≤L/h<3, 3≤L/h<4, 4≤L/h) were
considered.
For the columns in the database, the normalized stiffness ratio (Keff /Kg) was observed
to be strongly dependent on the axial load and aspect ratios as illustrated in Figure 3.6. The
trend line for each category was obtained using a linear least squares fit. The trend line in
Figure 3.6(a) is calculated using all of the data shown, which includes the corresponding
previous test data (i.e. test specimens in the ACI 369 database (Ghannoum et al., 2012) as
well as most of the tests recently conducted at the MAST lab. The trend lines for the
individual aspect ratio categories (Figure 3.6(b)) indicate that for columns with aspect
ratios greater than 2, normalized effective stiffness generally increases with axial load ratio,
and that the rate of increase further increases with increasing aspect ratio (L/h).
55
0.8
0.6
Keff/ Kg
0.4
Prev. Tests
0.2
MAST Tests
Trend Line
0
0 0.2 0.4 0.6 0.8 1
P / f'cAg
(a) 2 ≤ L/h < 3
1.2
1<L/h<2
2<L/h<3
1
3<L/h<4
L/h>4
0.8
Keff/ Kg
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
P / f'cAg
(b) Various L/h
Figure 3.6. Normalized Effective Stiffness vs. Axial Load Ratio
56
The normalized effective stiffness is shown relative to column cross-sectional
depth in Figure 3.7, and it suggests that normalized effective stiffness is not generally
affected by column size. Given the large spread of data in Figure 3.7, plots were created
for each column category depending on axial load and aspect ratios. The plots for two
categories, that contained MAST test data, are illustrated in Figure 3.8. It is observed that
including the MAST test results does not significantly change the observed trend. Among
the MAST test specimens, SP2 which was subjected to a loading protocol that resembles
that used for the rest of the columns in the database, exhibits a normalized stiffness that is
pretty close to the trend line. These plots confirm the lack of dependence between
normalized effective stiffness and the cross-sectional depth. Slight increasing or decreasing
1.5
Prev. Tests
1.25 MAST Tests

Trend Line
1
Keff / Kg
0.75
0.5
0.25
0
0 10 20 30 40
h (in)
Figure 3.7. Normalized Effective Stiffness vs. Cross-Sectional Depth
57
0.8
Prev. Tests
MAST Tests
Trend Line
0.6 SP2
Keff / Kg
0.4
0.2
0
0 10 20 30 40
h (in)
(a) 0 ≤ P/f'cAg < 0.2; 2 ≤ L/h < 3
0.8
Prev. Tests
MAST Test
0.6 Trend Line

Keff / Kg
0.4
0.2
0
0 10 20 30 40
h (in)
(b) 0.2 ≤ P/f'cAg < 0.3; 3 ≤ L/h < 4
Figure 3.8. Normalized Effective Stiffness vs. Cross-Sectional Depth
58
trends in some categories (i.e. the one shown in Figure 3.8(b) and others that are not shown
for brevity) is believed to be related to the effects of other parameters which are not held
constant, such as the reinforcement yield strength, longitudinal bar diameter, and
reinforcement ratio.
3.7. DRIFT CAPACITY AND DISPLACEMENT DUCTILITY
3.7.1. Calculation
In this study, the drift capacity (δ80) refers to the post-peak drift ratio corresponding
to a 20% loss of the peak lateral loading capacity. Accordingly, the displacement ductility
(μΔ) is defined as the ratio of drift capacity (δ80) to drift at yield (δy) calculated according
to the approach described in the previous section. Reported values for δ80 and δy by
Ghannoum et al. (2012) and Sivaramakrishnan (2010) were utilized to investigate the
influence of cross-sectional size on drift capacity and displacement ductility.
3.7.2. Influence of Axial Loading and Aspect Ratio
Similar axial load ratio and aspect ratio categories to those in the previous section
were employed to investigate the effect of cross-sectional dimension. Drift capacities,
shown in Figure 3.9(a) for 2≤L/h<3 which contains MAST test data, indicates a decreasing
trend with axial load ratio. This trend is also present for more slender columns (L/h >3)
according to the trend lines given in Figure 3.9(b). Yet, for stocky columns (L/h<2), the
trend line in Figure 3.9(b) indicates that axial load ratio does not have a significant effect
on the drift capacity. For slender columns (L/h>2), a larger aspect ratio corresponds to a
59
15
Prev. Tests
MAST Tests
12 Trend Line
SP2& SP7
Drift Capacity (%)

9
0
0 0.2 0.4 0.6 0.8 1
P / f'cAg
(a) 2 ≤ L/h < 3
7
1<L/h<2
6 2<L/h<3
3<L/h<4
Drift Capacity (%)
5 4<L/h<5
0
0 0.2 0.4 0.6 0.8 1
P / f'cAg
(b) Various L/h
Figure 3.9. Variation of Drift Capacity with Axial Load Ratio
60
smaller curvature demand to reach a desired drift ratio. But, it is accompanied by a more
pronounced secondary (P-Delta) moment that results in a reduction in the drift capacity
with increasing the axial load ratio. The drift capacity of the MAST test specimens exhibit
a vertical distribution as shown in Figure 3.9(a) which is believed to be due to the effect of
different loading protocols. In this figure, the column specimens subjected to fully
symmetric loading protocols with many reversals (i.e. SP2, and SP7 that were subjected to
loading protocols designed according to ACI 374-05 (ACI Committee 374, 2005)
guidelines) exhibit the lowest drift capacities and fall much closer to the trend line than do
the other MAST column specimens in the same category (2≤L/h<3).
Calculated values for displacement ductility exhibit more scattered patterns with
variations in axial load and aspect ratio than those for drift ratio, such that a general trend
could not be easily recognized and the plots are not shown here for brevity. This scatter
can be traced to uncertainty in the estimation of the yield displacement from
force-displacement plots of the specimens rather than measured yielding of the reinforcing
bars.
The drift capacity of the columns in the database are shown against their
cross-sectional depths in Figure 3.10. To minimize the effect of aspect ratio and axial load
ratio, the columns were categorized in the manner used in previous sections, however, only
the data for column categories containing MAST data are shown in Figure 3.10(a) and
Figure 3.10(b). The trend line in Figure 3.10(b), suggests that the drift capacity of columns
61
is not generally a function of their cross-sectional size. Also, further analysis of the
columns shown in Figure 3.10(a) reveals that the increasing trend in drift capacity with
column size is dominated by the series of tests conducted at the MAST lab. The difference
in the drift capacity of the MAST column specimens with similar cross-sectional depths
(i.e. 36 in. (914 mm)) is related to their distinct loading protocols, and loading protocol
was observed to have a significant effect on the drift capacity of columns. Similar
observations have been reported (Pujol, 2002). Limiting the MAST data to include only
the specimen with the symmetric cyclic loading protocol (i.e. column SP2 in Table 3.1),
modifies the trend line in Figure 3.10(a) such that it is independent of the cross-sectional
size. Therefore, the MAST tests reveal insensitivity of the drift capacity with variations in
the cross-sectional depth for columns subjected to similar cyclic loading protocols.
Displacement ductility is expected to follow the same relationship with cross-sectional
depth as does drift capacity, because the idealized drift at yield is independent h (not shown
here for brevity). Comparison of the displacement ductility for the specimens relative to
cross-sectional depth, shown in Figure 3.11, exhibits a similar relationship to that shown
in Figure 3.10 for drift capacity. Slight ascending or descending trends in the ductility of
these columns with h is believed to be related to the effect of the loading protocol, or to
slight changes in aspect ratio and axial load ratio within each category.
62
15
Prev. Tests
MAST Tests
12
All Tests
Drift Capacity (%)

9
0
0 10 20 30 40
h (in.)
(a) 0 ≤ P/f'cAg < 0.2; 2 ≤ L/h < 3
8
Prev. Tests
MAST Test
6 Trend Line
Drift Capacity (%)
0
0 10 20 30 40
h (in.)
(b) 0.2 ≤ P/f'cAg < 0.3; 3 ≤ L/h < 4
Figure 3.10. Variation of Drift Capacity with Cross-Sectional Depth
63
12
Prev. Tests
10 MAST Tests
All Tests
Disp. Ductility
8
0
0 10 20 30 40
h (in.)
(a) 0 ≤ P/f'cAg < 0.2; 2 ≤ L/h < 3
Prev. Tests
MAST Tests
6
Trend Line
Disp. Ductility
0
0 10 20 30 40
h (in.)
(b) 0.2 ≤ P/f'cAg < 0.3; 3 ≤ L/h < 4
Figure 3.11. Variation of Displacement Ductility with Cross-Sectional Depth
64
3.8. REINFORCING BAR BUCKLING
Buckling of reinforcing bars in an RC column is a complicated phenomenon that
depends on the interaction of cover concrete, transverse reinforcement, concrete core, and
loading history. Compression from applied forces on a RC column causes the concrete core
to expand laterally due to Poisson’s effect, dilatation and shear deformation, which in turn
mobilizes the confinement reinforcement that restrain the bars from buckling. Furthermore,
the cover concrete, if un-cracked, can also restrain the bars from buckling outward. The
longitudinal bars experience cyclic tension and compression during the evolution of a
seismic event and eventually yield under larger deformations. Upon load reversal and
before the cracks close, the longitudinal bars on one face of the column carry all of the
axial load. Accumulation of plastic deformation in tension elongates the longitudinal bars,
and when the elongation is combined with outward pressure from the concrete core and the
high axial loads from building weight, bar buckling can occur. Once buckled, the bars lose
a significant portion of their strength and stiffness and may eventually fracture due to large
plastic strain reversals at the location where a kink forms in response to buckling.
3.8.1. Characterization of Reinforcing Bar Buckling
Different researchers have employed a variety of approaches to define the onset of
buckling from experimental observations. Some researchers (Bayrak & Sheikh, 1996;
Saatcioglu & Grira, 1999) recognized the onset of buckling based on observations at or
near the end of the test. Brown et al. (2008) assumed an arbitrary lateral displacement of
the bar equal to 1% of the column diameter, equivalent to 0.2 in. (51 mm.) for their test
65
specimens, as the onset of buckling. On the other hand, Bournas and Triantafillou (2010),
considered a sudden increase in the compressive strain of the longitudinal bars, after
yielding, as a sign of bar buckling. Because there is no consensus for a unified method to
measuring lateral displacement of the reinforcing bars at the onset of buckling in
rectangular RC columns in the database, any general conclusion regarding the influence of
the cross-sectional size on the drift ratio at the onset of buckling would lack sufficiently
accurate supportive data. Instead, in this study, some observations from the tests conducted
by the authors at the MAST Lab are presented to provide an insight into the effect of
cross-sectional size on the mechanism of bar buckling.
For MAST tests, strain gages (Figure 3.12) were installed at a 6-in. (152.4-mm)
spacing along the height of four longitudinal bars in each column specimen. While fracture
of the longitudinal bars are typically followed by a marked drop in the load carrying
capacity as shown in Figure 3.13, bar buckling, as a local phenomenon, is barely observable
from strain histories of gages located far from the buckling zone. Therefore, the location
of the buckled sections (i.e. sections with a maximum lateral deformation) along the bar
were determined after each test and the strain gages right above and right below the buckled
sections were identified. While a sudden strain variation could be recognized from these
strain gages, investigation of the other strain gages revealed that regions of the bar away
from the buckling section could still be in their elastic range of constitutive behavior. Thus,
readings from the two strain gages located immediately above and below the buckled
section of the bars were utilized to determine the drift at the onset of buckling.
66
Figure 3.12. Location of Strain Gages for Cross Sections of MAST Test Specimens
Measured strain for a gage in specimen SP2 located at the proximity of a buckled bar
is illustrated in Figure 3.14. The circles and squares denote the largest strains in
compression and tension, respectively, in the preceding cycle after which a reduction in
the measured compressive strain is observed. These points coincide with drift ratios of
1.87% for specimen SP2 in the compression and tension directions of the bar, respectively
(Figure 3.14(b)). The circle corresponds to a cycle where the buckling is believed to start
67
300
200
Lateral Force (Kip)

100
0
-100
-200
-300
-6 -4.5 -3 -1.5 0 1.5 3 4.5 6
Drift(%)
Figure 3.13. Sudden Drops in Load Carrying Capacity due to Bar Fractures
as the compressive strain decreases in the following cycles. This definition of the onset of
buckling is consistent with the one by Bournas and Triantafillou (2010) in terms of a
reduction in the compressive strain in the loading cycles that follow bar buckling. It has
also been stated by other researchers that a reduction of the compressive strength due to
bar buckling is preceded by a large pre-strain (i.e., a large excursion in yield strain) in
tension (Brown et al., 2008; Massone, Polanco, & Herrera, 2014). This point is indicated
by a square in Figure 3.14. The horizontal span with low strain variations, between steps
20,000 and 25,000, in Figure 3.14(a) refers to a pause in testing before application of the
next level of displacement cycles. The next level of applied displacements was
accompanied by a sudden drop in the tensile strain as shown by a triangle, which can be
due to kinking of the bar following buckling and after which the measured strain data are
not reliable.
68
0.020 Onset of Buckling
Tension Excursion
Tension Drop
0.015
Strain (in./in.)
0.010
0.005
0.000
-0.005
10 15 20 25
3
Step (10 )
(a)
0.020 Onset of Buckling

Tension Excursion
Tension Drop
0.015
Strain (in./in.)
0.010
0.005
0.000
-0.005
-4 -3 -2 -1 0 1 2 3 4
Drift(%)
(b)
Figure 3.14. Characterization of the Onset of Bar Buckling
69
A criterion for determination of the approximate strain value to cause buckling can be
taken from the equation proposed by Pantazapoulou (1998) which relates the strain at the
onset of buckling to the bar diameter and its unsupported length, and it results in a strain
of 0.010 and 0.011 for No.8 and No.9 bars, respectively. However, validity of the estimated
strains based on Pantazapoulou’s equation depends significantly on the strain compatibility
condition in the specimens which can be violated in many cases due to cracking of the
specimens during construction, shipping, installation, or small pretest cycles. Therefore,
Pantazapoulou’s equation was not considered as a criterion to define the onset of bar
buckling, rather it was used as an approximate initial estimate of expected bar strains at
buckling.
For each column specimen, the drift ratios at the initiation of buckling for the two
strain gages closest to the buckled region on each reinforcing bar were determined using
the approach described above. The average of the drift ratios at buckling for all four
instrumented bars in the compression face of each column specimen was assumed to
represent the drift ratio at the onset of buckling (δb) as indicated in Table 3.3. Most of the
bars developed a kinked region at buckling which eventually fractured during subsequent
loading cycles due to low-cycle fatigue. Table 3.3 indicates that the drift ratios at bar
fracture (δF) can be significantly larger than the drift ratios at the onset of bar buckling
depending on the loading conditions. Berry and Eberhard (2005) proposed an equation to
estimate the drift ratio at the onset of bar buckling based on a statistical study of a dataset
of columns. However, the data used to calibrate their equation was based on visual
inspection of the specimens at or near the end of the tests to determine buckling of the bars.
70
As a consequence, their equation tends to overestimate the drift ratios at the onset of bar
buckling for column specimens that experienced many displacement cycles before fracture
of the bars (SP2, SP3, SP5, and SP7) as shown in Table 3.3. In fact, for these columns, the
estimated drift ratios according to their proposed equation (δcalc.) are closer to actual drift
ratios at fracture of the bars than the onset of buckling. Additionally, for column specimens
that experienced fewer loading cycles (SP1 and SP6 in Table 3.3), their proposed equation
tends to underestimate the drift ratio at the onset of buckling. The bar splices were located
away from the column-footing interface in these columns. Having mechanical bar splices
at or near the column base may result in a change in the drift ratio at the onset of buckling.
However, these splices are typically not recommended to be placed at the yielding zone
(i.e., near the column base) of RC columns in special moment resisting frames of Seismic
Design Categories C, D, E and F according to seismic provisions in chapter 21 of the ACI
318-11 (2011).
Table 3.3. Drift Ratios at the Onset of Bar Buckling and Fracture
Specimen δb (%) δf (%) δcalc.(%)
SP1 8.56 11.60 4.77

SP2 1.96 5.27 4.67
SP3 2.05 4.64 4.66
SP4 2.21 10.23 4.66
SP5 2.14 4.17 4.17
SP6 6.56 10.10 4.71
SP7 1.80 4.24 4.63
71
3.8.2. Influence of Column Section Size
Unlike flexural moment capacity, drift capacity, and ductility, observations from the
tests carried out at the MAST lab suggest that longitudinal bar buckling can be affected in
a significant way by the cross-sectional size of the columns. It is well accepted from
previous tests that confinement provided by transverse ties is required to prevent or
postpone bar buckling in rectangular RC columns. However, the MAST tests indicate that
larger reinforcing bars (No. 8 and No. 9) in columns made with normal strength concrete
can buckle in an in-plane mode (i.e. parallel to the face of the column) instead of outward
because the concrete cannot provide sufficient support to the bars parallel to the ties. In-
plane buckling in these bars occurred about the strong axis of the bars, which is contrary
to the common assumption that bars will only buckle about their weak axis.
Larger columns typically support a larger amount of axial load that requires larger bar
sizes. Compared to smaller columns, cover concrete is less effective in restraining larger
bars from buckling in bigger columns since it will crack at earlier stages of loading due to
the larger tensile forces carried by the longitudinal bars. This phenomenon makes columns
with larger cross-sectional dimensions more prone to bar buckling. Additionally, the
concrete surrounding the longitudinal bars is more susceptible to cracking and crushing in
bigger columns because it restrains larger longitudinal bars, which require much larger
restraint stresses to avoid buckling. Therefore, even if the bars are well restrained from
buckling outward by transverse ties, they are still prone to buckle sideways, and this
buckling mode can significantly reduce their load carrying capacity in the post-peak region
of the buckled bars. In-plane buckling of bars, as shown in Figure 3.15, was observed in
72
all tested specimens and it affected 36% of the buckled bars excluding those located at the
corners. The corner bars were excluded as their final buckled configuration was typically
a combination of outward and in-plane buckling. In many cases, no tie fracture or tie
opening was observed. However, after crushing of the concrete core, the bars were
observed to have clearly buckled parallel to the face of the column, and the buckling usually
occurred over a distance equal to two times the tie spacing with the middle tie offering no
restraint to the buckled bar.
Figure 3.15. In-Plane bar buckling and Fracture in MAST Column Tests
73
3.9. CONCLUSIONS
In this study, results from recent tests conducted at the MAST lab on full-scale RC
columns were combined with available dataset of RC columns to assess the effect of
cross-sectional size of columns on their seismic performance. A study on the parameters
representing seismic performance of RC columns confirms that effective stiffness, drift
capacity, and displacement ductility of RC columns can be significantly affected by axial
load and aspect ratios. In the case of drift capacity, the effect of axial load for stocky
columns with aspect ratio smaller than 2 is minor. However, for columns with larger aspect
ratios, the axial load ratio has an adverse effect on the drift capacity. This analysis also
confirms the sensitivity of the drift capacity and displacement ductility to loading history.
Test data analysis reveals that flexural moment capacity, drift capacity, and displacement
ductility of RC columns constructed with normal to moderately high strength concrete (2.5
ksi (17.24 MPA) ≤ f′c ≤ 10 ksi (68.94 MPA)) are not generally affected by the dimension
of their cross-sectional depths. However, observations from MAST tests imply that bar
buckling as an important seismic damage state can start at earlier drift ratios in larger
columns. These observations also suggests that despite bar fracture, which is typically
followed by a marked reduction in lateral load capacity, realization of the drift at the onset
of buckling requires proper sensor installation in the proximity of buckling zone. The
MAST tests revealed a mode of bar buckling seldom seen and never considered in
structural engineering of RC columns, the in-plane buckling mode in which the bars buckle
parallel to the face of the column rather than in the outward direction. In the observed cases
of in-plane bar buckling during the MAST tests, the bars deformed over a height equal to
74
two times the tie spacing (i.e. skipping the middle tie) and deflected sideways into the cover
on either side. This mode of buckling typically occurred about the strong axis of the bars
while the common assumption is that longitudinal bars buckle only about their weak axis.
From the MAST test observations, it is suggested that in-plane buckling initiates when the
cover concrete crushes and there is compression damage (softening) to the portion of the
core on either side of the buckled bars.
75
CHAPTER 4
IN‐PLANE BUCKLING OF LONGITUDINAL BARS IN

