See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245305027

Displacement-Based Design of Slender Reinforced Concrete Structural Walls—

Experimental Verification

Article  in  Journal of Structural Engineering · April 2004

DOI: 10.1061/(ASCE)0733-9445(2004)130:4(618)

CITATIONS READS

210 2,639

2 authors, including:

John W. Wallace
University of California, Los Angeles
164 PUBLICATIONS   3,270 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Drift Capacity of RC Walls with Special Boundary Elements View project

Utilizing Remote Sensing to Assess the Implication of Tall Building Performance on the Resilience of Urban Centers View project

All content following this page was uploaded by John W. Wallace on 09 November 2014.

The user has requested enhancement of the downloaded file.

 ACI STRUCTURAL JOURNAL TECHNICAL PAPER
Title no. 103-S21

Flexural Modeling of Reinforced Concrete Walls—

Experimental Verification
by Kutay Orakcal and John W. Wallace

pThis study presents detailed information on the calibration of a

nonlinear wall macromodel by comparing model results with
experimental results for slender reinforced concrete walls with
rectangular and T-shaped cross sections. Test measurements were
processed to allow for a direct comparison of the predicted and
measured flexural responses. Responses were compared at various
locations on the walls. Results obtained with the analytical model
for rectangular walls compare favorably with experimental responses
for flexural capacity, stiffness, and deformability, although some
significant variation is noted for local compression strains. For T-
shaped walls, the agreement between model and experimental
results is reasonably good, although the model is unable to capture
the variation of the longitudinal strains along the flange.
Fig. 1—Multiple-vertical-line-element model: (a) MVLEM
Keywords: flexure; stiffness; strain; walls. element; and (b) model of wall.

INTRODUCTION RESEARCH SIGNIFICANCE

Prediction of the inelastic response of reinforced concrete The use of structural walls for earthquake resistance is
(RC) structural walls and wall systems requires accurate, common; therefore, wall models that accurately capture
effective, and robust modeling and analysis tools that cyclic wall responses, yet are simple enough for design
incorporate important material characteristics and behavioral office applications, are needed. An MVLEM is able to
response features such as neutral axis migration, tension- capture important response features; however, detailed
stiffening, progressive gap closure, confinement, nonlinear comparisons between model results and experimental results
shear behavior, and the effect of fluctuating axial force on are not available. In this paper, detailed comparisons are
strength and stiffness. Effective analytical models should be made between results obtained with a MVLEM and results
relatively simple to implement and reasonably accurate in obtained in experimental studies. Comparisons allow not
predicting the hysteretic responses of RC walls at both local only for a better understanding of the inelastic behavior of
and global levels, as well as capturing the interaction of the RC walls, but also identify model capabilities and as well as
walls with other structural members. ways to improve the model.
Various phenomenological macroscopic models have been
proposed to incorporate such response features in predicting ANALYTICAL MODEL
the inelastic response of RC structural walls. The multiple- The model in Fig. 1(a) is an implementation of the generic
vertical-line-element model (MVLEM) proposed originally MVLEM for structural walls. A horizontal spring placed at
by Vulcano et al.1 has been shown to successfully capture the element center of rotation (at relative height ch) simulates
important response characteristics by the simplicity of a the shear response of the wall element. Flexural and shear
macroscopic model. Yet, the model has not been implemented modes of deformation of the wall element are uncoupled
into widely available computer programs and has not been (that is, flexural deformations do not affect shear strength or
sufficiently calibrated with and validated against extensive deformation), which is a very commonly used assumption. A
experimental data at both local and global response levels. structural wall is modeled as a stack of m elements, which
Given these shortcomings, a research project was under- are placed one upon the other (Fig. 1(b)). The flexural
taken to investigate and improve the MVLEM for RC wall response is simulated by a series of n uniaxial elements (or
systems, as well as to calibrate and validate it against macrofibers) connected to infinitely rigid beams at the top
experimental data. A description of the improved model, and bottom (for example, floor) levels. The primary
implementation of detailed cyclic constitutive relationships, simplification of the model involves applying the plane-
and the sensitivity of the model predictions to both model sections-remain-plane assumption (validated experimentally
and material parameters are presented by Orakcal et al.2 This by Thomsen and Wallace3,4 for the walls investigated in this
paper emphasizes the accuracy and limitations of the model study) in calculating the strain level in each uniaxial element.
by comparing model results with experimental results. The
study presented herein focuses on modeling and simulation ACI Structural Journal, V. 103, No. 2, March-April 2006.
MS No. 03-485 received July 15, 2004, and reviewed under Institute publication policies.
of the axial-flexural response; an improved model for shear Copyright © 2006, American Concrete Institute. All rights reserved, including the making
behavior as well as ways to incorporate axial-shear-flexure of copies unless permission is obtained from the copyright proprietors. Pertinent discussion
including author’s closure, if any, will be published in the January-February 2007 ACI
interaction will be presented in a follow-up paper. Structural Journal if the discussion is received by September 1, 2006.

196 ACI Structural Journal/March-April 2006

Kutay Orakcal is a postdoctoral researcher in the Department of Civil Engineering,
University of California, Los Angeles, Calif. (UCLA). He received his PhD in civil
engineering from UCLA in 2004. His research interests include behavior and modeling
of reinforced concrete elements and systems.

John W. Wallace, FACI, is a professor of civil engineering at UCLA. He is a member

of ACI Committees 318, Structural Concrete Building Code; 318-H, Seismic Provisions
(Structural Concrete Building Code); 335, Composite and Hybrid Structures; 369,
Seismic Repair and Rehabilitation; 374, Performance-Based Seismic Design of Concrete
Buildings; E 803, Faculty Network Coordinating Committee; and Joint ACI-ASCE
Committee 352, Joints and Connections in Monolithic Concrete Structures. His
research interests include response and design of buildings and bridges to earthquake
actions, laboratory and field testing of structural components and systems, and structural
health monitoring.

