Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Strength Degradation Model for Low-Rise

Reinforced Concrete Walls Derived from


Dynamic and Quasi-Static Tests
Julian Carrillo,a) M.EERI, and Sergio M. Alcocer,b) M.EERI

Results from a previous experimental program demonstrated that loading


rate, strength mechanisms associated with the failure mode, low-cycle fatigue,
and cumulative values of displacement, and dissipated energy strongly affect
the degradation properties of reinforced concrete (RC) walls for low-rise housing.
Thus, data obtained from quasi-static (QS) cyclic tests should not immediately be
assumed to represent a conservative lower bound on a specimen’s capacity.
Aimed at numerically correlating results measured during dynamic and QS-cyclic
testing, this paper proposes a strength degradation model. The model is readily
applicable to seismic design or assessment of performance of existing and new
structures. A seismic demand model is also proposed for correlating the intensity
and duration of a given earthquake-induced movement to parameters that define
the degradation model. Correlations are deemed useful for interpreting the results
of QS-cyclic tests and for calibrating hysteretic and behavioral analytical models
obtained from QS testing. [DOI: 10.1193/011713EQS008M]

INTRODUCTION
The seismic performance of structural systems and components is commonly assessed
through experiments using quasi-static (QS) and shake-table testing. In QS testing, the strain
rate and low-cycle fatigue effects still generate uncertainties about the reliability of this
method. In shake-table testing, drawbacks are basically related to the cost of testing, equip-
ment limitations, specimen size, and the associated scale effects, as well as the difficulty of
controlling movements during testing. Despite extensive experimental research effort in the
last three decades, some important issues about the influence of testing methods on structural
performance continue to be the subject of discussion and disagreement among researchers.
For instance, there is concern that QS testing, which forms the basis of current design values,
is not representative of the actual demand imposed by a seismic event, and thus, can lead to
nonconservative designs (Gatto and Uang 2003).
To identify and assess the main parameters affecting the strength degradation observed
during shake-table and QS-cyclic testing of reinforced concrete (RC) walls for low-rise hous-
ing, an experimental and analytical research program was carried out. The first phase of this
program (Carrillo and Alcocer 2013a) was aimed at identifying the causes of observed

a)
Research Professor, Departamento de Ingeniería Civil, Universidad Militar Nueva Granada, UMNG, Colombia,
Carrera 11 No. 101-80, Bogotá, Colombia
b)
Research Professor, Instituto de Ingeniería, Universidad Nacional Autónoma de México, UNAM, Ciudad
Universitaria, CP 04510, Mexico City, Mexico

197
Earthquake Spectra, Volume 31, No. 1, pages 197–214, February 2015; © 2015, Earthquake Engineering Research Institute
198 J. CARRILLO AND S. M. ALCOCER

differences between results measured during shake-table and QS-cyclic testing of wall speci-
mens. The performance of 12 walls was compared by means of crack propagation, failure
modes, hysteresis curves, loading rates, number of cycles, and cumulative parameters of
response. When the dynamic and QS cyclic responses were compared, strength degradation
properties were shown to be clearly dependent on loading rate, strength mechanisms asso-
ciated with failure modes, and low-cycle fatigue, as well as cumulative parameters such as
drift ratio and dissipated energy. Thus, it was found that data obtained from QS-cyclic tests
cannot always be safely assumed to represent a conservative lower bound on the performance
capacity of RC walls for low-rise housing subjected to earthquake-induced deformations.
Considering advantages, drawbacks, and differences between shake-table and QS testing,
a cost-effective strategy may be a combination of these methods so that adjustments or cor-
relations of behavior can be made. In this way, it may be possible to better estimate the
behavior of specimens subjected to real seismic actions. However, a literature review reveals
the lack of research on correlations of this type for RC walls in low-rise housing. For this
reason, the second phase of the research program, which is the focus of this paper, was aimed
at correlating results measured during shake-table and QS-cyclic testing using a strength
degradation model. The model proposed in this paper considers the increments of damage
associated with low-cycle fatigue, energy dissipation, and cumulative parameters (e.g., dis-
placement demand and dissipated energy).

EXPERIMENTAL PROGRAM
A two-story house with RC walls was used as the prototype in this study. Typical low-rise
concrete housing has walls with concrete strengths varying between 15 MPa and 20 MPa;
100-mm thickness; web steel reinforcement smaller than the minimum prescribed by the
American Concrete Institute’s ACI 318-11 building code (ρmin ¼ 0.25%; ACI 2011);
and, almost always, web shear reinforcement made of welded-wire fabrics. Walls using
web reinforcement smaller than the minimum prescribed by the ACI code were tested to
study their performance and, in particular, to measure their lateral strength and deformation
capacity, because this type of construction is common in some Latin American countries in
low- and medium-hazard seismic zones, where design is controlled by vertical actions or by
temperature effects on concrete. Test results of these walls confirmed that the minimum shear
wall reinforcement ratio prescribed by ACI 318-11 appears to be excessive for controlling
diagonal tension cracking (Carrillo and Alcocer 2012b).
In the experimental program, 12 isolated walls were studied: six full-scale prototype
walls tested under QS-cyclic loading and six reduced-scale (1∶1.25) models tested under
unidirectional shake-table excitations. Examined variables were wall geometry (solid
walls and those with door and window openings), type of concrete (normal and lightweight),
web steel ratio (0.125% and 0.250%), type of web reinforcement (mild-steel deformed bars
and cold-drawn welded-wire fabrics), and testing method (shake table and QS-cyclic). The
main features of the walls in the study are summarized in Table 1. A detailed description of
the experimental program and design conditions may be found in Carrillo and Alcocer
(2013a, 2013b).
During shake-table testing, models were subjected to three earthquake hazard levels
(magnitudes 7.1, 7.7, and 8.3) using records labeled CA-71, CA-77, and CA-83. Models
STRENGTH DEGRADATION MODEL FOR LOW-RISE RC WALLS DERIVED FROM DYNAMIC AND QS TESTS 199

