Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
TECHNICAL PAPER
INTRODUCTION
Shear failure of inadequately proportioned and detailed
columns is a leading cause of the damage and collapse of
older concrete buildings subjected to earthquakes. Assessment of the potential for shear failure requires an estimate
of expected earthquake ground shaking, expected dynamic
response of the building frame, and expected column capacity.
This study focused on the estimation of the building frame
dynamic response and column capacity. Analytical models
are developed, calibrated, and assessed by comparison with
the observed performance of a one-third-scale reinforced
concrete (RC) frame tested on an earthquake simulator. The
companion paper (Ghannoum and Moehle 2012) presents the
details of the experimental program and experimental results.
This study builds on important developments in analytical modeling and dynamic response simulation of RC in
recent years. The modeling and simulation are implemented
using OpenSees (McKenna et al. 2000), a software framework for developing applications to simulate the performance of structural and geotechnical systems subjected to
earthquakes. The software includes various types of finite
elements for modeling the nonlinear behavior of RC frame
elements subjected to earthquake loading, including the
force formulation fiber-section elements originally proposed
by Spacone et al. (1996). This element model takes advantage of nonlinear concrete and steel material properties also
incorporated in OpenSees.
More recently, Elwood (2004) introduced the limit state
material model for shear and axial failures of RC columns
with light transverse reinforcement. This model introduces
shear and axial springs attached in series with the fibersection elements, with limiting drifts associated with shear
and axial failure defining when these springs are activated.
It was of interest in this study to examine the efficacy of the
limit state model for a multi-story, multi-bay frame.
ACI Structural Journal/May-June 2012
403
SHAKE-TABLE TESTS
The test structure was bolted atop load cells fixed to the
earthquake simulator at the Richmond Field Station of the
University of California, Berkeley, Berkeley, CA. The test
structure was braced out of plane, and then lead weights
were attached to the beams to increase inertial effects and
column axial stresses. The resulting gravity axial load was
approximately 0.16Ag f c on the first-story interior columns
and 0.08Ag f c on the first-story exterior columns (Ag is the
column gross section area). Horizontal base motions representative of earthquake shaking were then imparted horizontally within the plane of the test structure. One shaking test
(designated the half-yield test) produced approximately
half the yield stress in the longitudinal steel of the first-story
columns. This test was followed by three high-intensity tests
that caused the partial collapse of the test frame. Only the
half-yield test and the first of the three high-intensity tests
(Test 1) are described in this paper. Additional details about
frame behavior during the strong shaking can be found in the
companion paper (Ghannoum and Moehle 2012).
ANALYTICAL MODEL
Beam and column elements
Force formulation fiber-section elements (Spacone et al.
1996) were used to model columns and beams with secondorder P-D effects included for columns. The fiber-section
force formulation is relatively simple to implement, as it
only requires section geometry and material properties as
input variables. These elements can account for axial load
variations while simulating the full range of cyclic flexural
behavior, including concrete cracking, steel yielding, and
concrete crushing.
Three sub-elements were used per column (Fig. 1). The
sub-elements at column ends had length h, where h is the
column section depth in the direction of loading. These
elements consisted of two fiber sectionsone at each
endwhich correspond to the two integration points of the
elements. The middle column elements were discretized
into five fiber sections. This configuration results in most,
if not all, of the inelastic rotations occurring within the
sub-elements at the column ends and produces an effective
analytical plastic hinge length of approximately h/2. Because
the underlying goal is to develop an analytical model that
requires the least amount of user calibration possible while
providing a high degree of accuracy, a more elaborate evaluation of the plastic hinge length was not deemed necessary,
and the simpler choice of h/2 was selected.
The beams were discretized into six force formulation
sub-elements containing two fiber sections each. These
sub-elements spanned between discrete loading points at
lead-weight attachment locations and out-of-plane bracing
connections. This approach produced sub-elements at the
ends of beams 12 in. (305 mm) long. The effects of beam
end element lengths on the model response were found to
be minimal.
Elastic shear deformations of frame elements were
introduced into the analytical model using zero-length
elastic spring elements at both ends of the columns and
beams (Fig. 1). The stiffness of these springs was given by
(5/6GAg)/(L/2), where G is the shear modulus, Ag is the gross
section area, and L is the element clear length. Shear deformations were minimal compared with flexural ones.
Zero-length springs (Elwood 2004) were introduced
at the base of the columns with wide hoop spacing
ACI Structural Journal/May-June 2012
s f s db
, s y
8ue
(1)
405
Ss =
y f y db
8ue
+ y
)( f
f y db
8u p
, s > y
(2)
where fy is the bar yield stress; fs is the bar stress at the interface; ey is the bar yield strain; es is the bar strain at the interface; db is the bar diameter; ue is the elastic bond stress; and
up is the plastic bond stress.