The stiffness properties ki and force-displacement relationships

of the uniaxial elements are defined according to constitutive
stress-strain relationships implemented in the model for
concrete and steel and the tributary area assigned to each
uniaxial element. The strains in concrete and steel are Fig. 2—Constitutive model for steel.
assumed equal (perfect bond) within each uniaxial element.
Although the model could be modified to incorporate slip of
reinforcing bars, the present model does not consider slip, in
part because the model was calibrated using tests results
where negligible slip was observed.3,4
The reinforcing steel stress-strain behavior implemented
in the wall model is the well-known nonlinear relationship of
Menegotto and Pinto5 (Fig. 2), as extended by Filippou et al.6
to include isotropic strain-hardening effects. The hysteretic
constitutive relation developed by Chang and Mander7
(Fig. 3) is used as the basis for the relation implemented for
concrete because it is a general model that provides the flexi-
bility to model the hysteretic behavior of confined and uncon-
fined concrete in both cyclic compression and tension, with
particular emphasis paid to the transition between crack opening
and closure. The material laws implemented in this study can be
controlled and calibrated to follow the relations developed by
Belarbi and Hsu8 or similar empirical relations to model tension Fig. 3—Constitutive model for concrete.
stiffening. Details of the analytical model and the material laws,
as well as the sensitivity of the analytical results to model and
material parameters are presented by Orakcal et al.2 Cyclic lateral displacements were applied to the walls using
a hydraulic actuator mounted horizontally to a reaction wall.
EXPERIMENTAL RESULTS The peak (reversal) lateral top displacement values and drift
Experimental results were obtained for two, approxi- levels applied to Specimens RW2 and TW2 are listed in
mately quarter-scale wall specimens tested by Thomsen and Table 1. Instrumentation was used to measure displace-
Wallace:3,4 one specimen with a rectangular cross section ments, loads, and strains at critical locations for each wall
(Specimen RW2) and another specimen with a T-shaped specimen (Fig. 5). Four wire potentiometers (WPs) were
cross section (Specimen TW2). The walls were 3.66 m tall mounted to a rigid steel reference frame to measure lateral
and 102 mm thick, with web and flange lengths of 1.22 m. displacements at 0.91 m intervals over the wall height. A
Detailing requirements at the boundaries of the wall specimens linear potentiometer was also mounted horizontally on the
were evaluated using the displacement-based design pedestal to measure any horizontal slip of the pedestal along
approach presented by Wallace.9,10 Well-detailed boundary the strong floor. Two additional linear potentiometers were
elements were provided at the edges of the walls over the mounted vertically at each end of the pedestal to measure rota-
bottom 1.22 m of each wall. A capacity design approach was tion caused by uplift of the pedestal from the strong floor.
used to provide sufficient shear capacity using ACI 318-89,11 Shear deformations were measured through the use of wire
Eq. (21-6), to resist the shear that develops for the probable potentiometers mounted on the bottom two stories (in an “X”
wall moment. Favorable anchorage conditions existed for configuration) of each specimen (Fig. 5). Sliding shear at the
the vertical reinforcement anchored within the pedestal at the wall-pedestal interface was not observed during the tests.
base of the walls with sufficient development length as well Axial (vertical) displacements at the wall boundaries were
as 90-degree hooks.3,4 Sample reinforcement details for a measured using two WPs mounted directly to the wall.
rectangular wall Specimen RW2 and for a T-shaped wall Linear variable differential transducers (LVDTs) were
Specimen TW2 are shown in Fig. 4. mounted vertically between the wall and the pedestal (Fig. 5),
The wall specimens were tested in an upright position. An over a gage length of 229 mm, at various locations along the
axial load of approximately 0.07Ag fc′ was applied at the top web and flange of each wall so that axial strains and section
of the walls by hydraulic jacks mounted on top of the load curvatures could be calculated. The rectangular wall was
transfer assembly and held constant throughout the duration instrumented with seven LVDTs spaced along the length of
of each test as is typical for isolated or weakly coupled walls. the wall, whereas the T-shaped wall had four LVDTs

ACI Structural Journal/March-April 2006 197

Table 1—Peak lateral top displacements at applied drift levels
Top Lateral drift
Load
Wall direction displacement* 0.1% 0.25% 0.5% 0.75% 1.0% 1.5% 2.0% 2.5%
(a) 2.9 7.2 16.1 24.5 33.1 50.6 67.9 86.2
Positive (b) 2.3 [81]† 5.9 [82] 13.0 [81] 20.1 [82] 28.5 [86] 45.5 [90] 62.1 [91] 79.8 [93]
(c) 1.2 [52]‡ 4.0 [68] 9.9 [76] 16.2 [80] 23.4 [82] 39.4 [86] 54.2 [87] 70.6 [88]
RW2
(a) –3.2 –7.6 –15.9 –24.2 –32.8 –49.8 –66.5 –83.8
Negative (b) –2.6 [83] –6.5 [85] –13.7 [86] –21.2 [88] –29.4 [90] –46.0 [92] –62.1 [93] –79.2 [94]
(c) –1.3 [50] –4.4 [67] –10.6 [77] –17.1 [80] –24.5 [83] –39.8 [86] –54.6 [88] –70.5 [89]
(a) 2.5 7.6 15.9 24.8 33.5 51.3 70.9 88.8
Positive (b) 2.1 [86] 6.6 [87] 13.7 [86] 21.7 [87] 29.9 [89] 46.5 [91] 63.3 [89] 80.2 [90]
(c) 1.0 [47] 4.5 [68] 10.5 [77] 17.9 [83] 25.5 [85] 41.2 [89] 56.5 [89] 72.2 [90]
TW2
(a) –2.5 –7.6 –15.9 –21.2 –32.7 –47.1 –64.9 –82.2
Negative (b) –1.6 [65] –4.3 [57] –9.7 [61] –13.5 [64] –21.7 [66] –36.6 [78] –49.6 [76] –66.2 [80]
(c) –0.4 [26] –1.8 [41] –5.9 [61] –9.4 [69] –16.3 [75] –30.9 [84] –42.6 [86] –58.7 [89]
*(a) = top displacement applied during testing, mm; (b) = top displacement with pedestal movement contribution subtracted, mm; and (c) = top displacement with pedestal movement and shear

