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

net/publication/320845424

New opensees models for simulating nonlinear flexural and coupled shear-
flexural behavior of RC walls and columns

Article  in  Computers & Structures · November 2017

DOI: 10.1016/j.compstruc.2017.10.010

CITATIONS READS

15 2,049

3 authors:

Kristijan Kolozvari Kutay Orakcal

California State University, Fullerton Bogazici University
16 PUBLICATIONS   169 CITATIONS    47 PUBLICATIONS   664 CITATIONS   

SEE PROFILE SEE PROFILE

John W. Wallace
University of California, Los Angeles
161 PUBLICATIONS   2,778 CITATIONS   

SEE PROFILE

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

Seismic Retrofit of Reinforced Concrete Shear Walls Designed and Constructed Prior to Enforcement of the Recent Seismic Design Codes 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 Kristijan Kolozvari on 18 December 2017.

The user has requested enhancement of the downloaded file.

 Computers and Structures 196 (2018) 246–262

Contents lists available at ScienceDirect

Computers and Structures

journal homepage: www.elsevier.com/locate/compstruc

New opensees models for simulating nonlinear flexural and coupled

shear-flexural behavior of RC walls and columns
Kristijan Kolozvari a,⇑, Kutay Orakcal b, John W. Wallace c
a
California State University, Fullerton, CA, USA
b
Bogazici University, Istanbul, Turkey
c
University of California, Los Angeles, CA, USA

a r t i c l e i n f o a b s t r a c t

Article history: This paper describes new model elements and material constitutive relationships implemented by the
Received 14 March 2017 authors into the widely-used open-source computational platform OpenSees (Open System for
Accepted 17 October 2017 Earthquake Engineering Simulation), aimed to enhance current nonlinear analysis and response assess-
Available online 3 November 2017
ment capabilities for reinforced concrete (RC) walls and columns. Classes added to the existing
OpenSees library include: (1) the Multiple-Vertical-Line-Element-Model (MVLEM) element with uncou-
Keywords: pled axial/flexural and shear responses, (2) the Shear-Flexure-Interaction-Multiple-Vertical-Line-Eleme
Reinforced concrete structural walls and
nt-Model (SFI-MVLEM) element with coupled axial/flexural and shear responses, (3) the Fixed-Strut-
columns
Analytical nonlinear modeling
Angle-Model (FSAM), which is a two-dimensional constitutive model for RC panel elements, (4) an
Shear behavior improved uniaxial constitutive model for concrete, and (5) an improved uniaxial constitutive model
Flexural behavior for reinforcing steel. Representative validation studies are also presented, where the analytical model
Shear-flexure interaction predictions are compared with results of quasi-static lateral load tests on selected RC column and wall
OpenSees specimens. Response comparisons reveal that the implemented models capture, with reasonable accu-
Earthquake engineering racy, the experimentally-observed behavior of the test specimens investigated. Based on the comparisons
presented, model capabilities are assessed and potential model improvements are identified.
Published by Elsevier Ltd.

1. Introduction bridge columns). Their role is to provide sufficient lateral strength

1.1. Background and aim
With implementation of performance-based methodologies in
modern seismic design codes and assessment guidelines, detailed
modeling and simulation of the nonlinear seismic behavior of rein-
forced concrete (RC) structural systems has recently gained much
importance. While modeling of the linear elastic response charac-
teristics of systems with complex geometry is no longer a signifi-
cant design challenge, reliable modeling approaches for robust
simulation of the nonlinear hysteretic behavior of RC structural
members are still needed.
RC structural walls and columns are often used as the primary
vertical structural members for resisting lateral loads imposed by
wind or earthquakes on buildings (e.g., core wall systems, wall-
frame dual systems, special moment frames) and bridges (e.g.,
⇑ Corresponding author.
E-mail addresses: kkolozvari@fullerton.edu (K. Kolozvari), kutay.orakcal@boun.
edu.tr (K. Orakcal), wallacej@ucla.edu (J.W. Wallace).
and stiffness to minimize nonlinear behavior and limit lateral dis-
placements during service-level earthquakes, as well as to provide
sufficient nonlinear deformation capacity (ductility) during a sev-
ere earthquake. Given the crucial role of walls and columns in
the seismic performance of RC structural systems, it is essential
that analytical models that are capable of capturing their impor-
tant response characteristics are available, for either design of
new structures or performance assessment of existing structures.
The most commonly-used modeling approach for nonlinear
response analysis of RC structures involves adaptation of fiber-
based models, with uncoupled axial/flexural and shear behavior,
for simulating the hysteretic behavior of walls and columns. Using
a plastic hinge model is typically deemed sufficient for beams, con-
sidering that axial loads in beams are negligible. In a fiber-based
model, nonlinear axial/flexural response of a wall or column is sim-
ulated using a series of uniaxial elements (or fibers), the behavior
of which is based on stress-strain relations for concrete and rein-
forcing steel (or simplified force-deformation relations), along with
the plane-sections-remain-plane assumption. Shear behavior of
the member is typically accounted for by using a horizontal spring

https://doi.org/10.1016/j.compstruc.2017.10.010
0045-7949/Published by Elsevier Ltd.
 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262
in the model element, with a specified shear force versus deforma-
tion (backbone) relation. Fiber models have been implemented
into various research-oriented (e.g., OpenSees, [21]) and
commercially-available computer programs (e.g., PERFORM-3D,
CSI; [35]), and have been widely used to simulate the nonlinear
behavior of RC walls in particular. Studies that compare model
and experimental results (e.g., [1,25]) show that conventional fiber
models provide reasonably accurate predictions of the flexural
response of walls. However, the inability of fiber models to account
for the interaction (coupling) between axial/flexural and shear
behavior is a significant drawback, as studies have shown that
uncoupled models tend to underestimate compressive strains in
concrete at the boundary regions of even relatively slender walls
with flexure-controlled responses [25], and overestimate the lat-
eral load capacity of walls with moderate aspect ratios [9]. Further-
more, using an uncoupled fiber model requires pre-definition of an
ad-hoc shear force versus deformation relation, which may intro-
duce bias in the analysis results. For example, Kolozvari and
Wallace [10] showed that uncoupled wall models with linear-
elastic shear force-deformation relationship with 0.5GAg (com-
monly adopted approach) tend to overestimate wall shear force
demand by approximately 30% and underestimate interestory drift
by approximately 50% within the plastic hinge region compared to
coupled models.
In terms of flexural and shear deformation contributions to
overall response and failure mode, RC walls and columns are typ-
ically characterized by their shear-span-to-depth (M/Vl) ratio.
Squat members – especially walls – with M/Vl ratios less than
approximately 1.0–1.5, generally exhibit shear-dominant behavior,
whereas well-designed slender members with aspect ratios greater
than 2.5–3.0 demonstrate flexure-controlled responses. The lateral
load behavior of walls and columns with intermediate M/Vl ratios
of 1.0–3.0 is generally influenced by both nonlinear flexural and
shear deformations. A number of experimental results obtained
from tests on RC walls (e.g., [19,34,38,39]) as well as columns
(e.g., [31–33]) have shown that nonlinear flexural and shear defor-
mations occur simultaneously, and shear deformations can be
influential on the response even when the overall response or fail-
ure mode of the member is not shear-governed. This indicates
inherent coupling between nonlinear flexural and shear responses
in RC members, which is commonly referred to as shear-flexure
interaction (SFI). It has been observed experimentally that SFI
can lead to increased compressive strain demands on walls and
columns, especially with intermediate M/Vl ratios, resulting in
reduced strength, stiffness, and deformation capacity compared
to behavior under pure bending. Therefore, it is important that
analytical models that can simulate this experimentally-observed
SFI behavior in RC members are available to researchers and
engineers.
A number of modeling approaches, with various levels of
sophistication and capabilities that incorporate SFI behavior in
RC walls (e.g., [6,7,11,18,20,28,30]) and columns (e.g., [4,44]), can
be found in the literature. An effective approach to capture SFI in
RC members using a fiber-based model formulation was first pro-
posed by Petrangeli et al. [30]. Massone et al. [18] adopted this
approach for analysis of RC walls under monotonic loading,
whereas Kolozvari et al. [11] extended it to address reversed cyclic
loading conditions, both using different constitutive modeling
approaches to represent the behavior of concrete under combined
normal and shear stresses. In the analytical model proposed by
Kolozvari et al. [11], the uniaxial fibers in a fiber-based model for-
mulation are replaced with RC panels, the biaxial behavior of
which is described using a two-dimensional constitutive model
with a fixed-crack-angle approach. Thereby, the effect of shear
deformations on the strains in concrete is directly incorporated,
and shear stresses developing in the model element evolve with
247
the normal stresses developing in concrete (in parallel and perpen-
dicular directions to cracks) and the shear stresses developing
along the cracks, based on the constitutive stress-strain relation-
ships adopted in the model (versus use of a pre-defined shear back-
bone relation). It has been shown that this model is capable of
capturing successfully the overall load-deformation behavior, non-
linear shear behavior, coupling of nonlinear flexural and shear
deformations, as well as local responses (e.g., strains and rotations)
for RC walls that experience significant SFI behavior under
reversed cyclic loading conditions [12].
Despite the significant number of analytical modeling
approaches available in the literature, only a small number of
model formulations are implemented in computational tools
(commercial and open-source) available to the broader engineer-
ing and research community. One of the most widely-used com-
putational platforms in structural/earthquake engineering is the
Open System for Earthquake Engineering Simulation (OpenSees,
[21]). OpenSees is an open-source software that includes a wide
range of features (e.g., model elements, materials models, solution
strategies, etc.) that can be used to conduct nonlinear analysis of
structural components and systems under earthquake ground
motions. Although OpenSees incorporates an extensive library of
model elements, only a few options are available for modeling of
RC walls and columns, where the displacement-based beam-
column element that follows a fiber-based formulation with
uncoupled axial/flexural and shear responses is being the most
commonly-used. However, the computational stability (conver-
gence), efficiency (rate of convergence), and accuracy of analytical
results obtained using the displacement-based beam-column ele-
ment have been shown to be sensitive to choice of number of inte-
gration points used along the element length [3,37]. Other models
available in OpenSees for simulation of RC walls and columns are
flexure-shear interaction displacement-based beam-column ele-
ment proposed by Massone et al. [18], strut-and-tie-based model
implemented by Panagiotou et al. [28], and shell element pro-
posed by Lu et al. [14]. The modeling approach implemented by
Massone et al. [18] is capable of simulating monotonic responses
only. A strut-and-tie (truss) approach developed by Panagiotou
et al. [28] is capable of capturing SFI, however, due to overlapping
areas of vertical, horizontal, and diagonal concrete struts in the
model, as well as the use of pre-define angles of concrete struts,
achieving accurate displacement responses over a broad range of
response amplitudes and wall configurations is a challenge. The
shell element implemented by Lu et al. [14] is a relatively recent
model based on a finite element formulation (not a macroscopic
model), which has not been extensively validated and is intended
to be used for RC walls only. In addition, material constitutive
models available in OpenSees, which are used to represent the
hysteretic stress versus strain behavior of concrete and reinforcing
steel, incorporate certain simplifications and occasional inconsis-
tencies in their formulation. In particular, existing uniaxial mate-
rial models for concrete are characterized with simplified
hysteretic rules, which represent material behavior in a crude
manner and do not allow capturing of cyclic concrete behavior
accurately (e.g., no gradual gap closure). Also, the most
commonly-used uniaxial material model for reinforcing steel is
subject to stress overshooting upon reloading after partial strain
reversal, which can lead to unreasonable predictions of hysteretic
loops of structural components subjected to cyclic or dynamic
loading. Finally, a reliable and efficient two-dimensional constitu-
tive model for representing the behavior of RC panel (membrane)
elements under plane-stress conditions is not available. Therefore,
there is still a considerable need for development and implemen-
tation of improved analytical models (model elements and mate-
rial constitutive models) to enhance current OpenSees
simulation capabilities associated with the nonlinear behavior of
248 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262

