SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

Recommendations for Modeling the Nonlinear Response of Flexural Reinforced Concrete Walls
Using Perform

Laura N. Lowes, Dawn E. Lehman, and Carson Baker

Synopsis: The PERFORM-3D software package is used commonly in engineering practice to conduct nonlinear
dynamic analyses of reinforced concrete walled buildings to their seismic response. However, few studies have
evaluated or improved on common modeling approaches for structural concrete walls. The research presented here
was conducted to establish best practices for modeling the full nonlinear response of walls exhibiting common flexural
failure modes. First, an experimental data set consisting of eight planar concrete walls was collected; these walls were
spanned a range of length-to-thickness ratios, shear stress demands, axial load ratios, and longitudinal reinforcement
configurations. For each wall specimen, a reference numerical model was created using typical modeling methods as
proposed by Powell. Comparison of simulated and measured cyclic response histories show that typical modeling
techniques result in relatively inaccurate simulation of cyclic response and very inaccurate simulation of drift capacity.
To improve the model accuracy, experimental data were used to determine appropriate values for the steel and concrete
material model cyclic response parameters. Experimental data and mathematical definitions for the concrete
compressive energy were used to develop recommendations for defining concrete post-peak stress-strain response to
achieve accurate, mesh-independent simulation of drift capacity. Finally, recommendations for the minimum number
of elements were examined. Comparison of simulated and measured cyclic response histories show that the new
modeling recommendation result in accurate, mesh independent simulation of cyclic response, including drift capacity.
Future work will evaluate the proposed modeling approach for asymmetric and flanged walls.

Keywords: structural walls, finite element modeling, seismic response, flexure

173
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

ACI member Laura Lowes is a Professor in Civil and Environmental Engineering. She received her BSCE from the
University of Washington in 1992, her MSCE and PhD from UC Berkeley, in 1993 and 1999, respectively. Her
research expertise lies in the area of nonlinear modeling of reinforced concrete structures. She has conducted extensive
research on reinforced concrete frames and walls. She is a member of ACI Committees 318C, 369 and 447.

ACI member Dawn Lehman is a Professor in Civil and Environmental Engineering. She received her BSCE from
Tufts University. After working as a practicing structural engineering in Boston MA, she went on to receive her MEng
in 1992 and her PhD in 1998 from UC Berkeley. Her research expertise lies in experimentation and computational
simulation of large-scale structures subjected to seismic loading. She has developed new structural systems for bridges
and buildings in regions of high seismicity. She is a member of ACI Committees 318H, 341, and 352.

Carson Baker received his MS in Civil Engineering in 2016 from the University of Washington. He is a practicing
project structural engineer with Coughlin Porter Lundeen in Seattle WA.

INTRODUCTION

PERFORM-3D is used commonly to assess the earthquake performance of buildings that employ slender, flexure-
controlled concrete to resist lateral loads. As such, modeling parameters for use with PERFORM-3D are required that
result in accurate estimation of the response of concrete walls under cyclic loading, including accurate estimation of
the stiffness, strength, hysteretic response, and deformation capacity. Verification that a structure will maintain lateral
loading carrying capacity when subjected to specified earthquake demands is a critical component of earthquake
performance assessment. Thus, accurate estimation of drift capacity is particularly important.

To better understand flexural wall failure mechanisms, which determine wall drift capacity, the results of previous
experimental investigations of walls are reviewed. Birely (2012) evaluated experimental data from over 60 laboratory
tests of slender (shear-span ratio greater than 2.0) wall specimens and found that failure, i.e. degradation in the lateral
capacity, of these walls is a result of one or more damage modes: (1) compression-buckling (CB) characterized by
compression damage to the boundary element including crushing of boundary-element concrete and buckling of
boundary element reinforcing steel (Fig. 1a), (2) buckling-rupture (BR) characterized by low-cycle fatigue of the
reinforcement, which includes buckling and subsequent fracture of longitudinal steel (Fig. 1b), and (3) shear. Fig. 2
shows the observed damage modes of the slender-wall database by Birely. As indicated in the plot, the majority of the
walls exhibited a compression-controlled failure mode (i.e., boundary element crushing). It is of note that all of the
walls that exhibited compression-controlled failure satisfied the ACI tension-controlled design limit (ACI 318 2014);
therefore, simply designing a wall to meet the tension-control criterion does not necessarily result in tension-controlled
failure. Thus, accurate estimation of drift capacity for walls exhibiting flexure-controlled response required accurate
estimation of the compression-controlled failure mechanism, including concrete crushing and reinforcement buckling.
Where most prior work has focused on accuracy in terms of strength, it is well known that flexural strength can be
accurately predicted using most programs and modeling approaches where drift capacity, needed for modeling RC
systems beyond maximum strength, is not. As such, this study focused on improving accuracy of models to predict
drift capacity. The recommendations are to improve modeling accuracy in terms of both of these global response
parameters.

EXPERIMENTAL DATA SET USED FOR MODEL CALIBRATION AND VALIDATION

A comprehensive literature review by Whitman (2015) identified 22 planar wall test specimens that exhibited flexure-
controlled response when tested in the laboratory under quasi-static lateral loading and that were appropriate for use
in model development. Wall specimens included in the Whitman data set if they met the following conditions: i) the
specimen was planar (rectangular) and subjected to in-plane flexure, shear and axial loading with lateral load applied
cyclically under displacement control, ii) specimens were constructed of normal weight concrete and normal strength
concrete and steel, iii) specimen failure resulted from deteriorating flexural response, including concrete crushing, bar
buckling and/or bar fracture, iv) wall specimen thickness exceeded 76 mm (3 in.), and v) data required to fully define
and evaluate a numerical model were available in the literature. Wall specimens were chosen from the Whitman study
for use in the current study i) if they exhibited a compression-controlled flexural failure and ii) to represent the full
range of boundary element detailing from no boundary element confinement to boundary element confinement
exceeding current ACI Code requirements by restraining every longitudinal bar. Eight wall specimens were chosen

174
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

from this data set for use in the current study. An eighth specimen was chosen from the Whitman data set for use in
the current study that exhibited a buckling-rupture type failure.