REINFORCED CONCRETE COLUMNS UNDER EXTREME
EARTHQUAKE LOADING3
4.1. SUMMARY
Subassemblage Testing (MAST) Laboratory of the University of Minnesota to investigate
their performance under extreme seismic events. The specimens were designed according
to seismic provisions of ACI 318-11 (ACI Committee 318, 2011) and incorporated
closely-spaced transverse hoops at their base. However, at large drift ratios during these
tests, longitudinal bars were observed to buckle parallel to the face of the columns with
transverse ties having little effect. This previously unobserved bar buckling phenomenon
is investigated numerically to gain a better understanding of the column characteristics that
affect it. First, a bar-spring mechanical model is used to understand the conditions needed
to prevent buckling of longitudinal bars by means of restraints with finite stiffness. Second,
3
Under review by Engineering Structures Journal.
76
a three-dimensional (3D) nonlinear finite element analysis of the lower portion of the
specimen subjected to monotonic loading was formulated in ABAQUS and validated with
test data. The analysis reveals that columns featuring larger cross-sectional dimensions and
longitudinal bar sizes and lower strength concrete are more prone to in-plane bar buckling.
Keywords: reinforced concrete; column, bar buckling, in-plane buckling, finite element
modeling
4.2. INTRODUCTION
Reinforcing bar buckling is a complex phenomenon that can reduce ductility of RC
columns and lead to significant stiffness reduction and strength loss. The complexity of
this phenomenon is attributed to the fact that the reinforcing bar buckling in RC columns
involves both material and geometric nonlinearities and depends on several parameters
including tie spacing and effectiveness, spalling of cover concrete, expansion of concrete
core, and loading history. Due to the complexity of the problem, many prior experimental
and analytical studies were carried out on isolated bars under compression to investigate
bar buckling behavior. Experimental investigations showed that the behavior of bars in
compression is generally different from that in tension (Dhakal & Maekawa, 2002; Bae,
Mieses, & Bayrak, 2005) and hence analytical formulations were proposed to capture the
constitutive relations of compression bars including buckling (Monti & Nuti, 1992; Gomes
& Appleton, 1997; Rodriguez et al., 1999). Additionally, the investigations revealed that
the post-yield behavior of reinforcing bars under compression is generally controlled by
the ratio of their unsupported length to bar diameter (s/db) as well as their eccentricity to
77
bar diameter ratio (e/db) (Bayrak & Sheikh, 2001; Bae et al., 2005). Considering the major
role of transverse ties in controlling the bar buckling, some research studies idealized the
bar buckling problem as a system of bar and springs with the springs representing the axial
stiffness introduced by transverse ties (Dhakal & Maekawa, 2002; Papia et al., 1988).
In all the former studies of bar buckling in RC columns, it is assumed that the bars will
buckle outward, that is perpendicular to the nearest column face, and the transverse ties are
essential in restraining them from buckling. However, recent tests carried out at the MAST
Lab in the Department of Civil, Environmental and Geo- Engineering of the University of
Minnesota on full scale RC columns subjected to extreme loading conditions similar to
what the columns experience during extreme seismic events revealed another buckling
mode in which the bars buckled parallel to the compression face of the columns (in-plane
bar buckling). In several instances, the ties were observed to still be effective in restraining
the bars from outward buckling. However, after extensive damage to the concrete
surrounding the bars, the reinforcing bars buckled over a height equal to twice the tie
spacing and exhibited large lateral deformations parallel to compression face of the
columns.
The main restraint against in-plane buckling is the concrete surrounding the bars,
especially the concrete residing between adjacent bars (i.e. the concrete “teeth”). This
concrete, while being within the column core, tends to be the portion of the core that gets
damaged earlier and to a greater extent during a load test. Extensive damage to the concrete
surrounding the bars would typically happen under large seismic loading conditions during
78
which concrete exhibits severe inelastic behavior. The unique control and loading
capabilities of the MAST Lab provided the opportunity to continue loading of the
specimens far beyond the extent of loading in previous tests on RC columns. During these
tests, the loading continued until the specimens lost more than 80% of their lateral load
capacity whereas in almost all previous tests, loading was terminated after a 20% strength
loss. In the authors’ opinion, such an extended loading regime provided the chance to
observe a failure mechanism in RC columns that was not addressed in previous tests.
In this study, first the required lateral restraint to prevent bar buckling is investigated
numerically using finite element (FE) modeling of isolated bars. In practice, the lateral
restraint can be provided by transverse ties in the case of outward deformation, or by
concrete surrounding the bars (i.e. the “teeth”) in the case of in-plane deflection. Results
from analysis of the isolated bars are then utilized in a nonlinear finite element model in
ABAQUS/Explicit to investigate the observed failure mechanism in the longitudinal bars
of tested column specimens. The model was validated with experimental data and
employed to explore the influence of several parameters that can affect in-plane buckling
of the bars.
4.3. EXPERIMENTAL PROGRAM
The columns tested at the MAST Lab represent the lower portion of typical columns
on the ground floor of a moment resisting frame in a high-rise building. The column
specimens were designed according to seismic provisions of ACI 318-11 (2011) and
featured two distinct cross-sectional dimensions: 36×28 in. and 28×28 in. (914×711 mm.
79
and 711×711 mm.). The specimens were built using normal-strength, normal-weight
concrete with 28-day nominal compressive capacity (f′c) of 5000 psi (34.47 MPa), ASTM
A706 (2003) Grade-60 No. 8 and No. 9 steel reinforcing bars confined by ASTM A615
(2003) Grade-60 No. 5 steel hoops. Transverse hoops were placed at close spacing in the
base of the columns to prevent outward bar buckling and ensure ductile behavior (Figure
4.1).
At the beginning of each test, a constant axial load was applied on top of the specimens
(Figure 4.1) and kept constant during the tests. To simulate behavior at near-collapse
conditions during extreme seismic excitation, the column specimens were then subjected
to either a monotonic or one of several progressively increasing displacement reversals
until they lost more than 80% of their lateral loading capacity. Further details regarding
these tests can be found elsewhere (Nojavan et al., 2015).
4.4. TEST OBSERVATIONS
When the lateral loading was applied to the column specimens, flexural cracks formed
on tensile face of the columns (Figure 4.1). Flexural cracking initiated at approximately
0.2% drift level, which was equal to 6 mm of the crosshead displacement. Additionally,
flexure-shear cracks formed in the lower portion of the other two faces of the columns and
longitudinal cracks were observed along the corners on the compression sides. Following
concrete cracking, the longitudinal bars started to yield at approximately 0.5% drift ratio
(15 mm of crosshead displacement). However, the columns exhibited further lateral load
80
Figure 4.1. Specimen Geometry and Detailing
capacity due to confining pressure from the transverse reinforcement and strain hardening
of longitudinal reinforcement. An increase in lateral load was accompanied by more crack
developments and propagations. Finally, the specimens started to lose flexural strength due
to crushing of the cover concrete and buckling of the longitudinal bars. The columns
exhibited severe strength loss by the end of the test (more than 80% of their peak lateral
load capacity) following the fracture of the buckled bars and extensive damage to the
concrete around the perimeter.
While it is commonly assumed that longitudinal bars in an RC column would buckle
in the outward direction (i.e. perpendicular to the compression face of the column),
investigation of the buckled bars at the end of the tests conducted at the MAST lab indicated
81
that this assumption is not necessarily true in all cases. Instead, it was observed in all tested
specimens that some of the bars buckled parallel to the compression face of the columns
(Figure 4.2). More specifically, when only the two middle bars on compression faces of
the columns in monotonic and cyclic uniaxial tests at the MAST Lab are considered, 10
out of 24 bars (i.e., 42% of the total observed buckling cases) revealed either a complete
or a dominant in-plane buckling mode. A buckling mode is considered to be completely
in-plane when no out of plane translation is observed in the buckled region whereas if a
small outward translation is present in conjunction with large in-plane translation, it is
considered a dominant in-plane buckling mode.
Another observation is related to the axis of bending in the buckled bars. Since
reinforcing bars incorporate ribs along their length, they have strong and weak axes of
bending. In practice, longitudinal bars are typically oriented in a way that their outward
deformation (i.e. deflection perpendicular to the column face) mobilizes flexural rigidity
corresponding to the weak axis while their in-plane deflection (i.e., deflection parallel to
the column face) occurs about their strong axis of bending. In many of the observed cases
of buckled bars in-plane buckling occurred about the strong axis of bending of the bars,
which is contrary to the common assumption, that is supported by experimental results
from isolated bars, that the bars would only buckle about their weak axis.
82
(a)
(b)
Figure 4.2. In-Plane Bar Buckling During Tests at the MAST Lab: (a) Column
Specimen under Monotonic Loading, (b) Column Specimen Subjected to Cyclic
Loading
83
4.5. ANALYTICAL INVESTIGATION USING BAR-SPRING
MODEL
A bar and spring model is used to investigate the demands on lateral restraints that
have finite stiffness to prevent buckling of longitudinal reinforcing bars. For outward (i.e.,
out-of-plane) buckling, these restraints are the hoops or ties that are also used as shear
reinforcement and reinforcement for confinement. For in-plane buckling, the restraints are
provided by the concrete that bears against the longitudinal reinforcement.
4.5.1. General Description of the Bar-Spring System
The buckling behavior of bars can be simply modeled with a bar-spring system as
illustrated in Figure 4.3 where springs represent the axial stiffness of the ties preventing
the bars from buckling. Dhakal and Maekawa (2002) employed such model to estimate the
required tie stiffness for a stable buckling mode and to estimate the buckling length of the
bars. Based on their investigation, Dhakal and Maekawa (2002) proposed Eq. 4.1 to
estimate the required stiffness (Kn) for the nth stable buckling mode as:
 4 EI
Kn  n (4.1)
s3
Where according to their definition, n = the number of tie spacings within the buckled
zone, s = tie spacing, EI = reduced flexural rigidity of the bars that depends on their yield
strength (fy), and αn= a constant that depends on the stable buckling mode. For the case
shown in Figure 4.3, in which the buckled length equals two tie spacings, the required tie
84
stiffness to prevent the bars from buckling is calculated as 33.1, 166.1, and 408.3 kip/in.
(5.8, 29.1, and 71.5 KN/mm) for No. 6, No. 9, and No. 11 bars, respectively, with αn given
as 0.1649 and assuming a yield strength (fy) of 74 ksi (510.2 MPa) for longitudinal bars of
the test specimens.
To further investigate the behavior of the buckled bars in the bar-spring system, a 3D
finite element model of the system represented in Figure 4.3 was built in ABAQUS using
2-node shear flexible beam-in-space (B31) elements representing the longitudinal bar and
a linear spring at its mid-height whose stiffness corresponds to the axial stiffness of the tie
leg. The length of the bar (l) was selected equal to two times the tie spacing that was used
in tested RC columns (i.e. s=5 in. (127 mm) and l=10 in. (254 mm.)). Behavior of the steel
bar was modeled using a trilinear curve fitting the stress-strain curve of longitudinal bars
in tested specimens with yield and ultimate strengths of 74 ksi and 112 ksi (510.2 MPa,
and 772.2 MPa), respectively (Figure 4.4) while the elastic stiffness of spring was taken as
constant during each analysis. A number of analyses with different bar sizes and values for
spring stiffness were conducted.
An initial imperfection was introduced to perturb the bar geometry and mobilize
buckling. In practice, this imperfection can be caused by misalignment or initial
crookedness of the bars. To minimize the effect of the initial imperfection on the buckling
load, a separate series of analyses were performed and the imperfection magnitude was
selected sufficiently small such that estimated critical loads had a maximum error of 1%
compared to the exact analytical solutions in the case of elastic bar. To simulate fixed
85
Figure 4.3. Bar-Spring System to Simulate Bar Buckling Behavior
120
100
Test
80
Stress (ksi)
Model
60
40
20
0
0 0.05 0.1 0.15 0.2 0.25
Strain (in./in.)
Figure 4.4. Stress-Strain Behavior of Longitudinal Bars
boundary conditions as assumed by Dhakal and Maekawa (2002), all degrees of freedoms
(DOFs) at the top and bottom of the bar (Points A and B in Figure 4.3) were fixed except
the vertical translation on the top (Point B). During each analysis, the vertical load on top
86
of the bar increased gradually using Modified Riks (arc-length) method (Crisfield, 1981;
Ramm, 1981; Powell & Simons, 1981) until the bar buckled.
4.5.2. Static Analysis Using Modified Riks (Arc-Length) Method
The modified Riks method was used to obtain static equilibrium solutions for this
geometrically and materially nonlinear problem with emphasis on the effect of buckling on
the descending branch of the force-displacement. The loading on the system is governed
by a single scalar parameter, namely the load proportionality factor (λ), an arc-length and
an iterative procedure that was used to determine a set of load and displacement at each
increment that satisfies static equilibrium. The method is neither displacement nor load
control, and the arc length includes components of both. The magnitude of loading at a
specified increment is defined as follows:
Ptotal = Po + λ (Pref – Po) (4.2)
where Po = the dead load (i.e., any load applied to the structure prior to the Static Riks
step), λ = the load proportionality factor that changes in each load increment, and Pref = the
reference load (i.e., the load that is applied during Static Riks step). A unit load was applied
to the bar during the Static Riks step (Pref =1). Since, there are no other loads prior to the
applied unit load, P0 was equal to zero, hence the magnitude of the load proportionality
factor, λ, represents the evolution of loading (Ptotal) during analysis.
87
4.5.3. Results from Bar-Spring Analysis
For each bar size in the study, a series of analyses were carried out with spring
stiffness as the variable parameter. The applied axial loads for such analyses on No. 9 bars
and various spring stiffness coefficients is represented in Figure 4.5, which indicates that
the magnitude of critical load increases with a rise in the spring stiffness. However, beyond
120
100
80
F (kips)
60 Kx=0
Kx=25
40 Kx=50
Kx=100
20 Kx=166
Kx=500
0
0 0.1 0.2 0.3 0.4
Lateral Disp. (in.)
Figure 4.5. FE Analysis Results for a No. 9 Bar
stiffness of 50 kip/in (8.8 KN/mm.) the rate of increase in critical load with addition of
stiffness drops significantly such that a spring that is 10 times stiffer would only slightly
increase the critical load (i.e. less than 3%). Adding more stiffness will eventually activate
higher buckling modes. A comparison of the estimated critical loads associated with
different spring stiffness with those having stiffness as proposed by Dhakal and Maekawa
(2002) reveals that their proposed equation overestimates the required spring stiffness to
88
prevent bar buckling (Table 4.1). Specifically, this analysis shows that a spring stiffness of
100 kip/in (17.5 KN/mm.) for No. 11 and 50 kip/in (8.8 KN/mm.) for No.9 bars would
suffice to deliver the axial load capacity within 1% and 2.5% of those obtained from their
proposed spring stiffness (i.e. 408 and 166 kip/in., respectively). In other words, providing
25% and 30% stiffness as of those estimated by Eq. 4.1 for No. 11 and No. 9 bars,
respectively, can prevent the first mode of bar buckling with only a small reduction in the
magnitude of the critical load (i.e., less than 3%).
The bar-spring system analogy can also be applied to the case of in-plane buckling of
the bars. While longitudinal bars in RC columns are restrained by transverse hoops to
prevent outward buckling, it is only the concrete surrounding the bars, as well as the friction
between the bars and transverse hoops, that prevent them from buckling sideways (i.e.,
parallel to the compression face of the column). In this case, the spring stiffness would
represent an initial stiffness provided by the surrounding concrete to restrain the bars from
buckling. The demand on concrete stiffness would increase with an increase in bar size
(Table 4.1). Assuming that the preventive effect of concrete on the in-plane bar buckling
can be idealized by a system of discrete springs, Table 4.1 implies that when concrete with
the same properties are used (i.e. similar spring stiffness) columns incorporating larger bar
sizes are more vulnerable to in-plane buckling as their reinforcing bars require stiffer
surrounding concrete (i.e. stiffer spring) to restrain the occurrence of in-plane buckling.
89
Table 4.1. Critical Loads from FE Analysis of Bar-Spring System
Bar Size
Kx (kip/in.)
No. 6 No. 9 No. 11
0 31.85 78.65 138.55
5 32.14 81.24 140.91
25 45.39 93.36 148.99
33 47.66* 99.00 152.80
50 48.79 108.52 165.79
100 49.23 110.96 173.76
166 111.16* 174.56
300 111.24 174.76
408 111.27 174.80*
500 111.28 174.81
800 174.84
* Represents buckling load calculated based on Eq. 4.1
4.6. FINITE ELEMENT MODELING OF RC COLUMN
SPECIMENS
A three-dimensional, nonlinear finite element model is described here and used to
simulate the observed in-plane buckling of the longitudinal reinforcement in the RC
column specimens tested in the MAST Lab. The model focuses on the portion of the
column adjacent to the base of the member, and it offers a more realistic representation of
this portion of the columns than does the bar-spring model in the preceding section. The
goal of the FE analysis is to simulate the phenomenon of in-plane buckling by means of a
qualitative assessment, and to give an indication of the variables that may affect it.
90
4.6.1. General Description of the FE Model
A 3D finite element model representing the test columns was built in ABAQUS to
investigate qualitatively the in-plane buckling of reinforcing bars in the tested specimens.
Since the in-plane bar buckling was observed during test of the first specimen (SP1) under
monotonic loading, SP1 was considered for model calibration of in-plane buckling
behavior. Also, to reduce analysis cost, only the lower portion of column specimen SP1
(25 in. above the column base) was modeled because in-plane buckling was observed in
this region (Figure 4.1).
The FE model was 25 in. (635 mm) tall and featured cross-sectional dimension of
36×28 in. (914×711 mm.) (depth × width) (Figure 4.6). Fixed boundary conditions were
applied to the nodes at the column base whereas all translational and rotational Degrees of
Freedom (DOFs) of the nodes on top of the column were constrained to a reference point
at the centroid of the cross section. Constraining the nodes to the reference point eliminates
singularity issues in the problem, while the loads are applied to the reference point.
Concrete was modeled using 8-node brick elements with reduced integration and hourglass
control (C3D8R). The maximum and minimum mesh size for concrete elements were 2.64
and 2.125 in. (67 and 54 mm.). Longitudinal bars were modeled with 2-node shear flexible
beam-in-space (B31) elements. The bars, except those on the compression face, were
defined within the concrete solid using embedment technique. In this approach, a perfect
bond between embedded elements (reinforcing bars) and the host element (surrounding
concrete) is assumed. The interaction between cracked concrete and longitudinal bars (i.e.,
91
CDP for core conc.
CDP for cover
Embedded Bars P
25 in.
Compression Face M (635 mm)
(b
Buckling Bars
Conc.
Transverse ties Ky Node
Conc. Kx
36 in. (914 Node s
mm) Ky
28 in. (672 Kx
mm) s
(a) K
Kx
Springs along Z direction Z Outward
are not shown for clarity. buckling
(c)
Direction
X
Y
In-plane
buckling
Figure 4.6. General Description of FE Model: (a) Modeled Section (b) Mesh of
Concrete Material (c) Modeling of Buckling Bars within Concrete Elements
bond-slip) is defined by adjusting the strain softening behavior of concrete in tension (i.e.,
tension stiffening). When embedded, all translational DOFs at the nodes of embedded
reinforcing bar elements are eliminated and these nodes are constrained to the interpolated
values of corresponding DOFs at adjacent nodes of the host (concrete) element. As a result,
bar buckling cannot be captured when the embedment technique is used since buckling
92
involves a sudden lateral deformation of the bars free from the surrounding concrete. Thus,
on the compression face where bar buckling is likely to occur, the bars were modeled as
independent B31 elements. These independent elements were connected to the reference
point on top of the specimen to move in accordance with the translation and rotation of the
specimen top. On the other hand, restraining effects of the ties and surrounding concrete
on the buckling bars were modeled using linear spring elements.
4.6.2. Estimation of Spring Stiffness
Given the proximity of most longitudinal bars to 90° bends in the ties, the spring
stiffness associated with ties were defined on the basis of the axial stiffness of tie legs as
follows:
Et At nl
K tie   (4.3)
Lt nb
where Et, At, and Lt are elastic modulus (200000 MPa), cross-sectional area and length of
the tie leg, respectively. Also, nl and nb represent the number of tie legs in the direction of
restraint and the number of bars that are restrained, respectively.
In the case of in-plane buckling, the restraining effect depends on the performance of
the concrete surrounding the bars. An analogy of the preventive concrete pressure on the
rebar to the force created in a discrete system of springs can be made in order to relate
spring stiffness to the characteristics of the physical problem. For the concrete to act as an
axial element, it is assumed that the concrete between adjacent longitudinal bars has begun
to crack off from the rest of the core (i.e., micro-crack generation reducing stress transfer
93
with the rest of the core) (Figure 4.7). In that case, the secant stiffness can be idealized as
follows:
Peff  eff Aeff