RC structures, which in-turn will advance analytical capabilities of improved uniaxial material models for concrete and reinforcing
researchers and engineers worldwide. steel.

1.2. Research objective and scope 2.1. Flexural model element (MVLEM)

A research project was initiated to implement a number of new The model element implemented in OpenSees for simulation of
features for nonlinear analysis of RC structural walls and columns flexure-dominated behavior of RC members is the two-
into the computational platform OpenSees, in order to address the dimensional Multiple-Vertical-Line-Element-Model (MVLEM;
shortcomings mentioned previously, as well as to validate the [26,43]. The axial/flexural response of the element is simulated
models using existing experimental data obtained from tests on by a series of uniaxial macro-fibers connected to rigid beams at
wall and column specimens. This paper describes the analytical the top and bottom levels, as illustrated in Fig. 2a. The relative
models implemented by the authors into OpenSees, including rotation between top and bottom boundaries of the model element
new model elements and material constitutive models, and pro- is concentrated at the center of rotation defined at a relative height
vides examples of model applications to nonlinear analysis of RC ch (Fig. 2a); a value of c ¼ 0:4 is commonly used [25,43]. Distribu-
walls and columns. tion of curvature is assumed to be constant along the element
The newly-implemented OpenSees classes described include height.
the following model elements and constitutive models (Fig. 1): Based on displacements at the element degrees of freedom
(1) MVLEM: macroscopic fiber-based element with uncoupled flex- fDg ¼ f D1 D2 . . . D6 gT (Fig. 2a), axial deformations of the m
ural and shear responses, (2) SFI-MVLEM: macroscopic fiber-based micro-fibers fuy g are calculated as:
element with shear-flexure interaction behavior, (3) FSAM: two- 8 9 2 38 9
dimensional (plane-stress) constitutive model for RC panel ele- > uy;1 > 0 1 x1 0 1 x1 > D1 >
>
> >
> >
> > >
< uy;2 = 6 7< =
ments, (4) ConcreteCM: hysteretic uniaxial material (stress- 6 0 1 x2 0 1 7 D2
x2
¼6 7 ð1Þ
strain) model for concrete, and (5) SteelMPF: hysteretic uniaxial >
> : >
> 4: : : : : : 5>> : >
>
>
: >
; >
: > ;
material model for reinforcing steel. OpenSees implementation of uy;m 0 1 xm 0 1 xm D6
the model elements and material models included not only devel-
opment of source codes, but also preparation of detailed user doc- The average normal strain in y-direction for j-th RC macro-fiber ey;j
umentation (e.g., manuals, examples, input files), all of which is is then calculated by dividing the axial deformation in y-direction
available through the OpenSees Wiki Page as well as the PEER uy;j by the model element height h:
report by Kolozvari et al. [11].
uy;j
An overview of model formulations is presented, followed by ey;j ¼ ð2Þ
h
comparison of the analytical results predicted by the models with
experimentally-measured responses obtained for various RC wall Finally, the resulting axial stresses in the macro-fibers are obtained
and column specimens characterized with a range of properties based on the fiber axial strains ey;j and the corresponding uniaxial
(e.g., cross-sections, reinforcing ratios, axial loads, etc.) for assess- constitutive material models used for concrete and reinforcing
ment of model capabilities and shortcomings. steel.
Shear response of the model element is simulated by a horizon-
2. Description of the analytical models tal spring placed at the relative height ch (Fig. 2a). Deformation of
the shear spring uSh is calculated from the element degrees of free-
This section provides descriptions of the analytical models dom as:
implemented in OpenSees as a part of this research effort, includ- uSh ¼ f 1 0 c  h 1 0 ð1  cÞ  h gfDg ð3Þ
ing model elements for simulation of flexure-dominated and
shear-flexure-interaction responses of RC members, as well as a The resulting spring shear force F Sh is then obtained based on the
plane-stress constitutive model for reinforced concrete, and spring deformation uSh and the implemented constitutive model

Fig. 1. New OpenSees classes.

 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262 249

Fig. 2. Implemented model elements: (a) MVLEM element and (b) SFI-MVLEM element.

(
for wall shear behavior, typically described by an ad-hoc X
m X
m X
m

force-deformation rule (e.g., linear-elastic behavior or an origin- fF e;int gD ¼ F Sh ;  F y;j ; F Sh ch  F y;j xj ; F Sh ; F y;j ;
j¼1 j¼1 j¼1
oriented hysteresis model with a nonlinear backbone relationship). )T
Since axial/flexural and shear responses are described indepen- X
m