Table 1. Main specimen characteristics

Web reinforcement
Concrete
No. Wall Testing type Geometry type ρh,v (%) Type
1 MCN50mC Shake-table Solid Normal 0.125 Welded-wire fabric
2 MCN50mD QS-cyclic
3 MCN100C Shake-table Solid Normal 0.250 Deformed bars
4 MCN100D QS-cyclic
5 MCL50mC Shake-table Solid Light 0.125 Welded-wire fabric
6 MCL50mD QS-cyclic
7 MCL100C Shake-table Solid Light 0.250 Deformed bars
8 MCL100D QS-cyclic
9 MVN50mC Shake-table Openings Normal 0.125 Welded-wire fabric
10 MVN50mD QS-cyclic
11 MVN100C Shake-table Openings Normal 0.250 Deformed bars
12 MVN100D QS-cyclic

were tested under progressively more severe earthquake actions, scaled up considering the
value of peak acceleration as the reference factor, until the final damage stage was attained. In
QS-cyclic testing, the loading protocol consisted of a series of increasing amplitude cycles.
For each increment, two cycles at the same amplitude were applied (cycles I and II).

KEY TRENDS IN TEST RESULTS


The main parameters affecting strength degradation during dynamic and QS-cyclic test-
ing were assessed. Wall performance was compared using failure modes, hysteresis curves,
loading rate, number of cycles, and cumulative parameters (e.g., drift ratio and dissipated
energy). Significant findings and observations of the testing program are summarized next.

FAILURE MODELS AND CRACK PROPAGATION


During dynamic testing, walls reinforced for web shear using welded-wire fabric and
50% of the minimum code-prescribed web steel reinforcement exhibited diagonal tension
(DT) failure. This failure mode was governed by concrete inclined cracking, plastic strains
in most of the web shear reinforcement, and subsequent wire fractures. Failure was brittle
because of the limited deformation capacity of the wire reinforcement itself. In contrast, walls
reinforced with deformed bars and the code-prescribed minimum web steel reinforcement
exhibited a mixed failure mode in which diagonal tension and diagonal compression
(DT-DC) was observed (i.e., yielding of most of the web steel reinforcement and, simulta-
neously, noticeable web crushing of the concrete). Although walls tested under QS-cyclic
loading exhibited comparable failure modes and cracking patterns, the number and length of
200 J. CARRILLO AND S. M. ALCOCER

cracks were greater than in walls tested under dynamic loading. These differences were
mainly credited to the strain rate.

HYSTERETIC BEHAVIOR, LOADING RATE, AND LOW-CYCLE FATIGUE


The main differences in hysteretic behavior of the specimens tested using dynamic
actions and those tested under QS-cyclic loading were associated with shear strength capa-
city. For instance, in walls that failed in DT, where the inelastic portion of the stress-strain
curve was almost nonexistent, the degradation effect was not readily apparent. In these walls,
the peak strength measured during dynamic testing was 7% higher, on average, than that
measured during QS-cyclic testing, in spite of the number of equivalent cycles, which
was, on average, nine times that in QS-cyclic testing. Thus, it was confirmed that the higher
shear strength observed in dynamic testing of specimens with DT failure was associated with
the loading rate effect; for instance, maximum velocities measured during dynamic loading
were approximately 640 times higher than those measured during QS-cyclic testing.
For most walls tested under QS-cyclic loading and failure in DT-DC, strength degrada-
tion occurred at drift ratios much larger than those associated with peak shear strength; more-
over, the degradation rate was very low. In contrast, strength degradation in dynamic tests
began as soon as the peak shear strength was reached, and the degradation rate was more
pronounced. In walls failing in DT-DC, peak shear strength measured during dynamic testing
was 14% lower than that measured during QS-cyclic testing because the number of equiva-
lent cycles was 15 times higher, on average. As a preliminary conclusion, the higher the
number of cycles, the more pronounced the strength degradation. Furthermore, a higher
degradation rate causes lower drift capacities, which are associated with a particular
shear strength value (i.e., shear strength that defines limit states). For example, when com-
paring dynamic and QS-cyclic results for walls failing in DT-DC, drifts at shear strength and
ultimate drift capacities were, on average, 33% and 45% lower, respectively, in dynamic
testing. It was obvious that low-cycle fatigue is a key explanation for the differences in
observed behaviors between the two failure modes.

