Professional Documents
Culture Documents
net/publication/327065847
CITATIONS READS
11 1,434
10 authors, including:
Matias Hube
Pontificia Universidad Católica de Chile
71 PUBLICATIONS 327 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Forecast Engineering: From Past Design to Future Decisions | ERASMUS+ Programme - Key Activity 2: Strategic Partnerships in Higher Education View project
All content following this page was uploaded by John W. Wallace on 17 June 2019.
Fig. 3—MVLEM formulations: (a) 2-D version (Orakcal et al. 2004); and (b) 3-D version (Fischinger et al. 2004).
tion points per story is typical), and corresponding areas of is a major disadvantage of the FWE and other fiber-based
the materials on the wall cross section. Nodal resultants are models available in the literature.
computed using an appropriate numerical integration rule
along the length of the element. FWE employs material regu- Multiple-Vertical-Line-Element-Models with and
larization to obtain cyclic predictions of the wall behavior without shear-flexural interaction
that are insensitive to mesh size, particularly in the softening The Multiple-Vertical-Line-Element-Model (MVLEM)
region (that is, loss of capacity) of the wall load-deformation was originally proposed by Kabeyasawa et al. (1983) and
response (Coleman and Spacone 2001); details on material later modified and extended by many researchers (Vulcano et
regularization used in FWE are presented by Vásquez et al. al. 1989; Fischinger et al. 1990; Orakcal and Wallace 2006;
(2016). Chowdhury and Orakcal 2013). In the original MVLEM
Shear flexibility is incorporated in the FWE via shear formulation, axial/flexural behavior of a wall segment is
sections (Fig. 2(b)), which is common with other force-based simulated using a number of uniaxial springs (macro-fibers)
beam-column elements. Specific to the FWE, the behavior connected to rigid beams at the top and the bottom of the
of the shear section is defined using backbone relationship wall segments, as illustrated in Fig. 3 for two- and three-di-
by Gérin and Adebar (2009) that is characterized with four mensional (2-D and 3-D, resepectively) versions of the
points: 1) cracking (γcr, τcr), where shear cracking capacity is model. The shear response of the model element is uncou-
calculated using ACI 318-14 Equation 22.5.6.1; 2) yielding pled from the axial and flexural responses, and is defined
(γy, τy), which is not reached under flexural compression using a horizontal spring (Fig. 3(a) and (b)); the location
dominated behavior; 3) failure (γf, τf), which is only reached of the spring also represents the assumed center of rotation
in the section with largest demand, due to localization of for the element. In the presented study, two conceptually
deformations; and 4) the maximum shear resistance under different MVLEM-FD and MVLEM-SS were employed:
flexural compression behavior (γflex, τflex). In the FWE, shear 1) MVLEM-FD defines the response using force-displace-
response is dependent on axial load only by incorporating ments approach, and 2) MVLEM-SS is based on stress-
the Modified Compression Field Theory (Vecchio and strain relationships.
Collins 1986) at the shear section level. However, shear MVLEM-FD—In the basic version of MVLEM-FD, the
behavior is uncoupled from the flexural behavior, which force-displacement relationships presented in Fig. 4(a) and
Fig. 5—SFI-MVLEM-FD accounting for inelastic shear and shear-flexural interaction in structural walls (vertical springs are
not shown).
(b) are used to define the flexural and shear response, respec- HSA mechanism in Specimens R2 and WSH6; for specimen
tively. MVLEM-FD was extended to account for axial-flex- descriptions, refer to the section “Experimental Date Used
ural-shear interaction (Rejec 2011) by introducing a hori- for Model Evaluation”. Factors that define pinching, damage
zontal shear spring on each vertical spring (macro-fiber), and degradation of the unloading stiffness in the Hysteretic
where the behavior of each horizontal and vertical spring model were specified based on typical shapes of hysteretic
is coupled (Fig. 5); this model is referred to as SFI-MV- loops (corresponding to HSA, HSD, and HSS mechanisms)
LEM-FD. Each of the horizontal springs takes into account as reported in the previous benchmark studies: α = 1.0, β =
three shear mechanisms: 1) dowel effect of vertical bars 1.5, γ = 1.05, δ = 0.5 (Fig. 4). More details about SFI-MV-
(HSD) described by relationships proposed by Dulacska LEM-FD can be found in Fischinger et al. (2014).
(1972) and Vintzeleou and Tassios (1987); 2) axial resis- MVLEM-SS—Another modification of the original
tance of horizontal/shear bars (HSS) defined using model by MVLEM to incorporate interaction between shear and flex-
Elwood and Moehle (2003); and 3) interlock of aggregate ural behavior was proposed by Kolozvari et al. (2015a,b;
particles in the crack (HSA) modeled using constitutive rela- Fig. 6) and implemented into publicly available version
tionships by Vecchio and Lai (2004). The current state of of the computational platform OpenSees (Kolozvari et al.