dently, there is no coupling between these responses in the MVLEM F Sh ð1  cÞh þ F y;j xj ð6Þ
j¼1
element.
The stiffness properties and force-deformation relationships of The MVLEM element implemented in Opensees is conceptually
the uniaxial elements are derived based on hysteretic stress- similar to the widely-used displacement-based beam-column ele-
strain relationships adopted for concrete and reinforcing steel, ment (DB-BC, [37]), with the following two simplifications: (1) the
and the tributary area assigned to each uniaxial element. For a MVLEM uses average strains and stresses developing within each
specified set of displacement components at the six nodal degrees macro-fiber over the element height to obtain the deformations
of freedom {D} of a MVLEM element, the element stiffness matrix and forces in the corresponding fiber, whereas the DB-BC element
relative to these degrees of freedom is: evaluates the strain-stress relationship at a number of integration
2 3 points over the element height and obtains deformations and
Xm X m X m
forces by integration of strains and stresses and (2) the distribution
6 ky;j  ky;j xj ky;j xj 7
6 j¼1 7 of wall curvature over the height of each model element is
6 j¼1 j¼1
7
6 Xm X m 7 assumed to be uniform, as opposed to linear distribution of curva-
T 6
6 2 2 2 7
27
½K e D ¼ ½b  6 k Sh c h þ k x2
y;j j kSh cð1  cÞh  k x
y;j j
7  ½b
ture assumed between element nodes of the DB-BC element formu-
6 j¼1 j¼1 7 lation. These simplifications minimize convergence problems
6 7
6 X m 7 encountered in using the beam-column element, which results in
4 symm: 2 2
kSh ð1  cÞ h þ 25
ky;j xj
improved numerical stability and computational efficiency of the
j¼1
MVLEM over the DB-BC element. Based on comparison of the
ð4Þ
MVLEM and the DB-BC model results presented in Section 3.3.2,
analysis times using the MVLEM are approximately 90% and 60%
where kSh is the stiffness of the horizontal spring in the model ele- of the analysis times using the DB-BC model with four and eight
ment, ky;j is the stiffness of the j-th uniaxial macro-fiber, and xj is its integration points, respectively. However, because of its simplifica-
distance to the central axis of the element. Matrix ½b denotes the tion of constant curvature assumed along element height, the
geometric transformation matrix converting the six element MVLEM may provide less accurate predictions of local deformation
degrees of freedom {D} to the element deformations of extension, responses (i.e., strains, rotations). Nevertheless, this can be
relative rotation at the bottom, and relative rotation at the top of improved by using several elements over the anticipated plastic
each wall element, and it is given as: hinge region (or the entire height) of the member. The MVLEM is
2 3 compatible with all uniaxial material models available in OpenSees
0 1 0 0 1 0 for concrete and reinforcing steel, and the accuracy of the analyti-
6 7
½b ¼ 4 1=h 0 1 1=h 0 0 5 ð5Þ cal response predictions (e.g., strain predictions, shape of the hys-
1=h 0 0 1=h 0 1 teretic loops, etc.) is directly related to the level of detail of the
constitutive material model selected for the analysis.

The internal (resisting) force vector of a MVLEM element is

assembled from the axial forces in the vertical directions F y;j of
all element macro-fibers, which are calculated based on the axial
stresses developing in concrete and reinforcing steel within each
macro-fiber as well as their corresponding tributary areas, together
with the shear force developing in the horizontal spring F Sh . The
resisting (internal) force vector of a single MVLEM element can
be expressed as:
2.2. Shear-flexure-interaction model element (SFI-MVLEM)
The Shear-Flexure-Interaction Multiple-Vertical-Line-Element
Model implemented in OpenSees (SFI-MVLEM [12,13]) captures
interaction between axial/flexural and shear responses in RC
structural members under cyclic loading. The SFI-MVLEM element
incorporates a two-dimensional constitutive RC panel behavior
250
described by the Fixed-Strut-Angle-Model (FSAM; [27,41]), into
the fiber-based formulation of the MVLEM (Fig. 2b), with RC pan-
els replacing its uniaxial macro-fibers. The FSAM represents a
two-dimensional (plane-stress) constitutive relationship that
relates the strain field imposed on a RC panel (ex , ey , and cxy ) to
the resultant of the stresses developing in concrete and reinforc-
ing steel, converted into smeared stresses in concrete (rx , ry , and
sxy ). Therefore, coupling between axial and shear modes of behav-
ior (normal and shear stresses) is captured at each RC panel
(macro-fiber) level, which further allows capturing of the interac-
tion between flexural and shear responses at the SFI-MVLEM ele-
ment level.
Similarly to the original MVLEM formulation, axial strains in
wall longitudinal direction on each RC panel ey;j are calculated from
the deformations at the element degrees of freedom {D} (Fig. 2b)
according Eqs. (1) and (2). Element shear deformation uSh is
obtained from Eq. (3) and converted to shear strain on each RC
panel element cxy;j , assuming uniform shear strain distribution
across the wall cross-section. Since shear stresses developing in
the model element evolve with the normal stresses developing in
concrete and the shear stresses developing along the cracks (based
on the behavior of the constitutive RC panel elements and the
stress-strain relationships adopted for the materials), the SFI-
MVLEM does not incorporate a shear spring or require definition
of a shear backbone relation, contrary to an uncoupled fiber-
based model formulation (e.g., MVLEM or the displacement-based
beam-column element).
Axial strains in the horizontal direction of each RC panel ele-
ment ex;j are obtained assuming that the resultant horizontal stres-
ses associated with steel and concrete (i.e., resultant smeared
stress on concrete, rx) is equal to zero. This assumption is incorpo-
rated in the SFI-MVLEM formulation by introducing additional
degrees of freedom defined in the horizontal direction for each
macro-fiber of the model element fdg ¼ f d1 d2 . . . dm gT
(Fig. 2b), which are kinematically independent from the six
degrees of freedom at the top and bottom boundaries of the ele-
ment {D}. Therefore, the coupling between axial and shear
responses of each macro-fiber is achieved through the constitutive
two-dimensional RC panel behavior. In OpenSees, these degrees of
freedom {d} are assigned to nodes that are generated and added to
the analysis domain automatically (i.e., without user intervention),
and are located at the centroid of each RC panel within the model
element. The degrees of freedom {d} are treated as external degrees
of freedom that contribute to the element tangent stiffness matrix
and the resisting force vector, and their values are updated at each
iteration and analysis/time step. Therefore, the total number of
degrees of freedom of each SFI-MVLEM element is increased to 6
+ m (where m is the number of macro-fibers in the model element),
compared to 6 in the original MVLEM formulation, which leads to
considerable decrease in computational efficiency (e.g., run-time)
of the SFI-MVLEM, compared to the uncoupled MVLEM. For any pre-
scribed strain field (ex;j , ey;j and cxy;j ), the axial stiffness in x and y
directions (kx;j and ky;j ) and the shear stiffness (kSh;j ) of the j-th RC
panel element are:
        