deformation contributions subtracted, mm.

†
[value in brackets] = (b)/(a), %.
‡
[value in brackets] = (c)/(b), %.

Fig. 4—Wall cross-sectional views and model discretization: (a) Specimen RW2;
and (b) Specimen TW2.

mounted along the web and five mounted along the outside CALIBRATION OF ANALYTICAL MODEL
face of the flange. Axial concrete strains within the boundary Geometry
regions of the specimens also were measured using Figure 4 displays discretization of the wall cross sections
embedded concrete strain gauges, and strains in the reinforcing for the analytical model, with eight uniaxial elements
steel were measured through the use of strain gauges at wall defined along the length of the wall (n = 8) for Specimen
base and first-story levels (Fig. 5). More detailed information RW2, and 19 uniaxial elements (n = 19) for Specimen TW2.
concerning the walls is presented in Thomsen and Wallace3,4 A refined configuration with eight uniaxial elements was
and Massone and Wallace.12 assigned for the flange Specimen TW2 (n = 12,…,19)

198 ACI Structural Journal/March-April 2006

 Fig. 5—Instrumentation on wall specimens.

because the neutral axis was expected to be within the flange

during loading subjecting the flange to compression. The
analytical model was discretized along wall height to allow
consistent strain comparisons between model and experimental
results at all locations where LVDTs, concrete strain gauges,
and wire potentiometers were provided. The height of the
model elements used for all local strain comparisons were set
equal to the gauge length of the instruments used. Accordingly,
16 MVLEM elements were used for the modeled walls (m = 16)
with eight elements along the first-story height: four
elements along the second story, and two elements along the
third and fourth stories each. A value of 0.4 was selected for
the parameter c defining the center of relative rotation for
each wall element, based on previous studies.1,2

Materials
Fig. 6—Calibration of steel constitutive model.
Steel stress-strain relation—The reinforcing steel stress-
strain relationship described by the Menegotto and Pinto5
model was calibrated to reasonably represent the experi- sensitivity of the model response with regard to these
mentally observed properties of the longitudinal reinforcement parameters is discussed in Orakcal et al.2
(No. 3 and No. 2 deformed bars) used in the experimental Concrete stress-strain relations—The monotonic envelope
study. The tensile yield strength and strain-hardening curves of the implemented concrete hysteretic stress-strain
parameters were modified according to the empirical relations relation for compression and tension allow control on the
proposed by Belarbi and Hsu8 to include the effect of tension shape of both the ascending and descending (that is, prepeak
stiffening on steel bars embedded in concrete. Figure 6 and postpeak) branches. The curves can be calibrated for
shows the calibrated analytical steel stress-strain relations in selected values of peak stress fc′ , strain at peak stress εc′ ,
tension and compression, as well as results of stress-strain elastic modulus Ec, and also by the parameter r defining the
tests (on bare bars) for the reinforcement used in the shape of the envelope curve, allowing for model refinement
construction of the wall specimens. The parameters used for (Fig. 3). The envelope curve used in the analytical model for
the calibration of the steel stress-strain relation for compression unconfined concrete in compression was calibrated using
(based solely on the results of tests on bare bars) and for results of monotonic stress-strain tests conducted at time of
tension (including modifications to account for tension testing on standard 152.4 x 304.8 mm cylinder specimens of the
stiffening) are provided in Table 2 and 3. concrete used in the construction of the walls (Fig. 7(a)). The
The calibration of parameters R0, a1, and a2 (accounting concrete tensile strength was determined from the relationship
for the cyclic degradation of the curvature coefficient R and ft = 0.31 f c′ (MPa), and a value of 0.00008 was selected for
the Bauschinger effect) requires results of cyclic stress-strain the strain εt at peak monotonic tensile stress (Fig. 7(b)), as
tests on reinforcing bars, which were not conducted as part suggested by Belarbi and Hsu8 based on a series of tests on
of the present experimental studies, and are typically not RC panels with concrete cylinder compressive strengths
available. The values R0 = 20, a1 = 18.5, and a2 = 0.0015, consistent with the compressive strength of concrete used for
proposed by Elmorsi et al.13 based on the experimental the construction of the wall specimens. The shape of the
results carried out by Seckin,14 were used in this study. The monotonic tension envelope was calibrated (via the parameter