(a)

BR CB Shear
Figure 2— Failure modes observed in laboratory tests of
slender concrete walls (Birely 2012)
(b)

Figure 1— Typical damage modes in slender walls

(Pugh et al. 2015)

 Table 1 list the specimens and provides design parameters of particular interest to this study. Table 2 provides
demand and response parameters. Whitman (2015) provides additional information of the specimens. Parameters
listed in Table 1 and 2 are defined as follows:
 Wall length, lw = the in-plane length of the wall.
 Cross-Sectional Aspect Ratio (CSAR) = lw/tw, where lw is the wall length and tw is the wall thickness.
 Shear span ratio = M/(V*lw), where M is the moment developed at the base of the wall, V is the shear developed
at the base of the wall. Note that the shear span ratio equals the wall aspect ratio if zero moment is applied at the
top of the wall.
 Est. No. of stories = number of stories estimated by assuming a story height equal to 12 times the wall width.
Note that PERFORM analyses employed a specified number of elements per story; for the top story this number
was adjusted based on the height of the top story.
 f'c = measured concrete compressive strength.
 fy,BE = measured tensile yield strength for boundary element longitudinal steel.
 ρlong = longitudinal reinforcement ratios for boundary element, web region and gross section.
 ρvol_BE = volumetric reinforcement ratio for wall boundary element confining reinforcement.
 Mander confinement ratio = The ratio of confined to unconfined concrete compressive strength, 𝑓𝑐𝑐′ /𝑓𝑐′ determined
using the confinement model by Mander et al. (1988) as presented by Paulay and Priestley (1992).
 Boundary element detailing indicates that confining reinforcement meets the ACI 318-14 Code requirements for
a special structural wall (SBE), an ordinary structural wall (OBE) or neither (NBE); no confining reinforcement
is provided for “unconfined” specimens.
 Axial load ratio (ALR) = P/(Agf’c), where P is the axial load at the base of the wall (including self-weight of the
specimen computed assuming a unit weight of 150 lb/ft3), Ag is the gross area of the wall and f’c is the measured
concrete compressive strength.

175
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

 Maximum shear stress demand vmax = Vb / Acv√𝑓𝑐′ , where Vb is the maximum base shear developed during the
test, Acv is the shear area, taken equal to the gross section area.
 Shear demand-capacity ratio = Vb/Vn, where Vb is the maximum base shear and Vn is the shear strength computed
per ACI 318 (2014) using measured concrete and steel strengths.
 Flexural strength ratio = Mb/Mn, where Mb is the maximum base moment developed during the test and Mn is the
nominal flexural strength of the wall corresponding to a compressive strain of 0.003 at the extreme fiber using
measured concrete and steel strengths (defined above.
 y = yield drift which is the drift at first yield of longitudinal reinforcement as determined from a section analysis
assuming zero shear deformation.
 u = drift capacity = drift at which the lateral load carrying capacity of the wall dropped to 80% of the maximum,
for drift demands larger than drift corresponding the maximum strength.
 Failure mode (FM) indicates the primary mechanism causing loss of lateral load carrying capacity: compression-
buckling, characterized by simultaneous crushing of concrete and buckling of longitudinal steel (CB) or buckling-
rupture (BR), characterized by buckling of reinforcing steel followed by bar rupture.

Table 1— Wall test specimens used for model development – design parameters
Wall Est. Mander
Wall CSAR 𝑀𝑏 𝑓𝑐′ 𝑓𝑦,𝐵𝐸 𝜌𝑙𝑜𝑛𝑔 𝜌𝐵𝐸 Boundary Boundary
Reference Length No. of Conf.
ID 𝑙𝑤 /𝑡 𝑉𝑏 𝑙𝑤 (ksi) (ksi) (%) (%) Element Element Detail
(in.) Stories Ratio

Dazio et
WSH4 79 13.3 2.28 2.5 5.93 84.6 0.85 N/A 1.0 Unconf.
al. (2000)

Oh et al.
WR0 59 10.0 2.00 1.7 4.77 65.1 0.62 N/A 1.0 Unconf.
(2002)

Oesterle et
R1 75 18.8 2.40 3.8 6.49 74.2 0.49 1.47 1.02 NBE
al. (1976)

Wallace
and
RW1 48 12.0 3.13 3.1 4.58 63.0 1.15 2.93 1.06 NBE
Thomsen
(1995)
Wallace
and
RW2 48 12.0 3.13 3.1 6.33 63.0 1.15 2.93 1.09 NBE
Thomsen
(1995)
RW-
A20- Tran
48 8.0 2.00 1.3 6.83 68.4 1.29 3.23 1.26 SBE
P10- (2012)
S38
RW-
A20- Tran
48 8.0 2.00 1.3 7.05 69.2 2.83 7.11 1.27 SBE
P10- (2012)
S63

Dazio et
WSH6 79 13.3 2.26 2.5 6.61 79.8 0.85 1.38 1.24 xSBE
al. (2000)

176
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

Table 2— Wall test specimens used for model development – demand and response parameters
𝑃 𝑉𝑏 𝑉𝑏 𝑴𝒃 𝚫𝒚 𝚫𝒖 Failure
Wall ID
𝐴𝑔 𝑓𝑐′ ′
𝐴𝑐𝑣 √𝑓 𝑐 𝑉𝑛 𝑴 𝒏 (%) (%) Mode
WSH4 0.06 2.8 0.62 1.06 0.29 1.60 CB
WR0 0.11 3.0 0.74 1.08 0.52 2.14 CB
R1 0.00 1.1 0.23 1.17 0.17 2.30 BR
RW1 0.11 2.6 0.50 1.07 0.48 2.26 CB
RW2 0.09 2.7 0.52 1.16 0.55 2.35 CB
RW-A20-P10-S38 0.07 3.8 0.81 1.26 0.55 3.18 CB
RW-A20-P10-S63 0.07 6.3 0.91 1.13 0.66 3.00 CB
WSH6 0.11 3.6 0.83 1.11 0.31 2.04 CB

PERFORM ELEMENT FORMULATIONS AND MATERIAL MODELS FOR MODELING WALLS

Nonlinear analysis of flexure-controlled concrete walls using PERFORM typically employs either the “shear wall
element” or the “general wall element”. The general wall element is intended for analysis of complex reinforced
concrete walls with irregular openings and was not considered in this study. This paper considers only the shear wall
element. The shear wall element is i) intended to represent the nonlinear response of concrete walls subjected to in-
plane axial and lateral loading, such as the planar wall test specimens presented in the previous section, ii) can be used
to simulate nonlinear flexural response, nonlinear shear response or both, though flexural and shear response models
are decoupled, and iii) is used commonly in practice.

PERFORM manuals and user guides provide detailed information about the shear wall element (see
www.csiamerica.com/); a brief discussion of the element formulation is provided here. The element is a displacement-
based 4-node macro-element with three rotational and three translational degrees of freedom per node. The macro-
element formulation combines three response models to simulate wall behavior: i) a fiber-type section model,
comprising linear elastic or nonlinear concrete and steel fibers, simulates in-plane flexural response, ii) a uniform
shear layer, with a one-dimensional linear elastic or nonlinear shear response model, simulates in-plane shear response,
and iii) a uniform linear-elastic plate-bending model represents out-of-plane response. Nodal displacements and
rotations are used to compute a constant in-plane shear deformation, a constant in-plane curvature and a constant axial
deformation; these three deformations determine the in-plane shear, moment and axial load developed in the wall.

