Comportamientos en Muros de Corte NEHRP
Comportamientos en Muros de Corte NEHRP
Comportamientos en Muros de Corte NEHRP
Conventionally, vertical beam-type elements are used for analysis of shear wall
structures. Beams may be formulated as “lumped plasticity” models with finite
plastic hinges or as “distributed plasticity” models with fiber elements. Plane
sections are enforced to remain plane within each fiber section, and axial and flexure
responses are typically uncoupled from shear behavior. In PERFORM-3D,
Nonlinear Analysis and Performance Assessment for 3D Structures (CSI, 2013c), the
wall elements are essentially of this fiber beam formulation.
Modeling of coupling beams is more straightforward, although assuming that the end
of the beam (at the face of the wall) is parallel to the centerline of the wall ignores the
local deformation (beam strain penetration) at the joint between the beam and the
wall.
5.1.3 Damping
The models used in the examples are more complex than are generally used in
current design practice; however, these studies indicate that the potential exists for
analytical tools to capture the complex behavior of concrete walls, and they also offer
insights related to the observations of concrete wall behavior presented in other
chapters.
5.2.1 Simulation of Overall Wall Buckling
The finite element simulation of the test was performed in LS-DYNA with the
concrete modeled as shell elements and the reinforcing bars modeled with fiber
integrated beam elements embedded into the concrete shells. LS-DYNA incorporates
a large deformation formulation, which automatically incorporates P-delta effects,
and an explicit solver which permits solution into the negative stiffness range.
The cyclic stress-strain characteristic of the steel material model for the element
fibers (in the absence of buckling) was compared with that measured by Rodriguez et
al. (1999). The experimental and modeled steel stress-strain relationships are
compared in Figure 5-2. The material model is therefore able to capture the cyclic
stress-strain behavior of the steel reasonably well, displaying the Bauschinger effect
on reversal of loading. While it tends to overestimate the reloading strength at lower
strain levels, it is accurate at higher strain levels.
A simulation of one of the buckling tests in Rodriguez et al. (1999), 16 mm (#5) bar
with s/d = 6, was conducted to investigate if the onset of buckling under the
experimental cyclic protocol could be replicated by analysis. The analysis model,
shown in more detail in Appendix F, was discretized into ten fiber elements along the
gauge length of the test specimen.
The analytical and experimental results were compared to verify that the model is
capable of capturing the initiation of experimentally determined buckling by
considering the difference in longitudinal strain ε 1 and ε 2 at opposite sides of the bar
over the gauge length. The comparison shows that the analytical model captures the
hysteretic stress-strain relationship at the two opposite sides of the test specimens
well.
Figure 5-3 Simulated effect of loading history for bars with s/d ratios of 2.5 and 8.
The results described above were used to investigate the possible influence of
reinforcing bar buckling on the capacity of the concrete wall boundary element of the
planar wall specimen PW4 tested and documented in Lowes et al. (2011). The main
reinforcing bars are restrained with crossties such that the s/d ratio is 4.
Figure 5-4 shows predicted axial force versus compressive strain characteristics of
the components of the whole boundary element (under uniform compressive strain)
for the following assumptions:
• Reinforcing bars in compression follow monotonic steel stress-strain.
• Reinforcing bars are previously cycled in tension and reenter compression with
minimum resistance.
• Reinforcing bars are previously cycled in tension and reenter compression with
maximum resistance.
Figure 5-5 shows the total resistance of the boundary element associated with the
three different assumptions for the behavior of reinforcing bars. In all cases, there is
net reduction in resistance as the unconfined cover concrete fails. Beyond that point,
the resistance in all three cases remains fairly constant. It is observed that while the
tangent stiffness remains positive when monotonic behavior of reinforcing bars is
assumed, stiffness becomes negative when the cyclic behavior envelopes are adopted
for compressive strains exceeding about 0.007.
Figure 5-5 Total axial force versus compressive strain in the boundary element
components of specimen PW4.
Material properties for the reinforcement and concrete (confined and unconfined)
were based upon the test results associated with wall specimen PW4, and are detailed
in Appendix F.
5.2.3.1 Results of Initial Simulation without Bar Buckling
An initial simulation was performed where the potential buckling of reinforcing bars
was suppressed. Figure 5-7 compares the base moment versus drift hysteresis of the
simulation with that of the test.