ACI Structural Journal/March-April 2006 199

r) to reasonably represent the average postcrack stress-strain response components using the methodology described by
relation proposed by Belarbi and Hsu8 that considers the Massone and Wallace.12 This methodology involves
effects of tension stiffening on concrete (Fig. 7(b)). processing the displacements measured by the WPs (in vertical,
The compression envelope used in the analytical model for horizontal, and “X” configurations as shown in Fig. 5) to sepa-
confined concrete was calibrated using the empirical relations rate the measured wall lateral displacements into flexural and
proposed by Mander et al.15 for the peak compressive stress shear deformation contributions. To compare model results
and the strain at peak compressive stress. The confined directly with “flexural components” of the measured responses,
concrete stress-strain behavior was manipulated based on the a linear elastic stiffness kH with a very large (~infinite) value
area, configuration, spacing, and yield stress of the transverse was assigned to the horizontal shear spring.
reinforcement (Fig. 6, 435 MPa for the 4.76 mm wire) in the
confined regions within the first story height (0 to 0.91 m) of
the test specimens. The postpeak slope of the monotonic
stress-strain relations implemented in the wall model for
confined and unconfined concrete were calibrated (via the
parameter εcr) to agree with the postpeak slope of the Saat-
cioglu and Razvi16 model for both confined and unconfined
concrete (Fig. 7(a)). The parameters used for the calibration
of the monotonic (envelope) stress-strain relations for
concrete in tension and for confined and unconfined
concrete in compression are presented in Table 2 and 3. The
parameters for unconfined concrete also were used for the
wall web as well as for cover concrete areas within the
boundary regions of the walls.
The hysteretic stress-strain rules defined by Chang and
Mander,7 modified slightly as noted by Orakcal,17 were used
to simulate the cyclic behavior of both confined and uncon-
fined concrete implemented in the wall model. The empirical
relationships derived by Chang and Mander7 (based on an
extensive column database) for key hysteretic parameters
(Esec , Epl , ∆f, and ∆ε in Fig. 3) were used in the present
model. Further details on the implemented hysteretic stress-
strain relationships for concrete can be found in Chang and
Mander7 and Orakcal et al.2
Shear force-deformation relation—The study presented
herein focuses on modeling of the flexural responses, thus
the experimental results for wall lateral displacements at top
and story levels were separated into flexural and shear

Table 2—Calibrated constitutive parameters for

concrete in tension and steel in compression
No. 3 reinforcing No. 2 reinforcing
Concrete in tension bar in compression bar in compression
t, Ec, σy, E0, σy, E0,
MPa εt GPa εcr r MPa GPa b MPa GPa b Fig. 7—Calibration of concrete constitutive model: (a)
2.03 0.00008 31.03 ∞ 1.2 434 200 0.02 448 200 0.02 compression; and (b) tension.

Table 3—Calibrated constitutive parameters for concrete in compression and steel in tension
RW2 TW2
Boundary Web Flange Flange Flange-web inter- Web Web
Material Parameter (confined) (unconfined) (confined) (unconfined) section (confined) (unconfined) (confined)
fc′ , MPa 47.6 42.8 43.9 42.8 43.9 42.8 57.1
εc′ 0.0033 0.0021 0.0024 0.0021 0.0024 0.0021 0.0056
Concrete in
compression Ec, GPa 31.03 31.03 31.03 31.03 31.03 31.03 31.03
εcr 0.0037 0.0022 0.0025 0.0022 0.0025 0.0022 0.0073
r 1.90 7.00 3.80 7.00 3.80 7.00 1.45
σy, MPa 395 — 395 — 395 — 387
No. 3 reinforcing
bar in tension E0, GPa 200 — 200 — 200 — 200
b 0.0185 — 0.0185 — 0.0185 — 0.02
σy, MPa — 336 — 336 — 356 —
No. 2 reinforcing
bar in tension E0, GPa — 200 — 200 — 200 —
b — 0.0350 — 0.0350 — 0.0295 —