The fiber-type section model is used to simulate the nonlinear flexural response of the wall cross section. The user
discretizes the wall cross section into a series of unconfined and confined concrete fibers overlaid with steel fiber;
typically, the area occupied by the reinforcing steel is “double counted” as it is assigned both to the steel fibers and
the concrete fibers. A one-dimensional nonlinear stress-strain material response model is assigned to the unconfined
concrete, the confined concrete and the reinforcing steel. PERFORM limits the number of fibers that can be assigned
to a single wall element; thus, multiple elements are often used along the in-plan length of the wall.

For assessment of wall performance under earthquake loading, nonlinear one-dimensional material response models
are used to define concrete and steel fiber response and, occasionally, to define shear response. Typically, nonlinear
material response is defined using the “YULRX” model (Fig. 3) to define the envelope to the stress-strain history; the
“YULRX” model can represent and envelope that is trilinear or trilinear to the point of strength loss. PERFORM does
not support deterioration of the YULRX envelope due to cyclic loading; strength deterioration due to cyclic loading
is implicitly included in the YULRX envelope. Thus, the YULRX envelope used for monotonic loading would be
different (stronger or with strength loss occurring at a larger deformation demand) than for cyclic loading. PERFORM
does support deterioration of unloading/reloading stiffness.

177
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

Figure 3— Action-deformation envelope for PERFORM ‘YULRX’ material model (PERFORM-3D user guide 2006)

INVESTIGATION AND CALIBRATION OF SHEAR WALL ELEMENT AND CONSTITUTIVE

MODELS

As suggested by the discussion above, building a PERFORM model of a reinforced concrete wall requires the engineer
to make a number of modeling decisions and input multiple model parameters. Data from the experimental tests listed
in Table 1 were used to investigate the impact on model accuracy of various modeling decisions and user-defined
model parameters and, ultimately, develop modeling recommendations to achieve accurate simulation of stiffness,
strength, cyclic response and deformation capacity. The subsections below first present an investigation of the
accuracy with which wall response is simulated using the “typical modeling approach” and then present investigations
of the impact on response simulation and recommendations for i) mesh refinement as defined by the number of
elements used along the length and up the height of the wall, ii) degradation of unloading and reloading stiffness due
to cyclic loading, and iii) regularization of a material response models to minimize mesh-sensitivity and provide
accurate stimulation of deformation capacity. Table provides modeling parameters values used in the investigation.

Evaluation of the Current Modeling Recommendations

Investigation of wall modeling using PERFORM started with current modeling recommendations, referred to herein
as the reference modeling approach. The reference model was characterized by i) one shear wall element used to
represent the entire length of the wall, ii) two shear wall elements used to represent the height of the first story, where
flexural yielding is expected, and one element used to represent the entire height of upper stories, iii) no degradation
in energy dissipation or unloading/reloading stiffness for steel, iv) no degradation in energy dissipation for concrete
in compression, and v) use of standard concrete and steel stress-strain models without regularization.

The reference model employed a level of mesh refinement that is consistent with recommendations for PERFORM
modeling found in the literature. Two elements were used vertically to represent the bottom story of the wall and one
element was used vertically to represent upper stories; each element spanned the entire in-plan length of the wall.
PERFORM limits the number of fibers that can be used for a single element to 16. As such, eight concrete fibers and
eight steel fibers were used. Concrete fibers were distributed as three per boundary element and two in web; steel
fibers were distributed as two per boundary element and four in the web.

Unconfined and confined concrete stress-strain models were defined and assigned, respectively, to web and boundary
element concrete fibers. The PERFORM 1D concrete model was employed with the YULRX model used to define
the tension and compression envelopes. Concrete elastic modulus in tension and compression was defined per ACI
318 as 57000√𝑓𝑐′ psi with 𝑓𝑐′ in psi. Table lists the stress (F*) and strain (D*) values used to define the YULRX

178
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

envelope for compression response, with confined concrete strength, 𝑓𝑐𝑐′ , defined per by Mander et al. (1988) and
concrete strain capacity defined per Paulay and Priestley (1992) as
𝜀𝑐𝑚𝑎𝑥 = 0.004 + 1.4𝜌ℎ 𝑓𝑦ℎ 𝜀ℎ𝑚 /𝑓𝑐𝑐′ Eq. 1
where 𝜌ℎ is the volumetric ratio of confining steel and 𝑓𝑦ℎ and 𝜀ℎ𝑚 are, respectively, the yield strength and strain at
maximum strength of the confining reinforcement. Fig. 4 shows typical compressive stress-strain envelopes for
unconfined and confined concrete.

Table 3— Concrete compression stress-stain material model parameters

DU DL DR
FY FU FR/FU
(in/in) (in/in) (in/in)
Unconfined 0.75𝑓𝑐′ 𝑓𝑐′ 0.001 0.002 0.00202 0.01
Confined 0.75𝑓𝑐𝑐′ 𝑓𝑐𝑐′ 0.200 0.004 0.00404 𝜀𝑐𝑚𝑎𝑥

Figure 4— Typical compressive stress-strain envelopes for unconfined and confined concrete

Reinforcing steel stress-strain models were defined and assigned, respectively, to web and boundary element steel
fibers. The PERFORM 1D Inelastic Steel model without buckling was employed. A single YULRX model was used
to define the stress-strain envelope for tension and compression. Measured steel material properties were used to
define the YULRX envelope; where strain hardening was not well defined by the available data, a hardening modulus
equal to 1.5% of the elastic modulus was assumed. Fig. 5 shows a typical reinforcing steel stress-strain response
envelope.

Figure 5— Typical stress-strain envelope for reinforcing steel

179
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

An elastic shear material was used for the reference model. Shear stiffness was defined equal to 0.1G cAg per the
recommendations of the ATC 72-1 (2010) with Gc = 0.4Ec per ACI 318 and Ag taken equal to the gross area of the
wall section.

Fig. 6 shows the simulated and measured response histories for Specimen RW1, with simulated response achieved
using the reference model. The most obvious observation is that the reference model results in a simulated drift at
onset of lateral strength loss that is approximately double that observed in the laboratory. Beyond this, the reference
model is observed to provide a relatively poor representation of cyclic response. In comparison with the measured
response history, the reference model does not simulate the gradual reduction in unloading stiffness observed in each
cycle or the graduate loss of reloading stiffness observed in subsequent cycles. Similar results were observed when
the reference model was used to simulate the response of other specimens.