Figure 5-7 Comparison of specimen PW4 cyclic test and simulation results
without bar buckling.
While the maximum flexural strength of the wall and the cyclic stiffness degradation
are fairly well represented, the onset of gross strength reduction is not predicted. In
the test, crushing of concrete and bar buckling occurred at the toe of one boundary at
a drift ratio of 0.75%, followed by similar failure at the other boundary toe in the first
cycle to 1% drift. Figure 5-8 shows the observed damage. The simulation predicted
that drifts in excess of 2.0% could be achieved without loss of resistance.
A detailed study was performed to investigate bar buckling in the confined boundary
region of shear wall specimen PW4. The reinforcement cage was modeled using
fiber beam elements (as for the Rodriguez et al. (1999) single bar tests), but because
the middle bar on the end face of the wall was not restrained by a hook, the entire
cage at the end of the wall was modeled. The reinforcement cage was restrained such
that the bars could not deflect into the core of the wall, but could deflect outwards, as
would be the case if the cover concrete had spalled. The modeled reinforcement cage
was subjected to the vertical strain history predicted for the toe of the wall in Section
5.2.3.1, above. Figure 5-9 shows the shape of the reinforcement cage after buckling.
The middle (untied) bar buckles first, followed by the corner bars.
In order to incorporate this bar buckling performance into the shell model of wall
specimen PW4, the parameters input to the bar buckling algorithm of the shell were
adjusted, as well as possible, to represent the hysteresis of the detailed reinforcement
cage model. This is described in Appendix F.
The LS-DYNA simulation of wall specimen PW4 was re-run with the effect of bar
buckling represented within the shell element. Figure 5-10 compares the base
moment versus drift hysteresis obtained from this simulation to that of the
experiment. It can be seen that in this improved model, which includes bar buckling
in the shell element, deterioration of strength is predicted once the wall is cycled
beyond 0.75% drift. This is directly attributable to the representation of bar buckling
because no other changes were made to the model. The simulations predict the cyclic
response well (the boundary fails at 1% drift in the simulation, and in the test, one
boundary fails at 1% drift and the other at 1.25% drift).
Figure 5-10 Comparison of specimen PW4 cyclic test and simulation results with
bar buckling.
This set of simulations shows that the phenomenon of bar buckling has a major effect
on the cyclic response of the wall, and that it is possible, in principle, to include bar
buckling effects in predictive models.
Located in Concepción, the Alto Rio building was designed in 2007, completed in
2009, and collapsed in the 2010 Maule earthquake. Studies conducted by engineers
and researchers have led to a range of conclusions regarding the potential cause of
collapse. The building was 15 stories tall with two basement levels supported by a
mat foundation on alluvial soil. Photographs of the building before and after the
earthquake are shown in Figure 5-12. Sections of the building are shown in Figure 5-
13.
The construction of Alto Rio was typical of modern high-rise residential buildings in
Chile comprising interconnected, lightly reinforced, thin concrete shear walls
supporting reinforced concrete floor slabs without beams. Models of these buildings
are likely to include three-dimensional assemblies of interconnected thin walls, with
many openings, discontinuities, irregularities, and consideration of coupling via floor
slabs. As a result, seismic response analysis of reinforced concrete wall buildings of
this type presents several analytical challenges.
Using the findings from the preceding sections, a case study continuum model of the
Alto Rio building was developed to simulate the observed damage and collapse
behavior of the building. Material property information was collected from available
design drawings and testing of samples from the building following the earthquake.
(a) (b)
Figure 5-13 Alto Rio Building: (a) longitudinal section; and (b) transverse section.
Figure 5-14 Plan view of LS-DYNA model slice for the Alto Rio building.
Figure 5-15 shows elevations of the principal walls modeled in the longitudinal and
transverse directions to illustrate the irregularities and discontinuities. Figure 5-16
shows the plan for the first and second floors. The typical shell element used in the
walls was 10 inches (250 mm) high by 8 inches (200 mm) wide.
Grid 13 Grid 17
Figure 5-15 LS-DYNA model showing elevations. Dark blue color indicates
corner elements that are lightly confined, light colors are unconfined.