ENERGY DISSIPATION
During dynamic testing, the mean value of normalized cumulative dissipated energy at
shear strength in the specimens with DT failure was two times higher than that of specimens
with DT-DC failure. To remove the scale factors in the results from the shake-table tests and
to consider the different values of the mechanical properties of materials, the shear force
times drift ratio at shear strength was used to normalize the cumulative dissipated energy.
The observed variations in cumulative dissipated energy were essentially related to the effect
of low-cycle fatigue on the strength mechanisms, which in turn were associated with the
different failure modes. For instance, when the failure mode was governed by concrete crush-
ing (i.e., DT-DC failures), stiffness and strength degradation rates increased and pinching of
hysteresis loops became more significant as the number of cycles increased. As a result,
hysteresis loops were narrower, and thus, cumulative dissipated energy was reduced—
from 8.5 (on average) for walls with DT failure to 4.2 (on average) for walls with
DT-DC failure. In contrast, when the failure mode was controlled by plastic yielding of
the steel reinforcement and its subsequent fracture (i.e., DT failures), the effect of the number
STRENGTH DEGRADATION MODEL FOR LOW-RISE RC WALLS DERIVED FROM DYNAMIC AND QS TESTS 201

of cycles on the normalized cumulative dissipated energy was smaller than that observed in
walls with DT-DC failure (8.5 versus 4.2, on average).
When analyzing the results for specimens tested using QS-cyclic loading only, the effect
of number of cycles on concrete crushing was not apparent. Test results revealed that loading
rate, low-cycle fatigue, and cumulative parameters strongly affect the stiffness and energy
dissipation capacity. For this reason, data obtained from QS-cyclic tests should not always be
safely assumed to represent a conservative lower bound on the performance capacity of RC
walls for low-rise housing subjected to earthquake-induced deformations.

PROPOSED STRENGTH DEGRADATION MODEL


Nearly all models that have attempted to include parameters that affect the degradation of
component properties or the degradation of the concrete itself have been developed, partially
or entirely, from results measured in experiments (Shkolnik 1996, Petryna et al. 2002,
Maekawa and El-Kashif 2004, Tsuno and Park 2004). Based on the measured responses
of six wall specimens tested under QS-cyclic loading and six wall specimens tested
under shake-table excitation, an empirical strength degradation model was developed in
this study. To numerically correlate dynamic and QS-cyclic shear strength degradation, mea-
sured cumulative parameters were studied. For each fully reversed cycle and for both testing
methods, the following parameters were calculated: drift ratio, R; cumulative drift ratio, Rcum ;
ratio between dynamic and QS-cyclic shear strength, V∕V s ; dissipated energy, E; cumulative
dissipated energy E cum ; and mean value of equivalent cycles for a given value of drift ratio, N.
Parameters E cum and Rcum were calculated as the summation of dissipated energy or the
maximum drift ratio experienced in a cycle plus the values of the preceding cycles, respec-
tively. Parameter V s is associated with the maximum value of the shear strength measured
during the first (I) or second (II) cycle. The number of equivalent cycles for a given range of
drift ratio, N k , was estimated by using Equation 1:
E cum-i
Nk ¼
EQ-TARGET;temp:intralink-;e1;62;307 (1)
Ej

Parameter N k refers to equivalent cycles because it represents the maximum value of the
ratios between cumulative energy dissipated in cycle i, E cum-i , and energy dissipated that is
associated with a range of drift ratio j, E j . Energy dissipated, E j , was calculated by using
Equation 2:

1X
n1

Ej ¼ E (2)
n1 i¼1 i
EQ-TARGET;temp:intralink-;e2;62;215

where E i is the energy dissipated in cycle i and n1 is the number of cycles associated with a
given range of drift ratio. A numerical example of calculating N k may be found in Carrillo
and Alcocer (2013a).
In lieu of attempting to determine the yield strain (and displacement) in RC walls gov-
erned by shear deformations, drift ratio was selected as a representative deformation para-
meter. Trends from experimental results were used for selecting functional forms of
202 J. CARRILLO AND S. M. ALCOCER

equations that describe the degradation model. An iterative nonlinear regression analysis was
carried out for deriving the empirical equations of this study. Investigation of existing trends
between residuals (prediction errors) and model parameters helped to improve the forms of
the equations.
The strength degradation model developed here attempts to characterize the relation
between dynamic and QS-cyclic shear strength, V∕V s . Test data were sorted into groups
according to the parameters that control the model’s behavior. It was observed that strength
degradation was remarkably influenced not only by wall geometry but also by failure mode
(i.e., concrete crushing or plastic yielding of steel reinforcement). Thus four wall categories
were identified: category 1, for solid walls that exhibited a mixed DT-DC failure mode
(specimens MCN100D and MCL100D); category 2, for walls with openings that exhibited
a mixed DT-DC failure mode (specimen MVN100D); category 3, for solid walls that exhib-
ited DT failure mode (specimens MCN50mD and MCL50mD); and category 4, for walls
with openings that exhibited DT failure mode (specimen MVN50mD).