Figure 6— Load-Displacement response for wall specimen rw1 as measured and as simulated using the reference
model

Cyclic response parameters for reinforcing steel

For reinforcing steel, unloading and reloading stiffnesses are defined by the energy dissipation factor and the stiffness
factor. The energy dissipation factor defines the extent to which stiffness loss reduces energy dissipation; it equals the
ratio of the energy dissipated during a stress-strain cycle for the material with stiffness loss to the energy dissipated
without stiffness loss. The stiffness factor varies from negative to positive 1.0 and determines whether unloading
stiffness or reloading stiffness or both are reduced. Fig. 7 shows four possible stress-strain cycles for reinforcing steel.
The dashed lines in Loops A-C show the same stress-strain cycle with no stiffness loss. The solid lines in Loops A-C
show three possible stress-strain cycles with stiffness loss. All three of the stress-strain cycles with stiffness loss were
generated using an energy dissipation factor of 0.5. Loops A-C, however, employ different stiffness factors: in Loop
A unloading stiffness is reduced (stiffness factor = -1), in Loop C reloading stiffness is reduced (stiffness factor = 1)
and in Loop B both unloading and reloading stiffness are reduced (stiffness factor = 0). Experimental data from tests
of reinforcing bars (Mat et al., 1976; Panthaki, 1991, Mander 1994) were used to determine the parameters that provide
a best-fit to the steel cyclic stress-strain data:

Energy dissipation factor providing a best fit to cyclic reinforcing steel data = 0.75
Stiffness factor providing a best fit to cyclic reinforcing steel data = 0.5

Fig. 9 shows measured and simulated response for two wall specimens in Table 1 (a,b) with default cyclic material
response parameters, (c) with the proposed reinforcing steel and concrete cyclic response parameters, and (d,e) with
the proposed reinforcing steel and concrete cyclic response parameters as well as material regularization discussed
below. These data show that the calibrated cyclic response parameters result in significantly improved simulation of
measured response; similar results were observed for other wall specimens.

180
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

Investigation of cyclic response parameters for reinforcing steel and concrete

The accuracy of simulated response histories can be improved by employing user-defined cyclic calibration factors
for reinforcing steel and concrete material models. Specifically, PERFORM allows the user to specify i) the extent to
which unloading and/or reloading stiffnesses deviate from that associated with an elastic-plastic material with linear
hardening and ii) if this deviation is constant throughout the load history or changes with strain demand. Different
approaches are used for reinforcing steel and concrete, which exhibit very different material response.

(a) (b) (c)

Figure 7— Three options for simulating stiffness loss under cyclic loading (Figure 5.14 from the PERFORM-3D
user guide 2006).

(a) (b)

(c)

Figure 8— Load-Deformation response as measured and as simulated using (a,b) default material models and (c,d,e)
recommended steel and concrete energy dissipation and stiffness factors (c) without and (d,e) with regularization of
material response)

181
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

(d) (e)

Figure 9 (continued)— Load-deformation response as measured and as simulated using (a,b) default material
models and (c,d,e) recommended steel and concrete energy dissipation and stiffness factors (c) without and (d,e)
with regularization of material response)

Cyclic response parameters for concrete

PERFORM allows users to calibrate the concrete cyclic response history in the compression. Under tensile loading,
concrete cracks at a relative low stress demand and then exhibits rapid strength loss; unloading is characterized by a
“damage” material response, with all inelastic deformation attributed to stiffness loss and unloading resulting in zero
residual deformation. This behavior is represented exactly by the PERFORM material model, and, thus, mechanisms
for user calibration of this response are not provided. Fig 19a shows a typical measured cyclic stress-strain history for
concrete subjected to compressive loading. These data show reduction in unloading and reloading stiffness with
increasing compressive strain demand; the blue regions in this figure represent energy dissipated during the unload-
reload cycle. PERFORM provides a mechanism for reducing reloading stiffness but not unloading stiffness; reduction
in reloading stiffness is defined by a user-specified relationship between energy dissipation and compressive strain.
Fig. 11. shows the relationship between the energy factor and the reloading stiffness. An energy factor of 1.0 results
in initiation of reloading at zero compressive strain, minimum reloading stiffness, and maximum energy dissipation;
an energy dissipation factor of 0.0 results in initiation of reloading at the maximum allowable compressive strain, a
maximum reloading stiffness and zero energy dissipation. Experimental data such as shown in Fig 10(a) were used to
determine appropriate energy dissipation values as a function of maximum compressive strain demand. Fig. 11 shows
data points and the proposed energy dissipation model; Table 4 provides proposed energy dissipation values for
PERFORM material states. Fig. 10 shows the simulated (b) stress-strain history compared with the measured history
(a). The proposed model provides a good representation of the measured strain at onset of reloading and reloading
stiffness. As stated above, the calibrated cyclic response parameters result in significantly improved simulation of
measured response (Fig. 9).

182
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

(a)

(b)

Figure 10— Measured (a) and simulated (b) concrete compression stress-strain history (Sihna et al., 1964)

Figure 11— Impact of perform energy dissipation factor on concrete compressive reloading stiffness

183
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

Figure 12— Measured and proposed concrete compressive energy dissipation factor as a function of compressive
strain as measured and as proposed

Table 4— Concrete Energy Dissipation Factors

Material Y U L R X
State (yield) (ultimate) (loss) (residual) (rupture)
Energy
1.0 0.4 0.4 0.1 0.1
Factor

Story 3

Story 2

Story 1

Reference 2 Elements 3 Elements 4 Elements

Model per Story per Story per Story

Figure 13— Shear wall element meshes used for reference model and vertical mesh refinement study (2, 3 and 4
elements per story)

184
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

Investigation of mesh refinement

Standard practice for modeling walls using the shear wall element is to use one or two elements per story and one or
two elements along the length of the wall. Classical finite element analysis requires multiple analyses using different
levels of mesh refinement to verify that a “converged” solution is achieved. For the current study, the results of
previous research were used to develop recommendations for mesh refinement along the length of the wall. A vertical
mesh refinement study was conducted to determine the impact of mesh refinement on estimated performance and
whether or not the standard practice for meshing results in acceptable accuracy. Fig. 13 shows mesh configurations
used in the mesh refinement studies.

Horizontal mesh refinement

A single PERFORM shear wall element simulates constant curvature, and thus a linear vertical strain distribution,
along the in-plan length of the wall. However, experimental test data and data from nonlinear continuum analyses
show the actual vertical strain distribution in a planar wall to be nonlinear once inelastic action occurs. Fig. 13 shows
the average vertical strain distribution at the bottom of planar wall RW2; these strain values are computed using the
measured vertical deformation over an 18 inch vertical gage length at the base of the wall, with measurements taken
at seven locations along the length of the wall (diamond-shaped data points on plot). Fig. 14 shows the vertical strain
distributions for two planar walls as simulated using nonlinear continuum analysis (ATENA) and as simulated using
a fiber-section model in which the plane-sections-remain-plane (PSRP) assumption is imposed such that there is a
linear strain distribution. A PERFORM analysis employing a single wall element at the base of the wall results in a
plane-sections-remain-plane assumption. One of the simulated planar walls has a low cross-sectional aspect ratio
(CSAR = lw/t) and low shear stress demand; the other has a high CSAR and high shear stress demand. Both strain
profiles suggest that the need for multiple elements along the length of the wall to simulate a multilinear vertical strain
distribution. Based on the data in, it is recommended that four shear wall elements be used along the length of the wall
with one element used to model each boundary element (vertical black line indicates location of web-boundary
element interface) and two elements used for the interior web region as shown in Fig. 13. All subsequent analyses use
this level of horizontal mesh refinement.