Ks   (4.4)
 eff  eff
where Δeff = the maximum lateral displacement of the bar when it buckles, σeff = the
concrete strength when the buckling starts, and Aeff = the effective contact area between
bars and surrounding concrete which prevents the bars from in-plane buckling.
The maximum lateral displacement at the onset of in-plane buckling (Δeff) depends
on many parameters (e.g. characteristic properties of the concrete surrounding the bars and
its performance as well as that of the longitudinal bars, loading history, core concrete
expansion and friction between ties and longitudinal bars) and its accurate measurement
requires a careful instrumentation and monitoring of the longitudinal bars during the test.
The bars which buckled in the in-plane mode during the MAST tests were observed to
have a maximum lateral displacement between db and 3db by the end of the tests. Therefore,
for the purpose of this study and because the onset of buckling is considered here, Δeff was
assumed to be equal to db. It is further assumed that the effective strength of concrete at
buckling (σeff ) would be only a fraction of the compressive strength of core concrete since
the concrete surrounding the bars will soften as the specimen is loaded in compression and
flexure. The state of strain at the onset of in-plane buckling is unknown, thus the stress
level is indeterminate. A lower bound estimate for the stress level is approximated
94
Figure 4.7. Preventive Effect of Surrounding Concrete from In-Plane Buckling after
Cover Spalling
assuming that the concrete is deformed well into the post-peak range of compression
behavior and its strength has degraded to the “residual strength” level that the Kent-Park
concrete model (Kent & Park, 1971) quantifies as 0.2λf′c, where λ is the factor that
represents the strength enhancement due to confining reinforcement. Thus, Eq. 4.4 can be
rewritten as follows:
Ks 
0.2f cs d b   0.2f s
c (4.5)
db
95
where βs = the dimension along the length of the bar in which the concrete is effective in
restraining the buckling bar and λ is suggested by Kent and Park as follows:
 s f yh
 1 (4.6)
f c
where ρs = volumetric ratio of the hoops and fyh = yield strength of the hoops. From
experimental observations, the in-plane buckling length was typically equal to two times
the tie spacing (2s). Therefore, assuming a half-sinusoidal shape over the observed
buckling length, the deflected shape of the bar (before buckling) is given by
( x)  x 
 sin   for 0  x  2 s and    max at x  s (4.7)
 max 2 s
The concrete is assumed to provide restraint over the portion of the buckled bar with a
lateral deflection  ≥ 0.5max, and at the location xo,  =0.5max which can be determined
as follows:
( x )   xo  xo 2 1
 0.5  sin  or  sin 0.5  1 3 (4.8)
 max 2 s  s 
The resulting effective length factor is
  (2s  2 x0 ) /(2s )  2(1  1 / 3)  4 / 3 (4.9)
when substituted in Eq. 4.5, the effective (axial) stiffness of concrete springs is as follows:
96
4
Ks  f cs (4.10)
15
In the case of SP1 with f'c = 4.86 ksi (33.5 MPa), fyh=60 ksi (413.7 MPa), s = 5 in. (127
mm), and ρs = 0.0148, λ would be 1.18 from Eq. 4.6 and the stiffness Ks will be equal to
7.65 kip/in. (1.35 KN/mm).
The behavior of the springs representing the axial stiffness of ties and restraining effect of
concrete, was modeled differently in tension and compression. In the inward direction (i.e.,
-Y direction in Figure 4.6(c)), the concrete is assumed incompressible and hence the bars
cannot bend inward whereas the tie stiffness will be activated when the bars deflect in the
outward direction (i.e., +Y direction). In the case of lateral deformation (i.e., parallel to the
column face), the behavior would depend on the geometry of the bar and transverse hoops.
For the bar shown in Figure 4.6(c), it is only the concrete stiffness that restrains the bar
from deflection in the -X direction (Kx1) whereas the tie legs would limit its deformation
in the +X direction (Kx2). Distinct spring stiffness corresponding to the deflection of this
bar in the +X and –X directions is represented in Figure 4.8. Additionally, the four bars
on the compression face (i.e., buckling bars) were connected to their adjacent concrete
elements using stiff linear springs along Z direction to simulate strong bond conditions as
was assumed for the embedded bars.
97
Figure 4.8. Spring Stiffness Parallel to the Column Face
4.6.3. Modeling of Concrete Behavior
Concrete behavior is modeled using Concrete Damage-Plasticity (CDP) implemented
in ABAQUS/Explicit. The CDP model assumes two main limit states for concrete as
cracking in tension and crushing in compression when concrete behaves in a brittle manner
(i.e., under low confining pressure). Damage associated with these two failure mechanisms
represents an isotropic stiffness degradation and is characterized independently in tension
and compression by a single scalar parameter. Damage parameters in tension (dt) and
compression (dc) are functions of hardening variables (also known as equivalent plastic
strains, ~t pl and ~cpl ). Hardening parameters represent the evolution of the yield surface and
stiffness reduction and are automatically calculated by ABAQUS from inelastic strains
(i.e., total strain minus the elastic strain corresponding to undamaged material state). The
CDP model uses the yield function proposed by Lubliner et al. (1989) with modifications
proposed by Lee and Fenves (1998) to capture behavior in tension and compression. The
98
potential plastic flow is based on the Drucker-Prager hyperbolic function and the flow rule
is non-associated.
For the RC columns modeled herein, the elastic modulus of concrete was estimated as
3,974 ksi (27400 MPa) for a concrete with compressive strength of 4.86 ksi (33.5 MPa)
according to Sec 8.5.1 of ACI 318-11 (2011). Except the dilation angle which was assumed
as 30°, the default values were used for the other parameters in the CDP model. These
default values are 1.16 for the ratio of the initial equibiaxial compressive yield stress to
initial uniaxial compressive yield stress, 2/3 for the ratio of the second stress invariant on
the tensile meridian to that on the compressive meridian, 0.1 for eccentricity and 0 for
viscosity.
The stress-strain behavior of concrete in compression was assumed linear up to 0.45f′c
and defined according to the model proposed by Kent and Park thereafter. A different
stress-strain behavior was considered for core concrete as compared to the one for cover
concrete as illustrated in Figure 4.9 to account for the confining effect of rectangular hoops.
The uniaxial tensile behavior of concrete was assumed linear up to a tensile yield stress
(ft0) of 0.522 ksi (3.6 MPa) calculated according to Sec 9.5.2.3 of ACI 318-11(2011). The
post-peak behavior of concrete in tension (tension stiffening) was defined by specifying
the fracture energy (Gf, the energy required to form a unit area of crack) as a material
property. This model assumes a linear softening behavior after cracking in tension until a
complete loss of tensile strength at a cracking displacement equal to ut0 = 2Gf /ft0. Although
suggested values for fracture energy of plain concrete are typically low, low values of this
99
parameter often lead to numerical issues. Moreover, fracture energy is much larger in
reinforced concrete due to the presence of distributed reinforcing bars (Mercan, Schultz, &
Stolarski, 2010). In this study, a value for fracture energy equal to 0.001 kip/in. (1.76×10-4
KN/mm) was used for the concrete. Also, since calibration was done using only the RC
column specimen under monotonic loading, damage parameters were not needed.
7
Cover Concrete
6
Compressive Stress (ksi)
Core Concrete
5
0
0 0.02 0.04 0.06 0.08
Strain (in./in.)
Figure 4.9. Stress-Strain Relationship of Concrete
4.6.4. Explicit Dynamic Analysis of the Model
The FE model of the bottom of RC column specimen SP1 was analyzed using the
Explicit Dynamic procedure in ABAQUS, which is an efficient way of solving dynamic
problems. This procedure is explicit in the sense that equilibrium equations at time t+Δt
are solved based entirely on solutions in time t using a central difference operator. Static
problems can be treated quite well using Explicit Dynamic analysis provided that the inertia
effects are low. Minimizing inertia effects can be achieved by either decreasing the mass
100
density of concrete or increasing the time over which the load is applied. Unlike Static Riks
(Arc-Length) method that is sensitive to parameters that define behavior of concrete and
encounters convergence issues, Explicit Dynamic analysis is more robust, due to
smoothing of the solutions by inertia effects, and can handle problems for which the former
method fails to converge. However, the central difference operator used in Explicit
Dynamic analysis is only conditionally stable. The stable time limit (Δt) for this operator
is related to the time required for a dilatational wave to pass through the smallest element
dimension in the model mesh. Thus, the stability limit is inversely proportional to the
highest eigenvalue in the system. By default, estimation of a stable time limit and time
incrementation scheme are automatically carried out in ABAQUS. The total number of
increments can be estimated as n=T/Δt where T represents total analysis time. While the
number of increments is typically much larger than that needed for the Static Riks analysis,
each increment is fairly inexpensive because it does not require solving a system of
nonlinear equations. However, computational cost of the analysis can significantly increase
by mesh refinement.
To minimize inertia effects during Explicit Dynamic analysis of the model, the loading
was applied smoothly. When the analysis time is low or the time increments are large,
oscillation will typically appear in the results due to inertia effects. To eliminate inertia
effects, it is usually enough to set the loading time 10 to 50 times larger than the
fundamental period of system (i.e. period associated with the lowest natural frequency of
the system) (Mercan et al., 2010). In this analysis, the lowest frequency was estimated as
19.41 Hz resulting in a natural period of 0.052 sec. Therefore, the total analysis time was
101
set to 2 sec. and subdivided into two equally spaced windows. During the first window,
the axial load was applied and increased as a ramp function to its final value of 756 kips
(3362.9 KN) at the end of the first loading window (i.e. 1 sec. from start point). The
magnitude of the axial load was kept constant during the rest of the analysis. Following the
first loading window, a top rotation was applied to the column whose magnitude increased
as a ramp function to a final rotation of 0.05 radian at the end of the step time (i.e. 2 sec).
The loading time during each window was equal to 1 sec., which is approximately 20 times
the fundamental period of the system. To estimate inertia effects and evaluate accuracy of
the analysis, total strain energy (EI) and total kinetic energy (EK) of the system were
calculated. Using the above loading rate, the ratio of the kinetic energy to total strain energy
(EK / EI) remained small (less than 0.1%) indicating that inertia effects are negligible.
4.6.5. Model Validation
The 3D model described in the previous section was analyzed using Explicit
Dynamic analysis and its results were compared against test observations. The FE model
could capture the incidence of in-plane buckling in the bars located on the compression
face of the column (Figure 4.10). The two middle bars represent in-plane buckling towards
the centerline of the column that resembles to what was observed during test of specimen
SP1. These middle bars were located at the corners of the middle hoop that restrained them
from in-plane buckling in the opposite direction as shown in Figure 4.2.
102
View 1
View 1 Compression
Figure 4.10. In-Plane Buckling of Bars in the FE Model
Variation of the base moment with rotation at a section located 25 in. (635 mm.) above
the column base during the test is represented in Figure 4.11 and compared to the top
moment in the FE model. To obtain the rotation at this section during the test, the curvature
profile of the lower portion of the column was estimated based on displacement readings
with the LVDTs located on the SE and NW faces of the column and at a distance equal to
12, 24, and 36 in. (305, 610, and 914 mm.) above the column base. The estimated curvature
profile was integrated to get rotations at 24 and 36 in. (610 and 914 mm.), which were then
interpolated to obtain the rotation at a distance of 25 in. (635 mm.) from the column base.
The test specimen exhibited larger rotations at the column base due to deflection at the
column-footing interface generated by bar slip of the bars within the footing, and this
behavior is not captured in the FE model because it does not include representation of the
103
35000
30000
Base Moment (kip.in)

25000
Test
20000
FEM Model
15000
10000
5000
0
0 0.01 0.02 0.03 0.04 0.05
Top Rotation (rad.)
Figure 4.11. Moment-Rotation Behavior of Tested and Modeled Specimens
footing. Additionally, the model underestimates the peak strength and strain-hardening
behavior of the tested specimen due to the idealization regarding spring stiffness
representing the restraining effect of the concrete. In other words, the spring stiffness due
to the effect of concrete surrounding the bars is estimated based on the assumption that
concrete has reached its residual capacity. This limit state has been chosen to ensure that
conditions for bar buckling are present during the FE analysis. The spring stiffness and
thus, the amount of restraint on the bars located on the compression face of the column, is
much larger at initial stages of the test and its magnitude deteriorates during loading of the
specimen. An underestimation of the initial restraining effect of the concrete on reinforcing
bars can lead to differences in strength and post-yield stiffness as observed in Figure 4.11.
However, for the purpose of a “qualitative” study of buckling behavior of the bars located
on the compression face, these differences are considered acceptable.
104
4.6.6. Parametric Study
The FE model discussed in previous section (benchmark model) was employed in a
parametric study to investigate the influence of several parameters on the in-plane buckling
of middle and corner bars. The parameters include concrete compressive strength,
longitudinal bar size and spacing, tie spacing, and cross-sectional dimensions. A list of the
models is given in Table 4.2. Maximum in-plane deformation of middle and corner bars of
the benchmark model were compared to corresponding values of the other models to
estimate the influence of the parameters under study.
Table 4.2. Parametric Study Matrix
Tie Spacing
Paramete Bar Size f'c (ksi) Bar Spacing (in.)
Model 2
b×h (in ) (in.)
r
6 9 11 4 5 6 3.5 5 8 5.9 7.9 11.9
Base
1
Model
2 Concrete
3 Strength
4
Bar Size 36×28
5
6 Bar
7 Spacing
8 Tie
9 Spacing
10 42×32 9 1/4
Column
Size
11 20×16 3 15/16
105
4.7. RESULTS AND OBSERVATIONS
The FE models listed in Table 4.2 were subjected to an axial load and top rotation
similar to those applied to the benchmark model. For each parameter, results from three
models (including the benchmark model) were compared.
4.7.1. Effect of Concrete Compressive Strength (f′c)
Models 2 and 3 in Table 4.2 were defined with properties similar to the benchmark
model except that they feature concrete compressive strengths (f′c) of 3.5 and 8 ksi (24.1
and 55.2 MPa), respectively. The spring stiffness associated with concrete surrounding the
bars (Ks) in Models 2 and 3 were calculated from Eq. 4.10 as 5.85 and 11.85 kip/in. (1.02
and 2.07 KN/mm), respectively. In-plane buckling was observed at the lower portions of
middle and corner bars in both models. The maximum in-plane displacement for corner
and middle bars of these two models are compared to those from the benchmark model in
Figures 4.12(a) and 4.12(b). As the concrete compressive strength (f′c) decreases, these
figures reveal that initiation of in-plane buckling occurs earlier and the bars end up with
larger lateral deformations. In other words, in-plane bar buckling is expected to start earlier
in RC columns that are built with normal or low strength concrete due to the reduced
effective stiffness associated with lower concrete strength. Moreover, an RC column with
a lower f′c is likely to undergo earlier cracking and crushing under the effect of axial and
flexural loads. Therefore, assuming that the outward buckling is completely prevented by
transverse hoops, the support from concrete to prevent in-plane deformation of the bars
would vanish faster in these columns, which reduces the resistance to in-plane buckling of
106
the longitudinal bars. Additionally, in-plane buckling would result in a larger deformation
of longitudinal bars embedded in concrete with lower f′c. Thus, the sudden reduction of
strength due to bar fracture (from kinking of the buckled region) can occur earlier in these
columns. As shown in Figure 4.13, the embedded bars in Model 3 that incorporates
concrete with f′c of 8 ksi (55.2 MPa), started to lose axial load capacity at a larger top
rotation than did bars in models with f′c of 3.5 and 5 ksi (24.1 and 34.5 MPa) (i.e., Model
2 and benchmark model, respectively).
0 0
-0.5 -0.5
Lateral Disp. (in.)
Lateral Disp. (in.)
-1 -1
3.5 ksi 3.5 ksi
5 ksi 5 ksi
-1.5 -1.5
8 ksi 8 ksi
-2 -2
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
Top Rotation (rad) Top Rotation (rad)
(a) (b)
Figure 4.12. Effect of Concrete Compressive Strength on Buckling Displacement in:
(a) Corner Bars, (b) Middle Bars
107
100 100
80 80
Axial Force (kips)