TRENDS IN EXPERIMENTAL RESULTS


Test data for R and V∕V s for the first three specimen categories and for five N values
(see Equation 1) are plotted in Figure 1. A square, triangle, circle, or diamond is used for
each N value. Fitted nonlinear regression curves are also shown in the graph. Because of the
limited displacement capacity of walls reinforced for web shear using the welded-wire fab-
ric available, the second (descending) branch of the strength degradation curve for cate-
gories 3 and 4 (DT failure mode) was not observed. Notably, in Figure 1c, there are
no data (squares) in the descending branch. For walls failing in DT but showing displace-
ment demands beyond peak shear strength (i.e., walls using wire reinforcement with higher
elongation capacity), the same strength degradation rate associated with categories 1 and 2
(DT-DC failure mode), respectively, was used. This was a conservative approach because
test results demonstrated that wall-strength degradation with a DT-DC failure is higher than
that with DT failure mode.
The trends shown in Figure 1 reproduce the relation between dynamic and QS-cyclic
envelopes of hysteresis curves. Shown in Figure 2b, for instance, are typical envelopes
of hysteresis curves (pull direction of displacement) associated with the earthquake records
used for shake-table testing of wall MCN100D and the envelope associated with the second
cycles of QS-cyclic testing. Other envelopes are given elsewhere (Carrillo and Alcocer
2013a). As an example, the dynamic envelopes undergo “softening” of the hysteresis
loops when compared to QS-cyclic envelopes (Figure 2b), mainly at low displacement
demands. Considering that wall behavior is governed by shear deformations, softening
occurred as a consequence of the pinching of the dynamic hysteresis loops. Pinching
increased considerably as cumulative damage increased; for instance, significant differences
are shown in Figure 2b when comparing the QS-cyclic envelope and the envelope associated
with earthquake record CA-83 at 75% (“83-75”).

MODEL DEVELOPMENT
From trends of experimental results, as well as from parameters used to measure model
accuracy, it was found that the model shown in Figure 2a properly describes strength
STRENGTH DEGRADATION MODEL FOR LOW-RISE RC WALLS DERIVED FROM DYNAMIC AND QS TESTS 203

Figure 1. Nonlinear regression analysis for the strength degradation model: (a1–a5) Solid walls
with DT-DC failure, (b1–b5) walls with openings and with DT-DC failure, and (c1–c5) solid
walls with DT failure.

degradation behavior. The strength degradation model could be divided into two branches
joined at drift ratio, R 0 . According to test data, R 0 represents the drift ratio associated with
peak V∕V s ratio for a given value of N.
To estimate the ratio of dynamic strength degradation to shear strength associated
with QS-cyclic behavior, V∕V s , a nonlinear model described by Equation 3 is proposed.
The bilinear model mainly depends on the number of cycles associated with a con-
stant value of N and on drift ratio R. Trends from experimental results shown in
204 J. CARRILLO AND S. M. ALCOCER

Figure 2. Strength degradation: (a) proposed model and (b) hysteresis envelopes for wall
MCN100D.

Figure 1 resulted in the following equation between strength degradation, V∕V s , and drift
ratio R:
V
EQ-TARGET;temp:intralink-;e3;41;449

¼ a3 Rb3 R < R0
Vs
V
¼ a4 Rb4 R ≥ R0 ð3Þ
Vs

where R 0 should be computed by using Equation 4:

R 0 ¼ c1 ln N þ c2 ≥ 0
EQ-TARGET;temp:intralink-;e4;41;364 (4)

where constants c1 and c2 depend on both wall geometry and type of shear failure mode.
Proposed values for these constants are presented in Table 2. In Equation 3, variables a3 and
a4 depend on ðV∕V s Þ0 and on drift R 0 and should be calculated using Equations 5 and 6,
respectively:

ðV∕V s Þ0
a3 ¼ (5)
R b03
EQ-TARGET;temp:intralink-;e5;41;284

a4 ¼ a3 R 0ðb3 b4 Þ
EQ-TARGET;temp:intralink-;e6;41;236 (6)

Table 2. Constants for the proposed strength degradation model

Failure Wall
Category type geometry c1 c2 d3 d4 e3 e4 f3 N cs FR
1 DT-DC Solid 0.138 0.840 0.002 0.891 0.024 0.49 1.335 5 0.181
2 Openings 0.129 0.855 0.004 0.762 0.327 0.47 1.343 5 0.136
3 DT Solid 0.142 0.995 0.018 0.396 0.0241 0.49a 1.199 3 0.079
4 Openings 0.137 1.053 0.032 0.319 0.3272 0.47a 1.214 3 0.113
a
Data not available; parameter taken from Categories 1 and 2, respectively.
STRENGTH DEGRADATION MODEL FOR LOW-RISE RC WALLS DERIVED FROM DYNAMIC AND QS TESTS 205

where ðV∕V s Þ0 is the initial strength ratio associated with R0 to be computed using
Equation 7. Drift ratio R0 represents the minimum drift ratio value to be used in the
model for ensuring mathematical stability and was purposely established as 0.005%.
ðV∕V s Þ0 ¼ f 3 þ g3 ln N
EQ-TARGET;temp:intralink-;e7;62;603 (7)

where constant f 3 depends on both type of shear failure mode and wall geometry. Proposed
values for f 3 are presented in Table 2. Variable g3 depends on f 3 and N cs and should be
calculated using Equation 8:
1  f3
g3 ¼ (8)
ln N cs
EQ-TARGET;temp:intralink-;e8;62;535

where N cs is the number of equivalent cycles observed in QS-cyclic testing. N cs values are
presented in Table 2. When using the strength degradation model, N should be between N cs
and the maximum number of steady cycles registered during dynamic testing
(∼300 ≥ N ≥ N cs ; see Figure 1). In Equation 3, variables b3 and b4 depend mainly on N
and should be computed using Equations 9 and 10, respectively.