CSAR = 12, vmax = 2.65f'c1/2

0.03
0.025 ATENA ATENA
PSRP PSRP
0.02
BE Location BE Location
0.015
0.01
Strain

0.005
0
-0.005
Drift = 0.50% Drift = 0.75%
-0.01
Drift = 1.00% Drift = 1.50%
-0.015 Drift = 2.00% Drift = 3.00%
-0.02
0 0.2 0.4 0.6 0.8 1
x/lw
Figure 14—Vertical strain profile for RW2 Figure 15— Simulated vertical strain profiles

Vertical mesh refinement

A single PERFORM shear wall element simulates a constant vertical strain over the height of the element; however,
the very nearly linear moment distribution that develops in a wall between floors suggests a linear strain distribution
and the need for multiple elements per story to represent this linear strain profile. Thus, a vertical mesh refinement
study was conducted; Fig. 13 shows the different meshes used in the study. Fig. 16 shows simulation results, from
which it can be concluded that vertical mesh refinement has minimal impact on simulated stiffness and strength but
that for the typical model, mesh size determines simulated deformation capacity. Similar results were observed for all
specimens listed in Table 1. Subsequent sections of this paper present modeling recommendations that achieve mesh-
objective simulation of response, such that mesh size does not determine simulated deformation capacity. Subsequent
analyses use different levels of vertical mesh refinement as identified in Table 6.

Regularization of material response: concrete and steel in compression

The simulated response histories presented in Fig. 16 shows that mesh size affects the drift at onset of significant
strength loss, which is subsequently referred to as drift capacity. This mesh sensitivity in simulation of drift capacity
has been observed by others (Coleman and Spacone 2001, Mohr 2007, Chiaramonte 2011, Pugh et al. 2015) using a
range of element formulations (beam-column, shell and solid) to simulate the response of a range of reinforced

185
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

concrete components (beams, columns, walls, and piles). Regularization of material response, using a measure of the
energy dissipated in post-peak region of the material response curve and a mesh-dependent length, has been used by
others (Coleman and Spacone 2001, Chiaramonte 2011, Pugh et al. 2015) to produce mesh-objective estimation of
response. This concept is employed here to develop recommendations for defining the concrete stress-strain response
curve used in the PERFORM wall element to achieve accurate, mesh-objective estimation of drift capacity in walls.

30% Strength Loss

Different levels of mesh

refinement results in
different displacement
capacities

Figure 16— Load-deformation response as measured and as simulated using different numbers of elements per story

Specimen WSH4 is classified as exhibiting a compression-buckling failure, with the measured strength loss shown in
Fig. 16 resulting from concrete crushing and simultaneous buckling of longitudinal reinforcement in the compression
region at the bottom of the wall. In the simulation, once concrete crushing results in reduced moment capacity for the
wall, deformation localizes in the failing/softening element at the bottom of the wall. Deformation in the softening
element at the base of the wall increases, but deformation in the elements above the critical bottom section drop as
these elements begin to unload. If a more refined mesh is used to simulate response, concrete crushing still localizes
in the bottom element, but now, because the reduced-height bottom element has the same stress-strain model as the
full-height element, the reduced-height element exhibits a substantially smaller deformation (deformation equals the
product of strain and element height) at a given element stress level / wall moment capacity. Thus, the more highly
refined mesh simulates initiation of strength loss at a lower drift level and more rapid strength loss with increasing
displacement demand.

Fig. 17 shows the mesh-dependent behavior observed in Fig. 16 at a global level and at the section level for wall
WSH4. These results were generated using the OpenSees (http:\\opensees.berkeley.edu) force-based beam-column
element and fiber-section model available in OpenSees with standard, unregularized material models; however,
similar results are obtained using PERFORM. Figs. 17(Figurea)-(c) show simulation results for different levels of
mesh refinement. Where mesh-refinement in PERFORM takes the form of using more elements along the height of
the wall, using the OpenSees beam-column element, mesh refinement takes the form of increasing the number of fiber
sections (integration points identified as I.P. in the figures) used along the length of the element. Fig 17a shows
measured and simulated load versus drift; the displacement at onset of strength loss and the rate of strength loss show
mesh-sensitivity similar to that shown in Fig. 16. Fig. 17(b) shows the simulated normalized maximum concrete
compressive strain distribution up the height of the wall prior to onset of strength loss. The data in this figure show
approximately the same distribution regardless of the level of mesh refinement; increasing mesh refinement provides
a smoother distribution and more accurate representation of the distribution over the wall height. Fig. 17(c) shows the
simulated normalized maximum concrete compressive strain distribution up the height of the wall at the onset of
strength loss. The data in this figure show very different distributions at the base of the wall with different levels of
mesh refinement. In this figure, increased mesh refinement results in substantially increased maximum concrete strain
at the bottom section of the wall where strength loss occurs. Above the bottom section, the strain distributions are not
affected by mesh refinement. It is this “localization” of material damage, characterized by increasingly large strain

186
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

demands, that results in increasingly rapid global strength loss. Behavior similar to that shown in Fig. 17 is observed
when the PERFORM wall elements are used to simulate response.

(a) Response of WSH4 as measured and as simulated using (b) PT. A: simulated maximum (c) PT. B: simulated maximum
OpenSees force-based beam-column elements and four concrete strain distribution concrete strain distribution
levels of mesh refinement (4, 5, 7 and 9 fiber sections per
element)

Figure 17— Simulated curvature and strain fields prior to and at failure (Pugh 2012)

Regularization of material response to mitigate mesh sensitivity has been used for many years in continuum analysis
of concrete structures. Regularization is based on observations from material tests that 1) damage localizes in a region
within the test specimen and that this region is approximately the same size regardless of specimen size, and 2) the
fundamental material response is stress versus deformation not stress versus strain. For concrete subjected to tensile
loading, damage localizes at the crack and the fundamental material response, which is consistent regardless of
specimen size, is crack stress versus crack width opening (a deformation measure). Concrete fracture energy, G f,
defined as the area under the softening portion of the tensile stress versus crack width opening history, is considered
a fundamental material property, and standardized tests exist for measuring this quantity (CEB-FIP 1990). Jansen and
Shah (1997) conducted compression tests of concrete cylinders of varying height and showed that damage localizes
in a crushing region that is approximately the same thickness regardless of cylinder height and that the softening
portion of the concrete compression stress versus deformation history is approximately constant regardless of
specimen height. Jansen and Shah (1997) developed recommendations for defining concrete crushing energy, G fc,
equal to the area under the softening portion of the compression stress versus deformation history (Fig. 18). Nakamura
and Higai (2001) conducted a study similar to that of Jansen and Shah (1997), found similar results and developed
recommendations for defining Gfc. Though, it should be noted that for normal weight concrete, Jansen and Shah
recommend a crushing energy of 0.143 kip/in while Nakamura and Higai recommend a crushing energy of 0.457
kip/in.