Axial Force (kips)
60 60
40 40
3.5 ksi 3.5 ksi
5 ksi 5 ksi
20 20
8 ksi 8 ksi
0 0
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
(a) (b)
Figure 4.13. Variation of Axial Loads of the Reinforcing Bars During FE Analysis in:
(a) Corner Bars, (b) Middle Bars
4.7.2. Effect of Longitudinal Bar Size and Spacing
As suggested by FE analysis of simple bar-spring models, larger bar sizes require
more restraining force from the surrounding concrete to resist buckling. However, as the
specimen is subjected to axial and flexural loads, longitudinal bars exhibit lateral
deformations leading to in-plane buckling when the concrete softens. Results from FE
analysis of models with different bar sizes (Models 4 and 5) are compared with the
benchmark model in Figures 4.14 and 4.15 for corner and middle bars in the compressive
face. As noted in these figures, larger bar sizes (i.e. No. 11 and No. 9) experience larger
final deformations in the buckling zone. Final lateral in-plane deformation for a No. 11 bar
is approximately 50% larger than that for a No. 6. However, all bar sizes start to lose their
axial strengths and buckle at approximately similar top rotations. Unlike bar size, the effect
of longitudinal bar spacing seems to be insignificant on the in-plane buckling of the bars.
108
A comparison of the results from Models 6, 1 and 7 with No. 9 longitudinal bars that were
spaced at 5.9, 7.9, and 11.9 in. (150, 201 and 302 mm), respectively, shows that in-plane
buckling initiates at similar loading level and leads to approximately the same lateral
deformations in these models. It should be noted that longitudinal bar spacing may have a
role in the extent of damage and stiffness deterioration in the concrete surrounding the bars
during the test. However, at a specified concrete damage state implemented in these FE
models where concrete reaches its residual compressive strength, bar spacing is shown to
be insignificant for in-plane buckling of the bars.
0 0
Lateral Disp. (in)
Lateral Disp. (in)
-0.5 -0.5
#11 #11
-1 -1
#9 #9
#6 #6
-1.5 -1.5
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
(a) (b)
Figure 4.14. Effect of Bar Size on Buckling Displacement in: (a) Corner Bars, (b)
Middle Bars
109
0 0
6 in. 6 in.
8 in. 8 in.
12 in. 12 in.
Lateral Disp. (in)

-0.5 -0.5
Lateral Disp. (in)
-1 -1
-1.5 -1.5
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
(a) (b)
Figure 4.15. Effect of Bar Spacing on Buckling Displacement in: (a) Corner Bars, (b)
Middle Bars
4.7.3. Effect of Tie Spacing
While transverse ties are effective in preventing outward buckling of bars, they may
also help the bars to resist in-plane buckling by enhancing concrete properties through
confinement. Closely spaced transverse hoops will reduce softening rate of concrete and
postpone its stiffness deterioration. Therefore, longitudinal bars in columns with closely
spaced ties can resist in-plane buckling longer as the surrounding concrete that is effective
in restraining them from in-plane buckling will degrade at a slower rate. However, the
beneficial effect of lateral ties in limiting the in-plane buckling will disappear once the
concrete crushes and loses significant amounts of strength. Three columns with transverse
ties spaced at 4, 5 and 6 in. (102, 127, and 152 mm.) (Models 8, 1 and 9) which were
designed in accordance with seismic provisions in ACI 318-11 (2011), are compared in
Figure 4.16. As the spacing of transverse ties increase, these columns lose their axial
110
capacity faster and undergo larger lateral deformations under the applied loads (Figure
4.16). However, these variations are minor since the concrete is considered to have
degraded to its residual capacity in these FE models.
0 0
4 in. 4 in.
5 in. 5 in.
Lateral Disp. (in)
6 in.
Lateral Disp. (in)

-0.5 -0.5 6 in.
-1 -1
-1.5 -1.5
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
(a) (b)
Figure 4.16. Effect of Tie Spacing on Buckling Displacement in: (a) Corner Bars, (b)
Middle Bars
4.7.4. Effect of Column Cross-Sectional Size
Models 10 and 11 in Table 4.2 represent typical perimeter moment frame columns in
multi-story buildings that satisfy ACI 318-11 (2011) seismic provisions. Model 10 is a
column featuring a 42×32 in. (1067×813 mm) cross-sectional dimension that is built with
4 No. 11 bars spaced at 9 ¼ in. (235 mm) on the compression face (i.e., 16 No. 11 bars
within the whole cross section) whereas Model 11 represents a column that is 20×16 in.
(508×406 mm) in cross section and includes 4 No. 6 bars spaced at 3 15/16 in. (100 mm) on
the compression side. Both columns incorporate No. 5 transverse hoops located at 5 in.
111
(127 mm) spacing along the columns and concrete with f′c of 5 ksi (34.5 MPa). The axial
load on the models prior to the application of top rotation was 1008, and 240 kips (4483.8
and 1067.6 KN) for models 10 and 11, respectively, which is equivalent to an axial load
ratio (P/f′cAg) of 15%. FE analysis of these models reveals that in-plane buckling of
longitudinal bars in Model 10 with larger bars embedded in a bigger cross section is more
critical than in Models 1 and 11. In-plane buckling of bars in the column with a bigger
cross section (i.e. Model 10) initiates earlier and these bars lose their axial load capacity
faster (Figure 4.17). Therefore, larger columns incorporating larger bar sizes appear to have
higher propensity to in-plane buckling than smaller columns that are constructed with
similar concrete properties but which incorporate smaller bar sizes.
0 0
Lateral Disp. (in)
-0.5 -0.5
Lateral Disp. (in)
42x32 42x32
-1 -1
36x28 36x28
20x16 20x16
-1.5 -1.5
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
(a) (b)
Figure 4.17. Effect of Cross-Sectional Size on Buckling Displacement in: (a) Corner
Bars, (b) Middle Bars
112
4.8. CONCLUSIONS
This paper presents results from a computational study of a failure mechanism
observed during tests of seven full scale RC columns at the MAST lab subjected to a
constant axial load and various lateral loading protocols. During failure, longitudinal bars
were observed to buckle parallel to the face of the columns with their surrounding concrete
completely crushed at the end of the tests. The loading capabilities at the MAST Lab
enabled testing of full scale column specimens at larger drift ratios than those applied
during previous tests under which the column specimens exhibited an unobserved mode of
failure due to in-plane buckling of reinforcing bars. Since concrete is believed to be the
main restraint for in-plane bar buckling, an analogy was made between the behaviors of
these bars to a system of bar-spring. Analysis of the bars with linear springs at their mid-
height using finite element techniques identified an upper-bound for elastic stiffness of the
springs to prevent bar buckling. It was observed that the required stiffness to prevent
buckling of the bars increases with bar size. Since the restraining effect of concrete
degrades during the tests, elastic stiffness of springs were estimated at a stage where
concrete has lost most of its compressive strength and thus the bars are prone to in-plane
buckling. On the other hand, large reinforcing bars can also exert greater force on their
surrounding concrete and accelerate its rate of damage. A 3D finite element model of the
lower portion of the specimen subjected to monotonic loading was developed in
ABAQUS/Explicit and its behavior was validated using test data. The model was then
utilized to study the influence of several parameters that can affect in-plane buckling of the
bars including concrete compressive strength, longitudinal bar size and spacing, tie
113
spacing, and cross-sectional dimensions of the columns. Results from these analyses reveal
that concrete with a larger f′c helps to postpone this mode of buckling and reduces rate of
axial strength loss in the bars. Additionally, larger bar sizes are recognized to be more
likely to experience in-plane buckling as their restraint demands are higher. However, the
effect of longitudinal bar spacing is insignificant in the stage where concrete has reached
its residual capacity. At this stage, reduction of tie spacing is found to have minor beneficial
effect on in-plane buckling of the bars. Finally, the specimens featuring larger
cross-sectional dimensions that incorporate larger bar sizes are shown to be more critical
to in-plane bar buckling. This final conclusion can be regarded as a reason for not reporting
the in-plane buckling in previous tests as the columns tested at the MAST Lab were larger
than almost all RC columns designed according to ACI 318 provisions.
114
CHAPTER 5
EFFECT OF LOADING PROTOCOLS AND ESTIMATION OF

DAMAGE INDICES IN REINFORCED CONCRETE COLUMNS
5.1. SUMMARY
Subassemblage Testing (MAST) Laboratory of the University of Minnesota under distinct
lateral loading protocols to large deformation demands. The loading protocols included a
single cycle monotonic, uniaxial and biaxial symmetric cyclic, and a near-collapse
protocols. To investigate the post-peak behavior of the specimens, loading on the
specimens was continued beyond the stopping point used in most of the previous tests (i.e.,
20% strength loss) until the specimens exhibited severe strength loss (i.e., approximately
80% reduction in the peak lateral load capacity) and stiffness degradation. The extent of
damage during each test was estimated by nine noncumulative and cumulative damage
index models. Additionally, the observed damage to the specimens during the tests was
utilized to define five visual damage categories. Calculated and observed damage along
with cyclic force-deformation response, energy dissipation, and stiffness reduction of the
115
specimens were considered to investigate the effect of the applied loading protocols.
Results of the study suggest that the biaxial loading can lead to a larger extent of damage
than similar protocols applied in one direction only. Also, a higher level of axial load will
cause a more severe damage state under similar lateral loading protocols. Finally, when
axial load ratio is similar during uniaxial tests, a combination of symmetric displacement
cycles followed by a monotonic push can lead to a more detrimental damage state than
purely symmetric cyclic, monotonic, or the near-collapse loading protocols.
5.2. INTRODUCTION
Structural performance of building and bridge elements can deteriorate when they are
subjected to seismic events. Large cyclic load reversals during an earthquake can result in
a reduction of the load-carrying capacity and stiffness of structural elements and cause
varying levels of damage. The extent of damage is related to structural properties and to
the characteristics of the loading history during an earthquake. In the case of an RC column,
damage can be in the form of minor tension cracks, cover spalling, yielding of reinforcing
bars or transverse hoops, bar buckling and fracture, and cracking and disintegration of the
core concrete.
Unlike traditional design approaches, recent performance-based design philosophy
anticipates the structure to dissipate energy during an earthquake and accepts certain levels
of damage in structures depending on the expected risk of occurrence of earthquakes during
their lifetime, and importance of the structure. Moreover, the ultimate capacity is associated
with the structure reaching a near-collapse limit state. A regular building structure may be
116
designed to experience little or no damage during minor earthquakes that occur frequently
during its lifetime (i.e., immediate occupancy limit state) while certain levels of structural
and nonstructural damage may be acceptable during less frequent medium earthquakes (i.e.
life safety limit state). For the design earthquake or a very severe earthquake with a large
return interval, even a non-repairable damage might be acceptable provided that the life-
safety is not compromised (i.e., the structure does not approach the near-collapse limit
state). On the other hand, accepted levels of damage during an earthquake or the design
objectives often lead to larger structural demands in more important structures (e.g.
hospitals, and bridges).
Structural design objectives in the performance-oriented seismic design philosophy
are defined based on accepted levels of damage (i.e., limit states) under distinct levels of
earthquake loads that are expected during lifetime of the structure. Estimated structural
demands (e.g. required strength, or stiffness) to meet certain structural objectives are then
compared against capacity of the structure (e.g. available strength, stiffness, or deformation
capacity). Evaluation of structural capacity, on the other hand, relies primarily on
numerical modeling tools that are calibrated against test results. In the case of RC columns,
these test results are used to estimate their lateral load and deformation capacity under
simulated seismic loadings.
While several tests have been conducted on RC columns, their performance,
specifically in the post-peak region is not completely understood. This lack of knowledge
in the softening regime of response for columns is mainly due to the fact that almost all of
117
the previous tests were terminated once specimens lost no more than approximately 20%
of their lateral load capacity. While an arbitrary definition of structural failure can be tied
to the aforementioned condition, RC columns can still survive collapse when loaded well
beyond this arbitrary point of failure. Therefore, a series of tests is needed to enhance
knowledge on softening behavior of RC columns when they lose most of their lateral load
capacity. Such an understanding is essential, especially for defining seismic collapse safety
limit states of RC columns in performance-based design.
Unlike RC bridge columns, there are limited previous experimental efforts that have
focused primarily on the effects of applied loading protocols on RC building columns and
during which similar test specimens were subjected to monotonic and cyclic loading
protocols. Considering the uncertainty in the amplitude and number of cycles that RC
columns in a building structure may sustain during an earthquake, their performance under
distinct simulated loading protocols will be beneficial in the estimation of their force-
deformation backbone capacity curves and characterization of damage during various
stages of loading.
Damage characterization is also an essential element for designing new structures
based on performance-oriented design as well as for initial assessment and retrofit decision
making after seismic events. The extent of expected damage is used to define different limit
states in performance-based design and to set structural objectives accordingly. These limit
states for building structures can be defined as experiencing minor or cosmetic damage
during frequent earthquakes or just remaining stable during catastrophic seismic actions.
118
In both cases, the limit states are related to the level of anticipated damage. Once a structure
experiences a seismic event, identification of the extent of observed damage will be
essential in the estimation of repairability of the structure and in preparation of retrofit
plans.
The state of damage in structures is often quantified by damage indices that are usually
scaled to be zero in the case of an undamaged structure, and unity in the case of collapse.
The incidence of collapse in structural elements is usually defined arbitrarily as a state of
loading under which the structural element loses a certain amount of its load carrying
capacity, typically 20% of the peak capacity. Several damage indices have been proposed
to quantify the extent of physical deterioration in structures and their residual capacity.
Most of the existing damage measures are described in terms of lateral deformation (i.e.
displacement, rotation, or curvature), stiffness, hysteretic energy, fatigue or a combination
of these parameters. However, application of different measures often leads to differing
results which makes recognition of the actual state of structural damage challenging.
In this paper, the effect of distinct loading protocols including monotonic, uniaxial
and biaxial symmetric cyclic, and a near-collapse protocol developed for this experimental
program on the performance of full scale RC columns is investigated. During these tests,
the column specimens were loaded beyond the common stopping point in previous tests
until they significantly lost their lateral load capacity and exhibited severe deterioration.
Observed damage during tests of the specimens is used to define damage categories and to
119
assess various damage index models. Observed and calculated damage values are then used
to study the effects of different lateral loading protocols.
5.3. PREVIOUS WORK ON DAMAGE INDICES
Several damage index models have been proposed to quantify the state of physical
deterioration and strength loss in structures under various loading conditions. Damage
indices are usually described locally, in the element level or globally, for the entire
structure. Member deformation (e.g. lateral displacement, rotation, or curvature), stiffness,
hysteretic dissipated energy, fatigue behavior, or a combination of these parameters are
typically used in expressing local measures of damage whereas global indices are often
calculated by weighted summation of damage in their components. Given that the
specimens in this experimental program represented the lower portion of the first-story
columns in a prototype RC moment frame, only local measures of damage are considered
here. The state of damage during loading depends on both the amplitude of the applied load
and the number of cycles. While noncumulative damage indices ignore the effect of
cycling, cumulative damage indices take the effect of both magnitude and number of cycles
into account usually by summation over all preceding cycles during a loading history. A
summary of the well-known damage index models is presented in the following section. A
comprehensive description of local and global damage indices is also presented by Chung
et al. (1987) and Williams and Sexsmith (1995).
120
5.3.1. Noncumulative damage indices
Perhaps the simplest measure to quantify damage in RC members are the ones
described in terms of ductility. Damage indices in terms of rotation and curvature ductility
were proposed by Banon et al. (Banon, Irvine, & Biggs, 1981):
 max  max   y
  1 (5.1)
y y
 max  max   y
  1 (5.2)
y y
where θmax and θy represent maximum and yield rotations, respectively and ϕmax and ϕy are
the corresponding curvatures. Banon (1980) also proposed a ductility definition based on
plastic rotation (θp), however he noted that θp may not increase as rapidly as the progression
of damage in the structural member, thus he suggested using the ductility based measures
in Eqs. 5.1 and 5.2. A similar equation can be expressed for lateral displacement ductility
as follows:
 max  max   y
   1 (5.3)
y y
where Δmax and Δy correspond to the peak displacement and displacement at yield,
respectively. While its simplicity is appealing, damage indices based on ductility introduce
some disadvantages. Calculated ductility terms based on Eqs. 5.1 and 5.2 are not scaled to
be zero for the undamaged members and one at their collapse state. Also, these damage
121
indices are affected by the magnitude of the yield deformation in the RC member which
can be defined based on the measured first yield in the system, an assumed equivalent
elastoplastic behavior or an elastoplastic energy absorption for the member, or a reduced
(secant) stiffness on an equivalent elastoplastic member response (R. Park, 1989).
Considering that rotations can occur due to shear deformation and bar slip, Banon et al.
(1981) indicated that an analytical estimation of the yield rotation assuming antisymmetric
bending and elastic stiffness tends to underestimate the true yield rotation and results in
rotation ductilities that are larger than experimental ones. Even if the yield deformations
are calculated correctly, two specimens with identical lateral force and deformation
capacity but different stiffness will exhibit significant differences in the calculated
displacement ductilities (μΔ ).
Lybas and Sozen (1977) proposed a damage index based on the ratio of the initial
tangent stiffness (Ko) to the reduced (secant) stiffness at maximum displacement of the
current cycle (Km):
Ko
DR  (5.4)
Km
The advantage of the above equation as compared to the ductility expressions is its
lack of dependency on yield displacement. However, analytical estimation of initial tangent
stiffness (Ko) may be difficult due to the presence of the shear deformation and bar slip.
Therefore, Banon et al. (1981) proposed replacing Ko with calculated stiffness of the
122
cracked member (Kcr) and modified Eq. 5.4 to obtain the flexural damage ratio (FDR) as
follows:
Kcr
FDR (5.5)
Km
where Kcr = 24EI/L3 in the case of antisymmetric bending, in which E = elastic modulus of
elasticity, I = moment of inertia, and L = member length. FDR seems to be a better indicator
of damage in the specimen compared to ductility terms as it takes the effects of strength
reduction and stiffness deterioration into account and is not sensitive to the displacement
at yield. However, 24EI/L3 does not represent a valid estimation for Kcr of a first-story
column since the top of the column is likely to rotate even if the rotation at the column base
is very small due to a very stiff foundation (which is not always the case).
Roufaiel and Meyer (1987) included secant stiffness corresponding to the failure (Kf)
and modified the flexural damage ratio as follows:
K m  K o