b3 ¼ d 3 N d4
EQ-TARGET;temp:intralink-;e9;62;429 (9)

b4 ¼ e3 ln N þ e4
EQ-TARGET;temp:intralink-;e10;62;404 (10)
An iterative nonlinear regression analysis for each data group of N and wall category was
performed for calculating constants c1 , c2 , d 3 , d 4 , e3 , e4 , and f 3 (Table 2). Dashed lines in
Figure 1 represent the upper and lower limits of dynamic strength degradation in relation
to QS-cyclic shear strength. To calculate upper and lower strength degradation curves,
Equation 7 was modified as follows:
ðV∕V s ÞoðenvÞ ¼ ðV∕V s Þo ð1  FRÞ
EQ-TARGET;temp:intralink-;e11;62;314 (11)

where FR represents the mean value of coefficients of variation of ratios between predicted
values, using Equations 3 to 8, and test data. The proposed values of FR for the four wall
categories are presented in Table 2. The continuous line of the degradation model in Figure 1
represents the average trend in the test data (i.e., FR ¼ 0). Considering that design and
assessment processes involve prediction of maximum and minimum probable capacities,
respectively, the upper and lower limits of Equation 8 can be used for seismic design
and for performance assessment of low-rise RC walls, respectively.

MODEL ACCURACY
The assessment of accuracy is important to ascertain model validity and robustness,
despite the fact that the model is empirical. In this endeavor, two commonly used measures
for goodness of fit of the results of the nonlinear regression analyses were computed
(Benjamin and Cornell 1970): the correlation coefficient (r) and the standard error of the
residuals (SE). These correlation quantities are commonly used to establish the accuracy
of nonlinear regression analysis (Benjamin and Cornell 1970) because they qualify the over-
all fit of data with the nonlinear regression—(the r factor) as well as the deviation from such
206 J. CARRILLO AND S. M. ALCOCER

Table 3. Accuracy of the proposed strength degradation model

Failure Wall
Category type geometry Parameter Value Mean
1 DT-DC Solid N 21 53 99 176 248 —
r 0.889 0.861 0.845 0.802 0.826 0.845
SE 0.051 0.061 0.069 0.089 0.116 0.077
2 DT-DC Openings N 21 55 91 195 286 —
r 0.932 0.874 0.934 0.945 0.890 0.915
SE 0.044 0.056 0.071 0.128 0.113 0.082
3 DT Solid N 21 55 89 185 299 —
r 0.795 0.801 0.898 0.747 0.826 0.813
SE 0.035 0.033 0.043 0.041 0.044 0.039
4 DT Openings N 21 55 105 — 281 —
r 0.827 0.841 0.933 — 0.940 0.885
SE 0.046 0.041 0.037 — 0.064 0.047

analysis (the SE value). To evaluate and qualify the accuracy of fitted nonlinear regression
curves used for predicting strength degradation, it is important to establish the ideal char-
acteristics with which a fitted curve should comply. First, r should be equal to or very close to
1.0—for instance, higher than 0.8. Also, SE should be small—for instance, lower than 0.1.
Computed r and SE values are presented in Table 3 (the parameters are also shown in
Figure 1). As summarized in the table, the proposed strength degradation model showed
very good agreement with the test data. For the lowest accuracy condition, as an example,
mean r was higher than 0.81 and mean SE was lower than 0.08. This good agreement can be
partially credited to the use of the same data from which constants were derived.

FAILURE MODE AND WALL GEOMETRY EFFECTS ON STRENGTH


DEGRADATION
Strength degradation curves for the four wall categories were calculated using Equations 3
through 10 and are shown in Figure 3. Considering that they indicate dynamic strength degra-
dation in terms of QS-cyclic shear strength, a horizontal dashed line with V∕V s ¼ 1.0 repre-
sents the QS-cyclic results. According to the results shown in Figure 3, dynamic strength is
larger than that measured under QS-cyclic testing only when N is roughly equal to or lower
than 6.
For comparison purposes, the results for N ¼ 15 are shown in Figure 3e. To evaluate the
effect of failure mode type, wall categories 1 and 3 (black lines: solid walls with DT-DC and
DT failure modes, respectively) and categories 2 and 4 (gray lines: walls with openings with
DT-DC and DT failure modes, respectively) are compared. Figure 3e shows that, for walls
with DT-DC failure (continuous line), strength degradation of the second branch takes place
at drift ratios smaller than those in walls with DT failure (dashed line). Indeed, for a given N
value, peak shear strength in walls with DT-DC failure is smaller than that expected in walls
with DT failure. This observation coincides with test data in which strength degradation was
STRENGTH DEGRADATION MODEL FOR LOW-RISE RC WALLS DERIVED FROM DYNAMIC AND QS TESTS 207

Figure 3. Strength degradation curves: (a) solid walls with DT-DC failure, (b) walls with open-
ings and with DT-DC failure, (c) solid walls with DT failure, (d) walls with openings and with DT
failure, and (e) N ¼ 15.

found to be higher in walls that exhibited a failure mode governed by concrete crushing (i.e.,
DT-DC failures) than in walls that showed a failure mode controlled by plastic yielding of
reinforcement and its subsequent fracture (i.e., DT failures).
To assess the effect of wall geometry, wall categories 1 and 2 are compared with wall
categories 3 and 4 in Figure 3e. There are no significant differences in the first branch of
strength degradation curves. However, for solid walls, strength degradation in the second
branch was higher than that in walls with openings (i.e., smaller R 0 value and higher degra-
dation rate). Notably, walls with openings for structural purposes are made of wall segments
with height-to-length ratios smaller than that of the overall wall itself. Previous studies evi-
denced more pronounced strength degradation in squat walls than slender walls. For instance,
Hidalgo et al. (2002) indicated that strength degradation increases as the height-to-length
ratio of walls decreases. Slender wall segments in walls with openings are subjected to a
more steady flexural behavior because shear deformations are less pronounced, and thus,
strength degradation is significantly lower. This trend is commonly observed during testing
of components subjected to earthquake loads. For example, strength degradation of a com-
ponent governed by flexural deformations is lower than that of a component governed by
shear deformations.