Figure 18— Concrete compression stress versus deformation and definition of concrete crushing energy, Gfc (image
from Jansen and Shah (1997))

187
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

Given the stress versus deformation history in Fig. 18Figure and the knowledge that stress versus deformation (Gfc)
in the post-peak regime is constant regardless of mesh size, regularized concrete stress-strain response models for
unconfined and confined concrete can be created for use in PERFORM (Fig. 19). Since there is no localization of
damage prior to achieving maximum strength, standard models may be employed to define the pre-peak portion of
the curves. To define the post-peak response, unconfined concrete is assumed to have no residual strength (strength at
point R is zero) and confined concrete is assumed to have residual strength equal to 20% of the confined compressive
strength (strength at Rcc = 0.2fcc). Strain at onset of residual strength is defined for unconfined concrete as
𝐺𝑓𝑐
( )
𝑓𝑐′ 𝐿𝑒𝑙𝑒𝑚
𝜀𝑅 = 𝜀𝑢 = 𝜀𝑜 − +2 ′ Eq. 2
𝐸𝑐 𝑓𝑐

and for confined concrete as

5 𝐺𝑓𝑐𝑐
′
0.8𝑓𝑐𝑐 ( )
3 𝐿𝑒𝑙𝑒𝑚
𝜀𝑅 𝑐𝑐 = 𝜀𝑢𝑐𝑐 = 𝜀𝑜𝑐𝑐 − + ′ Eq. 3
𝐸𝑐𝑐 𝑓𝑐

where 𝜀𝑅 is the strain at point R, which is often defined as and 𝜀𝑢 in concrete material models, 𝜀𝑜 is the compressive
strain at maximum concrete strength, 𝑓𝑐′ is the concrete compressive strength, 𝐸𝑐 is the concrete elastic modulus at
zero strain, 𝐺𝑓𝑐 is the concrete crushing energy, 𝐿𝑒𝑙𝑒𝑚 is the height of the wall element and the subscript cc is added
in Eq. 3 to indicate confined concrete.

10
Ucc,Lcc
9 Unconfined
8 Ycc Confined
7
U,L
Stress (ksi)

6
5
4
Y 𝐺𝑓𝑐𝑐
3
𝐿𝑒𝑙𝑒𝑚 Rcc
2
𝐺𝑓𝑐
1
𝐿𝑒𝑙𝑒𝑚 R
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
Strain (in/in)
Figure 19— Regularized concrete compression stress-strain YULRX envelopes for use with PERFORM

Review of experimental data (Birely 2012) shows that the majority of walls tested in the laboratory exhibit strength
loss due to simultaneous concrete crushing and reinforcement buckling (classified as a compression-buckling failure
in Table 2). To simulate the failure mode, it is necessary to simulate compression strength loss in reinforcing steel due
to buckling, and to achieve mesh-objective results, it is necessary to regularize the compressive strain at onset of steel
strength loss. Instead of using the PERFORM steel model with buckling, the simple buckling model proposed by Pugh
et al. (2015) for use in simulation of wall response was employed. This model assumes that once concrete has reached
residual compressive strength there is minimal restraint of bar buckling and reinforcing steel loses compressive
capacity. Thus, longitudinal reinforcing steel fails in compression due to buckling at 𝜀𝑅 , defined by Eq. 2, if the steel
is surrounded by unconfined concrete and or 𝜀𝑅 𝑐𝑐 , defined by Eq. 3, if the steel is surrounded by confined concrete.

Response data for walls WSH4 and WR0, which were constructed entirely of unconfined concrete and exhibited
strength loss due to concrete crushing, were used to determine an appropriate value for the crushing energy of
unconfined concrete, 𝐺𝑓𝑐 . A series of PERFORM analyses of the WSH4 and WR0 specimens were conducted using

188
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

the modeling recommendations presented above; for each analysis a different 𝐺𝑓𝑐 value, corresponding to a different
concrete u value, was used. For each analysis the simulated displacement at 20% strength loss was recorded. Fig. 20
shows the results of these analyses. Based on these results

Gfc = 0.5 kip/in Eq. 4

is recommended for use in regularizing unconfined concrete response in reinforced concrete walls. This value
approximately equal to the value recommended by Nakamura and Higai (2001) for regular strength concrete and is
consistent with recommendations by Pugh et al. (2015), 0.078𝑓𝑐′ kip/in. with 𝑓𝑐′ is ksi, for analysis of walls using
fiber-type beam-column elements in OpenSees. Fig. 21 shows the simulated response of wall specimen WSH4 with
regularization of concrete compression response using Eqs. 2 and 4 and simulation of bar buckling at 𝜀𝑅 per Eq. 2; the
data show accurate simulation of drift capacity and no mesh sensitivity.
∆𝑠𝑖𝑚𝑢𝑙𝑎𝑡𝑒𝑑 Τ∆𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑

𝐺𝑓𝑐 (kip/in)

Figure 20— Ratio of simulated to measured drift capacity as a function of unconfined concrete crushing energy,
Gfc, for specimens WSH4 and WR0 modeled using 2 and 4 elements per story

30% Strength Loss 30% Strength Loss

Without regularization of
With regularization of
material response, different
material response, All
levels of mesh refinement
levels of mesh refinement
results in different
result in the same, correct
displacement capacities
displacement capacity

Figure 21— Load-deformation response as measured and as simulated using regularized material models and different
numbers of elements per story

The approach used to determine unconfined concrete crushing energy was used also to determine confined crushing
energy, 𝐺𝑓𝑐𝑐 . A series similar analyses were performed on the five walls in Table 1 that exhibited strength loss due to
concrete crushing. Models were constructed using a confined concrete material model for the confined area of the