200 ACI Structural Journal/March-April 2006

 ANALYTICAL RESULTS AND COMPARISON
WITH TEST RESULTS
The analytical model was implemented in Matlab18 to
allow for a comparison between experimental and analytical
results. A displacement-controlled nonlinear analysis
strategy was selected to correlate the model results with
results of the drift-controlled cyclic tests subjected to
prescribed lateral displacement histories at the top of the
walls (Table 1). Before analysis, the lateral top displacement
history applied during testing and the measured lateral story
displacement histories for each specimen were processed to
remove displacement contributions resulting from shear and
pedestal movement to allow for a direct comparison of the
measured and predicted flexural responses. Measurements
obtained in horizontal and vertical linear potentiometers
mounted on the pedestals (Fig. 5) were used to remove
displacements caused by pedestal rotation (caused by uplift)
and pedestal sliding in the direction of the applied load.
To process the test data to remove shear deformation Fig. 8—Wall RW2, measured versus predicted load-
contributions, several steps were required. First, wall shear deformation responses.
deformations at the bottom two stories of each specimen
were calculated using data from WPs mounted on the bottom measured response reasonably well. The lateral load capacity
two stories of each specimen (Fig. 5), using the procedure and the lateral stiffness of the wall are well-represented for
recommended by Massone and Wallace12 (which allows most of the lateral drift levels. The rectangular wall specimen
separating of flexural and shear deformations). Nonlinear was observed to experience yielding of the longitudinal
shear deformations were concentrated within the first-story boundary reinforcement at a nominal (applied) drift level of
height of the walls, where shear yielding was observed approximately 0.75%,3 corresponding to a flexural drift level of
together with flexural yielding, despite the adequate shear approximately 0.45%. The prediction of the wall lateral load
capacity provided in design of the wall specimens. The capacity at this drift level (0.75% nominal) is accurate; the
measured lateral load-shear deformation response in the prediction corresponds to 98 and 93% of the measured
second story was observed to follow an approximately linear values for loading in the positive and negative directions,
elastic relationship; thus, the shear deformations in the third respectively. Cyclic properties of the response, including
and fourth stories of the wall specimens were estimated for stiffness degradation, hysteretic shape, plastic (residual)
the entire loading history using the linear elastic shear stiffness displacements, and pinching behavior are well represented in
values derived for the second-story level. The story shear the analytical results; therefore, the cyclic properties of the
deformations were then removed from the lateral displacement implemented analytical stress-strain relations for steel and
measurements at the top and story levels of the walls. concrete produce good correlation for global response.
Table 1 lists the peak (reversal) lateral top displacement The degradation in the measured lateral capacity and stiffness
values applied to Specimens RW2 and TW2 during testing, of the wall specimen during the second cycle to lateral flexural
together with corresponding lateral top displacement values drift level of approximately 2% in the positive direction is
obtained after subtracting the contributions due to pedestal due to buckling of the longitudinal reinforcement within the
movement and shear deformations. A considerable reduction boundary element of the specimen. This behavior is not
is observed in the top displacement values due to the contri- represented by the analytical model because the model does
butions of both pedestal movement and shear deformations. not consider reinforcing bar buckling. The underestimation
The analytical models for Specimens RW2 and TW2 were of the wall capacity at intermediate drift levels (for example,
subjected to the modified (reduced) top displacement histories 0.5 to 1.5% drift) can be attributed to the inability of the
determined using the procedures outlined in the previous analytical stress-strain yield asymptote for steel in tension to
paragraph. The measured axial load histories applied on the model the curved strain-hardening region observed in the
wall specimens during testing, as measured by load cells stress-strain tests for the No. 3 longitudinal reinforcing bars
during testing, were applied to the analytical models (on (Fig. 6), as well as the uncertainty in the calibration of the
average, approximately 7% of the axial load capacity for cyclic parameters governing the implemented steel stress-
RW2 and 7.5% for TW2, with a variation of approximately strain relation (R0, a1, and a2) and the parameters associated
±10%). The impact of significant axial load variation on wall with concrete tensile strength ( ft and εt). Possible recalibration
response (for example, for coupled walls) is an important of the associated material parameters would increase the
consideration, and the analytical model is capable of gener- predicted wall capacity and result in improved correlation for
ating responses under varying axial load, as it incorporates these intermediate drift levels.2 Such recalibration, however,
axial-flexural response interaction. Comparisons between would moderately impair the correlation for the wall
model predictions of the flexural responses and test results capacity at other drift levels as well as the cyclic properties
are summarized in the following paragraphs. of the wall response including plastic (residual) displacements
and pinching. The values used previously for these material
Rectangular wall, RW2 parameters have been found to give the best overall correlation
Figure 8 compares the measured and predicted lateral between predicted and measured responses for this study.
load-top flexural displacement responses for the rectangular The reader should refer to the paper by Orakcal et al.2 for
wall Specimen RW2. The analytical model captures the information on the sensitivity of model results to a realistic

ACI Structural Journal/March-April 2006 201

 Fig. 11—Wall RW2, concrete strain measurements by
LVDTs versus predictions.

Fig. 9—Wall RW2, lateral displacement profiles.

flexural deformations experienced within the plastic hinge
region of the wall.
Measured and predicted responses at specific locations are
compared in Fig. 11, which plots the average concrete strains
measured by the seven LVDTs over a 229 mm gauge length
at the base of the wall (Fig. 5), at applied peak positive top
displacement (top displacement reversal) data points, for
selected drift levels applied during testing. Similar trends
were observed in the results for other drift levels and also for
peak negative top displacement data points. Results shown
in Fig. 11 illustrate that the analytical model predicts the
tensile strain profile reasonably well, but significantly
underestimates the compressive strains. The compressive
strain predictions at the location of the outmost LVDT
correspond to 63, 55, and 30% of the measured values for
wall nominal drift levels of 0.5, 1, and 2%, respectively. The
accuracy in the prediction of the position of the neutral axis
Fig. 10—Wall RW2, lateral displacements and rotations at is reasonably good; the analytically predicted and experi-
first story. mentally obtained neutral axis positions vary by no more
than 5% of the wall length (Fig. 11).
level of uncertainty associated with calibration of these The larger measured compressive strains in concrete at the
parameters. base of the wall may be due to stress concentrations induced
Figure 9 shows a comparison of the lateral flexural at the wall-pedestal interface (the bottom of the LVDTs were
displacements of the wall, at peak top displacement (top mounted on the pedestal) due to the abrupt change in geom-
displacement reversal) data points for each drift level, etry (for example, Craig19). The concrete strain measure-
measured by the horizontal WPs at the first-, second-, and ments may also have been influenced by the attachment of
third-story levels, (Fig. 5), with the results of the analysis. the LVDTs to the wall specimen, as the LVDTs were affixed
The analytical model provides a good prediction of the wall to blocks glued to the wall surface; the measurements may
lateral displacement profile, and the distribution of deforma- not represent average compressive strains experienced along
tions along wall height. The reader should note that the drift the thickness of the wall specimen when spalling is
levels noted in the legend of Fig. 9, and in all subsequent observed. Finally, the larger compressive strains may be
figures, correspond to “nominal” drift levels applied to the partially due to the nonlinear shear response that the wall
wall during testing versus actual drift levels (that is, specimen experienced within the first-story height. Preliminary
measured top displacement modified to remove contributions analysis results using a modified MVLEM with an implemented
from pedestal and shear deformations). methodology for coupling shear and flexural displacements,
Figure 10 plots the measured and predicted lateral flexural based on biaxial constitutive relationships for concrete with
displacement histories at the first-story height (0.91 m) and compressive strain-softening, yields compressive strains
the rotations accumulated over the bottom 0.76 m of the wall larger than those predicted with the flexural model used herein.
(rotation measurements were obtained by calculating the Figure 12 compares measured responses for a specific
difference in the axial [vertical] displacements measured by gauge—an embedded concrete strain gauge with gauge
the two WPs mounted at wall ends [Fig. 5] and dividing the length of 83 mm (Fig. 5)—with results obtained with the
difference in the axial displacements by the distance between model. The concrete strain gauge data were available up to
the potentiometers). Again, very good agreement between data point number 330 (0.75% drift level), at which time the
the experimental and analytical results is observed, indi- gauge failed. The analytical prediction again underestimates
cating that the model successfully predicts the nonlinear the measured compressive strains and overestimates the