PROPOSED SEISMIC DEMAND MODEL


To calculate V∕V s ratios of a low-rise RC wall subjected to a particular earthquake
record, the first step is to estimate, for a given value of drift R, the number of equivalent
cycles, N, induced by the design earthquake. Thus, a model was developed for correlating
seismic demand characteristics and parameters that define the degradation model. Initially, a
parameter that represented seismic demand characteristics was selected. Several parameters
were assessed based on their distinctive advantages and drawbacks. Peak ground acceleration
and ground displacement are basic parameters for force- or displacement-based seismic
208 J. CARRILLO AND S. M. ALCOCER

design, respectively, and are commonly specified by codes. However, these parameters fail to
directly include the cumulative earthquake-induced energy because they represent only peak
values of earthquake intensity.

REPRESENTATION OF SEISMIC DEMAND


The Arias intensity factor, I A , has been shown to be an accurate estimator of damage,
particularly for predicting the response of short-period structures (Travasarou et al. 2003)
such as low-rise housing. Therefore, I A was selected to represent peak value and earthquake
loading history in terms of intensity and duration (Arias 1970). Arias intensity is a measure of
potential earthquake destructiveness associated with an acceleration time history. Factor I A is
calculated for a given direction through a point by means of Equation 12:
ðt0
π
IA ¼ a2 ðtÞdt (12)
2g
EQ-TARGET;temp:intralink-;e12;41;500

where a is the ground acceleration along a given direction, t 0 is the total duration of the
earthquake, and g is the acceleration due to gravity. To consider the effect of the entire series
of earthquake records applied during shake-table testing, cumulative Arias intensity was used
in this study: IA-cum . This cumulative parameter represents the Arias intensity of both the
earthquake record being analyzed and the preceding records. The I A was calculated from
acceleration records measured at the base of specimens during shake-table testing; cumula-
tive IA-cum values are shown in Table 4.

MODEL DEVELOPMENT
The relation between IA-cum and N (Equation 1) for five ranges of R in specimens tested
under dynamic loading is plotted in Figure 4. The mean drift value at each range of drift ratio
is shown in the figure. Fitted nonlinear regression curves are also shown in the graph.
Considering the functional forms of the proposed degradation model as well as trends of
experimental results, the model shown in Figure 5a was found to be adequate for representing
the relation between earthquake demand and parameters that define the degradation model.
Equation 13 is proposed to estimate the N value from IA-cum , which is expressed in m∕s.

Table 4. Cumulative Arias intensity values for the prototype house

IA-cum (m∕s)
Earthquake
record MCN50mD MCL50mD MVN50mD MCN100D MCL100D MVN100D
CA-71, 50% 0.40 0.35 0.35 0.36 0.37 0.35
CA-71, 100% 1.91 1.83 1.86 1.87 1.90 1.88
CA-77, 75% 5.70 5.29 5.59 5.71 5.47 5.58
CA-77, 100% 12.41 11.36 12.48 12.21 11.71 12.05
CA-83, 75% —a — — 30.14 28.52 31.10
a
The record was not applied to the specimen (the preceding record returned a failure).
STRENGTH DEGRADATION MODEL FOR LOW-RISE RC WALLS DERIVED FROM DYNAMIC AND QS TESTS 209

Figure 4. Nonlinear regression analysis for the seismic demand model.

 x
I A-cum

EQ-TARGET;temp:intralink-;e13;62;429 (13)
I A1

where I A1 is the Arias intensity for N ¼ 1 for a given story drift (Figure 4) and x is a fitted
parameter. From experimental trends, Equations 14 and 15 are proposed for calculating these
two parameters.

I A1 ¼ 1.25R 1.75


EQ-TARGET;temp:intralink-;e14;62;356 (14)

x ¼ 1.05R0.01
EQ-TARGET;temp:intralink-;e15;62;323 (15)

Parameters I A-1 and x depend on the story drift of the equivalent cycles, R—for instance, story
drift at a given performance level. An iterative nonlinear regression analysis for each data
group of R was used for calculating the constants of Equations 14 and 15. Although the
seismic demand model is strongly related to the measured wall response, three earthquake

Figure 5. Seismic demand: (a) Proposed model and (b) seismic demand curves.
210 J. CARRILLO AND S. M. ALCOCER