189
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

boundary elements and an unconfined concrete material model elsewhere. Concrete material response was regularized
using Eqs. 2 and 3. Equation 4 was used to define unconfined concrete crushing energy and a range of confined
concrete crushing energies were used. Reinforcing steel in the unconfined web region of the wall and in confined
boundary elements was defined, respectively, to fail in compression at 𝜀𝑅 defined by Eq. 2 and at 𝜀𝑅 𝑐𝑐 defined by Eq.
3. For each of the walls with confined concrete, analyses were done with two different levels of mesh refinement. The
results of these analyses (Fig. 22) show 𝐾 = 𝐺𝑓𝑐𝑐 Τ𝐺𝑓𝑐 required to achieve accurate estimation of deformation capacity
ranging from 1.0 to 2.5. Given 1) that experimental data show and traditional confined concrete material models
predict increasing concrete strain capacity with increasing confinement and 2) the wide range of confinement detailing
for the walls in Table 1, this variation in the 𝐾 is not surprising. Fig. 23 shows 𝐾 plotted versus 𝑓𝑐𝑐′ Τ𝑓𝑐′ , from which it
can be concluded that the confined concrete crushing energy increases with increasing confinement. Based on the data
in Fig. 23, the following model is recommended for defining 𝐺𝑓𝑐𝑐

𝐺𝑓𝑐𝑐 f′
𝐾= = 1 < 5 ( cc′ − 0.85) < 2.5 Eq. 5
𝐺𝑓𝑐 fc

which is equivalent to

f′
𝐺𝑓𝑐𝑐 = 0.5 < 2.5 ( cc′ − 0.85) < 1.25 Eq. 6
fc

where 𝑓𝑐𝑐′ is the confined concrete compressive strength defined by Mander et al. (1988).
∆𝑠𝑖𝑚𝑢𝑙𝑎𝑡𝑒𝑑 Τ∆𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑

𝐾 = 𝐺𝑓𝑐𝑐 Τ𝐺𝑓𝑐

Figure 22— Ratio of simulated to measured drift capacity as a function Gfcc for specimens in Table 1 with confined
concrete exhibiting a compression-buckling failure
𝐾 = 𝐺𝑓𝑐𝑐 Τ𝐺𝑓𝑐

𝑓𝑐𝑐′ Τ𝑓𝑐′

Figure 23— Ratio of simulated to measured drift capacity as a function Gfcc for specimens in Table 1 with confined
concrete exhibiting a compression-buckling failure

190
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

RECOMMENDATIONS FOR MODELING FLEXURAL WALLS USING PERFORM

The above recommendations for modeling flexural walls using PERFORM are summarized in Table 6. These
recommendations were used to simulate the response of the eight walls listed in Table 1. Simulation results were
evaluated on basis of the accuracy and precision with which maximum strength, displacement at yield, and
displacement capacity were simulated. These response quantities are defined as follows: i) maximum strength is
defined by the maximum base shear achieved in either loading direction, ii) yield is defined by the moment, and
corresponding base shear, at which a fiber-type model of the base section, assuming zero shear strain, indicates first
yield of longitudinal steel, and iii) displacement capacity is defined as the displacement at the point of applied load at
which strength dropped to 80% of maximum strength. Table provides response quantities for each wall and statistics
for the data set; Figs.24 (a)-h) show measured and simulated response histories. The data in Table and Fig. 24 support
the following conclusions and observations:

 Strength and displacement capacity are accurately and precisely simulated using the proposed modeling
recommendations.
 Displacement at yield is slightly underestimated and results have a relatively high level of uncertainty. This is
consistent with the results of other studies (e.g. Pugh et al. 2015) and likely results from the fact that in the
laboratory wall test specimens often exhibit significant shrinkage cracking that increases measured yield
displacement (Palermo and Vecchio, 2002) and is not typically captured in simulations.
 Displacement capacity is accurately estimated for specimen R1, which exhibited a buckling-rupture (BR) failure
in the laboratory. This is despite the fact that the model simulates strength loss due to concrete crushing and
reinforcement buckling but not fracture of reinforcement following buckling. Accurate simulation of
displacement capacity for walls exhibiting compression-buckling (CB) and buckling-rupture (BR) failure modes
using models that simulate only strength loss due to concrete crushing and bar buckling is consistent with previous
research (Pugh et al., 2015). Previous research shows also similar drift capacities for walls exhibiting CB and BR
failure modes. These results suggest that fracture of previously buckled reinforcement occurs very shortly after
buckling that results in significant strength loss.

Table 5— Ratio of simulated to measured response quantities

𝑉𝑚𝑎𝑥,𝑠𝑖𝑚 ∆𝑦𝑖𝑒𝑙𝑑,𝑠𝑖𝑚 ∆𝑢,𝑠𝑖𝑚
Specimen
𝑉𝑚𝑎𝑥 ∆𝑦𝑖𝑒𝑙𝑑 ∆𝑢
WSH4 1.00 0.93 1.00
WR0 0.98 0.61 0.88
R1 1.03 1.41 1.21
RW1 1.00 0.77 1.03
RW2 1.02 0.55 1.11
RW-A20-P10-S38 0.95 0.82 1.00
RW-A20-P10-S63 1.13 0.93 0.99
WSH6 0.97 1.09 1.07
Average 1.02 0.89 1.03
C.O.V. 0.06 0.31 0.10

191
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

(b)
(a)

Figure 24— Measured (black) and simulated (blue) response histories for walls listed in table 1; simulations employ
recommended modeling parameters, which are listed in the “validation study” column of Table 6

(c) (e)

(d) (f)

192
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

(g) (h)

Figure 24 (continued)— Measured (black) and simulated (blue) response histories for walls listed in Table 1;
simulations employ recommended modeling parameters, which are listed in the “validation study” column of Table 6

LIMITATIONS ON APPLICATION OF THE MODEL

Comparison of response as measured and as simulated shows that the proposed modeling recommendations result in
a model that is appropriate for use in assessing the earthquake performance of walls exhibiting flexure-controlled
response, with either compression- or tension-controlled failure mechanisms. Recent research by Whitman (2015)
provides guidance on wall design characteristics that can result in walls exhibiting a compression-shear failure rather
than a flexure-controlled failure. Whitman conducted nonlinear continuum-type analysis of planar walls with a range
of design characteristics, including walls with high cross-sectional aspect ratios (CSAR = ratio of wall length to wall
thickness) and high shear stress demands. Whitman found that walls with high cross-sectional aspect ratios and high
shear demands exhibit a compression-shear failure characterized by crushing of concrete at the boundary element-
web interface. Fig. 25 shows the range of CSARs and shear stress demand for which a compression-shear rather than
a flexure-controlled response could be expected for (a) walls with boundary element detailing that meets current ACI
Code requirements and (b) walls with boundary element confinement that extends to the neutral axis depth.

Given the results of the Whitman (2015) study and that the PEFORM model simulates only a flexural failure
mechanism, the PEFORM model is appropriate only for use in assessing the earthquake performance of walls that
expected to exhibit flexural failure per Fig. 25.