Nonlinear seismic response analyses were performed to explore the sensitivity of the
response to assumptions regarding the following:
• Material properties
• Inclusion of the vertical component of ground motion (in addition to the biaxial
horizontal components)
• Effect of the basement structure
• Level of intrinsic damping
The specified concrete grades for the building correspond to cylinder strengths f' c of
2.9 ksi (20 MPa) and 3.6 ksi (25 MPa). Initial simulations assumed a single concrete
material strength for the entire building of 4.4 ksi (30 MPa), allowing for an
overstrength factor, Ω, of 1.25 to 1.50 on the nominal strengths. The external faces
of the basement box were assumed to be effectively rigid.
Simulations were performed in which the exterior of the basement box was subjected
to either the triaxial Concepción ground motion time histories or to the biaxial
horizontal components only. In both cases, the simulation predicted concrete
crushing failure, which was initiated where the concentration of compression strain
occurs in the extreme fiber of the walls at ground level. The compression failure
zone does not spread vertically, but propagates across the width of the wall in
successive cycles, during which the sway period of the building elongates from 0.6
seconds to 2.2 seconds. The crushing is driven by both the effect of the weight of the
building acting on whatever area of concrete is in compression, and the lateral
seismic demand in axial, flexural, and shear response.
Figure 5-17 illustrates vertical strain and stress distributions in the wall on Grid 13
before and after concrete crushing during the response cycle approximately 11
seconds into the motion record for the triaxial simulation. Tensile stresses and strains
are positive. In this cycle, a horizontal band of (essentially unconfined) concrete is
predicted to crush (i.e., compression resistance becomes zero at strain greater than
0.005) on the east side of the wall at Grid 13 at ground level. The stress distribution
after crushing shows that the highest compression stress has moved from the extreme
edge of the wall to a point 3.7 m (12 feet) inwards of this; the entire zone outwards of
this has fully unloaded. Figure 5-18 illustrates shear stress distribution in the wall on
Grid 13 before and after concrete crushing.
The shear stress distribution in Figure 5-18 just prior to crushing shows that, above
the first story, more shear is taken in the east wall that the west wall. This is
consistent with the fact that this wall carries a higher vertical stress at this time (about
0.9 ksi, 6 MPa) than the west wall (approximately zero). The peak shear stress in the
east wall of less than 0.6 ksi (4 MPa) implies a principal compressive stress of 1.2 ksi
(8 MPa) and principal tensile stress of 0.3 ksi (2 MPa), which is close to the tensile
strength of the concrete. With the stiff basement assumption, there is clear evidence
that the shear is transferring to the grade level slab for reaction into the soil.
Figure 5-17 Vertical strain and stress distributions in the wall on Grid 13 before
and after concrete crushing.
Figure 5-18 Shear stress distribution in the wall on Grid 13 before and after
concrete crushing.
After the ground level concrete band on the east side of the wall on Grid 13 has
crushed, the highest shear stress occurs in the region of solid wall in the first story
below a series of stacked openings where the two upper walls connect, which is a
Figure 5-19 shows the total axial force in the wall on Grid 13 under uniaxial, biaxial
(no vertical component), and tri-axial simulation cases. Except in the uniaxial
excitation case, a major reduction in load in this wall occurs between 11 and 15
seconds into the shaking due to the crushing effect described above, occurring first at
the east, and then at the west end of the wall in the first story (with the vertical load
transferring to the longitudinal corridor walls). The observation of compression
damage over only a short height of the wall is predictable for the strain softening
failure associated with largely unconfined concrete and low reinforcement ratio. It is
also consistent with the observed damage on many of the buildings that remained
standing after the event.
Figure 5-19 Total axial force in the wall on Grid 13 under uniaxial, biaxial (no
vertical component), and tri-axial simulation cases.
For the uniaxial excitation case, there is concrete crushing at the top of the wall in the
first story at the location where a reduction of wall area occurs; the crushing
propagated to some degree, but did not cause collapse. Crushing in the biaxial and
triaxial excitation cases contributes to a gross degradation of the overall lateral
stiffness of the structure, and the elongation of fundamental period.
Figure 5-20 shows the response history of the transverse roof deflection ratio,
indicating that the period increases from about 0.6 seconds to 2.2 seconds just prior
to collapse. The elongation of the period beyond 1.3 seconds makes the building
susceptible to the sharp peak in the horizontal pseudo-acceleration spectra in the 1.5
to 2 second period range. This makes the building substantially more vulnerable to
damage, and ultimately leads to the rocking-induced collapse of the structure.