each spring component depends on deformations/displace- 2018). The proposed model, called the shear-flexure interac-
ments at the effective cracks of the element that are linked tion MVLEM (SFI-MVLEM-SS), incorporates biaxial RC
to the current displacements of the element nodes, which panel behavior into a two-dimensional formulation of the
enables coupling between axial/flexural and shear behavior MVLEM-SS implemented by Orakcal el al. (2004). Axial-
at the model element level. In the SFI-MVLEM-FD model shear coupling is achieved at the panel (macro-fiber) level,
formulation used in this study, constitutive behavior of the which allows coupling of axial/flexural and shear responses
shear resisting mechanisms (HSA, HSD, and HSS) are at the model element level. The constitutive RC panel model
represented with the Hysteretic material model available in implemented in the SFI-MVLEM-SS is the OpenSees mate-
OpenSees for Specimens RW-A15-P10-S78 and S6, while rial FSAM, which is an extension of the Fixed Strut Angle
the ShearSlip material (Kante 2005) is used to model the Model (Ulugtekin 2010, Orakcal et al. 2012; Fig. 6(b)).
Fig. 7—Hysteretic rules controlling response of springs in SFI-MVLEM-SS element: (a) strain-stress behavior of concrete; and
(b) strain-stress behavior of steel. (Note: 1 MPa = 0.145 ksi.)
The FSAM incorporates several strain-stress mechanisms concrete and steel are applied (Fig. 8(b)). The NLTM is
to represent the behavior of RC material including: 1) a implemented in the publicly available computational plat-
biaxial concrete behavior based on the fixed-strut model that form OpenSees as element Truss2 coupled with uniaxial
employs OpenSees uniaxial material ConcreteCM (Fig. 7(a)) material ConcretewBeta. In this study, the wall boundary
based on Chang and Mander (1994) formulation (Fig. 6(c)); elements (when present) are modeled using continuous
2) a uniaxial constitutive relationship for steel represented by nonlinear fiber-based beam-column elements (Spacone et al.
OpenSees material SteelMPF that follows the Menegotto and 1996), while remaining vertical and horizontal elements are
Pinto (1973) formulation (Fig. 7(b)) and is applied along the modeled as truss elements with fiber sections. The layout of
directions of horizontal and vertical reinforcement (Fig. 6(e)); diagonal truss elements resembles, but does not necessarily
3) a friction-based constitutive model used to capture shear match, the principal compressive stress trajectories when
aggregate interlock effects along concrete cracks (Fig. 6(d)); approaching the ultimate load, where inclination angle of the
and 4) a linear-elastic relationship representing dowel action diagonals proposed by Lu and Panagiotou (2014) is used.
on reinforcement (Fig. 6(f)). The shear spring is removed Strut angles used in this study for the wall specimens consid-
from the original MVLEM-SS formulation because the shear ered are as follows: 44.9 degress for RW2, 55.2 degrees for
stiffness and strength of each macro fiber (and therefore the SP4, 63.4 degrees for R2, 59.0 degrees for WSH6, and 48.7
model element) evolve according to the implemented consti- degrees for S6; for specimen descriptions, refer to the section
tutive RC panel behavior. Therefore, explicit definition of “Experimental Date Used for Model Evaluation”.
shear modeling parameters is not necessary. Each panel of concrete diagonals comprises two four-node
elements, where two of the nodes connect concrete diagonal
Nonlinear Truss Model elements to the main truss, while the others allow connecting
The Nonlinear Truss Model (NLTM) proposed by Panag- a virtual strain-gauge element to monitor the strains normal
iotou et al. (2012) is based on the strut-and-tie (truss) to the diagonal of interest (Fig. 9(a)). The capacity of the
approach, where a structural wall is modeled using nonlinear diagonal concrete truss elements is reduced as a function of
vertical, horizontal and diagonal truss elements connected transverse tensile strain, which allows coupling between the
at nodes along which cyclic uniaxial material laws for element compressive stress-strain behavior with the tensile
shear-resisting mechanisms along concrete cracks. Differently, section “Description of macroscopic modeling approaches,”
the NLTM incorporates axial/flexural and shear interaction via including discretization of model geometry, calibration of
a strut-and-tie modeling approach. material models, and load application.
Fig. 10—Vertical and horizontal model discretization of Specimen SP4: (a) P3D-SW; (b) FWE; (c) SFI-MVLEM-FD; (d)
SFI-MVLEM-SS; and (e) NLTM. (Note: Units are in.; 1 in. = 25.4 mm.)
flexural response of the wall more precisely (Fig. 10(c)), ness, and reloading towards the same strain and stress from
whereas length of the web segments was defined based on which unloading initiated.
the distances between the longitudinal web reinforcement The concrete model used in the FWE was calibrated
(two to four bars per macro-fiber). A similar yet simplified using an ascending parabola followed by a linear softening
approach for cross-sectional discretization was employed branch in compression (Scott et al. 1982), whereas tensile
in the SFI-MVLEM-SS and NLTM models, where the two strength was neglected, which corresponds to the well-
outer panels represent the confined wall boundaries (only known Concrete01 material model in OpenSees. Although
one per boundary) and the remaining equal-width panels Concrete01 assumes zero tension capacity of concrete,
represented the wall web (Fig. 10(d) and (e)). previous studies with FWE (Vásquez et al. 2016) showed
that this approach provides reasonable predictions of hyster-
Calibration of material models etic behavior of RC walls. For concrete cover and unconfined
Calibration of material parameters, which define the concrete, slope of the descending branch was determined
flexural and shear response characteristics of the wall using the concrete crushing energy proposed by Nakamura
models based on either material strain-stress or constitutive and Higai (2001), and a residual strength of zero was speci-
force-deformation behavior, is described in the following fied. Parameters of the constitutive relationship for confined
sections. concrete were calibrated according to model proposed by
Models based on material strain-stress behavior— Scott et al. (1982), with a residual strength corresponding to
Concrete stress-strain relationship: In the P3D-SW 20% of the peak strength.