Table 5. Accuracy of the seismic demand model

Parameter Value Mean


R (%) 0.06 0.17 0.30 0.47 0.54 —
r 0.969 0.981 0.966 0.953 0.868 0.918
SE 0.176 0.094 0.134 0.092 0.182 0.154

records with different characteristics were used during shake-table testing so that a broad
range of Arias intensity values was used for developing the seismic demand model.
Similarly to degradation models, the dashed lines in Figure 4 refer to the upper and lower
limits of the number of equivalent cycles and are computed as Equation 16:
I A1ðenvÞ ¼ I A1 ð1  FNÞ
EQ-TARGET;temp:intralink-;e16;41;489 (16)

where FN represents the mean value of the coefficients of variation of the ratios between
predicted values using Equations 13–15 and test data (FN ¼ 0.267). Again, the solid line
corresponds to the average trend of the test data (FN ¼ 0). The upper and lower limits of
Equation 16 can be used, respectively, for seismic design or performance assessment of low-
rise RC walls. Seismic demand curves calculated from model Equations 13–15 are shown in
Figure 5b. As shown, for a given Arias intensity, the number of equivalent cycles decreases
as the story drift increases. The proposed seismic demand model is useful for correlating the
seismic demand characteristics and the parameters that define the strength degrada-
tion model.

MODEL ACCURACY
The values of the statistical parameters used to measure goodness of fit of the results of
the nonlinear regression analysis for five R values are presented in Table 5 (these parameters
are also shown in Figure 4). As summarized in the table, the proposed seismic demand model
exhibits a suitable correlation with test data—that is, mean r was higher than 0.92, and mean
SE was lower than 0.16. Logically, this agreement can be traced to the use of the same data
from which constants were derived.

APPLICATION OF THE PROPOSED MODELS


As indicated, the proposed strength degradation model can be used for seismic design or
assessment of the seismic performance of RC walls for low-rise housing for a given seismic
demand. To apply the methodology, seismic demand should be initially characterized by
IA-cum . Then, the N induced by the earthquakes should be defined for a given R. Code limits
or acceptance limits proposed by Carrillo and Alcocer (2012a) may be selected as input for R.
A case study is presented to outline the application of the proposed models. It is intended for
seismic design—that is, the degradation model is associated with the upper limits of FR and
FN in Equations 11 and 16, respectively.
The proposed strength degradation model estimates the entire envelope of dynamic
hysteresis curves departing from a model calibrated using QS-cyclic test results or an
envelope of hysteresis curves measured during QS-cyclic testing. FEMA 356 (FEMA
STRENGTH DEGRADATION MODEL FOR LOW-RISE RC WALLS DERIVED FROM DYNAMIC AND QS TESTS 211

2000) and ASCE-41 (ASCE 2007) are examples of design guidelines that specify shear
force–displacement relationships derived from QS test results. Using the envelopes of
hysteresis curves measured during QS-cyclic testing in the experimental program reported
here and using the proposed strength degradation model, envelopes of dynamic strength
for some arbitrarily selected N values have been derived and are shown in Figure 6.
According to Cardenas et al. (1980), shear strength provisions for ordinary walls
(see Chapter 11 of ACI 2011) provide designs that assume strength equal to or higher
than that calculated for reversals of loading. However, the results in Figure 6 show that
shear failure mode, low-cycle fatigue, and cumulative damage may remarkably affect low-
rise RC wall shear strength. Taking into account that equations specified in ACI 318-11,
Chapter 11 (2011), were calibrated through QS-cyclic test results (see the dashed lines
labeled “QS-cyclic” in Figure 6), shear strength capacity observed during an earthquake
may decrease drastically with the number of cycles.
The proposed strength degradation model may also be utilized to estimate the envelopes
of dynamic behavior from the envelopes of hysteresis curves or behavioral models associated
with QS-cyclic behavior. For instance, envelopes of hysteresis curves associated with the
earthquake records used for shake-table testing were calculated using the seismic demand
model and the envelopes of hysteresis curves measured during QS-cyclic testing of RC
walls. These envelopes are shown in Figure 7. Predicted envelopes of all earthquake records
and measured envelopes are also shown in the graph. (The graphs are drawn to distinct scales
to better show results.)

LIMITATIONS OF THE PROPOSED MODEL


The applicability of the proposed predictive equations is limited to walls with
the characteristics of those tested as part of the research program reported here: typical
RC walls for one- to two-story economic housing, including solid walls and walls with

Figure 6. Predicted envelopes of dynamic hysteresis curves: (a1–a2) walls with DT failure and
(b1–b2) walls with DT-DC failure.
212 J. CARRILLO AND S. M. ALCOCER

Figure 7. Comparison of predicted and measured envelopes of dynamic hysteresis curves asso-
ciated with shake-table testing: (a1–a6) specimen MCL50mD (DT failure) (a5 is not shown
because the preceding record returned a failure), (b1–b6) specimen MCN100D (DT-DC failure),
and (c1–c6) specimen MVN100D (DT-DC failure).
STRENGTH DEGRADATION MODEL FOR LOW-RISE RC WALLS DERIVED FROM DYNAMIC AND QS TESTS 213

door and window openings, boundary element thickness equal to web thickness, response
governed by shear deformations, f c0 between 15 MPa and 25 MPa, axial stress <0.03f c0 ,
web steel ratio smaller ≤0.25%, wall web reinforcement made of deformed bars or
welded-wire meshes, and equal horizontal and vertical web reinforcement. Such models
are also applicable to walls made of normal-weight (19 ≤ γ ≤ 21 kN∕m3 ) and lightweight
(15 ≤ γ ≤ 19 kN∕m3 ) concretes. Although three earthquakes records with a broad range of
cumulative Arias intensity (0.3531.10 m∕s) were used during shake-table testing, the
models apply only to seismic demands with characteristics similar to those observed
in the epicentral region of Mexico. However, the rationale and methodology from
which they have been developed can be used to develop strength degradation models
of other structural systems under different seismic loading.