@F x @F x @ rx @ ex @ rx
kx;j ¼ ¼   ¼
@ux j @ rx j @ ex j @ux j @ ex j bj
        
@F y @F y @ ry @ ey
ky;j ¼ ¼   ¼
@uy j @ ry j @ ey j @u y j
    !   !
@F H @F H @ sxy @ cxy
kSh;j ¼ ¼   ¼
@uH j @ sxy j @ cxy @uH j
j j
K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262
where ð@ rx =@ ex Þj , ð@ ry =@ ey Þj and ð@ sxy =@ cxy Þj are the diagonal
terms of the stiffness matrix defined for each RC panel element,
bj is the panel width in horizontal x-direction, h is the model ele-
ment height in vertical y-direction, t is the wall thickness, and
Ax;j and Ay;j are tributary areas of the vertical and horizontal faces
of j-th RC panel element. The total shear stiffness of a SFI-MVLEM
element (kSh ) is calculated as a sum of the shear stiffness of all
RC panels within one model element. The SFI-MVLEM element stiff-
ness matrix is then assembled as a combination of the element
stiffness sub-matrix ½K e D , which corresponds to the six degrees
of freedom D at the top and the bottom of the model element (sim-
ilar to MVLEM stiffness matrix given in Eq. (4)), and the sub-matrix
½K e d corresponding to the m extensional degrees of freedom {d}.
Since the six degrees of freedom {D} are kinematically indepen-
dent from extensional degrees of freedom {d} (Fig. 2b), the element
stiffness matrix becomes a block matrix consisting two sub-
matrices ½K e D and ½K e d , expressed as:
2 3
½K e D j ½0
6 7
½K e  ¼ 4    5 ð10Þ
½0 j ½K e d
As well, given that the horizontal (extensional) degrees of free-
dom at each macro-fiber {d} are also independent from each other,
the element stiffness matrix relative to these degrees of freedom is
a diagonal matrix consisting of horizontal stiffness values for each
RC panel macro-fiber kx;j (Eq. (7)) as:
½K e d ¼ diag ½ kx;1 kx;2 : : kx;m m;m ð11Þ
Similarly, the corresponding internal force vector of the model
element can be expressed as:
8 9
< fF e;int gD >
> =
fF e;int g ¼  ð12Þ
>
: >
;
fF e;int gd
where the sub-vector fF e;int gD is the force vector relative to the
degrees of freedom {D} derived from the resulting axial forces in
the RC panel elements in the longitudinal direction of the model
element (Fy,j), together with the element shear force FSh, calculated
in the same manner as for the original MVLEM formulation (Eq. (6)).
Since a shear spring does not exist in the SFI-MVLEM model element
formulation, the internal shear force FSh is obtained as the sum of
the resulting shear forces Fxy,j developing in all of the RC panel ele-
ments within the model element. Sub-vector fF e;int gd is the element
force vector corresponding to the m extensional degrees of freedom
{d}, and is given by:
fF e;int gd ¼ f F x;1 F x;2
T
: : : F x;m g ; ð13Þ
where Fx,j (j = 1, . . . , m) are the resultant horizontal forces (with
zero magnitude at equilibrium) corresponding to the extensional
degrees of freedom dj.
Previous studies (e.g., [18]) have shown that that three major
assumptions implemented in the SFI-MVLEM formulation (i.e.,
assumptions of zero resultant horizontal stress, uniform shear strain
 distribution, and the plane-sections-remain-plane) are reasonable
Ax;j for cantilever walls with aspect ratios greater than 1.0, suggesting
 ð7Þ
that application of the proposed modeling approach is suitable for
 both slender and moderate-aspect-ratio walls and columns.
@ ry Ay;j
 ð8Þ
@ ey j h 2.3. Plane-stress constitutive model for reinforced concrete (FSAM)
@ sxy Ay;j The Fixed-Strut-Angle Model (FSAM, [27], Fig. 3), which is,
 ð9Þ implemented in OpenSees as an nDMaterial model, is a two-
@ cxy h
dimensional (plane-stress) constitutive relationship for simulating
 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262
Fig. 3. Behavior and modeling parameters of the constitutive RC panel model FSAM: (a) Strain-stress field, (b) concrete biaxial behavior, (c) concrete shear aggregate interlock,
(d) steel behavior, and (e) dowel action on reinforcement.
the behavior of RC panel elements under generalized, in-plane,
reversed-cyclic loading conditions. The constitutive behavior of
the FSAM is defined via hysteretic uniaxial stress-strain relation-
ships for concrete (Fig. 3b – along crack directions) and for rein-
forcing steel (Fig. 3d – along reinforcement directions). Although
the concrete stress–strain relationship adopted in the FSAM is fun-
damentally uniaxial in nature, it also incorporates biaxial softening
effects including compression softening [42] and hysteretic biaxial
damage [17]. In the baseline formulation of the FSAM [41], shear
aggregate interlock along cracks and dowel action on reinforcing
steel bars are neglected. In this case, crack directions coincide with
principal stress directions in concrete, zero shear stress develops
along the cracks, and the principal stress directions in concrete
(along which the biaxial stress-strain relationship for concrete is
applied) are assumed to be fixed along the crack (strut) directions,
while the principal strain directions rotate with the applied strain
field. Before formation of the first crack in concrete, the FSAM for-
mulation assumes that the principal strain directions on the panel
element coincide with the principal stress directions in concrete.
The first crack in concrete develops in perpendicular direction to
the principal tensile strain, implying formation of the first ‘‘fixed
strut”, when the principal tensile strain first exceeds the cracking
strain value of concrete. The second crack is assumed to develop
in perpendicular direction to the first crack, when the tensile strain
in the first strut direction exceeds the concrete cracking strain,
resulting in formation of the second ‘‘fixed strut”. During further
loading stages, these two ‘‘fixed struts” in concrete work in either
tension or compression, depending on the strain field applied on
individual the panel element.
On top of this baseline formulation of the FSAM, which origi-
nally assumes zero shear stress along crack surfaces [41], various
behavioral models can be implemented in the FSAM for represent-
ing shear aggregate interlock behavior along cracks and dowel
action on reinforcing steel bars. In the present FSAM formulation
implemented in OpenSees, a friction-based elasto-plastic constitu-
tive model for shear aggregate interlock effects in concrete ([27],
Fig. 3c) and a linear-elastic constitutive model for dowel action
on reinforcing steel bars ([11–13], Fig. 3e) are incorporated to rep-
resent the shear resisting mechanisms along cracks. The uniaxial
constitutive models by Chang and Mander [2] for concrete and
Menegotto and Pinto [22] for reinforcing steel were originally used
in development of the FSAM; however, any one of the uniaxial
models for concrete and steel available in OpenSees can be used
in the FSAM formulation implemented. The FSAM is also used to
define the constitutive behavior of the RC panel elements
251
(macro-fibers) used in the formulation SFI-MVLEM element
described in the previous section.
2.4. Uniaxial constitutive model for concrete (ConcreteCM)
The uniaxial hysteretic stress-strain relationship implemented
in OpenSees for concrete is the constitutive model developed by
Chang and Mander [2], which is a refined, rule-based, generalized,
and non-dimensional constitutive model that allows detailed cali-
bration of the monotonic and hysteretic model parameters and can
simulate the hysteretic behavior of both confined and unconfined,
ordinary and high-strength concrete, in both cyclic compression
and tension. The model addresses important behavioral features,
such as continuous hysteretic behavior under cyclic compression
and tension, progressive stiffness degradation associated with
smooth unloading and reloading curves at increasing strain values,
and gradual crack closure effects.
The compression envelope curve of the model by Chang and
Mander [2] (Fig. 4a) is defined by the initial tangent slope Ec,
the peak coordinate (e’c, f’c), a parameter r from Tsai [40] equation
defining the shape of the envelope curve, and a parameter e cr to
define normalized (with respect to e’c) strain where the envelope
curve starts following a straight line, until zero compressive
stress is reached at the spalling strain, esp. The tension envelope
curve is analogous to that of the compression envelope and is
defined using a similar set of envelope parameters for tension;
however, the tension envelope curve is shifted to a new origin
that is based on the unloading strain from the compression envel-
ope (Fig. 4a).
In order to define the hysteretic properties of the model, statis-
tical regression analyses were performed by Chang and Mander [2]
on an extensive experimental database. Based on the regression
analyses, empirical relations were developed for key hysteretic
parameters, such as those for secant stiffness (Esec) and plastic stiff-
ness (Epl) upon unloading from, and stress and strain offsets (Df
and De) upon return to the compression and tension envelopes
(Fig. 4b). Upon each unloading from the compression envelope,
the origin of the tension envelope is shifted based on the unloading
strain from the compression envelope (eun–), and the unloading
strain from the tension envelope (eun+) is re-evaluated so that it
corresponds to a tension strain ductility equal to a previously-
experienced compression strain ductility upon unloading from
the compression envelope, or a previously-experienced tension
strain ductility, whichever is greater. In terms of modeling general-
ized hysteretic behavior, the constitutive model uses smooth
252 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262

Fig. 4. Uniaxial material model ConcreteCM: (a) Compression and tension envelope curves, and (b) hysteretic parameters.

connecting curves (with slope continuity) for unloading and
reloading between the compression and tension envelopes, as well
as for partial unloading and reloading between the connecting
curves.
The original formulation of the constitutive model by Chang and
Mander [2], incorporating modifications by Orakcal [24] to remedy
minor numerical instabilities (see also [11]), is implemented in
OpenSees as uniaxialMaterial ConcreteCM. The newly-
implemented ConcreteCM model incorporates the smooth unload-
ing/reloading curves defined in the original formulation of the
Chang and Mander [2] model, as opposed to the existing Con-
crete07 model, which also an adaptation of the Chang and Mander
[2] model in OpenSees, but with simplified multi-linear unloading
and reloading rules. To illustrate the differences between the
newly-implemented ConcreteCM and the existing Concrete07, a
representative hysteretic response comparison obtained using
the two models is presented in Fig. 5a. In addition to user-
defined envelope parameters for defining the compression and
tension envelopes of the constitutive model, an optional input
parameter named as gap (with a value 0 or 1) is introduced in
the ConcreteCM model to allow control over the plastic stiffness
upon unloading from the tension envelope (Epl+), which influences
the intensity of gap closure (more gradual versus less gradual) in
the hysteretic stress-strain behavior of concrete. The effect of this
parameter on a sample hysteretic response of ConcreteCM material
model is presented in Fig. 5b. This allows control of the level of
pinching in the lateral load-displacement behavior of a RC mem-
ber, as illustrated by Kolozvari et al. [11].
2.5. Uniaxial constitutive model for reinforcing steel (SteelMPF)
The uniaxial hysteretic constitutive model for reinforcing steel
proposed by Menegotto and Pinto [22], as extended by Filippou
et al. [5] to include isotropic strain hardening effects, is imple-
mented in OpenSees as uniaxialMaterial SteelMPF, for simulating
the hysteretic stress-strain behavior of reinforcing steel bars.
Although similar material model formulations for reinforcing steel
are already available in OpenSees (e.g., Steel02), the formulation of
SteelMPF brings several distinctive features compared to the exist-
ing models. For example, the implemented SteelMPF model allows
definition of different yield stresses and strain hardening ratios for
compression and tension, which allows consideration of tension
stiffening effects on only the tensile stress-strain behavior of rein-
forcing bars embedded in concrete. As well, the model allows cal-
ibration of isotropic strain hardening parameters independently
for compression and tension. Furthermore, SteelMPF considers