(a) Boundary Element Meets ACI Code (2014) Requirements (b) Boundary Element Depth Extends to Neutral Axis

Figure 25— Average shear stress demand and cross-sectional aspect ratios resulting in flexure-controlled versus
flexure-shear failure modes for planar walls with boundary elements (a) meeting current ACI Code requirements and
(b) that extend to the neutral axis depth

193
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

SUMMARY AND CONCLUSIONS

The results of previous research by the authors and students at the University of Washington as well as experimental
data characterizing the response of planar walls were used to develop recommendations for modeling planar flexure-
controlled concrete walls using the PERFORM-3D (www.csiamerica.com/) software package. Recommendations
address mesh refinement as well as the definition of material response. Regularization of the concrete compression
response using concrete crushing energy and element height is recommended to achieve accurate, mesh-objective
simulation of the drift capacity of walls. Data show that without regularization of material response, simulated drift at
onset of strength loss is a function of mesh size. Equations are provided that define unconfined concrete crushing
energy and confined concrete crushing energy as a function of confinement. Recommendations include also the
appropriate range of application of the model, which excludes walls with large cross-sectional aspect ratios and walls
subjected to high shear stress demands. Comparison of simulated and measured response for a series of eight planar
walls, representing a range of design parameters with the recommended range of application of the model, shows that
the proposed modeling recommendations result in accurate and precise simulation of wall strength and displacement
capacity. Comparison of simulated and measured response shows that yield displacement is slightly under estimated
(average ratio of simulated to measured yield displacement in 0.89) and shows a relatively high level of uncertainty
(cov for ratio of simulate to measured yield displacement is 31%); this is consistent with the results of previous
research by the authors and others and attributed to variability of shrinkage cracking observed in laboratory test
specimens.

Table 6— Model parameters used for simulations

Simulation Used to Investigate
Reference Validation
Model Cyclic Cyclic Mesh Mesh Study
Response Response Refinement Refinement
Wall specimen used for study RW1 WSH4 WSH4 WSH4 WSH 4 All Walls
Figs. 22 (a) –
Figure(s) presenting results Fig. 9(b) Fig. 9(e) Fig. 16 Fig. 21
Figure 6 (c)
Number elements horizontally 1 1 1 4 4 4
Number of elements vertically in
2 2 2 2, 3, 4 2,3,4 4
yielding story
Number of elements vertically in
1 1 1 2, 3, 4 2,3,4 1
non-yielding stories
Concrete energy dissipation factor - - Table Table Table Table 4
Steel energy dissipation factor - - 0.75 0.75 0.75 0.75
Steel unloading/reloading
- - 0.5 0.5 0.5 0.5
stiffness factor
Unconfined concrete crushing 0.5 kip/in 0.5 kip/in 0.5 kip/in
- - -
energy, Gfc (Eq. 4) (Eq. 4) (Eq. 4)
Confined concrete crushing Eq. 5 & Eq.
- - - - -
energy, Gfcc 6
Buckling of steel modeled No No No No Yes Yes

194
 SP-339: Performance-Based Seismic Design of Concrete Buildings: State of the Practice

REFERENCES

1. ACI Committee 318, and American Concrete Institute. 2011. Building Code Requirements for Structural
Concrete (ACI 318-11) and Commentary. Farmington Hills, MI: American Concrete Institute.
2. Applied Technology Council (ATC) (2010). “Modeling and acceptance criteria for seismic design and analysis
of tall buildings,” Prepared by Applied Technology Council for the Pacific Earthquake Engineering Research
Center, Redwood City, California, Tech. Rep. Report No. PEER/ATC-72-1.
3. Birely, A.C. (2012), “Seismic performance of slender reinforced concrete walls,” Ph.D. Dissertation, Dept. of
Civil and Environmental Engineering, University of Washington, Seattle, WA.
4. Chiaramonte, M.M. (2011). “An analysis of conventional and improved marginal wharves” MS Thesis.
University of Washington, Seattle. 217 p.
5. Coleman, J. and Spacone, E. (2001). “Localization issued in force-based frame elements,” Journal of Structural
Engineering, ASCE 127(11): 1257–1265.
6. Comite Euro-International du Beton (1993) CEB-FIP Model Code 1990.
7. Dazio, A., Beyer, K. Bachmann, H. (2009). “Quasi-Static Cyclic Tests and Plastic Hinge Analysis of RC
Structural Walls,” Engineering Structures 31(7): 1556–71.
8. Jansen D., Shah, S. (1997). “Effect of length on compressive strain softening of concrete,” ASCE Journal of
Engineering Mechanics 123(1): 25–35.
9. Mander, J.B., Priestley, M.J.N., Park, R. (1988). “Observed Stress-Strain Behaviour of Confined Concrete,”
Journal of Structural Engineering, ASCE 114(8): 1827-1849.
10. Mehta, P.K., Monteiro, P.J.M. (1993). Concrete Structure, Properties and Materials. Prentice-Hall.
11. Mohr, D. (2007). “Nonlinear analysis and performance based design methods for reinforced concrete coupled
shear walls,” MS thesis, Dept. of Civil and Environmental Engineering, University of Washington, Seattle, WA,
2007.
12. Nakamura, H., Higai, T. (2001). “Compressive Fracture Energy and Fracture Zone Length of Concrete,”
Modeling of Inelastic Behavior of RC Structures Under Seismic Loads, Eds. Shing, P.B. and Tanabe, T., Reston,
VA: ASCE.
13. Oesterle, R., Aristizabal-Ochoa, A, Carpenter, J., Russell, H., Corley, W. (1976). Earthquake resistant structural
walls - tests of isolated walls, Portland Cement Association/National Science Foundation, Washington, D.C.,
Tech. Rep. No. NSF/RA-760815.
14. Oh, Y.-H., Han, S.W., Lee, L.-H. (2002). “Effect of Boundary Element Details on the Seismic Deformation
Capacity of Structural Walls.” Earthquake Engineering & Structural Dynamics 31(8): 1583–1602.
15. Palermo, D., Vecchio, F. (2002). “Behavior and analysis of reinforced concrete walls subjected to reversed cyclic
loading,” Department of Civil Engineering, University of Toronto, Toronto, ON, Tech. Rep. ISBN 0-7727-7553-
02, Publication No. 2002-01, 2002
16. Paulay, T., Priestley, M.J.N. (1992). Seismic Design of Reinforced Concrete and Masonry Buildings. John Wiley
and Sons.
17. Pugh, J.S., Lowes, L.N., Lehman, D.E. (2015). “Nonlinear line-element modeling of flexural reinforced concrete
walls,” Engineering Structures 104: 174-192.
18. Tran, T.A. (2012). “Experimental and Analytical Studies of Moderate Aspect Ratio Reinforced Concrete
Structural Walls.” Ph.D. Dissertation, University of California, Los Angeles.
19. Wallace, J.W., Thomsen J.H., (1995). “Seismic Design of RC Structural Walls. Part II: Applications.” Journal of
Structural Engineering 121(1): 88–101.
20. Whitman Z. (2015). “Investigation of seismic failure modes in flexural concrete walls using finite element
analysis,” MS thesis, University of Washington, Seattle

195