FDR   (5.6)
K f  K o
where the positive signs (+) denote the loading direction. A similar equation is expressed
for loading in the opposite direction and the flexural damage ratio is taken as the maximum
of the calculated values in positive and negative directions. Assuming a bilinear moment-
curvature response for the member, Ko will be equivalent to the secant stiffness at yielding.
123
5.3.2. Cumulative damage indices
The state of damage in RC members depends not only on the magnitude of the applied
loads, but also on the number of cycles they sustain. Repeated loading at large deformations
will result in the accumulation of plastic damage in the RC member and expedites strength
loss and stiffness deterioration due to the phenomenon known as low cycle fatigue.
Therefore, cumulative damage indices that include the effect of damage accumulation in
the preceding cycles have been proposed. Cumulative damage indices are typically
expressed in terms of accumulation of plastic deformation or strain in a form of low cycle
fatigue formulation, reduction of stiffness, normalized dissipated energy, or a combination
of these parameters.
5.3.2.1. Deformation- and Strain-based cumulative damage indices
To take the effect of cycling into account, Banon et al. (1981) attempted to extend his
proposed equation for rotation ductility (Eq. 5.1) and introduced normalized cumulative
rotation term:
NCR 
 max  y
(5.7)
y
A similar form can be expressed in terms of cumulative displacement:
NCD 
 max  y
(5.8)
y
124
While NCR can represent damage due to cumulative fatigue-type loading, Banon et al.
(1981) observed a large scatter in the results at the state of specimen failure.
Stephens and Yao (1987) modified the equation which was previously proposed by
Yao & Munse (1962) for steel structures and expressed a damage index that accounts for
the accumulation of plastic deformations due to cyclic loading on RC structural elements:
n
D   Di (5.9)
i 1
where D is the damage due to n cycles and:
1br
  pi 
Di    (5.10)
  
 f 
where Di = damage associated with the ith cycle, Δδpi+ = incremental positive plastic
deformation of the ith cycle (Figure 5.1), Δδf = the value of Δδpi+ single cycle test resulting
in the specimen failure, b is recommended as 0.77, and r is ratio of the plastic deformation
in the positive direction (Δδpi+) to that in the negative direction (Δδpi-) during ith cycle
described as:
 pi
r (5.11)
 pi
125
Figure 5.1. Definition of Plastic Deformations in Stephens-Yao Model
While the damage index based on Stephens’s model includes the effect of loading
reversals, it ignores the effect of sequence of cycles that can affect the extent of damage.
Additionally, when results from monotonic tests are not available, an estimation of Δδf
depends on the selection of the drift ratio at which failure is expected to occur.
Wang and Shah (1987) developed a cumulative displacement based damage model in
an exponential form assuming that the rate of damage accumulation is proportional to the
existing state of damage in the structure:
e sDi  1i
D s (5.12)
e 1
where
i
Di  c (5.13)
i f
126
where Δi = maximum imposed displacement in the ith cycle, and Δf = displacement capacity
under monotonic loading. A value of 0.1 was recommended by Wang and Shah (1987) for
c while they suggested taking s equal to 1.0 for a well reinforced beam-column joint and -
1.0 for a poorly reinforced joint that is prone to shear failure.
RC members are likely to sustain large displacement reversals during an earthquake
that causes large plastic strains to be accumulated over cycles and eventually result in the
member failure. Jeong and Iwan (1988) assumed a prescribed relationship for a member
that fails due to application of Nf cycles of loading at a given displacement ductility, μ:
N f s  C (5.14)
where s and C are positive empirical constants and are suggested to be taken as 6 and 416,
respectively. Their proposed equation is similar to the classical Coffin (1954)-Manson
(1953) relationship for low-cycle fatigue failure in which the number of cycles leading to
collapse is related to the accumulated plastic strains. Assuming that damage accumulation
is a linear function of the number of cycles with a constant amplitude (ni) and using the
simple rule proposed by Palmgren (1924) and Miner (1945), Jeong and Iwan (1988)
expressed the damage index as:
ni 1
D   ni is (5.15)
i N fi C i
where ni = number of cycles at the ith cycle, Nfi = number of cycles that cause collapse at
the ith cycle, and μi = displacement ductility corresponding to the ith cycle. A damage index
127
of 1 corresponds to failure. Their proposed damage index does not account for the effect
of loading sequence, however considering its simplicity, it can provide a qualitative
estimate of damage in an RC member.
More sophisticated fatigue-based damage indices have also been proposed in which
the number of cycles that cause collapse is related to the plastic or total strains in the
reinforcing bars. By conducting tests on reinforcing bars, Mander et al. (1994) related the
fatigue life of reinforcing bars to their plastic strains as a classical form of Coffin (1954)-
Manson (1953)relationship:
 p  0.08(2N f )0.5 (5.16)
where ϵp = plastic strain in the reinforcing bars, and Nf = number of cycles causing collapse.
Since estimation of plastic strains can be difficult due to the Bauschinger’s effect, a similar
form of this equation was also proposed based on the total strain (Mander & Cheng, 1995):
 t  0.08(2 N f ) 0.333 (5.17)
The damage index is calculated using Miner’s rule (1945):
1
DI   (5.18)
i 2 N fi
where Nfi = number of cycles causing collapse at the ith cycle.
128
Kunnath et al. (1997) correlated Eqs. 5.16 and 5.17 to the results from their tests on
bridge piers under constant-amplitude loading:
 p  0.065( N2 f )0.436 (5.19)
 t  0.06( N2 f )0.360 (5.20)
where N2f =number of cycles to failure. Kunnath et al. (1997) indicated that their proposed
modified equations are more suitable than those derived by Mander an Cheng (1995) for
RC sections as they consider the actual fatigue behavior of reinforcing bars in a RC section
and include damage due to shear and axial stresses as well as loss of confinement.
However, their model could not predict failure in two out of five specimens under cyclic
displacement reversals and in any of the six column specimens under random earthquake
excitations. During their tests, Kunnath et al. (1997) also observed that under low
amplitude displacement cycles, failure is generally controlled by loss of lateral support than
low cycle fatigue of reinforcing bars. Therefore, they suggested that a model that includes
damage corresponding to confinement deterioration as well as low cycle fatigue would be
better suited for RC columns.
To include the effect of lateral restraint provided by cover concrete on the performance
of longitudinal bars, Lehman and Moehle (2000) developed a two-phase damage model
based on the Coffin (1954)-Manson (1953). The first phase in their proposed model
considers the damage in the cover concrete whereas the second phase includes low cycle
fatigue failure in reinforcing bars. The damage index in each phase is defined based on
129
Miner’s rule. Before cover spalling, the damage index is only related to the state of strain
in cover concrete. Once, the cover concrete is completely failed (i.e., Dc =1), damage index
will be represented by fatigue of longitudinal bars.
 c 5
( N f ) c  33( ) (5.21)
 csp
 s 5.5
( N f ) s  0.08( )  0.92 (5.22)
 su
1
Dc   for Dc  1 (5.23)
i ( N f ) ci
1
Ds   for Dc  1 (5.24)
i ( N f ) si
where
(Nf)c = number of cycles to degrade cover concrete
(Nf)s = number of cycles to failure after cover spalling
ϵc = compressive strain in the extreme fiber of core concrete
ϵcsp = compressive strain at cover spalling
ϵs = tensile strain of the longitudinal bar
ϵsu = ultimate tensile strain in the extreme longitudinal bar
Dc = damage corresponding to cover spalling
130
Ds = damage corresponding to low cycle fatigue in the reinforcing bars after cover
spalling
ϵcsp and ϵsu were estimated based on experimental results whereas ϵc and ϵs were
predicted analytically. The two-phase damage index could predict failure in 5 out of 6
column specimens tested by the authors and all of the columns tested by Kunnath et al.
(1997).
Through finite element (FE) analysis of RC columns, Erduran and Yakut (2004)
introduced three categories of damage based on the maximum crack width that is caused
by loading of the specimens. The crack width was calculated based on measured tensile
strains of longitudinal reinforcing bars according to the equation proposed by Frosch
(1999). The maximum calculated crack widths of 0.2, 1, and 2 mm (0.008, 0.039, 0.079
in.) were selected to define three states of damage as negligible, light, and moderate and
were assigned damage scores of 0.5%, 7.5%, and 30%. The fourth (i.e., heavy) damage
state was related to the ultimate ductility capacity and was assigned a damage score of
90%. The ultimate ductility capacity of columns were calculated based on a normalized
parameter ρs / (P/Po) in which ρs represents the volumetric ratio of transverse reinforcement
and P/Po denotes the ratio of the applied axial load to the axial load capacity (P/Po). The
authors had shown formerly that the normalized parameter ρs / (P/Po) significantly affects
the ductility of RC columns.
131
Based on the observed relationship between the drift ratio and damage scores in each
ductility category, Erduran (Erduran & Yakut, 2004) proposed an equation for the damage
state at a target drift ratio (δ):
D ( )  f ( ) g ( ) (5.25)
where f(δ) and g(δ)are defined as:
1
f ( )  1  (5.26)
 b
exp( )
aCsC fy
   
g ( )  0.51  cos( ) if c (5.27)
 cCsC fy  CsC fy

g ( )  1 if c (5.28)
CsC fy
where a,b ,and c are constants that depend on the ductility levels (i.e. magnitude of ρs /(
P/Po)) and Cs and Cfy are correction factors for yield strength of longitudinal bars and
slenderness, respectively.
While their proposed damage index showed high correlation with damage curves in
ATC-40 (1996), it involves several arbitrary constant values and damage scores. Also,
three out of four of the defined damage categories (i.e., negligible, low, and moderate) rely
only on crack widths and hence invalid once cover spalling occurs. Their proposed
equation for determination of the fourth (i.e., heavy) damage state which is based entirely
132
on the calculation of the ultimate ductility is noncumulative and ignores the effects of low
cycle fatigue and loading history. Therefore, while their proposed equation might be useful
for design practices, its ability to provide an estimation of the damage state in RC columns
under large cyclic load reversals is questionable.
5.3.2.2. Energy-based cumulative damage indices
Dissipated energy during loading has also been used as a measure for state of damage
in RC members by researchers. Gosain et al. (1977) proposed a work index (Iw) to estimate
damage in RC beams and columns with dominant failure mode in shear:
n
pi  i
Iw   (5.29)
i 1 py y
where
n=number of cycles in which p/py ≥ 0.75
pi, Δi= lateral load at peak displacement and corresponding peak displacement during
the ith cycle
py, Δy= lateral load and corresponding displacement at yielding of reinforcing bars
Only cycles satisfying lateral load ratios with p/py ≥ 0.75 were considered. As the upper
bound p/py was estimated as 1.25, the authors also suggested simplification of the work
index as the summation of normalized deflections (i.e., Δi/Δy). As also indicated by the
authors, the proposed work index for a specimen that is loaded two times to a normalized
deflection of 10 would be equal to that for a specimen that is loaded four times to a
133
normalized deflection of 5 while the former would likely experience more damage. Thus,
the damage index as proposed in Eq. 5.29 did not correlate well with the experimental
database that the authors used. Therefore, they proposed modifications to account for the
effect of shear span and axial load.
A more sophisticated cumulative energy index was developed by Kratzig and
Meskouris (1987). In their proposed method, dissipated energy terms were calculated
separately for the primary half-cycles (PHCs) and follower half-cycles (FHCs) in the
positive and negative directions (Figure 5.2). A PHC was defined as the half cycle with a
displacement larger than that in the previous cycles whereas a FHC was considered as any
following half cycles of equal or smaller deformation amplitudes. For the positive
deformation of the specimen, the accumulated damage is defined as:
D 
E 
pi   Ei
(5.30)
E f   Ei
where
E+pi = energy in a PHC with a positive deformation amplitude
E+i= energy in a FHC with a positive deformation amplitude
E+f= energy in a monotonic test to failure for loading in the positive direction
Accumulated damage during negative deformation part is calculated in a similar approach
and yields the overall damage as:
D  D  D  D D (5.31)
134
Their proposed model includes both the damage due to large displacement excursions
as well as the fatigue-type damage due to repeated loading at smaller displacement
amplitudes. Calculated damage indices could properly predict failure for nine beams in
their study.
Figure 5.2. Definition of Primary and Follower Half Cycles (Kratzig & Meskouris, 1987)
5.3.2.3. Combined cumulative damage indices
A popular damage index was developed by Park and Ang (1985) assuming a linear
combination of normalized displacement to characterize noncumulative effects and
normalized dissipated energy to characterize cyclic loading effects:
D
m

 dE (5.32)
u Fy  u
where
135
Δm = maximum displacement
Δu = ultimate displacement capacity under monotonic loading
Fy = yield strength
dE= dissipated energy increment
β = strength deterioration factor due to cyclic loading
By regression analysis of 261 tests on RC beams and columns, Park and Ang (1985)
proposed an equation to estimate β :
l P 
  (0.447  0.073  0.24  0.31l )0.7 s (5.33)
d f cbd
where
l = length of specimen
d = distance from extreme compression fiber to the centroid of the longitudinal tension
reinforcement
l/d = shear span ratio, replaced by 1.7 if l/d < 1.7
P/f′cbd = normalized axial stress, replaced by 0.2 if P/f′cbd < 0.2
ρl = longitudinal reinforcement ratio (%), replaced by 0.75 if ρl < 0.75
ρs = confinement reinforcement ratio (%).
The authors subsequently proposed a damage classification based on observed data as
represented in Table 5.1 (Park, Ang, & Wen, 1987). Williams and Sexsmith (1995)
indicated difficulties in the estimation of ultimate displacement capacity (Δu) and strength
136
deterioration factor (β). They also criticized the damage scale for being nonlinear with the
severe damage state corresponding to the calculated damage index of magnitudes beyond
0.4.
Kunnath et al. (1992) proposed an alternative form of Park-Ang (1985) damage index
based on moment and curvature instead of load and displacement which was later used by
Stone and Taylor (1993) for investigation of the damage state in circular bridge piers.
However, Stone and Taylor (1993) indicated that for the case of simple structures such as
cantilever bridge piers, the results using moments and curvature were similar to those
obtained by using the force and displacement in the original Park-Ang (1985) equation.
Assuming a damage index equal to unity at ultimate displacement capacity, Eq. 5.32
can be written as:
m E
1  h (5.34)
u Fy u
where Eh= dissipated energy during loading. Under collapse condition due to monotonic
loading (i.e., when Δm = Δu), Eq. 5.34 results in the total dissipated energy to be zero.
However, during monotonic loading to failure, the specimen will exhibit inelastic behavior
and dissipate energy. To correct the issue, Chai et al. (1995) suggested removing the plastic
strain energy due to monotonic loading.
137
Table 5.1. Damage Classification for Park-Ang Model
Damage
Index Value (D) Description of the Observed Damage
Category
D < 0.1 Slight localized cracking
minor cracking throughout, partial

0.1 ≤ D < 0.25 Minor
crushing of concrete in columns
0.25 ≤ D < 0.4 Moderate extensive cracking, localized spalling
extensive crushing of concrete, disclosure

0.4 ≤ D < 1 Severe
of reinforcing bars
D≥1 Collapse Collapse
5.4. EXPERIMENTAL PROGRAM
Seven full-scale RC columns were tested as part of this research. All specimens except
SP5 featured a 36×28 in. cross section with 16 No. 9 longitudinal bars and transverse hoops
made of No. 5 bars and spaced at 5 in. along the height of the specimens. Specimen SP5
featured a smaller cross section (28×28 in) and incorporated 12 No.8 reinforcing bars
which were confined by transverse hoops at 5 in. spacing. The columns were subjected to
distinct lateral loading protocols as summarized in Table 5.2.
Each test started by application of an axial load. The magnitude of the applied axial
loads was set to 756 and 1176 kips, equivalent to an axial load ratio (P/f′cAg) of 0.15 and
0.3 for larger (i.e. 36×28 in.) and smaller (i.e. 28×28 in.) columns, respectively. While
keeping the axial load level constant, the specimens were subjected to lateral loading
protocols in the form of displacement excursions as represented in Figures 5.3 to 5.9.
138
Application of the lateral loadings on specimens continued until either the actuators
reached their stroke capacity or the specimens lost more than 80% of their peak lateral load
capacity in either direction.
The first specimen was subjected to a monotonic displacement until the stroke capacity
of the actuators exhausted at 15.61 in. of the crosshead displacement. To estimate the
reserve capacity of the specimen, the monotonic loading in the positive direction was then
followed by loading in the opposite direction and formed a single cycle (Figure 5.3). The
specimen response during the first test was utilized to estimate the peak deformation
capacity of the columns and to calibrate controllers of the loading crosshead at the MAST
Lab.
The loading protocol on specimens SP2 and SP5 were designed following the
procedures in the ACI 374-05 (2005). These loading protocols incorporated a series of
progressively increasing displacement cycles. The magnitudes of displacement cycles were
gradually increased with an approximate growth of 25% to 50 % in the drift ratio (Figures
5.4 and 5.7 ). Three cycles were applied in each drift level and were followed by a small
cycle at a magnitude equal to one third of that in the preceding cycle group. Initial cycles
were set to drift ratios within the linear elastic behavior of the specimens. Cyclic
displacement reversals on specimens SP2 and SP5 followed the aforementioned pattern
until these specimens significantly lost their peak lateral load capacity. These tests stopped
after 56 and 52 cycles and following the application of cycles at 8.91% and 7.13% drift
ratios for specimens SP2 and SP5, respectively.
139
Table 5.2. Applied Lateral Loading to the Specimens
Description of Loading δy δFmax δult Fy Fmax Fult Fult /

Specimen
Loading Protocol Direction (%) (%) (%) (kips) (kips) (kips) Fmax
Pos 0.80 3.10 12.09 238 290 212 0.73