Fig. 5. Implemented concrete material model ConcreteCM: (a) Cyclic behavior compared with Concrete07 and (b) gap closure behavior.
 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262 253

degradation of the cyclic curvature parameter R for strain reversals
in both pre- and post- yielding regions of the hysteretic
stress-strain behavior, which may provide more accurate predic-
tions of the yield capacity of RC structural members under cyclic
loading (discussed in Section 3.2), whereas Steel02 considers the
cyclic curvature degradation after formation of post-yield strains
only. Fig. 6a compares strain-stress responses obtained using
SteelMPF and Steel02 for a strain history that includes two strain
reversals before yielding, calculated at a boundary element at the
bottom of a slender RC wall specimen (RW2, [38], Fig. 8a) sub-
jected to reversed-cyclic lateral loading. Details on the wall speci-
men, the analytical wall model used, and analysis results are
presented in Sections 3.1 and 3.2. It can be observed from the
Fig. 6a that the difference between the two radii of the curved tran-
sitions between the elastic and post-yield stress-strain responses is
significant, which can considerably influence the predictions of the
overall wall load-deformation behavior as illustrated in Fig. 9b
(Section 3.3). Furthermore, strain-stress relationships obtained
using the existing Steel02 model in OpenSees show that when
reloading occurs after partial (small-amplitude) unloading in a
RC model element (possibly caused by dynamic loading or stress
re-distribution due to concrete cracking or local crushing), the
Menegotto and Pinto [22] formulation tends to produce artificial
stress overshooting in the cyclic stress-strain behavior of reinforc-
ing steel. This overshooting effect is not behavioral and causes non-
physical hardening in the stress-strain behavior upon reloading
after partial unloading, which has also been acknowledged by
Filippou et al. [5]. This anomaly, which is illustrated in Fig. 6b for
the Steel02 model, leads to overestimation of steel stresses pre-
dicted by Steel02 upon return from partial unloading and produces
a stress-strain curve that does not represent the physical constitu-
tive behavior of reinforcing steel. The overshooting effect observed
in Steel02 has been remedied in SteelMPF via manipulating the
model formulation so that reloading behavior after partial unload-
ing cannot overshoot the previous loading loop in the cyclic stress-
strain behavior. Comparison of strain-stress relationships obtained

Fig. 6. Steel02 versus SteelMPF: (a) Cyclic degradation of curvature parameter R, and (b) stress overshooting upon strain unloading/reloading.

Fig. 7. Model discretization for specimens (a) 2CLH18 [15] and (b) H/D(3)-T/M(0.0)/1.32% [36].
254 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262

Fig. 8. Model discretization for specimens (a) RW2 [38] and (b) B8 [23].

Fig. 9. Comparison of load-displacement responses for specimen RW2 [38]: (a) Experimental versus analytical results obtained using ConcreteCM and SteelMPF and (b)
Analytical results obtained using Concrete02, Concrete07, Steel02 and SteelMPF.

using the newly-implemented SteelMPF and the existing Steel02
models, for a strain history that includes low-amplitude unloading
followed by reloading, is presented in Fig. 6b.
3. Experimental validation of implemented models
3.1. Test specimens used for model validation
The MVLEM and SFI-MVLEM models implemented in OpenSees
were used to predict the behavior of two RC column specimens,
a bridge column specimen with a circular cross-section [36] and
a rectangular building column [15], as well as four RC wall speci-
mens [23,38,39] selected from the literature. The column and wall
specimens selected for the validation study are characterized with
a wide range of characteristics, including various shear-span-to-
depth ratios (1.5, 2.0, 3.0, and 3.2), horizontal and vertical reinforc-
ing ratios (low to moderate), axial load levels (from 0 to 7% of Agf0 c)
and failure modes (flexural, shear and combined shear/flexure fail-
ure). All specimens were tested under constant axial load and a
reversed-cyclic lateral displacement history applied at the top.
Important properties of column and wall specimens considered
are summarized in Tables 1 and 2, respectively.
3.2. Model calibration
Analytical models of the column and wall specimens were gen-
erated in OpenSees using the MVLEM and SFI-MVLEM model ele-
ments. The specimens were discretized with a number of model
Table 1
Properties of column specimens.

Spec. Size H Span to depth Vertical reinf. qv Horiz. reinf. qh fy f0 c N N/Agf0 c Failure mode
mm mm ratio (%) (%) MPa MPa kN
(in) (in) (ksi) (ksi) (kips)
1. / 635 1829 3.0 12 D25.4 mm 2.10 D12.7 mm@70 mm 1.32 450.9 25.78 592 0.070 Flex.
(/ 24) (72) (12 #8) (#4@2.75 in.) (65.4) (3.74) (1 3 3)
2. 457  457 2946 3.2 8 D25.4 mm 1.94 D9.5 mm@457 mm 0.07 330.9 33.09 503 0.073 Flex/Sh
(18  18) (1 1 6) (8 #8) (#3@18 in.) (48.0) (4.80) (1 1 3)

K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262

Table 2
Properties of wall specimens.

Spec. L H t H/L Boundary reinf. qbound. Web reinf. qweb fy f0 c N N/Agf0 c Failure modea
mm mm mm (%) (%) MPa MPa kN
(in) (in.) (in.) (ksi) (ksi) (kips)
RW2 1219 3658 102 3.00 8 D9.5 mm 2.93 D6.35 mm@190 mm 0.33 430.0 42.81 381.6 0.072 CB
(48) (1 4 4) (4) (8 #3) (#2@7.5 in.) (62.4) (6.21) (85.8)
B8 1905 4470 Web: 2.34 12 D19 mm 3.67 Horizontal: h: 1.38 447.5 47.23 1192.6 0.089 WC
(75) (1 7 6) 102 (4) (12 #6) D9.5 mm@102 mm v: 0.29 (64.9) (6.85) (268.1)
Bound: (#3@4 in.)
305 (12) Vertical:
D6.35 mm@229 mm
(#2@9 in.)
RW-A20-P10-S63 1219 2438 152 2.00 4 D12.7 mm + 4 D19 mm 6.06 D9.5 mm@127 mm 0.61 474.6 48.61 659.7 0.073 CB
(48) (96) (6) (4 #4 + 4 #6) (#3@5 in.) (68.8) (7.05) (148.3)
RW-A15-P10-S78 1219 1829 152 1.50 4 D19 mm + 4 D15.9 mm|(4 #6 + 4 #5) 7.11 D9.5 mm@152 mm 0.73 474.6 55.78 663.2 0.064 DC
(48) (72) (6) (#3@6 in.) (68.8) (8.09) (149.1)
a
CB = Crushing/buckling in boundary, WC = Web crushing, DC = Diagonal compression.