The degradation of the strength of the structure is reflected in the moment and shear
force histories in the wall on Grid 13, as illustrated in Figure 5-21 and Figure 5-22. It
can be seen that the effective period of the building increases slightly more under
triaxial excitation, perhaps reflecting slightly greater extent of concrete crushing.
Figure 5-23 shows the time history of the predicted total base shear ratio at grade
level. The maximum base shear ratio is 0.38, and is dominated by the first mode
component, the period of which increases during the event.
Figure 5-23 Total base shear ratio history at grade level of the wall on Grid 13.
Eventually, as vertical load-carrying capacity is lost at both ends of the wall at Grid
13, the building topples under gravity. Figure 5-24 shows a close-up of the damage
(a) (b)
(c)
Figure 5-24 Predicted damage: (a) overall view; (b) close-up of damage pattern;
and (c) alternate view of damage pattern.
Using test results on material samples taken from the structure after collapse, a
further set of simulations was performed assuming concrete with f' c = 5.2 ksi
(43 MPa) and reinforcing bars with f y = 70 ksi (480 MPa) and f u = 100 ksi (720
MPa).
The performance under triaxial excitation using measured properties was almost
identical to the performance of the simulations using expected properties (with f' c =
4.4 ksi; 30 MPa), up to 11 seconds into the record. After the eleventh second, the
increase in concrete strength substantially reduced the degree of crushing that
The sensitivity of the predicted outcome to the assumption regarding the basement
was explored under triaxial excitation, assuming that the basement box provided no
lateral support. In this analysis, the motions were applied to the bottom of the
basement only. No attempt was made to study soil-structure interaction.
Analysis with measured properties resulted in no significant zones of failure and the
slice model did not collapse.
The analysis was repeated with expected properties. As shown in Figure 5-25,
significant concrete crushing occurred in the walls at the bottom basement level, but
the slice model did not collapse.
Before concrete significant crushing After significant concrete crushing
Vertical strain
Vertical stress
Figure 5-25 Vertical strain and stress distributions in the wall on Grid 13 before
and after concrete crushing.
These simulations indicate that the assumption of a stiff basement is probably more
realistic since the following were observed: (1) a greater susceptibility to collapse;
(2) a collapse mode consistent with the observed damage; and (3) negligible damage
predicted below ground level, consistent with observed damage.
Nonlinear response history analysis of a slightly different slice of the Alto Rio
building was undertaken by Tuna and Wallace (2014) using the PERFORM-3D
software. This model represents the region centered around Grids 11 and 13. The
walls cantilevered from the lowest basement level, thus no restraint from the
basement was considered. Figure 5-26 illustrates the model.
Fiber beam cross sections assumed uniaxial stress versus strain relations for concrete
and steel having expected material strengths of f' c = 4.7 ksi (32.5 MPa) and f y = 70 ksi
(491 MPa). These values were 1.3 and 1.17 times the design strengths, respectively.
Models were developed both with and without slab coupling, which affects the
natural period of the model. More details of this study are provided in Appendix F.
Figure 5-27 shows the predicted roof drift ratio history, and Figure 5-28 shows the
base shear ratio history. It was observed that significant reduction in strength
commenced at 20 seconds (1.3% roof drift) and convergence failed at 25 seconds.
Figure 5-28 Base shear ratio time-history (Tuna and Wallace, 2014).
Although the models do not represent exactly the same slice of the building, broad
comparisons can be made to the LS-DYNA analysis results for the flexible basement
case. The following observations were made:
• Peak shear demands in the PERFORM-3D model are lower. This may be due in
part to the higher damping in the PERFORM-3D model. The intended Rayleigh
damping was 2.5% (between 0.2T and 1.0T) compared to 1% (frequency
independent) in the LS-DYNA model. However, significantly higher damping
will have been generated in PERFORM-3D because the effective response period
is about 1.6 seconds, which is more than twice the basic first mode period, and
well outside the intended Rayleigh damping range.
• Degradation of strength starts at 11 seconds in the LS-DYNA model and at 20
seconds in the PERFORM-3D model. This may be due to the difference in
damping, but may also be due to the plane-sections-remain-plane constraints on
the walls in PERFORM-3D, which suppress strain concentrations that might lead
to the initiation of crushing. Peak drifts in LS-DYNA are slightly higher (2.8%
compared to 2.4%).