SP1 Monotonic
Neg -4.00 -11.43 -239 -172 0.72
Symmetric cyclic (ACI Pos 0.65 1.07 6.91 258 314 58 0.19
SP2 374) Neg -0.70 -1.09 -6.91 -235 -278 -53 0.19
SP3 374)+monotonic push Neg -0.74 -3.51 -10.84 -235 -278 -64 0.23
SP4 374)+monotonic push
140
Neg -0.76 -1.51 -10.85 -241 -288 -21 0.07
SP5 374) Neg -0.60 -1.08 -5.53 -137 -168 1 -0.01
Pos 0.90 3.44 10.43 232 280 52 0.19
SP6 Near-collapse
Neg -0.71 -1.09 -11.62 -214 -257 -198 0.77
Pos 0.91 3.52 5.53 227 279 57 0.21
SP7 Biaxial
Neg -0.98 -3.52 -5.53 -218 -268 -33 0.12
Specimens SP3 and SP4 initially sustained displacement cycles that were designed
according to the same pattern as of those described for specimens SP2 and SP5. However,
these cycles were followed by monotonic push in positive and negative displacement
directions after applications of 5.67% and 3.66% drift cycles for specimens SP3 and SP4,
respectively (Figures 5.5 and 5.6). Unlike all other specimens, the loading protocol on
specimen SP6, the near-collapse protocol, incorporated symmetric and asymmetric cycles
followed by a final monotonic displacement push (Figure 5.8). The near-collapse loading
protocol was developed from time history analysis on low- and high-rise buildings
subjected to far-field earthquake motions.
The biaxial loading protocol on specimen SP7 was designed based primarily on the
same pattern as of that for SP2, but along both X and Y directions. Each full cycle in this
loading protocol can be represented by load paths A and B as noted in Figure 5.9(a) and
Figure 5.9(b) to form a full cycle at a target drift level (Figure 5.9(c)). The lateral loading
on the specimen started by application of a displacement along the positive Y direction
similar to that initially applied to specimen SP2. The specimen was then subjected to the
target displacement along X direction while it was unloaded along Y direction. The resulted
applied displacements along negative X and Y directions completed a full cycle at a desired
drift level. The aforementioned pattern continued to create gradually increasing
displacements along both directions. Loading of the specimen terminated after 52 full
cycles and following the application of 7.13% drift cycles (Figure 5.9(d)).
141
12
8
4
Drift (%)
0
-4
-8
-12
0 1 2 3
Cycle
Figure 5.3. Applied Loading Protocol to Specimen SP1
4
Drift (%)
-4
-8
0 10 20 30 40 50 60
Cycle
12
8
4
Drift (%)
0
-4
-8
-12
0 10 20 30 40 50 60
Cycle
142
12
8
4
Drift (%)
0
-4
-8
-12
0 10 20 30 40 50
Cycle
4
Drift (%)
-4
-8
0 10 20 30 40 50 60
Cycle
12
8
4
Drift (%)
0
-4
-8
-12
0 2 4 6 8 10
Cycle
143
(a) (b)
0.25 6
2
Drift Y (%)
Drift Y (%)
0 0
-2
-4
-0.25 -6
-0.25 0 0.25 -6 -1 4
Drift X (%) Drift X (%)
(c) (d)
144
5.5. MEASURED RESPONSE OF COLUMN SPECIMENS
Application of the lateral loads to the specimens were followed by initiation of flexural
cracking on the tension face of the columns at approximately 0.25% drift ratios. Once the
loading continued (in the case of SP1) or the loading direction was reversed (in the case of
cyclic tests), flexural cracks were formed on the other face as well. Flexural cracking was
accompanied by formation of flexure-shear cracks on the other two faces of the columns
that were parallel to the loading direction. Longitudinal bars started to yield at
approximately 0.5% drift ratios. Due to cracking of the cover concrete and yielding of the
reinforcing bars, the initial stiffness of the columns degraded and they exhibited inelastic
behavior. Progression of damage was accompanied by further stiffness degradation in the
specimens. However, the specimens could still sustain larger magnitudes of lateral load
due to the confining pressure of the hoops and strain hardening of the longitudinal bars.
After reaching the peak lateral load capacity, further loading of the specimens was followed
by progression of damage to the core concrete, yielding of transverse ties and buckling and
fracture of longitudinal bars that caused the specimens to lose strength and stiffness.
Lateral displacement at the top of the specimens (i.e., interface of the top block and
loading crosshead) and the applied lateral loads to the specimens were recorded during
each test. The force-drift ratio and force-deformation response of the specimens is
illustrated in Figures 5.10 to 5.17 . To estimate the yield displacement and corresponding
yield force for each specimen, a secant line was extended from the origin to a point on the
force-displacement cyclic envelope corresponding to a lateral load of 70% of the peak
capacity (0.7Fmax ). The secant line was further extended to intersect with the horizontal
line at Fmax (Point A in Figure 5.18). The yield displacement (Δy) is defined as the
145
horizontal ordinate of point A, and the corresponding yield force (Fy) is determined by
extending a vertical line down from Point A until it intersects the force-deformation cyclic
envelope at Point B. This method has been previously applied by other researchers (Bae &
Bayrak, 2008; Sivaramakrishnan, 2010) to estimate the yield displacement and yield force.
Magnitudes of yield displacement and yield force of each specimen along with those at
peak strength and ultimate loading stage of each specimen are represented in Table 5.2.
Except for the test of SP1 which was terminated after exhausting the full stroke
capacity of the actuators of the loading crosshead at an ultimate drift ratio (δult) equal to
12.1%, the rest of the specimens exhibited significant strength loss by the end of the tests
(Figures 5.11 to 5.17). A residual strength ratio (Fult/Fmax) of 73%, 19%, 23%, 7%, 0%,
19%, and 12% was measured at the end of the tests for specimens SP1 to SP7, respectively
(Table 5.2). Severe strength loss in the specimens was accompanied by extensive damage
to the reinforcing bars and the core concrete.
146
Figure 5.10. Force-Deformation Behavior of Specimen SP1
147
148
149
Figure 5.16. Force-Deformation Behavior of Specimen SP7 along Y Direction
Figure 5.17. Force-Deformation Behavior of Specimen SP7 along X Direction
150
Figure 5.18. Estimation of the Yield Displacement and Yield Force
5.6. DAMAGE ANALYSIS OF TESTED SPECIMENS
The state of damage during loading of the specimens was estimated using three non-
cumulative and five cumulative damage indices. The noncumulative damage indices
include displacement ductility, and the models proposed by Lybas (Lybas & Sozen, 1977),
and Roufaiel (Roufaiel & Meyer, 1987) which are expressed by Eqs. 5.3, 5.4, and 5.6,
respectively. The cumulative damage indices that were used in this study include models
proposed by Banon (Banon et al., 1981), Stephens (Stephens & Yao, 1987), Wang (Wang
& Shah, 1987), Kratzig (Kratzig & Meskouris, 1987), and Park (Park and & Ang, 1985),
and these indices are quantified using Eqs. 5.8, 5.11, 5.12, 5.30 and 5.32. Among these
indices, displacement ductility and estimated damage indices based on Lybas and Banon
models are not scaled. Thus, their results, which will be presented in the next section, are
not directly comparable to the rest of the models.
151
Estimated damage values for each specimen at the end of the tests are included in
Table 5.3. The column specimens under the loading protocols exhibited various strength
loss at the end of the tests (Table 5.2) so that calculated damage values for various
specimens were not comparable. Therefore, to study the effects of applied loading
protocols, final damage values should be calculated at a similar state (e.g. 20% or 50%) of
strength loss. The computed damage to specimen SP2 based on five damage models is
represented in Figure 5.19. The vertical dashed line in this figure corresponds to the state
of 20% strength loss in the specimen. At this state, RC columns often require repair to re-
establish their capacity and hence this state is commonly considered as the failure of
specimens in terms of their useful limit. In this study, the state of 20% strength loss is
referred to as the Limit of Useful Capacity at 80% strength (LUC80). In the case of
specimen SP2, a strength reduction of 9% and 31% was recorded following the application
of peak positive displacements during the 49th and 50th cycles, respectively. Therefore,
LUC80 was estimated to occur between these successive cycles.
Among all calculated damage indices for specimen SP2 in Figure 5.19, the one
proposed by Wang shows a slow progression until its final magnitude of 1.01 at the end of
the test. However, the corresponding damage index at LUC80 (i.e. intersection of the
damage line with vertical dashed line related to LUC80), is estimated as 0.6 representing
a damage state which is far below failure. Models by Roufaiel and Kratzig suggest a rapid
progression of damage at early stages of loading while their estimated damage values
remain approximately constant towards the end of the test. The two models estimate a
damage index of 0.93 at LUC80. The models by Stephens, and Park suggest a medium rate
of damage progression until the 40th cycle. These models suggest a rapid increase in
152
Table 5.3. Estimated Damage Values at the End of Each Test
Column Test Specimen

Damage Index
Model
SP1 SP2 SP3 SP4 SP5 SP6 SP7
Stepehens-Yao 0.465 1.679 0.823 0.826 1.688 0.491 1.106

Wang-Shah 0.041 0.466 0.317 0.246 0.455 0.082 0.342
Roufaiel-Meyer 0.975 1.000 1.007 1.019 1.036 1.012 1.008
Kratzig-Emskouris 0.973 0.892 0.934 0.940 0.915 0.852 0.812
Park-Ang 1.010 1.453 1.475 1.309 1.220 0.957 1.459
Banon et al. 15.148 141.377 87.523 72.364 139.687 24.496 76.136
Lybas-Sozen 17.035 46.779 17.480 64.799 31.012 51.817 24.062
Ductility 15.148 10.586 15.324 15.434 10.535 11.645 6.086
damage progression towards the end of the tests which is reasonable because the damage
is most likely to occur at an accelerating rate with drift as verified during the tests at the
MAST Lab. Stephens and Park models estimate failure after 42 cycles, which over-predicts
the failure of the specimens defined on the basis of 20% strength loss.
Since the loading protocols for most of the tested column specimens at the MAST Lab
were continued until they exhibited severe strength loss, failure of the specimens was
defined based on 50% strength loss (i.e., LUC50) to better estimate the state of damage at
later stages of loading. For specimen SP2, LUC50 is located between cycles 50 and 51
(Figure 5.20). As represented in Figure 5.20 shows that while the Wang model still
underestimates the extent of damage, the Stephens model correctly predicts failure of SP2.
The Roufaiel model also suggests a damage value of 0.96 at LUC50, thus it properly
represents a stage that is quite close to failure. The Roufaiel model also shows that it is
more deeply affected by accumulation of plastic deformations due to an increase in the
magnitude of applied displacements than by cycling at a constant displacement, and the
153
estimated damage index values are closer to the ones by the Kratzig model than the rest of
the models. However, Kratzig model suggests a smoother evolution of damage through the
entire loading protocol. The model by Park still over-predicts damage to specimen SP2.
Defining failure based on LUC50 results in an overall reduction in the calculated damage
values by Park model and thus a lower rate of over-prediction of damage. The model by
Park shows to increase in accuracy as the degree of damage increases.
Calculated damage indices for specimens SP3 and SP4 are shown in Figures 5.21 and
5.22. The progression of damage for specimens SP3 and SP4 was similar to that of SP2
until application of the final monotonic push (i.e., the last cycle for SP3 and the last two
cycles for SP4). Large monotonic displacements that were applied to specimens SP3 and
SP4 at the end of their loading protocols resulted in a sudden increase in the magnitudes of
estimated damage indices based on the Park model. The final monotonic push also resulted
in a large increase in the damage indices calculated by the Stephens, Roufaiel, and Kratzig
models for specimen SP4 which had exhibited less strength loss during preceding cyclic
loading excursions compared to SP3. The ultimate strength loss for the positive loading
direction of SP3 was less than 50% (Table 5.2), and hence the ultimate damage indices
were extrapolated to obtain the estimated values at 50% strength loss. At the intersection
with LUC50 line, calculated damage values by the Stephens and Kratzig models in the case
of SP3, and the Roufaiel and Kratzig models for SP4, were closer to the unit value (DI =
1) than those calculated based on the rest of the models. Estimated damage indices by Park
are highly affected by the final large displacement cycles. Considering the load-
deformation response of specimens SP3 and SP4 (Figures 5.12 and 5.13) confirms that
these specimens lost most of their strengths during the last cycles that carry, by far, the
154
majority of the damage. Therefore, calculated damage indices by Park model seems to be
closer to the actual performance of column specimens than those calculated by other
models. The model by Wang, on the other hand, was not sensitive to the final displacement
excursions.
The evolution of damage indices in specimen SP5 were approximately similar to that
in specimen SP2 since both specimens were subjected to progressively increasing
displacement cycles until failure. As the monotonic test on specimen SP1 was only
representative of specimens incorporating larger cross-sectional dimensions (i.e., 36×28
in.), the displacement corresponding to monotonic failure of specimen SP5 that featured a
smaller cross-sectional dimensions (i.e., 28×28 in.) was estimated by comparison of the
ultimate drift ratios of specimens SP2 and SP5. In specimen SP5, the state of 50% strength
loss (i.e., LUC50) was estimated to occur between cycles 49 and 50 (Figure 5.23).
Calculated damage indices from the Roufaiel, Kratzig, and Park models suggest a better
estimate of failure in specimen SP5 with corresponding damage values of 0.99, 0.91, and
1.11 at LUC50. On the other hand, estimated damage values by the Stephens and Wang
models were over- and under-predictive, respectively.
Except the Wang model, calculated damage indices by the rest of the models for
specimen SP6 are significantly affected by the applied monotonic push (Figure 5.24). The
results from the Roufaiel, Park, and Kratzig models were better in terms of failure
prediction than those based on Stephens and Wang models, which were under-predictive.
Finally, the evolution of damage indices is specimen SP7 followed an approximately
similar pattern as of that in specimen SP2 (Figure 5.25). Among all calculated damage
values for specimen SP7, those by the Roufaiel model indicated a better estimation of
155
failure at LUC50. Results from the Stephens and Kratzig models seem to underestimate
damage to specimen SP7. However, results from Park model are more realistic as the
dissipated energy term in this model were calculated based on total energy that was
dissipated during loading of the specimen along X and Y directions while the other models
ignore the damage caused by loading along X direction. For these models to be realistic
indicators of damage in specimen SP7, calculated damage values in both directions should
be combined. Qiu et al. (Qiu, Li, Pan, & Qian, 2002) proposed four methods to include the
effect of biaxial loading in the estimation of damage values by Park model. One of their
proposed method was to include the total dissipated energy along both loading directions
which was used to estimate damage values in specimen SP7. For seven column specimens
under biaxial loading in their study, the authors showed that results of all four proposed
methods were very close.
156
Figure 5.19. Calculated Damage Indices for Specimen SP2 Based on LUC80
157
158
159
5.7. CORRELATION OF CALCULATED AND OBSERVED

DAMAGE
It is essential for damage models to be calibrated against damage observed in
structures so that they can provide a measure of structural integrity after seismic events and
be used for retrofit decision making and seismic loss estimation. The results of such
calibrations are typically provided in tables that relate levels of observed damage to the
corresponding estimated damage values. Such calibration was carried out by Park and Ang
(1985) against a large number of experimental data. The Applied Technology Council
(ATC-40, 1996) also defines damage states based on visual observations after earthquakes
to help conduct seismic evaluations and preparing retrofit plans.
160
To better identify the correlation of estimated damage values for the different models
with observed damage to the specimens tested at the MAST Lab, the state of damage in
the test columns were categorized based on visual observation. Five visual damage
categories were defined by considering the extent of required repair and close to what were
proposed by Park et al. (1987) and Williams et al. (1997). At peak displacements
corresponding to the first and third cycles at a given drift level for each column specimen,
eight still images were captured at two elevations. These pictures were analyzed to define
the five visual damage categories that are summarized in Table 5.4. Also, images
representative of the beginning of visual damage categories II, “Light”, III, “Moderate”,
IV, “Severe”, and V, “Failure”, for specimen SP2 are shown in Figures 5.26 to 5.29. The
start of visual damage category I, “Negligible”, corresponds to the undamaged state of the
specimens and is not shown in these Figures. The measured responses of the specimens
during the tests indicate that the first three visual damage categories (“Negligible”, “Light”,
“Moderate”) were accompanied by little or no strength loss, the column specimens
exhibited noticeable reductions in lateral load capacities at the onset of the “Severe” (IV)
and “Failure” (V) visual damage categories (Table 5.5).
Corresponding damage index values at the beginning of visual damage categories
II-V were calculated using the damage models discussed in the previous section (Figures
5.30 to 5.34). Additionally, damage values based on the Lybas model (Lybas & Sozen,
1977), Banon’s normalized cumulative displacement model (Banon et al., 1981), and
displacement ductility are also computed (Figures 5.35 to 5.37). The corresponding
damage values are also included in Table 5.6. Among these models, calculated damage
values by the Stephens, Wang, and Park models are clearly distinct for each damage
161
Table 5.4. Visual Damage Categories Based on Test Observations
Damage
Category Description of the Observed Damage
State
None or minor hairline flexural or flexural-
I Negligible
shear cracks
Widespread cracking throughout and
II Light
initiation of cover spalling
Widespread cracks > 1/8 in. wide and cover
III Moderate
spalling around the corners
Extensive wide cracking and cover spalling
IV Severe throughout the column and/or single bar
exposure
Extensive cover spalling and bar exposure,
V Failure
buckling/fracture of longitudinal bars
category. However, the Wang model suggests damage values that are barely larger than
0.32 at failure. On the other hand, estimated damage values by the Park model
corresponding to the start of visual category V for “Failure” are greater than unity in four
out of seven specimens. The Roufaiel and Kratzig models estimate damage values that are
able to capture failure while their results seem to over-predict early damage in the
specimens due to cycling. The model proposed by Lybas results in damage values that are
not well distributed for different damage categories. Estimated damage values based on the
Lybas model also seem to be significantly affected by large displacements that were
applied during monotonic test (i.e., test of specimen SP1) and the last cycles of loading on
specimens SP3, SP4, and SP6 while insensitive to accumulated damage over cycles.
Calculated displacement ductilities and normalized cumulative displacements (i.e., damage
index values computed using the Banon model) are also separated over different visual
damage categories with the latter more evenly distributed. However, both of these
measures are not scaled which makes interpretation of their estimated values difficult.
162
Figure 5.26. Observed Damage at the Beginning of “Light” Damage Category
163
Figure 5.27. Observed Damage at the Beginning of “Moderate” Damage Category
164
Figure 5.28. Observed Damage at the Beginning of “Severe” Damage Category
165
Figure 5.29. Observed Damage at the Beginning of “Failure” Damage Category
166
Table 5.5. Measured Strength Loss at the Onset of Visual Damage Categories
Damage Category
Specimen
Light Moderate Severe Failure
SP1 0 0 3.1 *
SP2 1.5 1.6 3.4 7.3
SP3 0.0 0.3 2.5 27.4
SP4 0.0 0.0 5.2 80.3
SP5 1.0 4.8 9.8 24.0
SP6 0.0 0.0 13.5 81.4
SP7 0.0 0.0 0.0 35.9
*Test was terminated prior to achieving this damage category
167
1
Stepehens‐Yao
0.8
Damage Index
0.6
0.4
0.2
0
Figure 5.30. Estimated Damage Index Values Corresponding to Different Visual
Damage Categories Based on Stephens Model
0.4
Wang‐Shah
0.3
Damage Index
0.2
0.1
0
Damage Categories Based on Wang Model
168
1.2
Roufaiel‐Meyer
1
Damage Index
0.8
0.6
0.4
0.2
0
Damage Categories Based on Roufaiel Model
1.2
Kratzig‐Meskouris
1
Damage Index
0.8
0.6
0.4
0.2
0
Damage Categories Based on Kratzig Model
169
1.6
1.4 Park‐Ang
1.2
Damage Index
1
0.8
0.6
0.4
0.2
0
Damage Categories Based on Park Model
80
Lybas‐Sozen
60
Damage Index
40
20
0
Damage Categories Based on Lybas Model
170
100
Banon et al.
80
Damage Index
60
40
20
0
Damage Categories Based on Banon Model
18
Ductility
15
Damage Index
12
0
Damage Categories Based on Displacement Ductility
171
Table 5.6. Estimated Damage Values for Various Visual Damage Categories
Column Test Specimen

Damage Index Model
Visual Damage Category II - Light

Stepehens-Yao 0.032 0.063 0.065 0.111 0.003 0.020
Wang-Shah 0.037 0.054 0.054 0.068 0.009 0.037
Roufaiel-Meyer 0.345 0.465 0.443 0.607 0.220 0.139
Kratzig-Emskouris 0.300 0.382 0.390 0.525 0.088 0.245
Park-Ang 0.087 0.136 0.138 0.170 0.072 0.114
Banon et al. 8.777 13.483 13.623 22.760 1.218 3.605
Lybas-Sozen 1.510 1.756 1.763 2.444 1.095 1.156
Visual Damage Category III - Moderate

Stepehens-Yao 0.200 0.192 0.222 0.313 0.111
Wang-Shah 0.104 0.104 0.118 0.133 0.076
Roufaiel-Meyer 0.688 0.649 0.702 0.767 0.469
Kratzig-Emskouris 0.605 0.580 0.629 0.714 0.454
Park-Ang 0.298 0.287 0.361 0.330 0.273
Banon et al. 34.857 29.918 33.975 46.590 14.743
Lybas-Sozen 3.064 2.619 3.177 3.908 1.829
Visual Damage Category IV - Severe

Stepehens-Yao 0.465 0.484 0.469 0.340 0.549 0.003 0.218
Wang-Shah 0.041 0.189 0.189 0.194 0.205 0.045 0.118
Roufaiel-Meyer 0.975 0.803 0.780 0.992 0.875 0.970 0.634
Kratzig-Emskouris 0.973 0.763 0.744 0.934 0.819 0.820 0.594
Park-Ang 1.010 0.631 0.603 1.217 0.560 0.903 0.486
Banon et al. 15.148 64.001 54.876 56.930 69.529 12.851 25.483
Lybas-Sozen 17.035 4.681 4.033 17.071 6.156 11.156 2.584
Visual Damage Category V - Failure