202 ACI Structural Journal/March-April 2006

Fig. 12—Wall RW2, concrete strain gauge measurements
versus predictions.

measured tensile strains (which are believed to be unreliable,

as discussed as follows). Given the differences in the Fig. 13—Wall TW2, measured versus predicted load-defor-
measured and predicted results, additional comparisons were mation responses.
made using strain histories measured by the LVDTs located
on either side of the embedded concrete strain gauge. Results
obtained with the two LVDTs straddling the embedded
concrete strain gauge were used to estimate the strain at the
location of the embedded concrete gauge using linear
interpolation. Results are presented up to data point number
555, at which time the readings in the LVDT closest to the
wall edge became unreliable.
It is observed that the analytical tensile strain predictions
are in good agreement with the LVDT measurements in
tension, as mentioned in the discussion of Fig. 11. The readings
for the embedded concrete strain gauge are reasonably close
to those obtained with the LVDTs for compressive strains,
but the embedded gauge fails to correctly measure the tensile
strains (it appears that bond between the embedded concrete
strain gauge and the surrounding concrete is insufficient to
accurately measure the large tensile strains that develop at
wall boundaries). Similar data trends were observed for the
concrete strain gauge located at the opposite wall boundary
of Specimen RW2. Overall, the results indicate that the
analytical model underestimates the compressive strains but Fig. 14—Wall TW2, lateral displacement profiles.
predicts reasonably well the magnitude and variation of the
wall tensile strain histories, and that tensile strain measurements
for the embedded concrete gauge are not reliable. TW2. The correlation of the analytical and experimental
The underestimation of the compressive stresses does not results for the lateral load-top flexural displacement
apparently have a significant influence on the prediction of response (Fig. 13), story displacements (Fig. 14), and first-
the global flexural response (Fig. 8 to 10) for the modeled story displacement and rotations (Fig. 15) for TW2 resemble
wall and loading history used in this study. The flexural those for RW2, when the T-shaped specimen is subjected to
response of the wall is ductile and nonlinearity of the displacements in the positive direction (when the wall flange
response is dominated by yielding of the longitudinal steel. is in compression). Therefore, the model provides a reasonably
However, underestimating the concrete compressive strains good prediction of the response for a T-shaped wall with the
will result in the inability of the analytical model to accurately flange in compression, and the same conclusions noted for
simulate strength degradation (and ultimately, failure) of a RW2 apply to TW2. This result implies that the plane section
wall due to crushing of concrete. Furthermore, modeling of assumption, which assumes the entire flange is effective in
strength degradation and failure also must consider damage compression for all drift levels, is appropriate. For negative
localization effects, in which case the size of the elements displacements (when the wall flange is in tension), however,
used to model the wall will significantly influence the the analytical model overestimates (by up to 37%) the lateral
analytical response. The current model does not consider load capacity of the wall (Fig. 13), underestimates (by up to
damage localization and may not accurately simulate 21%) the lateral displacements at the first-story level (Fig. 14
strength degrading responses. and 15), and overestimates (by up to 25%) the rotations over
the bottom 0.76 m of the wall (Fig. 15). The reason for these
T-shaped wall, TW2 discrepancies between the analytical and experimental results is
Figure 13 to 18 compare the analytical model predictions the nonlinear tensile strain distribution experienced along
with experimental results for the T-shaped wall specimen, the flange and the web of the wall specimen during testing.

ACI Structural Journal/March-April 2006 203

Fig. 15—Wall TW2, lateral displacements and rotations at
first story.

Fig. 17—Wall TW2, concrete strain measurements by

LVDTs versus predictions.

Fig. 16—Wall TW2, measured concrete strain distributions

along wall flange versus predictions.

The concrete strains (LVDT readings along the bottom

229 mm of the wall) measured along the wall flange at peak
top displacement data points are plotted and compared with Fig. 18—Wall TW2, concrete strain gauge measurements
analytical results in Fig. 16 for selected drift levels. The versus predictions.
measured tensile strains follow a nonlinear distribution
along the width of the flange (Fig. 16), which cannot be be modified to account for the variation of longitudinal tensile
captured with the analytical model, which is based on a plane strain along the wall flange (for example, by implementation
section assumption that produces a uniform tensile strain of a nonlinear strain distribution relationship such as proposed
distribution along the flange. Because of this assumption, flange by Pantazopoulou and Moehle20).
tensile strains are overestimated, leading to overestimation Results presented in Fig. 16 also reveal that the analytical
of the lateral load capacity of the wall when the flange is in model for the wall Specimen TW2 significantly underestimates
tension (Fig. 13). The experimentally observed nonlinear the compressive strains in concrete (as was the case for
tensile strain distribution along the flange is also the reason RW2). Concrete strains measured by the LVDTs along the
for overestimation of the first-story rotations and the under- length of the specimen web (including the web-flange inter-
estimation of the first-story displacements (Fig. 15) by the section) are plotted for peak top displacement points for
analytical model. The first-story displacements predicted by selected lateral drift levels in Fig. 17. The distribution of
the model are lower (despite overestimation of the inelastic concrete tensile strains along the web is predicted with
rotations due to large tensile strains) because the length of reasonable accuracy; however, the prediction of the position
the plastic hinge region (the height over which steel yielding of the neutral axis is relatively poor for negative displacements
is observed) is larger in the analytical model than that due to the inability of the model to consider the observed
experienced by the wall specimen during testing. In contrast, nonlinear tensile strain distribution along the flange. Further-
the measured compressive strains along the flange (Fig. 16) more, the measured tensile strains in concrete (Fig. 17) tend
are approximately uniform for all applied drift levels, resulting to decrease suddenly at the web-flange intersection due to
in fairly good analytical prediction of the global response for the change in cross section geometry. This is again not
positive displacements. For an improved prediction of T-shaped considered by the analytical model, which assumes a linear
wall response when wall flange is in tension, the model should distribution of strains along the web of the wall, resulting in