CONCLUSIONS
An experimental and analytical research program was carried out to identify and assess
the main response parameters affecting strength degradation of low-rise RC walls subjected
to shake-table and quasi-static cyclic loads. The effect of these parameters in modifying
degradation properties was evaluated. Tests results indicated that loading rate, low-cycle
fatigue, and cumulative parameters (such as drift ratio and energy dissipated) strongly affect
n strength degradation properties. It was found that data obtained from quasi-static cyclic
tests may not always be safely assumed as a conservative lower bound of the strength
and displacement capacity of RC walls for low-rise housing subjected to earthquake-induced
deformations.
The main response parameters obtained through shake-table and quasi-static cyclic
testing were assessed. Based on trends in experimental results and an iterative nonlinear
regression analysis, empirical equations were proposed for estimating the strength degra-
dation of low-rise RC walls. A seismic demand model was also developed for correlating
the intensity and duration of a given earthquake to the parameters that define the degra-
dation model.
The applicability of the equations presented here is limited to walls with the character-
istics of those found in one- to two-story concrete housing. Although equations were devel-
oped from a limited set of experimental data, it is thought that models are useful to obtain
performance trends. The proposed degradation model is readily applicable to the seismic
design and assessment of seismic performance of existing and new structures. It can also
improve understanding of results from QS-cyclic and dynamic testing. It is the authors’
opinion that the methodology discussed here can be easily applied to other type of structural
elements or subassemblages.

ACKNOWLEDGMENTS
The authors wish to express their gratitude and sincere appreciation to Grupo CEMEX
for financing this research study and to the staff and students of the Shake-Table Laboratory
of the Instituto de Ingeniería at UNAM for their extensive assistance in the experimental
program. The first author also expresses his gratitude to the Universidad Militar Nueva
Granada (UMNG, Colombia) for financing research activities through project number
214 J. CARRILLO AND S. M. ALCOCER

IMP-ING-1574. The views expressed in this paper are solely those of the authors and do not
necessarily reflects the views of the sponsor.

REFERENCES
American Concrete Institute (ACI) 2011. Building Code Requirements for Structural Concrete
(ACI 318-11) and Commentary, ACI Committee 318, Farmington Hills, MI.
American Society of Civil Engineers (ASCE), 2007. Seismic Rehabilitation of Existing Buildings
(ASCE/SEI 41-06), Reston, VA.
Arias, A., 1970. A measure of earthquake intensity, in Seismic Design for Nuclear Power Plants,
(R. J. Hansen, ed.), MIT Press, Cambridge, MA, 438–483.
Benjamin, J., and Cornell, C., 1970. Probability, Statistics and Decision for Civil Engineers,
McGraw-Hill, New York, NY.
Cardenas, A., Russell, H., and Corley, W., 1980. Strength of low-rise structural walls, in Rein-
forced Concrete Structures Subjected to Wind and Earthquake Forces, Publication SP-63,
American Concrete Institute, Detroit, MI, 221–241.
Carrillo, J., and Alcocer, S., 2012a. Acceptance limits for performance-based seismic design
of RC walls for low-rise housing, Earthquake Engineering and Structural Dynamics 41,
2273–2288.
Carrillo, J., and Alcocer, S., 2012b. Seismic performance of concrete walls for housing subjected
to shaking table excitations, Engineering Structures 41, 98–107.
Carrillo, J., and Alcocer, S., 2013a. Experimental investigation on dynamic and quasi-static beha-
vior of low-rise RC walls, Earthquake Engineering and Structural Dynamics 42, 635–652.
Carrillo, J., and Alcocer, S., 2013b. Shear strength of reinforced concrete walls for seismic design
of low-rise housing, ACI Structural Journal 110, 415–425.
Federal Emergency Management Agency (FEMA), 2000. Prestandard and Commentary for the
Seismic Rehabilitation of Buildings, FEMA-356, Washington, D.C.
Gatto, K., and Uang, Ch., 2003. Effects of loading protocol on the cyclic response of woodframe
shear walls, Journal of Structural Engineering 129, 1384–1393.
Hidalgo, P., Ledezma, Ch., and Jordan, P., 2002. Seismic behavior of squat reinforced concrete
shear walls, Earthquake Spectra 18, 287–308.
Maekawa, K., and El-Kashif, K., 2004. Cyclic cumulative damage of reinforced concrete in post-
peak regions, Advanced Concrete Technology 2, 257–271.
Petryna, Y., Dfanner, F., Stangenberg, F., and Kratzig, T., 2002. Reliability of reinforced concrete
structures under fatigue, Reliability Engineering and System Safety 77, 253–261.
Shkolnik, I., 1996. Evaluation of dynamic strength of concrete from results of static tests, Journal
of Engineering Mechanics 122, 1133–1138.
Travasarou, T., Bray, J., and Abrahamson, N., 2003. Empirical attenuation relationship for Arias
intensity, Earthquake Engineering and Structural Dynamics 32, 1133–1155.
Tsuno, K., and Park, R., 2004. Prediction method for seismic damage of reinforced concrete
bridge columns, Structural and Earthquake Engineering 21, 97–111.
(Received 17 January 2013; accepted 9 September 2014)

You might also like