model, calibration of the unconfined concrete stress-strain In the SFI-MVLEM-SS model, the monotonic envelope of
relationship was based on as-tested material properties the stress-strain model for unconfined concrete in compres-
reported, whereas the confined concrete relationship was sion proposed by Chang and Mander (1994) (OpenSees
calibrated to follow the Saatcioglu and Razvi (1992) model. material ConcreteCM; Kolozvari et al. 2018) was calibrated
For both cases, trilinear relationships with strength loss to agree with material properties obtained from cylinder tests
(Fig. 1(c)) were used to approximately represent concrete on day of testing (when available) by matching the compres-
strain-stress behavior; with an initial slope of Ec up to a sive strength, the strain at compressive strength, initial tangent
stress of 0.4fc′, a second linear segment up to maximum modulus, and the shape parameter defining the monotonic
stress of fc′, a plateau at fc′, and a linear strength loss segment stress-strain curve proposed by Tsai (1988). The post-peak
following the plateau. This monotonic trilinear relationship slope of the strain-stress curve was calibrated to match the
forms the envelope for the cyclic strain-stress behavior, monotonic envelope proposed by Saatcioglu and Razvi
characterized by unloading parallel to the initial elastic stiff- (1992). The stress-strain envelopes for confined concrete
Fig. 16—Flexural and shear load-deformation responses for Specimen RW-A15-P10-S76. (Note: 1 mm = 0.0394 in.; 1 kN =
0.2248 kip.)
small magnitudes of the shear deformations measured, under- Local responses
estimation of shear deformations by SFI-MVLEM-SS model Figure 17 depicts comparison of analytically predicted
for this specimen did not affect the prediction of the overall and experimentally obtained vertical (longitudinal) strain
specimen hysteretic behavior. profiles along the wall base for Specimens RW2 and SP4,
A good agreement between the experimental results and the corresponding to drift levels of 0.5, 1.0, and 2.0%. The
response simulated by the NLTM is observed for both spec- experimental strains are measured using vertical LVDTs
imens in terms of relative contribution of shear and flexural over a gauge length of 230 and 355 mm (9 and 14 in.) for
deformations to total lateral displacement, as well as their Specimens SP4 and RW2, respectively, whereas the analyt-
hysteretic response shape. For Specimen RW2 (Fig. 15(e.1) ical strain predictions are obtained from the bottommost
and (e.2)), the model overpredicts shear deformations by element of each wall model considered, according to model
22% on average comparing to experimental data, whereas discretization described in the section “Mesh information”.
for Specimen SP4 (Fig. 16(e.1) and (e.2)), the ultimate shear All of the macroscopic modeling approaches considered
displacement after strength loss is underpredicted by 16% in this study, except the NLTM, employ the plane-sections-
in the positive direction and 40% in the negative direction remain-plane kinematic assumption. Therefore, analytically
(29% in average). obtained strain profiles for the plane-section models can
be defined with only two points at wall ends, with a linear
distribution of vertical strains in between. However, as can
also be observed in Fig. 17, experimentally measured strain
2 Table A.1 Comparison of simulated and measured response quantities based on experimental and
Model 0.69 204 2.16 248.3 248.3 306 1206 410 298 1095 410
SW‐P3D
Model 0.57 31 3.15 33.0 33.0 122 1830 7625 54 610 7625
Test 0.45 173 1.08 189.0 180.5 548 693 2766 386 693 874
Model 0.35 165 1.40 204.5 199.0 987 1410 7050 470 1175 914
Test 0.90 44 5.00 50.0 32.0 88 ‐ ‐ 49 ‐ ‐
FWE
R2
Model 0.50 186 1.09 186.0 167.5 1120 2135 2545 372 853 665
Test 0.90 44 5.00 50.0 32.0 88 ‐ ‐ 49 ‐ ‐
R2
Model 0.56 28 3.15 35.3 35.3 183 1650 16500 50 393 2357
Test 0.45 173 1.08 189.0 180.5 548 693 2766 386 693 874
SFI‐MVLEM‐SS
Model 0.29 165 1.44 204.5 193.0 1234 1410 9870 568 998 1337
Test 0.90 44 5.00 50.0 32.0 88 ‐ ‐ 49 ‐ ‐
R2
Model 0.58 29 3.19 33.0 33.0 166 1450 4350 51 483 1088
Test 0.45 173 1.08 189.0 180.5 548 693 2766 386 693 874
Model 0.33 165 1.44 184.9 157.5 1097 1974 2468 509 998 1045
NLTM