204 ACI Structural Journal/March-April 2006

overestimation of tensile strains at web-flange intersection The present model provides a flexible platform to address
(Fig. 17), as well as other aforementioned discrepancies these issues and to assess the influence of various material
between the predicted and measured responses when the and wall attributes on the nonlinear response of slender RC
flange of the wall is in tension. structural walls, as well as a practical platform for imple-
The longitudinal strain history for concrete predicted by menting further improvements. Implementation of the model
the analytical model is compared to results obtained with into a computational platform (refer to Reference 21, for
embedded concrete strain gauges (Fig. 5), as well as to example) will provide design engineers improved analytical
results obtained by linear interpolation between LVDTs capabilities to model the behavior of structural walls and their
located at the wall boundary (Fig. 5), in Fig. 18. It is again interaction with other structural elements, which is essential
evident that the analytical model underestimates the for the application of performance-based design.
compressive strains, but predicts the tensile strain history
reasonably well. The concrete strain gauge measurements ACKNOWLEDGMENTS
The work presented in this paper was supported by funds from the National
are in reasonable agreement with the LVDT readings for Science Foundation under Grants CMS-9632457 and CMS-9810012, as
compressive strains, but the embedded gauges do not accurately well as in part by the Earthquake Engineering Research Centers Program of
measure tensile strains. the National Science Foundation under NSF Award No. EEC-9701568
through the Pacific Earthquake Engineering Research (PEER) Center. The
test results used in this study were conducted by J. H. Thomsen, now with
SUMMARY AND CONCLUSIONS SGH, Arlington, Mass. The assistance of UCLA PhD students L. Massone
The intent of this paper was to convey detailed information and M. Melek is greatly appreciated. Any opinions, findings, and conclusions
on the calibration of the multiple-vertical-line-element or recommendations expressed in this material are those of the authors and
do not necessarily reflect those of the National Science Foundation.
model (MVLEM) and present comprehensive correlation
studies between the analytically predicted and experimentally
NOTATION
observed behavior of slender RC walls with rectangular and Ag = gross concrete area
T-shaped cross sections at various response levels. State-of- b = steel strain hardening ratio
the-art, robust constitutive relationships implemented into db = diameter of reinforcing bar
the MVLEM for concrete and reinforcing steel were calibrated E0 = elastic modulus for steel
Ec = elastic modulus for concrete
based on the mechanical properties of the materials used in Epl = cyclic plastic unloading modulus for concrete
the construction of the walls modeled, and by parameters Esec = cyclic secant unloading modulus for concrete
previously verified by other researchers (for example, for fc′ = peak concrete compressive stress (concrete compressive
confinement, tension stiffening, and cyclic stress-strain strength)
behavior of steel and concrete). Wall test results were ft = peak concrete tensile stress (concrete tensile strength)
R = cyclic curvature coefficient for steel
processed to allow direct comparisons between experimental R0, a1, a2 = steel parameters defining degradation of cyclic curvature
and analytical results at various response levels and locations (Bauschinger’s effect)
(for example, story displacements, rotations over the first- r = parameter defining shape of monotonic stress-strain
story level, and average strains). curve for concrete
∆f = cyclic reloading stress offset for concrete
Overall, the MVLEM, as used in this study, was shown to ∆ε = cyclic reloading strain offset for concrete
be an effective modeling approach for the flexural response ε0 = steel strain at intersection of elastic and yield asymptotes
prediction of slender RC walls, as the model provides good εc′ = monotonic strain at peak concrete compressive stress
predictions of the experimentally observed responses (wall εc0 = concrete strain at which cyclic tension envelope originates
εcr = concrete strain where monotonic stress-strain relation
lateral load capacity and lateral stiffness at varying drift starts following straight line
levels, yield point, cyclic properties of the load-displacement εpl = cyclic plastic concrete strain
response, displacement profile, average rotations and εr = steel strain at reversal point
displacements over the region of inelastic deformations [that εt = monotonic strain at peak concrete tensile stress
εy = steel yield strain
is, within first-story height], position of the neutral axis, and σ0 = steel stress at intersection of elastic and yield asymptotes
wall tensile strains). σr = steel stress at reversal point
The authors recommend the modeling approach, constitutive σy = steel yield stress
relationships, and calibration methodology used in this study ξ = absolute steel strain difference between current asymptote
intersection point and strain at previous reversal point
for a reliable prediction of the flexural response for slender
RC walls, with the following limitations: 1) the model
REFERENCES
significantly underestimates the compressive strains and 1. Vulcano, A.; Bertero, V. V.; and Colotti, V., “Analytical Modeling of
thus may not be accurate in simulating strength degradation RC Structural Walls,” Proceedings of the 9th World Conference on Earth-
and failure of walls due to crushing of concrete; 2) the quake Engineering, V. 6, Tokyo-Kyoto, Japan, 1988, pp. 41-46.
model, as adopted, does not consider the nonlinear tensile 2. Orakcal, K.; Wallace, J. W.; and Conte, J. P., “Nonlinear Modeling
and Analysis of Reinforced Concrete Structural Walls,” ACI Structural
strain distribution observed along the flange of T-shaped Journal, V. 101, No. 3, May-June 2004, pp. 688-698.
walls and thus is not accurate in predicting the T-shaped wall 3. Thomsen, J. H., and Wallace, J. W., “Displacement-Based Design of
response when the flange of the wall is in tension; 3) Reinforced Concrete Structural Walls: An Experimental Investigation of
although the model considers axial-flexural response Walls with Rectangular and T-Shaped Cross-Sections,” Report No. CU/
CEE-95/06, Department of Civil Engineering, Clarkson University, Postdam,
coupling, and thus would be expected to be effective for N.Y., 1995, 353 pp.
cases of varying axial load, present correlation studies were 4. Thomsen IV, J. H., and Wallace, J. W., “Experimental Verification of
made for walls under approximately constant axial load; and Displacement-Based Design Procedures for Slender Reinforced Concrete
4) the modeling methodology, as applied in this paper for Structural Walls,” Journal of Structural Engineering, ASCE, V. 130, No. 4,
2004, pp. 618-630.
correlation studies, is intended to simulate only flexural 5. Menegotto, M., and Pinto, E., “Method of Analysis for Cyclically
response and effects of shear-flexure interaction on possible Loaded Reinforced Concrete Plane Frames Including Changes in Geometry
shear failure mechanisms are not considered. and Non-Elastic Behavior of Elements Under Combined Normal Force and