For a bilinear steel stress-strain relation, the bar stressversus-slip relation derived from Eq. (1) and (2) is parabolic
in both the elastic and plastic regions (Fig. 5). This relation
can be linearized in both regions without much loss in accuracy (Fig. 5). The bilinear stress-slip relation is then defined
by the yield stress fy, yield slip Sy, and hardening ratio (ratio
of plastic modulus to elastic modulus). From Eq. (1), the
longitudinal bar slip at yield is
Sy =
y f y db
(3)
8ue
fc
Joints
Joints were modeled as rigid elements. This approach
produced reasonably accurate analytical results, as shown
in the following, which suggests that the bar-slip model
described previously accounted for much of the deformation
associated with the joints.
Material properties
The concrete material model was a Kent-Scott-Park
concrete material with degraded linear unloading/reloading
stiffness according to Karsan-Jirsa (Scott et al. 1982). The
envelope relation was calibrated to the stress-strain relation
obtained from the cylinder tests. Confinement effects on core
concrete were estimated from Mander et al. (1988). Other
input concrete properties were: 1) concrete tensile strength
of 7.5 fc psi (0.62 fc MPa); and 2) concrete tensile softening slope of E0/5, where E0 is the concrete material tangent
modulus at zero strain = (2f c/ec) (where ec is the strain at f c).
The longitudinal reinforcing steel material model was a
Giuffre-Menegotto-Pinto (Menegotto and Pinto 1973) steel
material with isotropic strain hardening. Measured material
properties were used for the uniaxial behavior, except that
the yield strength was increased to account for strain-rate
effects. Maximum longitudinal reinforcement strain rates
recorded from strain gauges were found to be in the range
of 0.05 to 0.2 (1/s) in Test 1. Based on these rates and data
by Malvar (1998), the yield stress for longitudinal steel was
increased to 1.25 times the measured static yield stress.
The strain-rate-adjusted steel stresses produced calculated
sectional responses that more closely matched experimental
results than did calculations using the static properties. The
modulus of elasticity Es was not modified for strain-rate
effects (Malvar 1998). The strain-hardening ratio was 0.01.
Damping
Damping was introduced as stiffness and mass proportional
Rayleigh damping and was based on the first and second
modes of vibration of the structure. The lowest modes were
used to minimize the generation larger than defined mass
proportional damping. Modal frequencies used for damping
calculation were obtained from eigenvalue analysis updated
at each time increment. The tangent stiffness evaluated at
each time increment was also used in evaluating the damping
matrix. The damping ratio was taken as 2% of critical based
on laboratory test observations (Ghannoum 2007). Updated
tangent stiffness and modal properties were used to minimize the introduction of erroneous damping when the structure sustained large inelastic deformations (Charney 2008;
Petrini et al. 2008; Priestley and Grant 2005).
Analytical
Initial uncracked
Initial uncracked
0.31
0.35
0.32
0.34
0.101
0.115
0.107
0.114
0.069
0.08
0.068
0.073
406
of 12
fc psi (1
of 7.5
fc psi (0.62
Fig. 8First-story horizontal drift ratios versus axial loads: half-yield test.
ACI Structural Journal/May-June 2012
407
Fig. 9First-story critical column horizontal drift ratios versus shear: halfyield test.
and then the recorded base motion for Test 1. Figure 10 shows
characteristic global response results. The analytical model
reproduces with reasonable accuracy the main characteristics of the top- and first-level drift and base shear histories up
to the initiation of shear degradation in Column B1. Mainly
due to discrepancies between when shear failure occurred
in Column B1 and when it was triggered by the analytical
model, larger errors were observed in the analytical results
beyond that shear failure point. The analytical model also
adequately represents the observed first-story elastic stiffness and plastic shear levels (Fig. 12).
At the column element level, good agreement between the
analysis and experiment is achieved by the proposed analytical model up to the initiation of shear failure in Column B1
(Fig. 13). Both elastic stiffness and plastic shear levels are
reasonably estimated by the model, which supports the 25%
increase in the longitudinal steel yield strength for strainrate effects. Figure 14 compares the measured and calculated column end rotations. A close correlation between
the measured and calculated results is noted again up to the
initiation of shear failure in Column B1.
During Dynamic Test 1, Column B1 sustained severe shear
damage and strength degradation, along with gradual but
partial axial failure. Column A1 did not experience apparent
shear or axial degradations, although it did sustain minor
shear cracking. The analytical model, however, showed
shear failures of both Columns A1 and B1, with axial failure
initiation in Column B1. The analytical shear failures of
Columns A1 and B1 occurred a half cycle prior to the experimentally recorded shear failure initiation of Column B1
(Fig. 13). Columns A1 and B1 sustained shear failures in the
analytical model at drift ratios of 2.4% and 2.2%, respectively. These drifts are lower than the experimental drift ratio
at shear failure of Column B1 of 3.2%.
The shear spring attached to Column B1 appears to model
the shear degrading slope of that column fairly well, even
though the failure in the analysis occurred half a cycle prior
to the actual failure. After significant shear degradation, the
measured shear behavior of Column B1 appears to oscillate
randomly about a low residual value owing to the complex
geometry of the failed section and the continually varying
applied axial load. The shear failure model simulates that
behavior by limiting shear strength to a constant residual
shear value.