• It is not clear what the failure mechanism is in the PERFORM-3D model because
the analysis terminated due to convergence difficulties.
Based on the results of wall simulation studies herein, the following observations are
made regarding considerations for analytical modeling of reinforced concrete wall
buildings:
• The lateral response of gravity loaded reinforced concrete walls is significantly
different under cyclic loading conditions than monotonic loading. Brittle
concrete crushing failure is possible even in walls that would be considered
ductile under monotonic loading.
• Reinforcing bar buckling is observed in cyclic wall tests, even with ACI 318
conforming details, and can be a decisive factor in the seismic response.
Reinforcing bar buckling can occur as the unconfined concrete cover spalls (i.e.,
when compressive strains exceed about 0.004). The initiation of buckling and
post buckling resistance of a reinforcing bar cage is affected by s/d ratio and
previous (tensile) loading history (such as described in Chapter 2 of this report).
• In relatively thin walls, the maximum compression resistance of the boundary
zone is likely to occur as soon as the cover concrete spalls, irrespective of how
well confined the core is. This can lead to major instantaneous loss of strength if
the wall is supporting force-controlled gravity load.
• Overall wall buckling can be caused by concrete crushing, or by application of
compression to a wall that has been subjected to significant tensile strain, such
that the ensuing compression is resisted primarily by the previously stretched
reinforcing bars (such as described in Chapter 3 of this report).
• All of the above effects can be analytically replicated using a large-deflection
finite element solver with reinforcing bars modeled using a refined mesh of fiber
beams and advanced (e.g., layered sandwich type) nonlinear shell elements to
model the concrete. The steel material model must include Bauschinger type
hysteresis.
• For practical purposes, it is not feasible to model every reinforcing bar with
multiple fiber beam elements between restraining ties to simulate buckling
explicitly. The possibility of a phenomenological algorithm for the reinforcing
bar, including buckling (calibrated against tests and detailed fiber beam models)
for use within a multi-layer reinforced concrete shell element was demonstrated.
Further work is required to develop a generalized algorithm.
• The initiation of concrete spalling is a function of the local peak compression
strain. Spalling permits bar buckling, and will often trigger significant loss of
strength and negative stiffness. These studies and test measurements show that
the plane-sections-remain-plane assumption will significantly underestimate peak
compressive strain in the critical regions of planar shear walls, and the same is
Based on the case study simulation of damage observed at the Alto Rio building, and
results obtained from studies by other researchers, the following observations are
made:
• LS-DYNA analyses using shell elements with expected material properties
simulate overall building behavior that is consistent with the observed collapse of
the Alto Rio building, and these analyses indicate that the collapse was governed
by flexure-induced (and gravity driven) concrete crushing. However, because
concrete crushing is a brittle behavior, the extent of concrete crushing observed
in analysis is sensitive to modeling assumptions, such as the concrete strength at
critical locations and ground motions at different times and in different
directions.
• Once crushing initiates, it propagates extensively in a single load cycle. This is
an example of in-cycle strength degradation. The gravity load being supported
by the concrete at the point of crushing (in combination with flexure) is a force-
controlled action, which is a particularly effective driver for propagation.
• LS-DYNA simulations and the assessments of others predict severe damage and
reduction of lateral resistance at roof drifts of 1% to 1.3%, but collapse does not
occur until drifts exceed 4%. The building has a redundant gravity-load carrying
system with many walls, which the simulations indicate continue to carry load.
These results are consistent with the observation that many heavily damaged
buildings did not collapse.
• There is evidence of high shear stress beneath vertically aligned series of wall
openings (such as described in Chapter 4 of this report) but the LS-DYNA
simulation indicates that this was not the main factor leading to collapse in this
case. One of the reasons may be that almost all of the flexure above the first
story is taken in just one of the two walls, which is the one with the higher
compression. This is consistent with the concern of Kohrangi et al. (2012) that
the shear transferred by coupling could fully yield the small amount of tensile
reinforcement in the tensile side wall.
• The effective period of the building nearing collapse is much longer than the
nominal natural period. For such cases, the usual period range over which
Rayleigh damping is to be targeted needs to be reconsidered, otherwise the