ACI Structural Journal/March-April 2006 205

Bending,” Proceedings, IABSE Symposium on Resistance and Ultimate 13. Elmorsi, M.; Kianush, M. R.; and Tso, W. K., “Nonlinear Analysis of
Deformability of Structures Acted on by Well-Defined Repeated Loads, Cyclically Loaded Reinforced Concrete Structures,” ACI Structural Journal,
Lisbon, 1973, pp. 15-22. V. 95, No. 6, Nov.-Dec. 1998, pp. 725-739.
6. Filippou, F. C.; Popov, E. G.; and Bertero, V. V., “Effects of Bond 14. Seckin, M., “Hysteretic Behavior of Cast-in-Place Exterior-Beam-
Deterioration on Hysteretic Behavior of Reinforced Concrete Joints,” Column-Slab Sub-assemblies,” dissertation, University of Toronto, Toronto,
EERC Report No. UCB/EERC-83/19, Earthquake Engineering Research Ontario, Canada, 1981, 266 pp.
Center, University of California, Berkeley, Calif., 1983, 184 pp.
15. Mander, J. B.; Priestley; M. J. N.; and Park, R., “Theoretical Stress-
7. Chang, G. A., and Mander, J. B., “Seismic Energy Based Fatigue
Strain Model for Confined Concrete,” Journal of Structural Engineering,
Damage Analysis of Bridge Columns: Part I—Evaluation of Seismic
ASCE, V. 114, No. 8, 1988, pp. 1804-1826.
Capacity,” NCEER Technical Report No. NCEER-94-0006, State University
of New York, Buffalo, N.Y., 1994, 222 pp. 16. Saatcioglu, M., and Razvi, S. R., “Strength and Ductility of Confined
8. Belarbi, H., and Hsu, T. C. C., “Constitutive Laws of Concrete in Tension Concrete,” Journal of Structural Engineering, ASCE, V. 118, No. 6, 1992,
and Reinforcing Bars Stiffened by Concrete,” ACI Structural Journal, pp. 1590-1607.
V. 91, No. 4, July-Aug. 1994, pp. 465-474. 17. Orakcal, K., “Nonlinear Modeling and Analysis of Slender Reinforced
9. Wallace, J. W., “A New Methodology for Seismic Design of Reinforced Concrete Walls,” dissertation, University of California, Los Angeles, Calif.,
Concrete Shear Walls,” Journal of Structural Engineering, ASCE, V. 120, 2004, 224 pp.
No 3, 1994, pp. 863-884. 18. The Math-Works, Inc., “Matlab,” Natick, Mass., 2001.
10. Wallace, J. W., “Seismic Design of Reinforced Concrete Shear
19. Craig, R. R., Mechanics of Materials, John Wiley and Sons, New
Walls; Part I: New Code Format,” Journal of Structural Engineering,
York, 2000, 752 pp.
ASCE, V. 121, No. 1, 1995, pp. 75-87.
11. ACI Committee 318, “Building Code Requirements for Reinforced 20. Pantazopoulou, S. J., and Moehle, J. P., “Simple Analytical Models
Concrete (ACI 318-89) and Commentary (318R-89),” American Concrete for T-Beams in Flexure,” Journal of Structural Engineering, ASCE, V. 114,
Institute, Farmington Hills, Mich., 1989, 353 pp. No. 7, 1988, pp. 1507-1523.
12. Massone, L. M., and Wallace; J. W., “Load-Deformation Responses 21. Pacific Earthquake Engineering Research Center, “OpenSees—Open
of Slender Reinforced Concrete Walls,” ACI Structural Journal, V. 101, System for Earthquake Engineering Simulation,” University of California,
No. 1, Jan.-Feb. 2004, pp. 103-113. Berkeley, Calif., http://opensees.berkeley.edu/OpenSees/developer.html.

206 ACI Structural Journal/March-April 2006

View publication stats