255
256 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262
elements along the specimen height (n) and a number of RC macro-
fibers along the specimen cross-section (m), based on the specimen
geometry and the configuration of vertical reinforcement. Fig. 7
illustrates discretization of the models for the column specimens,
with seven elements along the height (n = 7) and seven macro-
fibers (m = 7) across the cross-section of the rectangular column
specimen 2CLH18 ([15]; Fig. 7a), and with six elements along the
height (n = 6) and five macro-fibers (m = 5) for the circular column
specimen H/D(3)-T/M(0.0)/1.32% ([36]; Fig. 7b). Model discretiza-
tion for the wall specimens was conducted such that the wall
cross-section is represented with a number of approximately
equal-width panel macro-fibers, where one RC panel is assigned
to each wall boundary and the remaining panels are used for the
web, while model element height is chosen such that aspect ratio
of the panel macro-fibers is approximately 1.0, as illustrated in
Fig. 8. Discretization of the model for the rectangular wall speci-
men RW2 [38], with six macro-fibers along wall length (m = 6)
and fifteen model elements along the wall height (n = 15) is shown
in Fig. 8a, whereas model discretization for the barbell-shaped wall
specimen B8 [23], with six macro-fibers along the wall length (m =
6) and thirteen model elements along the wall height (n = 13), is
presented in Fig. 8b. The reinforcement ratio in vertical and hori-
zontal directions for each macro-fiber (RC panel) was obtained
based on the areas of corresponding reinforcing bars and concrete
within the macro-fiber. For the circular column specimen, the rein-
forcement ratio in the horizontal direction was obtained based on
the area of transverse reinforcement defined for each macro-fiber
multiplied by the cosine of the angle between the orientation of
the transverse reinforcement in that particular macro-fiber and
the direction of loading (e.g., E-W, Fig. 7b).
The parameters of the uniaxial material models for concrete and
reinforcing steel were calibrated to match as-tested material prop-
erties (where available), or using empirical relationships presented
in the literature, as described by Orakcal and Wallace [25]. At con-
fined regions (e.g., confined wall boundaries), the monotonic
envelope of the compressive stress-strain relationship of concrete
was calibrated using the confinement model proposed by Mander
et al. [16], based on the amount and configuration of the transverse
reinforcement provided. A linear-elastic shear force-deformation
relationship, with effective shear stiffness equal to 0.5GAg (cracked
shear stiffness commonly used in practice) was used for calibration
of the shear spring in the MVLEM. In the SFI-MVLEM, there is no
need to define a shear stiffness or a shear backbone relation, since
shear behavior of the model is coupled with the stress-strain
behavior of concrete (as well as shear aggregate interlock along
cracks and dowel action on the reinforcement) at the model ele-
ment level. Constant axial load and cyclic lateral displacement his-
tory matching test conditions were applied at the top node of the
analytical models, and displacement-controlled analysis was per-
formed to simulate the cyclic response of test specimens.
3.3. Comparison of MVLEM predictions with experimental data
3.3.1. Response predictions for a flexure-controlled wall and sensitivity
to material models
Fig. 9a compares the lateral load versus top displacement
responses measured during the test and predicted by the MVLEM,
using the newly-implemented material models for concrete (Con-
creteCM) and reinforcing steel (SteelMPF). As shown in the figure,
the analytical model captures, reasonably well, the experimentally
measured lateral load versus top displacement response of the
slender wall specimen RW2 [38], which showed flexure-
dominated behavior. Cyclic properties of the response, including
stiffness degradation, hysteretic shape, plastic (residual) displace-
ments, and pinching behavior are reasonably well represented in
the analytical results; therefore, cyclic characteristics of the imple-
mented stress–strain relationships for concrete and reinforcing
steel are suitable for obtaining accurate global response predic-
tions. The lateral capacity of the wall is predicted very closely for
most lateral drift levels. The underestimation of the wall capacity
at intermediate drift levels in the negative loading direction (e.g.,
0.5–1.5% drift) can be attributed to the inability of the yield asymp-
tote in the stress–strain model for steel in tension to represent the
curved strain-hardening region observed in the stress–strain tests
for the #3 longitudinal reinforcing bars, as well as uncertainties in
calibration of the cyclic parameters governing the implemented
steel stress–strain model (R0, a1, and a2 of SteelMPF) and the
parameters associated with concrete tensile strength (f t and et of
ConcreteCM).
Furthermore, Fig. 9b compares MVLEM predictions using exist-
ing material models in OpenSees for concrete and steel, in order to
compare the capabilities of the newly-implemented material mod-
els to the existing ones. The existing concrete models considered
in the comparisons include the commonly used stress-strain rela-
tionship proposed by Yassin [45] (Conrete02), which is character-
ized with the compression envelope by Kent and Park [8], multi-
linear unloading/reloading relationships, a linear tension stiffening
branch and sudden gap closure, as well as a simplified version of
the Chang and Mander model ([2], Concrete07) with multi-linear
unloading/reloading branches (Fig. 5a). The existing material
model considered for reinforcing steel is the existing Menegotto
and Pinto [22] formulation (Steel02, Fig. 6). It can be observed from
Fig. 9b that using Steel02 results in overprediction of the yield
capacity of the wall during the loading first cycle to yield displace-
ment, compared to both model results obtained using SteelMPF
and experimental results, because the curvature parameter R in
the Steel02 model formulation does not degrade during loading
cycles prior to yielding of the boundary reinforcement, as also
illustrated in Fig. 6a. Fig. 9b further shows that when Concrete02
is used in the analysis, the pinching characteristics of the response
are significantly overestimated due to simplified and abrupt gap
closure behavior implemented in the Concrete02 formulation. In
addition, when Concrete07 is used to simulate the stress–strain
behavior of concrete, the overall lateral load capacity of the wall
is underestimated by approximately 10%, and shape of the hys-
teretic loops is characterized with almost no pinching due to the
simplified unloading/reloading rules implemented in the Con-
crete07 adaptation of the Chang and Mander [2] model. Overall,
material models available in OpenSees for concrete and reinforc-
ing steel incorporate shortcomings that can lead to overestimation
of yield capacity and overestimation or underestimation of the
hysteretic energy dissipation capacity of RC walls. The material
models implemented in OpenSees for concrete (ConcreteCM) and
steel (SteelMPF) as part of this research study provide the opportu-
nity to obtain improved predictions of the hysteretic lateral load –
deformation responses of RC members subjected to reversed cyclic
loading.
3.3.2. Response predictions for elements with shear-flexure interaction
behavior
Comparison of MVLEM predictions with experimentally-
measured responses for a rectangular column specimen 2CLH18
[15] and a wall specimen RW-A15-P10-S78 [39] are presented in
Fig. 10. The responses of these specimens are also simulated with
the widely-used displacement-based beam-column element (DB-
BC, [37]) model available in OpenSees, which is conceptually sim-
ilar to the MVLEM as discussed in Section 2.1, and analysis results
are compared with test data in Fig. 11. The DB-BC element model
was calibrated similarly to the MVLEM in terms of geometric dis-
cretization and material properties, and 2 (or 4) integration points
per model element were used for the analyses. Column specimen
 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262 257

Fig. 10. Experimental versus MVLEM results for load-displacement responses of specimens: (a) 2CLH18 [15] and (b) RW-A15-P10-S78 [39].

Fig. 11. Experimental versus DB-BC model results for load-displacement responses of specimens: (a) 2CLH18 [15] and (b) RW-A15-P10-S78 [39].