Stepehens-Yao 0.827 0.823 0.826 0.835 0.491 0.727
Wang-Shah 0.287 0.317 0.246 0.278 0.082 0.253
Roufaiel-Meyer 0.891 1.007 1.019 0.927 1.012 0.904
Kratzig-Emskouris 0.857 0.934 0.940 0.871 0.852 0.781
Park-Ang 1.087 1.475 1.309 0.788 0.957 1.161
Banon et al. 92.678 87.523 72.364 91.326 24.496 57.878
Lybas-Sozen 7.666 17.480 64.799 8.908 51.817 6.134
172
5.8. EFFECT OF LOADING PROTOCOLS
One of the primary goals of this study was to characterize the applied lateral loading
protocols to the specimens in terms of severity of observed damage. Identification of the
worst loading scenario that a specimen is likely to experience during an earthquake will be
essential in the generation of capacity curves and determination of structural demands for
performance-based design. While all of the loading protocols applied to the specimens
tested at the MAST Lab resulted in significant strength reduction and stiffness
deterioration, the column specimens subjected to these loading protocols exhibited distinct
behavior in terms of ultimate displacement ductility, rate of softening, energy dissipation
and damage propagation.
To investigate the effect of the loading protocols, the cyclic load-deformation
envelopes for each column specimen were defined by connecting the peak lateral
displacements at the first cycle of each drift level. Such calculation for specimen SP2 is
shown in Figure 5.38. Comparison of the cyclic envelopes indicates that among all seven
specimens, specimen SP5, which had a larger axial load ratio (P/f′cAg = 0.3), and specimen
SP7, which was subjected to biaxial loading, exhibited the smallest drift capacities (Figure
5.39). Specimen SP2 also reached its ultimate strength at a drift level smaller than those
for specimens SP3, SP4, and SP6 which lost most of their strength during the final
monotonic push at the end of their loading protocols. The displacement during the
monotonic push applied to these specimens resulted in rapid strength reduction and
stiffness deterioration, while the softening occurred gradually for specimens subjected to
only cyclic load reversals (SP2, SP5, and SP7). The effect of the lateral loading protocols
on the initial stiffness and peak load capacity of the specimens was insignificant.
173
The extent of damage to the specimens due to loading can be characterized by the
energy that they dissipated during the tests. Accumulated dissipated energy during loading
of the specimens is shown in Figure 5.40. Since the strength loss at the end of the test was
different for each of the specimens, estimated energy dissipations of specimens SP2, SP4,
SP5, SP6 and SP7 were truncated at the state of 50% strength loss (i.e., LUC50) so that the
ultimate magnitudes for these specimens can be comparable. On the other hand, the energy
dissipation of specimen SP1 and SP3 at LUC50 were obtained by extrapolation, as these
specimens exhibited a final strength loss that was less than 50% of the peak lateral load
capacity during loading of the specimens in the positive direction. As illustrated in Figure
5.40, column SP7 showed larger magnitudes of energy dissipation than did the other
specimens, but reached a final value equal to that for specimen SP2. Calculated energy
dissipation for specimen SP7 were obtained by summation of the corresponding values
during loading of the specimen along X and Y directions. Specimen SP5 that featured a
smaller cross-sectional dimension dissipated less energy than all other specimens except
SP6 which dissipated energy mainly over one cycle. Energy dissipation in specimen SP6
was basically quite similar to that of specimen SP1 which was under monotonic loading.
The ultimate monotonic push that was applied to specimens SP3 and SP4 resulted in a rapid
increase in the dissipated energy. However, at LUC50, both of these specimens exhibited
less dissipated energy than specimen SP2 that sustained more cyclic displacement
excursions.
174
Figure 5.38. Generation of Cyclic Envelope for Specimen SP2
Figure 5.39. Force-Deformation Cyclic Envelopes for All Tests
175
Figure 5.40. Calculated Energy Dissipation until Occurrence of 50% Strength Loss
Energy dissipation during loading was accompanied by the evolution of damage in the
specimens. Calculated damage values based on the Lybas (Lybas & Sozen, 1977) model
are shown in Figure 5.41. Specimen SP5 shows a higher rate of stiffness reduction during
cyclic loading reversals. However, the applied monotonic displacement to specimens SP1,
SP3, SP4, and SP6 results in major stiffness deterioration such that the calculated damage
values at 50% strength loss (i.e., LUC50) of these specimens are significantly larger than
those in cyclic tests. Estimated damage indices by the Stephens, Roufaiel and Kratzig
models for all specimens, which are given in Figures to 5.44, suggest a higher extent of
damage in specimen SP5 than specimen SP2 that experienced similar displacement
excursions, but under a lower axial load ratio. On the other hand, the monotonic push which
was applied during loading of specimens SP3, SP4, and SP6 also caused a major increase
176
in the calculated damage values by the Roufaiel and Kratzig models. Specimen SP1 which
was under monotonic loading as well as specimen SP6 which experienced the least number
of cycles exhibit final damage values based on the Roufaiel model that are close to those
of all other specimens except specimen SP3. The Kratzig model, on the hand, suggests a
smaller damage value for SP6, and the Stephens model does not predict failure for SP6 at
all. Calculated damage values by the Stephens, Roufaiel and Kratzig models do not capture
the extent of damage in specimen SP7, which was subjected to the biaxial loading protocol,
when only Y-direction response is used for estimating the damage indices. To obtain a
more realistic value for specimen SP7, damage indices should be calculated in both loading
directions and their effects combined. However, the combination of damage that occurs
from loading in two orthogonal directions is challenging, especially if a unit value is
expected at failure. Among the damage models considered, the one proposed by Park can
inherently capture the effect of loading in two directions when the total energy dissipation
during loading in both directions is taken into account. Calculated damage values obtained
from the Park model (Figure 5.45) suggest a higher extent of damage in specimen SP7 than
all other specimens except SP3.
177
Figure 5.41. Damage Values by Lybas Model until Occurrence of 50% Strength Loss
Figure 5.42. Damage Values by Stephens Model until 50% Strength Loss
178
Figure 5.43. Damage Values by Roufaiel Model until 50% Strength Loss
Figure 5.44.Damage Values by Kratzig Model until Occurrence of 50% Strength Loss
179
Figure 5.45. Damage Values by Park Model until Occurrence of 50% Strength Loss
5.9. CONCLUSIONS
The extent of damage during tests of seven RC columns subjected to distinct loading
protocols was estimated using nine noncumulative and cumulative damage models. In
addition, observed damage to the specimens during each test was categorized into five
visual damage categories and used as a measure to validate the calculated damage indices.
Among all damage index models that were studied, the model by Park and Ang (1985)
provided an overall better estimate of actual damage in specimens. The observed and
calculated damage were used along with performance characteristics of the columns based
on force-deformation envelope, and energy dissipation to assess the applied loading
protocols. The results of this study indicate that the column specimen subjected to biaxial
loading (i.e., SP7) experienced more damage during loading. For this specimen, results
180
from Park and Ang (1985) model were closer to the actual state of damage in the specimen
as it can directly take the effect of biaxial loading into account provided that the calculated
energy dissipation values include the hysteretic energy dissipated during loading in both
directions. The study also shows that column specimen SP5, which had a similar loading
protocol as of that for SP2, but under a larger axial load ratio, sustained a more severe
extent of damage. In addition, the monotonic push displacement that was applied following
a number of cyclic displacement reversals resulted in a rapid strength reduction in the
specimens and a higher extent of damage as of that in the case of pure cyclic or monotonic
displacements.
181
CHAPTER 6
SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS
6.1. SUMMARY
A series of seven full-scale reinforced concrete columns were tested at the MAST Lab
as a part of this study to increase the understanding of the performance of typical columns
in high-rise buildings that are designed using current code requirements and then subjected
to extreme seismic events. The columns featured cross-sectional dimensions that were
larger than almost all columns tested previously under simulated seismic loadings and were
subjected to distinct lateral loading protocols including a monotonic five cyclic uniaxial
and biaxial, and a near-collapse loading protocol. To investigate the post-peak behavior of
the columns, loading of the specimens continued beyond the common stopping point
during previous tests. During these applied loading protocols, the specimens exhibited
significant strength reduction (i.e., 80%) and stiffness deterioration.
Results from these tests were considered along with those from previous tests on RC
columns to study the effects of cross-sectional size of column specimens on their seismic
performance. To this end, five parameters were considered necessary to characterize
182
column performance, and these include moment capacity, effective stiffness, drift capacity,
displacement ductility, and longitudinal bar buckling. With the test data and these five
parameters, the effect of cross-sectional depth of columns on each parameter was studied
in detail.
The loading capabilities at the MAST Lab enabled the testing of full-scale column
specimens at larger drift ratios than those applied during previous tests, and under these
conditions, the column specimens exhibited a previously unobserved mode of failure which
features in-plane buckling of the longitudinal reinforcing bars. A 3D FE model of the lower
portion of the columns was analyzed in addition to that of isolated bars to investigate the
in-plane buckling phenomenon that occurred in 6 of 7 tests at the MAST Lab. A parametric
study was carried out to investigate the effect of concrete compressive strength,
longitudinal bar size and spacing, tie spacing, and column cross-sectional on the buckling
behavior of reinforcing bars.
Damage to the specimens during each test was quantified by several cumulative and
noncumulative damage index models. In addition, the damage to the specimens was
categorized by visual test observations and used as a measure to assess the calculated
damage indices. Calculated and observed damage to the specimens were then considered
along with the force-deformation cyclic envelope, and hysteretic energy dissipation during
each test to study the effect of loading protocols.
183
6.2. CONCLUSIONS
Experimental and analytical investigation of the seismic performance of the full-scale
RC columns in this research leads to the following conclusions:
1) A study of the parameters defining the seismic performance of RC columns
confirms that effective stiffness, drift capacity, and displacement ductility of
RC columns can be significantly affected by axial load and aspect ratios. The
effect of axial load for stocky columns with aspect ratio smaller than 2 is minor.
However, for columns with larger aspect ratios, an increase in the axial load
ratio and aspect ratio is accompanied by an increase in the effective stiffness
while it has an adverse effect on the drift capacity and displacement ductility.
2) It was also confirmed that the drift capacity and displacement ductility of RC
columns are sensitive to the characteristics of the lateral loading protocols so
that an increase in the number of loading cycles results in a reduction of drift
capacity and displacement ductility. In addition, RC columns exhibit a lower
drift capacity under biaxial loading condition than under a uniaxial loading
protocol.
3) Flexural moment capacity, normalized effective stiffness, drift capacity, and
displacement ductility of RC columns constructed with normal strength to
moderately high strength concrete (2.5 ksi (17.24 MPA) ≤ f′c ≤ 10 ksi (68.94
MPA)) are not generally affected by the dimension of their cross-sectional
depths.
4) The MAST tests revealed a mode of longitudinal bar buckling that has not been
reported and is not considered in the structural design of RC columns: An in-
184
plane buckling mode was observed in which the bars buckle parallel to the face
of the column as opposed to the presumption that they would only buckle in
the outward direction.
5) In-plane buckling of bars typically occurred about the strong axis of the bars
during tests at the MAST Lab while the common assumption is that
longitudinal bars buckle only about their weak axis.
6) Results from FE modeling of the lower portion of the columns suggest that
concrete with a larger nominal compressive strength (f′c ) helps to postpone
in-plane buckling of the bars.
7) In addition, larger longitudinal bar sizes are recognized to be more likely to
experience in-plane buckling as their restraint demands are higher. However,
the effect of bar spacing is insignificant.
8) Unlike outward buckling, in-plane buckling of reinforcing bars is not
controlled by tie-spacing. Rather, it is the concrete surrounding the bars that
restrains the bars from in-plane buckling.
9) The specimens featuring larger cross-sectional dimensions that incorporate
larger bar sizes are shown to be more critical to in-plane bar buckling.
10) An analysis of the damage evolution during tests of RC columns at the MAST
Lab under distinct loading protocols reveals that columns will exhibit more
damage during cyclic biaxial loading protocol as compared to uniaxial loading
conditions.
11) All damage index models that were considered in this research, except the one
proposed by Park and Ang (1985), are described for uniaxial loading. The
185
model proposed by Park and Ang (1985) can directly include the effect of
biaxial loading if the hysteretic energy dissipated during loading in both
directions is considered.
12) The study also shows that column SP5, which had a similar loading protocol
to that of SP2, but under a larger axial load ratio, experienced more damage.
13) Finally a monotonic push displacement that was applied following a number
of cyclic displacement reversals resulted in a rapid strength reduction in the
specimens. Among all specimens, SP7 and SP3 which sustained a biaxial
loading protocol and a uniaxial symmetric cyclic followed by a monotonic
push, respectively exhibited more damage. On the other hand, the loading
protocols with the smallest number of cycles (i.e., the monotonic and the near
–collapse loading protocols) were the least damaging ones. The symmetric
cyclic loading protocol that was applied to specimen SP2 was in the middle in
terms of severity of resulted damage.
6.3. RECOMMENDATIONS FOR FUTURE RESEARCH
The current study recommends that the following topics be investigated in the future
studies:
1) The in-plane buckling of reinforcing bars needs to be investigated
experimentally in more detail. The effects of concrete compressive strength,
longitudinal bar size and spacing, tie spacing and overall cross-sectional size
of the columns require verification by conducting tests for which in-plane
buckling is the main focus of the study.
186
2) A more detailed FE analysis to include the effects of deterioration of concrete
surrounding the bars would be beneficial for a better estimation of the
sensitivity of in-plane bar buckling to the relevant parameters.
3) Methods to prevent in-plane buckling of longitudinal bars need to be studied.
Crossties with 180 degree end hooks may be helpful in restraining the bars
from in-plane buckling.
187
REFERENCES
ACI Committee 318. (2011). Building Code Requirements for Structural Concrete (ACI
318-11) and Commentary. Farmington Hills, MI: American Concrete Institute.
ACI Committee 318. (2014). Building Code Requirements for Structural Concrete (ACI
318-14) and Commentary. American Concrete Institute.
ACI Committee 374. (2005). Acceptance Criteria for Moment Frames Based on Structural
Testing and Commentary (ACI 374.1–05). American Concrete Institute, Detroit,
MI, USA.
Advisory Committee on Earthquake Hazards Reduction (ACEHR). (2008). Effectiveness
of the National Earthquake Hazards Reduction Program (A Report from the
Advisory Committee on Earthquake Hazards Reduction) (p. 34).
American Society of Civil Engineers (ASCE). (2000). Prestandard and commentary for
the seismic rehabilitation of buildings (FEMA-356). Federal Emergency
Management Agency, Washington, D.C.
American Society of Civil Engineers (ASCE). (2007). Seismic Rehabilitation of Existing
Buildings (ASCE/SEI 41-06) (ASCE 41-06 Supplement No.1). American Society
of Civil Engineers, Reston, VA.
Applied Technology Council (ATC). (2009a). Effects of Strength and Stiffness
Degradation of Seismic Response (FEMA P440A). Federal Emergency
Management Agency, Washington, D.C.
Applied Technology Council (ATC). (2009b). Qualification of Building Seismic
Performance Factors (FEMA P695). Federal Emergency Management Agency,
Washington, D.C.
188
ASTM A 370-03a. (2003). Standard Test Methods and Definitions for Mechanical Testing
of Steel Products. ASTM International, West Conshohocken, PA.
ASTM A706/A706M-03a. (2003). Standard Specification for Low-Alloy Steel Deformed
and Plain Bars for Concrete Reinforcement. ASTM International, West
Conshohocken, PA.
ASTM C 39C/C39M. (2003). Standard Test Method for Compressive Strength of
Cylindrical Concrete Specimens. ASTM International, West Conshohocken, PA.
ASTM Standard A615/A615M-03a. (2003). Specification for Deformed and Plain Carbon-
Steel Bars for Concrete Reinforcement. ASTM International, West Conshohocken,
PA.
ATC-40. (1996). Seismic evaluation and retrofit of concrete buildings. Applied
Technology Council, Redwood City, CA.
Bae, S., & Bayrak, O. (2008). Seismic Performance of Full-Scale Reinforced Concrete
Columns. ACI Structural Journal, 105(2), 123–133.
Bae, S., Mieses, A. M., & Bayrak, O. (2005). Inelastic buckling of reinforcing bars. Journal
of Structural Engineering, 131(2), 314–321.
Banon, H. (1980). Prediction of seismic damage in reinforced concrete frames.
Massachusetts Institute of Technology.
Banon, H., Irvine, H. M., & Biggs, J. M. (1981). Seismic damage in reinforced concrete
frames. Journal of the Structural Division, 107(9), 1713–1729.
Bayrak, O., & Sheikh, S. A. (1996). Confinement steel requirements for high strength
concrete columns. In Proceedings of the 11th World Conference on Earthquake
Engineering.
189
Bayrak, O., & Sheikh, S. A. (2001). Plastic hinge analysis. Journal of Structural
Engineering, 127(9), 1092–1100.
Berry, M. P., & Eberhard, M. O. (2005). Practical performance model for bar buckling.
Journal of Structural Engineering, 131(7), 1060–1070.
Berry, M. P., Parrish, M., & Eberhard, M. O. (2004). PEER structural performance
database user’s manual (version 1.0). University of California, Berkeley.
Bournas, D. A., & Triantafillou, T. C. (2010). Bar buckling in RC columns confined with
composite materials. Journal of Composites for Construction, 15(3), 393–403.
Bresler, B., & Gilbert, P. H. (1961). Tie requirements for reinforced concrete columns. ACI
Journal Proceedings, 58(11), 555–570.
Brown, W. A., Lehman, D. E., & Stanton, J. F. (2008). Bar buckling in reinforced concrete
bridge columns. PEER Report 2007/11, Pacific Engineering Research Center,
University of California, Berkeley, CA.
Chai, Y. H., Romstad, K. M., & Bird, S. M. (1995). Energy-based linear damage model for
high-intensity seismic loading. Journal of Structural Engineering, 121(5), 857–
864.
Chung, Y. S., Meyer, C., & Shinozuka, M. (1987). Seismic damage assessment of
reinforced concrete members.
Coffin, (LF) Jr. (1954). A study of the effects of cyclic thermal stresses in ductile metals.
In ASME (Vol. 76, pp. 931–950).
Crisfield, M. A. (1981). A fast incremental/iterative solution procedure that handles “snap-
through.” Computers & Structures, 13(1), 55–62.
190
Dhakal, R. P., & Fenwick, R. C. (2008). Detailing of plastic hinges in seismic design of
concrete structures, 105(6), 740–749.
Dhakal, R. P., & Maekawa, K. (2002). Reinforcement Stability and Fracture of Cover
Concrete in Reinforced Concrete Members. Journal of Structural Engineering,
128(10), 1253–1262.
Elwood, K. J., & Eberhard, M. O. (2009). Effective stiffness of reinforced concrete
columns. ACI Structural Journal, 106(4), 476–484.
Erduran, E., & Yakut, A. (2004). Drift based damage functions for reinforced concrete
columns. Computers & Structures, 82(2), 121–130.
Esmaeily-Gh, A., & Xiao, Y. (2002). Seismic behavior of bridge columns subjected to
various loading patterns (PEER Report 2002/15). Pacific Earthquake Engineering
Research Center, College of Engineering, University of California, Berkeley.
French, C. W., Schultz, A. E., HAJJAR, J. F., Shield, C. K., Ernie, D. W., Dexter, R. J., …
Wan, C. P. (2004). Multi-axial subassemblage testing (MAST) system: Description
and capabilities. In Proceedings of the 13th World Conference on Earthquake
Engineering, Vancouver, Canada.
Frosch, R. J. (1999). Another look at cracking and crack control in reinforced concrete.
ACI Structural Journal, 96(3), 437–442.
Ghannoum, W., Sivaramakrishnan, B., Pujol, S., Catlin, A. C., Fernando, S., Yoosuf, N.,
& Wang, Y. (2012). ACI 369 rectangular column database. Network for Earthquake
Engineering Simulation (database), Dataset, DOI, 10, D36688J50.
191
Gilbersten, N. D., & Moehle, J. P. (1980). Experimental Study of Small-Scale R/C
Columns Subjected to Axial and Shear Force Reversal. Structural Research Series,
(481).
Gomes, A., & Appleton, J. (1997). Nonlinear cyclic stress-strain relationship of reinforcing
bars including buckling. Engineering Structures, 19(10), 822–826.
http://doi.org/10.1016/S0141-0296(97)00166-1
Gosain, N. K., Jirsa, J. O., & Brown, R. H. (1977). Shear requirements for load reversals
on RC members. Journal of the Structural Division, 103(7), 1461–1476.
Haselton, C. B., Liel, A. B., Lange, S. T., & Deierlein, G. G. (2008). Beam-column element
model calibrated for predicting flexural response leading to global collapse of RC
frame buildings (PEER Report 2007/3). Pacific Earthquake Engineering Research
Center, College of Engineering, University of California, Berkeley.
Ingham, J. M., Liddell, D., & Davidson, B. J. (2001). Influence of loading history on the
response of a reinforced concrete beam. Bulletin of the New Zealand National
Society for Earthquake Engineering, 34(2), 107–124.
Jaradat, O. A., McLean, D. I., & Marsh, M. L. (1998). Performance of Existing Bridge
Columns under Cyclic Loading? Part 1: Experimental Results and Observed
Behavior. ACI Structural Journal, 95(6).
Jeong, G. D., & Iwan, W. D. (1988). The effect of earthquake duration on the damage of
structures. Earthquake Engineering & Structural Dynamics, 16(8), 1201–1211.
Kent, D. C., & Park, R. (1971). Flexural members with confined concrete. Journal of the
Structural Division, 97(7), 1969–1990.
192
Kratzig, W. B., & Meskouris, M. (1987). Nonlinear seismic analysis of reinforced concrete
frames. Earthquake Prognostics, 453–462.
Kreger, M. E., & Linbeck, L. (1986). Behavior of reinforced concrete columns subjected
to lateral and axial load reversal. In Proceedings of the Third US National
Conference on Earthquake Engineering (pp. 24–28).
Kunnath, S. K., El-Bahy, A., Taylor, A. W., & Stone, W. C. (1997). Cumulative seismic
damage of reinforced concrete bridge piers. In Technical Report NCEER. US
National Center for Earthquake Engineering Research.
Kunnath, S. K., Reinhorn, A. M., & Lobo, R. F. (1992). IDARC version 3.0: a program for
the inelastic damage analysis of reinforced concrete structures.
Lee, J., & Fenves, G. L. (1998). Plastic-damage model for cyclic loading of concrete
structures. Journal of Engineering Mechanics, 124(8), 892–900.
Lehman, D. E., & Moehle, J. P. (1998). Influence of longitudinal reinforcement ratio on
column response. In Eleventh European Earthquake Engineering Conference.
Lehman, D. E., & Moehle, J. P. (2000). Seismic performance of well-confined concrete
bridge columns. Pacific Earthquake Engineering Research Center.
Lubliner, J., Oliver, J., Oller, S., & Onate, E. (1989). A plastic-damage model for concrete.
International Journal of Solids and Structures, 25(3), 299–326.
Lybas, J. M., & Sozen, M. A. (1977). Effect of Beam Strength and Stiffness on Dynamic
Behavior of R/C Coupled Walls. Department of Civil Engineering, University of
Illinois.
193
Mander, J. B., & Cheng, C.-T. (1995). Replaceable hinge detailing for bridge columns. In
Proceedings of the National Seismic Conference on Bridges and Highways:
“Progress in research and practice.” San Diego, CA.
Mander, J. B., Panthaki, F. D., & Kasalanati, A. (1994). Low-cycle fatigue behavior of
reinforcing steel. Journal of Materials in Civil Engineering, 6(4), 453–468.
Mander, J. B., Priestley, M. J. N., & Park, R. (1984). Seismic Design of Bridge Piers.
Research Report 84-2, Department of Civil Engineering, University of Canterbury,
Christchurch, New Zealand.
Manson, S. S. (1953). Behavior of materials under conditions of thermal stress.
Massone, L. M., Polanco, P., & Herrera, P. (2014). Experimental and Analytical Response
of RC Wall Boundary Elements. Presented at the 10th U.S. National Conference
on Earthquake Engineering, Anchorage, AK.
Mehanny, S. S., Kuramoto, H., & Deierlein, G. G. (2001). Stiffness Modeling of
Reinforced Concrete Beam-Columns for Frame Analysis. ACI Structural Journal,
98(2), 215–225.
Mercan, B., Schultz, A. E., & Stolarski, H. K. (2010). Finite element modeling of
prestressed concrete spandrel beams. Engineering Structures, 32(9), 2804–2813.
Miner, M. A. (1945). Cumulative damage in fatigue. Journal of Applied Mechanics, 12(3),
159–164.
Monti, G., & Nuti, C. (1992). Nonlinear cyclic behavior of reinforcing bars including
buckling. Journal of Structural Engineering, 118(12), 3268–3284.
Moyer, M. J., & Kowalsky, M. J. (2003). Influence of tension strain on buckling of
reinforcement in concrete columns. ACI Structural Journal, 100(1), 75–85.
194
National Institute of Standards and Technology (NIST). (2009). Research Required to
Support Full Implementation of Performance-Based Seismic Design (NIST GCR
09-917-2). National Institute of Standards and Technology Building and Fire
Research Laboratory, Gaithersburg, Maryland.
Nojavan, A., Schultz, A. E., Chao, S.-H., Haselton, C. B., Simasathien, S., Palacios, G., &
Liu, X. (2014). Preliminary Results for NEESR Full-Scale RC Column Tests under
Collapse-Consistent Loading Protocols. Presented at the 10th U.S. National
Conference on Earthquake Engineering, Anchorage, AK.
Nojavan, A., Schultz, A. E., Haselton, C., Simathathien, S., Liu, X., & Chao, S.-H. (2015).
A New Data Set for Full-Scale Reinforced Concrete Columns under Collapse-
Consistent Loading Protocols. Earthquake Spectra, 31(2), 1211–1231.
Palacios, G. (2015). Performance of full-scale ultra-high performance fiber-reinforced
concrete column subjected to extreme earthquake-type loading and effect of surface
preparation on the cohesion and friction factors of the aashto interface shear
equation. The University OF Texas at Arlington.
Palmgren, A. (1924). Die lebensdauer von kugellagern. Zeitschrift Des Vereins Deutscher
Ingenieure, 68(14), 339–341.
Pantazopoulou, S. J. (1998). Detailing for reinforcement stability in RC members. Journal
of Structural Engineering, 124(6), 623–632.
Papia, M., Russo, G., & Zingone, G. (1988). Instability of longitudinal bars in RC columns.
195
Park, R. (1989). Evaluation of ductility of structures and structural assemblages from
laboratory testing. Bulletin of the New Zealand National Society for Earthquake
Engineering, 22(3), 155–166.
Park, Y.-J., & Ang, A. H.-S. (1985). Mechanistic seismic damage model for reinforced
concrete. Journal of Structural Engineering, 111(4), 722–739.
Park, Y.-J., Ang, A. H.-S., & Wen, Y.-K. (1987). Damage-limiting aseismic design of
buildings. Earthquake Spectra, 3(1), 1–26.
Paulay, T., & Priestley, M. J. . (1992). Seismic Design of Reinforced Concrete and Masonry
Buildings (1 edition). New York: Wiley-Interscience.
Powell, G., & Simons, J. (1981). Improved iteration strategy for nonlinear structures.
International Journal for Numerical Methods in Engineering, 17(10), 1455–1467.
Pujol, S. (2002, August). Drift capacity of reinforced concrete columns subjected to
displacement reversals. Purdue University.
Qiu, F., Li, W., Pan, P., & Qian, J. (2002). Experimental tests on reinforced concrete
columns under biaxial quasi-static loading. Engineering Structures, 24(4), 419–
428.
Ramm, E. (1981). Strategies for tracing the nonlinear response near limit points. Springer.
Rodriguez, M. E., Botero, J. C., & Villa, J. (1999). Cyclic stress-strain behavior of
reinforcing steel including effect of buckling. Journal of Structural Engineering,
125(6), 605–612.
Roufaiel, M. S., & Meyer, C. (1987). Analytical modeling of hysteretic behavior of R/C
frames. Journal of Structural Engineering, 113(3), 429–444.
196
Saatcioglu, M., & Grira, M. (1999). Confinement of reinforced concrete columns with
welded reinforced grids. ACI Structural Journal, 96(1), 29–39.
Scribner, C. F. (1986). Reinforcement buckling in reinforced concrete flexural members.
ACI Journal Proceedings, 83(6), 966–973.
Sivaramakrishnan, B. (2010). Non-linear modeling parameters for reinforced concrete
columns subjected to seismic loads. The University of Texas at Austin.
Stephens, J. E., & Yao, J. T. (1987). Damage assessment using response measurements.
Stone, W. C., Cheok, G. S., & Stanton, J. F. (1995). Performance of hybrid moment-
resisting precast beam-column concrete connections subjected to cyclic loading.
ACI Structural Journal, 91(2), 229–249.
Stone, W. C., & Taylor, A. W. (1993). Seismic performance of circular bridge columns
designed in accordance with AASHTO (NIST Building Science Series No. 170).
Gaithersburg, MD: National Institute of Standards and Technology.
Suda, K., Murayama, Y., Ichinomiya, T., & Shimbo, H. (1996). Buckling behavior of
longitudinal reinforcing bars in concrete column subjected to reverse lateral
loading. In Proc., 11th World Conf. on Earthquake Engineering.
Takemura, H., & Kawashima, K. (1997). Effect of loading hysteresis on ductility capacity
of reinforced concrete bridge piers. Journal of Structural Engineering, 43, 849–
858.
Wang, M.-L., & Shah, S. P. (1987). Reinforced concrete hysteresis model based on the
damage concept. Earthquake Engineering & Structural Dynamics, 15(8), 993–
1003.
197
Williams, M. S., & Sexsmith, R. G. (1995). Seismic damage indices for concrete structures:
a state-of-the-art review. Earthquake Spectra, 11(2), 319–349.
Williams, M. S., Villemure, I., & Sexsmith, R. G. (1997). Evaluation of seismic damage
indices for concrete elements loaded in combined shear and flexure. ACI Structural
Journal, 94(3).
Yao, J. T. P., & Munse, W. H. (1962). Low-cycle axial fatigue behavior of mild steel.
ASTM Special Technical Publication, 338, 5–24.
Yarandi, M. S. (2007). Seismic retrofit and repair of existing reinforced concrete bridge
columns by transverse prestressing. University of Ottawa, Canada.
Zhou, X., Higashi, Y., Jiang, W., & Shimizu, Y. (1985). Behavior of reinforced concrete
column under high axial load. Transactions of the Japan Concrete Institute, 7, 385–
392.
198
APPENDIX A
CONSTRUCTION, INSTRUMENTATION, AND