The axial failure model assumes that column axial failure
initiates when residual shear strength is reached. The
ACI Structural Journal/May-June 2012
Av f yt d
s
6[0.5] f
Nu
c
0.8 Ag
+ k
+ 1+
M
6[3.45e-3 ] fcAg
Vd
(4)
s
3
1
1 P
1
=
+ 4t
L 100
500 [ 42 ] fc 40 Ag fc 100
Sasani (2007) used a Bayesian parameter estimation technique to develop a probabilistic drift capacity model for RC
columns failing in shear. Sasani (2007) observed that the
drift ratio DR at shear failure increased with an increasing
volumetric transverse reinforcement ratio rs and shear spansection height ratio a/h and decreased with an increasing
axial load ratio (ho = P/Ag f c). Sasani (2007) also differentiated between the case of single and double curvature column
deformations through the factor q, which equals 1.0 for
columns deformed in single curvature and 0.85 for columns
deformed in double curvature. Shear failure is defined when
the shear resistance capacity drops to 80% of the maximum
shear. The model, which is intended to produce a mean
response estimate, is given by
(5)
0.18
DR = 0.74
s o
(6)
(0 < e
<1
( 13 e
<2
(7)
(9)
Figure 15 overlays the various model shear failure initiation points on the test results of Columns A1 and B1 for
Test 1. Table 2 lists the measured and calculated drift ratios
at shear failure and the ratios of calculated-to-measured
drift ratios. The measured drift ratio at shear failure is
defined as the absolute value of the drift ratio at which shear
resistance drops to 80% of the maximum shear. Note that
Column A1 sustained shear cracking for the largest positive
displacement excursion, although reversal of deformations at
the peak apparently protected the column from shear failure.
In contrast, nominal shear failure for Column B1 occurred
during the subsequent negative drift excursion, resulting in
clearly visible shear distress in the column (refer to Ghannoum and Moehle [2012]).
All four models produced conservative estimates of drifts
at shear-failure initiation. The models of Kato and Ohnishi
(2002) and Sasani (2007) produced the most accurate estimates of the nominal shear-failure drifts. The Kato and
Ohnishi (2002) model requires the input of drift ratio at yield
which, for this study, was taken directly from data measured
during the test, whereas the results for the Sasani (2007)
model were based entirely on data external to the shaketable test.
where
m cp 2
D
je 3en
p
=
L m cp
4
D j 2 3 5en 3
e
a
h
CONCLUSIONS
A detailed analytical model was developed to simulate the
dynamic behavior of a three-story, three-bay, RC frame up
to the point of severe shear-strength degradation and axial
failure of critical columns. A new zero-length fiber-section
representation of longitudinal bar slip from anchorage
regions was introduced. Good correlation between measured
and calculated load-deformation behaviors was observed at
(8)
Column B1
Measured or calculated
Calculated/measured
Measured or calculated
Calculated/measured
Measured
3.65
4.66
2.39
0.65
2.18
0.47
3.39
0.93
3.90
0.84
1.90
0.52
1.82
0.39
Sasani (2007)
3.89
1.07
4.00
0.86
Drift ratio at nominal shear failure is absolute value of drift ratio at which shear resistance capacity drops to 80% of maximum shear.
From Test 2.
411
the frame system and element levels. The model was capable
of simulating the important aspects of the measured dynamic
response from low deformation levels to well beyond flexural yielding. The following observations and recommendations are derived from this analytical exercise:
1. Force formulation fiber-section elements and zero-length
fiber-section bar-slip models can reproduce RC frame element
flexural behavior with high fidelity, as they are able to adapt
to variations in frame element boundary conditions (that is,
end restraints and axial loads). Such analytical models are
relatively simple to implement, as they only require section
geometry and material properties as input variables.
2. A bilinear longitudinal bar stress-versus-slip relation that
is based on a bi-uniform anchorage bond stress provided a
good basis for bar-slip-induced rotation modeling. An elastic
anchorage bond stress of 12
fc psi (1
REFERENCES
ACI Committee 318, 2008, Building Code Requirements for Structural
Concrete (ACI 318-08) and Commentary, American Concrete Institute,
Farmington Hills, MI, 473 pp.
ASCE/SEI Committee 41, 2007, Seismic Rehabilitation of Existing
Structures (ASCE/SEI 41/06), American Society of Civil Engineers,
Reston, VA, 428 pp.
Berry, M. P., 2006, Performance Modeling Strategies for Modern
Reinforced Concrete Bridge Columns, PhD dissertation, University of
Washington, Seattle, WA.
Charney, F. A., 2008, Unintended Consequences of Modeling Damping
in Structures, Journal of Structural Engineering, ASCE, V. 134, No. 4,
pp. 581-592.
Elwood, K. J., 2004, Modeling Failures in Existing Reinforced Concrete
Columns, Canadian Journal of Civil Engineering, V. 31, pp. 846-859.
Elwood, K. J., and Eberhard, M. O., 2009, Effective Stiffness of
Reinforced Concrete Columns, ACI Structural Journal, V. 106, No. 4,
July-Aug., pp. 476-484.
412