2CLH18 exhibited predominantly flexural behavior (with flexural
deformations comprising about 80% of total lateral displacement,
where shear and other deformations contributed by approximately
20%) until shear-flexure failure occurred at a displacement ductil-
ity level of 5.1. The behavior of wall specimen RW-A15-P10-S78
was characterized with significant contribution of nonlinear shear
deformations (approximately 40%) to total lateral displacement
and overall shear-flexure interaction behavior. Results presented
in Fig. 10 show that the MVLEM is not capable of accurately simu-
lating the load-displacement response of the column and the wall
specimens investigated, since their behavior is influenced by both
nonlinear shear and flexural deformations. For both the column
and the wall specimen, the initial stiffness, unloading/reloading
stiffness values, yield strength, and hysteretic energy dissipation
capacity are considerably overestimated by the MVLEM. Fig. 11 fur-
ther reveals that the load–displacement responses obtained using
the DB-BC model for the column and the wall specimen are very
similar to the results obtained using the MVLEM, in terms of stiff-
ness, capacity, and hysteretic shape of the response, which demon-
strates the very similar response prediction capabilities of the two
modeling approaches. Therefore, modeling approaches where flex-
ural and shear responses are assumed uncoupled, such as the
MVLEM and DB-BC element, or any other commonly used fiber-
based model formulation (e.g., shear wall element in Perform 3D
[29]), lack the ability to accurately represent the behavior of RC
components where shear deformations are substantial and interac-
tion between shear and flexural responses occurs, which is typi-
cally observed in RC columns and walls with shear-span-to-
depth ratios between approximately 1.0 and 3.0. As presented in
the following section, the SFI-MLVEM model implemented in Open-
Sees as part of this study overcomes this significant shortcoming of
uncoupled modeling approaches.
258 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262
3.4. Comparison of SFI-MVLEM predictions with experimental data
The SFI-MLVEM also was calibrated and validated for the RC col-
umn and wall specimens described in Section 3.1. The following
sections provide details on validation of the implemented SFI mod-
eling approach against experimental data at both global and local
response levels.
3.4.1. Response predictions for reinforced concrete colummns
Fig. 12 compares the experimentally-measured and
analytically-predicted load-displacement responses obtained using
the SFI-MVLEM for the two column specimens considered. It can be
observed from the figure that the analytical model predicts reason-
ably well the lateral load capacities of the columns at almost all
drift levels, as well as the hysteretic characteristics of the observed
load-displacement responses including cyclic stiffness degradation
and pinching characteristics. Lateral loads are underestimated by
approximately 5% at small and intermediate drift levels for Speci-
men #2 [36]. Initial stiffness of both column specimens is slightly
overestimated by the model, whereas unloading stiffness values at
increasing drift levels are predicted with reasonable accuracy.
Although the overall load-displacement behavior of the specimens
is well-predicted by the SFI-MVLEM, strength loss observed in the
experiments due to specific flexural or shear modes of failure
(excluding concrete crushing) is not always captured in the analyt-
ical results. This shortcoming is mainly associated with the inabil-
ity of the model to simulate certain failure mechanisms such as
buckling of reinforcing bars observed in the specimen tested by
Shanmugam [36] (since rebar buckling is not implemented in the
model formulation) or sliding shear failure along cracks (due to
the simplicity of the constitutive relationships implemented in
the model formulation to represent shear stress transfer mecha-
nisms across the cracks). These model limitations may lead to
overestimation of the drift capacity of columns experiencing such
failure modes. Future studies will focus on model improvements
to address these issues.
3.4.2. Response predictions for reinforced concrete walls
The SFI-MLVEM implemented in OpenSees also was used to pre-
dict the response of four representative wall specimens (Table 2),
with varying aspect ratios (1.5, 2.0, 2.6, 3.0), axial load levels (0%,
Fig. 12. Experimental versus SFI-MVLEM results for load-displacement responses of specimens: (a) 2CLH18 [15] and (b) H/D(3)-T/M(0.0)/1.32% [36].
7.5% and 10%), reinforcement configurations, and cross section
geometries (rectangular, barbell). Test results were compared to
model predictions in terms of lateral load versus top displacement
responses for all specimens, as well as lateral load versus shear and
flexural deformation components of top displacement and distri-
bution of vertical strains at the base of the wall, for the wall spec-
imens investigated.
3.4.2.1. Load-displacement responses. Fig. 13 compares the
experimentally-measured and analytically-predicted lateral load
versus top displacement responses for the four wall specimens
considered (RW2, [38]; B8, [23]; RW-A20-P10-S63 and RW-A15-
P10-S78, [39]. It can be observed from the figure that the analytical
model captures, reasonably well, the lateral load and stiffness char-
acteristics of all specimens at most of the applied drift levels. Lat-
eral loads on the specimens are overestimated at drift levels lower
than 0.5% for wall specimens with aspect ratios of 1.5 and 2.0, indi-
cating a moderate overprediction of lateral stiffness for moderate-
aspect ratio specimens, whereas for drift levels between 0.5% and
2.0%, analytically-predicted lateral loads are within the ±10% range
of the experimentally-measured values. In addition, hysteretic
characteristics of the response, including stiffness degradation,
plastic (residual) displacements, and pinching behavior are well-
represented in the model results for the moderate-aspect ratio wall
specimens with aspect ratios of 1.5 and 2.0, whereas pinching is
slightly overestimated for the remaining two relatively slender
wall specimens with aspect ratios of 2.6 and 3.0. Similarly to
results obtained for RC columns, significant strength degradation
observed in the tests at large drift levels, caused by buckling of lon-
gitudinal boundary reinforcement, or sliding shear failure adjacent
to the wall-foundation interface (observed during testing of speci-
men RW-A15-P10-S78), were not captured by the analytical model
(since present model formulation does not incorporate these fail-
ure mechanisms), resulting in overestimation of wall drift capacity.
3.4.2.2. Local responses. Comparisons of the experimentally-
measured and analytically-predicted flexural and shear deforma-
tion components of the lateral load versus top displacement
response of specimen RW-A15-P10-S78 are presented in Fig. 14.
The analytical model successfully captures the nonlinear flexural
and shear deformations and their coupling throughout the entire
 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262
Fig. 13. Experimental versus SFI-MVLEM results for load-displacement response of specimens: (a) RW2 [38], (b) B8 [23], (c) RW-A20-P10-S63 [39] and (d) RW-A15-P10-S78
[39].
cyclic loading history of the wall. As revealed in both experimental
and analytical results, flexural and shear yielding occur almost
simultaneously at a lateral load level of approximately 800 kN.
As well, the model successfully reproduces the shapes of the load
versus flexural and shear displacement responses, with flexural
response characterized by no pinching or low-level pinching, and
shear behavior characterized by a highly pinched load-
deformation response. While the analytical model reasonably cap-
tures the initial and cyclic flexural stiffness of the wall at all drift
levels, shear stiffness is overestimated at drift levels lower than
0.5%. The magnitudes of nonlinear flexural and shear deformation
components predicted by the model generally match the
experimentally-measured values throughout the loading history.
However, as depicted in Fig. 14b, the analytical model underesti-
mates the experimentally-measured shear deformations during
the second loading cycle to 3.0% drift in both directions, because
of its inability to capture the sliding shear failure adjacent to the
wall-foundation interface, which was also the reason for lateral
strength degradation in the experimentally-observed behavior.
259
A representative comparison between experimentally-
measured and analytically-predicted profiles of vertical strains
(along wall length) at the bottom of wall specimen RW-A20-P10-
S63, corresponding to lateral drift levels in the positive loading
direction of 0.5%, 1.0% and 2.0%, are presented in Fig. 15. Similar
correlation between model and test results was observed for other
drift levels as well as drifts in the negative loading direction. The
experimentally-measured average vertical strains in concrete are
obtained from four vertical LVDTs located at 10.2 cm (4 in.), 33.0
cm (13 in.), 88.9 cm (35 in.), and 11.8 cm (44 in.) along the length
of the wall specimen, over an LVDT gauge length of 35.6 cm (14
in.) (from 2 in. to 16 in. height from the wall base); whereas the
analytically-predicted strain profiles are obtained using the aver-
age vertical strains in along the bottommost SFI-MVLEM element
of 30.5 cm (12 in.) length.
It can be observed from Fig. 15 that the plane-sections-remain-
plane assumption incorporated analytical model formulation rea-
sonably approximates the experimentally-measured vertical strain
profiles at drift levels of 0.5% and 1.0%. At these drift levels, both
260 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262

Fig. 14. Experimental versus SFI-MVLEM results for load-deformation responses of specimen RW-A15-P10-S78 [39]: (a) flexure and (b) shear.

Fig. 15. Experimental versus SFI-MVLEM results for vertical strain profiles at wall base for specimen RW-A20-P10-S63 [39].