INSTALLATION OF SPECIMENS
A.1. CONSTRUCTION
The specimens were constructed in an upright position at the University of Texas at
Arlington (UTA) and shipped to the MAST Lab for testing. Reinforcement details for
perimeter and space frame colum specimens are shown in Figs. A.1 to A.4. As illustrated
in Figure A.5 the reinforcing cage was built horizontally and placed into the concrete forms
prior to pouring of the concrete. The specimens were constructed along with a footing
block, and a top block to be connected to the MAST strong floor and loading crosshead,
respectively (Figure A.6).
199
Figure A.1. Reinforcement Details for Perimeter Frame (PF) Column Specimens
(Designed and created by the author and reported by (Palacios, 2015))
200
Figure A.2. Reinforcement Details for Perimeter Frame (PF) Column Specimens
201
Figure A.3. Reinforcement Details for Space Frame (SF) Column Specimens
202
Figure A.4. Reinforcement Details for Space Frame (SF) Column Specimens
203
(a)
(b)
Figure A.5. (a) Building Reinforcing Cage, (b) Placing the Cage in the Forms
204
Figure A.6. Forming Column and Top Block
A.2. INSTRUMENTATION
Prior to pouring concrete, strain gages were installed on reinforcing bars, transverse
hoops, and into the concrete. To install the strain gages on reinforcing bars, first the desired
location of the strain gage on the bar was marked. The surface of the bar at the marked
location was grounded, and sanded with 400 grit sandpaper to provide a smooth surface
for the strain gage (Figure A.7). The smoothed surface of the bars were cleaned (Figure
A.8). The gages were oriented so that pulling their wires out of the concrete would not
require the wires to fold back and cause any damage to the gage wires. The gages were
taped to the bar. The location of the strain gages were then coated with layers of
polyurethane, nitrile rubber, moisture sealing electrical tape to protect the gages.
205
In addition, aluminum anchors were placed in the specimens prior to pouring the
concrete (Figure A.9). These anchors were used when connecting the external sensors (i.e.,
LVDTs, string pots and tiltmeters) to the specimens at the MAST Lab (Figures A.10 and
A.11).
Figure A.7. Grinding and Cleaning of Bars Prior to Strain Gage Installation
206
Figure A.8. Attaching the Strain Gage to the Prepared Surface of the Bar
207
Figure A.9. Installation of an Aluminum Anchor for External Instrumentation
208
Figure A.10. Installation of Vertical LVDTs on SE Face of the Column
209
Figure A.11. Installation of LVDTs and String Pots on NE Face of the Column
210
A.3. SHIPPING AND INSTALLATION
To protect the specimens from tensile cracking, the specimens were post tensioned
using four threaded rods, each 1 in. in diameter. Lifting and tilting of the specimens was
carried out using 2 ½ in. holes that were designed in the footing and top blocks. Loading
of the specimens at UTA, specimen shipping, unloading and tilting up the specimens at the
MAST Lab are represented in Figures A.12 to A.16. At the MAST Lab, the specimens
were placed on top of a three-piece spacer block that was constructed at the University of
Minnesota so that the minimum required height condition for the loading crosshead is
satisfied (Figures A.17, and A.18).
211
Figure A.12. Post Tensioning of the Specimen Prior to Shipping
212
Figure A.13. Lifting the Specimen Before Shipping
Figure A.14. Delivery of the Specimen to the MAST Lab
213
Figure A.15. Unloading the Specimen at the MAST Lab
214
Figure A.16. Tilting Up the Specimen at the MAST Lab
215
Figure A.17. Three-Piece Spacer Block
216
Figure A.18. Installation of the Specimen on Top of the Spacer Blocks
217

Performance of Full-Scale Reinforced Concrete Columns Subjected To Extreme Earthquake Loading

Uploaded by

Copyright:

Available Formats

Performance of Full-Scale Reinforced Concrete Columns Subjected To Extreme Earthquake Loading

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Performance of Full-Scale Reinforced Concrete Columns Subjected To Extreme Earthquake Loading

Uploaded by

Copyright:

Available Formats

Performance of Full-Scale Reinforced Concrete Columns

Subjected to Extreme Earthquake Loading

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

CMMI-10041633. I would like to acknowledge the Department of Civil, Environmental

and Geo-Engineering of the University of Minnesota for the Departmental Fellowship

foremost, I would like to express my extreme gratitude to my advisor, Prof. Arturo E.

part of this project and for his comments.

at UTA in construction of the specimen is also appreciated.

my parents and my wife

Subassemblage Testing (MAST) Laboratory of the University of Minnesota to investigate

peak behavior of the specimens at near-collapse conditions, hence loading on the

exhibited severe strength loss and stiffness degradation.

to study the effects of cross-sectional size on parameters representing seismic performance

of columns including moment capacity, effective stiffness, drift capacity, displacement

loading protocols on the performance of tested column specimens.

List of Tables ................................................................................................................... IX

List of Figures ................................................................................................................... X

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

1.1. Background ................................................................................................... 1

1.2. Research Objectives and Summary .............................................................. 4

1.3. Thesis Organization ...................................................................................... 5

CHAPTER 2 A NEW DATASET FOR FULL-SCALE REINFORCED

2.1. Summary ....................................................................................................... 8

2.2. Introduction ................................................................................................... 9

2.3. Test Setup.................................................................................................... 13

2.4. Specimens ................................................................................................... 16

2.5. Materials ..................................................................................................... 18

2.6. Instrumentation ........................................................................................... 20

2.7. Loading Protocols ....................................................................................... 25

2.8. Test Observations........................................................................................ 28

2.9. Future Use of the Data Set .......................................................................... 36

2.10. Conclusions ................................................................................................ 38

CHAPTER 3 INFLUENCE OF CROSS-SECTIONAL SIZE ON SEISMIC

3.1. Summary ..................................................................................................... 40

3.3. Research Significance ................................................................................. 44

3.4. Experimental Investigation and Data Selection .......................................... 45

3.5. Flexural Moment Capacity ......................................................................... 51

3.5.1. Calculation ..................................................................................... 51

3.5.2. Influence of Column Section Depth h ............................................. 52

3.6. Effective Stiffness ....................................................................................... 53

3.6.1. Calculation ..................................................................................... 53

3.6.2. Influence of Axial Loading and Aspect Ratio ................................. 55

3.6.3. Influence of Column Section Depth h ............................................. 57

3.7. Drift Capacity and Displacement Ductility................................................. 59

3.7.1. Calculation ..................................................................................... 59

3.7.2. Influence of Axial Loading and Aspect Ratio ................................. 59

3.7.3. Influence of Column Section Depth h ............................................. 61

3.8. Reinforcing Bar Buckling ........................................................................... 65

3.8.1. Characterization of Reinforcing Bar Buckling ............................... 65

3.8.2. Influence of Column Section Size ................................................... 72

3.9. Conclusions ................................................................................................. 74

CHAPTER 4 IN-PLANE BUCKLING OF LONGITUDINAL BARS IN

4.1. Summary ..................................................................................................... 76

4.2. Introduction ................................................................................................. 77

4.3. ExperimentaL Program ............................................................................... 79

4.5. Analytical Investigation Using Bar-Spring Model ..................................... 84

4.5.1. General Description of the Bar-Spring System .............................. 84

4.5.2. Static Analysis Using Modified Riks (Arc-Length) Method............ 87

4.5.3. Results from Bar-Spring Analysis................................................... 88