tensile and compressive strains at wall boundaries are predicted
with maximum 15% deviation from test results. However, at the
drift level of 2.0%, the experimentally-observed strain relaxation
at the wall boundary in tension is not represented in the model
results. The model generally overestimates tensile strains at the
wall boundaries at large drift levels, due to its plane-sections-
remain plane assumption. However, as also observed in Fig. 15,
the model provides reasonable predictions of the compressive
strains (overestimated by approximately 10%) and the depth of
the neutral axis on the wall cross-section.
4. Summary and conclusions
This paper provides information on analytical models imple-
mented by the authors into the widely-used and open-source com-
putational platform OpenSees, for simulating the response of
reinforced concrete structural walls and columns under cyclic
loading. Five new classes were added to existing OpenSees library
of model elements and materials, including: (1) the Multiple-Verti
cal-Line-Element-Model (MVLEM) with uncoupled axial/flexural
and shear responses, (2) the Shear-Flexure-Interaction Multiple-V
ertical-Line-Element-Model (SFI-MVLEM), which considers interac-
tion between nonlinear axial/flexural and shear responses, (3) the
Fixed-Strut-Angle-Model (FSAM), which is a plane-stress constitu-
tive model for RC panel (membrane) elements, (4) an improved
uniaxial material model for concrete (ConcreteCM), and (5) an
improved uniaxial material model for reinforcing steel (SteelMPF).
Descriptions of the implemented model formulations are provided,
with particular emphasis on distinguishing features between the
newly-implemented models and the existing models in OpenSees.
Additional information on the models and user manuals for the
new OpenSees classes can also be found on the OpenSees Wiki
web page.
 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262
The analytical models implemented into OpenSees were vali-
dated against experimental results obtained from tests on two
reinforced concrete columns with circular and rectangular cross
sections that experienced flexural and shear-flexural failures,
respectively, as well as four reinforced concrete wall specimens
characterized with a range of characteristics including aspect
ratios (1.5, 2.0. 2.4, 3.0), axial loads (between 0.0 and 0.10Agf0 c)
and failure mechanisms (flexural, web-crushing, diagonal com-
pression). Comparisons between experimentally-measured and
analytically-predicted load-displacement responses reveal that
both the MVLEM and the SFI-MVLEM models are capable of captur-
ing, with reasonable accuracy, the cyclic load-displacement
response attributes of slender (flexure-dominated) wall specimens,
including their lateral load capacity, lateral stiffness, cyclic stiffness
degradation, and pinching characteristics. However, for the
moderate-aspect-ratio walls considered, where interaction
between shear and flexural responses is pronounced, only the
SFI-MVLEM can capture the load-displacement behavior accurately,
and the MVLEM tends to over-predict the lateral stiffness and lat-
eral load capacity of such walls. In addition, it has been observed
that newly implemented material models for concrete (Con-
creteCM) and steel (SteelMPF) provide improved predictions of
the wall load-deformation behavior in terms of yield capacity, cyc-
lic stiffness degradation and pinching characteristics comparing to
results obtained using existing OpenSees material models. A major
shortcoming of the MVLEM and SFI-MVLEM model formulations,
identified based on the response comparisons presented, is the
inability of the models to capture strength degradation observed
in the experimental results due to failure modes associated with
buckling (or fracture) of reinforcing steel bars, sliding shear failure
along cracks, or lateral instability of the compression boundary,
since the implemented models do not incorporate these failure
mechanisms in their formulation.
Future research can focus on applying the models implemented
in this study to conduct improved seismic response simulation
studies on building or bridge components or systems, using Open-
Sees. Additional studies will focus on development and implemen-
tation of more robust constitutive models to represent shear
transfer mechanisms along cracks, development and implementa-
tion of rebar buckling and fracture behavior into the constitutive
relationship for reinforcing steel, and constitutive modeling of the
sliding shear mechanism above the base of a wall, all of which
would improve model capabilities towards capturing failure modes
associated with these mechanisms. Finally, the implemented con-
stitutive models can be used in other formulations, such as develop-
ment of a robust three-dimensional finite-element modeling
approach for simulating the responses of walls with non-
rectangular cross sections under bi-directional lateral loading.
Acknowledgments
This work was supported by the State of California’s Transporta-
tion Division - Earthquake Risk Reduction Fund through the Pacific
Earthquake Engineering Research Center (PEER) Transportation
Systems Research Program and by the National Science Foundation
under Grants CMMI-0825347 and CMMI-1208192. Any opinions,
findings, and conclusions expressed herein are those of the authors
and do not necessarily reflect those of the sponsors.
References
[1] Applied Technology Council. PEER/ATC 72–1: modeling and acceptance criteria
for seismic design and analysis of tall buildings; 2010.
[2] Chang GA, Mander JB. Seismic energy based fatigue damage analysis of bridge
columns: Part I - Evaluation of seismic capacity NCEER Technical Report No.
NCEER-94-0006. Buffalo (NY): State University of New York; 1994.
261
[3] Du K, Sun J, Xu W. Evaluation of section and fiber integration points in fiber
model. In: Proceedings of 15th world conference on earthquake engineering,
Lisbon, Portugal.
[4] Elwood KJ. Shake table tests and analytical studies on the gravity load collapse
of reinforced concrete frames [PhD Dissertation]. Berkeley: University of
California; 2002. p. 419.
[5] Filippou FC, Popov EP, Bertero VV. Effects of bond deterioration on hysteretic
behavior of reinforced concrete joints Report EERC 83-
19. Berkeley: Earthquake Engineering Research Center, University of
California; 1983.
[6] Fischinger M, Rejec K, Isakovic T. Modeling inelastic shear response of RC walls.
In: Proceedings of 15th world conference on earthquake engineering, Lisbon,
Portugal.
[7] Jiang H, Kurama Y. Analytical modeling of medium-rise reinforced concrete
shear walls. ACI Struct J 2010;107(4):400–10.
[8] Kent DC, Park R. Flexural members with confined concrete. J Struct Div, ASCE
1971;97(ST7):1969–90.
[9] Kolozvari K. Analytical modeling of cyclic shear-flexure interaction in
reinforced concrete structural walls PhD Dissertation. Los
Angeles: University of California; 2013.
[10] Kolozvari K, Wallace JW. Practical nonlinear modeling of reinforced concrete
structural walls. ASCE Struct J 2016;142(12). https://doi.org/10.1061/(ASCE)
ST.1943-541X.0001492.
[11] Kolozvari K, Orakcal K, Wallace JW. Shear-flexure interaction modeling for
reinforced concrete structural walls and columns under cyclic loading
Technical Report PEER 2015/12. Berkeley (USA): Pacific Earthquake
Engineering Research Institute; 2015.
[12] Kolozvari K, Orakcal K, Wallace JW. Modeling of cyclic shear-flexure
interaction in reinforced concrete structural walls. I: theory. ASCE J Struct
Eng 2015;141(5):04014135.
[13] Kolozvari K, Tran T, Orakcal K, Wallace JW. Modeling of cyclic shear-flexure
interaction in reinforced concrete structural walls. II: experimental validation.
ASCE J Struct Eng 2015;141(5):04014136.
[14] Lu XZ, Xie LL, Guan H, Huang YL, Lu X. A shear wall element for nonlinear
seismic analysis of super-tall buildings using OpenSees. Finite Elem Anal Des
2015;98:14–25.
[15] Lynn CA, Moehle PJ, Mahin AS, Holmes TW. Seismic evaluation of
existing reinforced concrete building columns. Earthq Spectra 1996;12
(4):715–39.
[16] Mander JB, Priestley MJN, Park R. Theoretical stress-strain model for confined
concrete. ASCE J Struct Eng 1988;114(8):1804–26.
[17] Mansour MY, Hsu TC, Lee JY. Pinching effect in hysteretic loops of R/C shear
elements. ACI Struct J 2002;205:293–321.
[18] Massone LM, Orakcal K, Wallace JW. Modeling flexural/shear interaction in RC
walls, ACI-SP-236, deformation capacity and shear strength of reinforced
concrete members under cyclic loadings. Farmington Hills (MI): American
Concrete Institute; 2006. p. 127–50 [Paper 7].
[19] Massone LM, Wallace JW. Load – deformation responses of slender reinforced
concrete walls. ACI Struct J 2004;101(1):103–13.
[20] Massone LM, Orakcal K, Wallace JW. Modeling of squat structural walls
controlled by shear. ACI Struct J 2009;106(5):646–55.
[21] McKenna F, Fenves GL, Scott MH, Jeremic B. Open system for earthquake
engineering simulation (OpenSees). Berkeley (CA): Pacific Earthquake
Engineering Research Center, University of California; 2000.
[22] Menegotto M, Pinto E. Method of analysis for cyclically loaded reinforced
concrete plane frames including changes in geometry and non-elastic behavior
of elements under combined normal force and bending. In: Proceedings, IABSE
symposium, Lisbon, Portugal.
[23] Oesterle R, Aristizabal-Ochoa J, Fiorato A, Russel H, Corley W. Earthquake
resistant structural walls-tests of isolated walls: phase II. Skokie (IL): Portland
Cement Association; 1979. p. 327.
[24] Orakcal K. Nonlinear modeling and analysis of slender reinforced concrete
walls [Ph.D. Dissertation]. University of California Los Angeles; 2004.
[25] Orakcal K, Wallace JW. Flexural modeling of reinforced concrete walls -
experimental verification. ACI Struct J 2006;103(2):196–206.
[26] Orakcal K, Conte JP, Wallace JW. Flexural modeling of reinforced concrete
structural walls - model attributes. ACI Struct J 2004;101(5):688–98.
[27] Orakcal K, Massone LM, Ulugtekin D. Constitutive modeling of reinforced
concrete panel behavior under cyclic loading. In: Proceedings, 15th world
conference on earthquake engineering, Lisbon, Portugal.
[28] Panagiotou M, Restrepo JI, Schoettler M, Kim G. Nonlinear cyclic truss model
for reinforced concrete walls. ACI Struct J 2012;109(2):205–14.
[29] Perform 3D, 3D performance based design software. Computers and Structures
Inc.
[30] Petrangeli M, Pinto PE, Ciampi V. Fiber element for cyclic bending and shear of
RC structures. I: theory. ASCE J Eng Mech 1999;125(9):994–1001.
[31] Priestley MJN, Benzoni G. Seismic performance of circular columns with low
longitudinal reinforcement ratios. ACI Struct J 1996;93(4):474–85.
[32] Pristley MJN, Seible F, Calvi GM. Seismic design and retrofit of bridges. New
York (NY): John Wiley; 1996.
[33] Saatcioglu M, Ozcebe G. Response of reinforced concrete columns to simulated
seismic loading. ACI Struct J 1989;86(1):3–12.
[34] Sayre B. Performance evaluation of steel reinforced shear walls MS Thesis. Los
Angeles: University of California; 2003.
[35] Seismostruct. Seismosoft Inc. – Earthquake Engineering Software Solutions.
262 K. Kolozvari et al. / Computers and Structures 196 (2018) 246–262
[36] Shanmugam SP. Seismic behavior of circular reinforced concrete bridge
columns under combined loading including torsion PhD
Dissertation. Missouri University of Science and Technology; 2009.
[37] Taucer FF, Spacone E, Filippou FC. A fiber beam-column element for seismic
response analysis of reinforced concrete structures Rep. No. UCB/EERC-91/
17. Berkeley (CA): Earthquake Engineering Research Center, College of
Engineering, University of California; 1991.
[38] Thomsen JH, Wallace JW. Displacement-Based Design of Reinforced Concrete
Structural Walls: An Experimental Investigation of Walls with Rectangular and
T-Shaped Cross-Sections Report No. CU/CEE-95/06. Postdam (New
York): Department of Civil Engineering, Clarkson University; 1995.
[39] Tran TA, Wallace JW. Cyclic testing of moderate-aspect-ratio reinforced
concrete structural walls. ACI Struct J 2015;112(6):653–65.
[40] Tsai WT. Uniaxial compressional stress-strain relation of concrete. ASCE J
Struct Eng 1988;114(9):2133–6.
View publication stats
[41] Ulugtekin D. Analytical modeling of reinforced concrete panel elements under
reversed cyclic loadings M.S. Thesis. Turkey: Bogazici University, Istanbul;
2010.
[42] Vecchio FJ, Collins MP. Compression response of cracked reinforced concrete.
ASCE J Struct Eng 1993;83(2):219–31.
[43] Vulcano A, Bertero VV, Colotti V. Analytical modeling of RC structural walls.
Proceedings, 9th world conference on earthquake engineering, Tokyo-Kyoto,
Japan, vol. 6. p. 41–6.
[44] Xu SY, Zhang J. Hysteretic shear-flexure interaction model of reinforced
concrete columns for seismic response assessment of bridges. Earthquake Eng
Struct Dynam 2010;40(3):315–37.
[45] Yassin MHM. Nonlinear analysis of prestressed concrete structures under
monotonic and cyclic loads PhD Dissertation. Berkeley: University of
California; 1994.