You are on page 1of 11

Engineering Structures 136 (2017) 430–440

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Comparison of static pushover and dynamic analyses using RC building


shaking table experiment
Shuang Li a,b,⇑, Zhanxuan Zuo a,b, Changhai Zhai a,b, Lili Xie b,c
a
Key Lab of Structures Dynamic Behavior and Control of the Ministry of Education, Harbin Institute of Technology, Harbin 150090, China
b
School of Civil Engineering, Harbin Institute of Technology, Harbin 150090, China
c
Institute of Engineering Mechanics, China Earthquake Administration, Harbin 150080, China

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

Article history: The main objective of this study is to assess the accuracy and applicability of static pushover analysis
Received 14 June 2016 (POA). The POA is widely used to estimate the structural seismic response due to its simplicity.
Revised 11 January 2017 However, the accuracy of this method still needs to be evaluated due to the inherent limitations in its
Accepted 14 January 2017
theory, especially for the evaluation of structural collapse capacity problems. Meanwhile, the collapse
Available online 30 January 2017
tests are also lacking to verify the accuracy of POA when it is used to assess the structural collapse capac-
ity. In this paper, the accuracy and applicability of the POA are evaluated through comparison with the
Keywords:
dynamic time history analysis (THA) which validated by a ductile RC frame complete collapse shaking
Pushover analysis
Dynamic time history analysis
table test. The study compared the top displacement, the inter-story drift ratio and the curvature of col-
Shaking table test umn ends. It is found that the POA tends to greatly underestimate the structural responses when the
RC frame structure is severely damaged and close to the collapse state. The errors increase with the development
Seismic collapse capacity of structural damage, where the maximum error could reach more than 60%. The POA may provide incor-
rect judgement on the occurrence of collapse. Prudent use of POA is needed while assessing the structural
seismic collapse capacity.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction range and can also be used to highlight the potential weak areas
in a structure.
The responses of structural systems under strong ground shak- Owing to its simplicity and acceptable accuracy in most cases,
ing generally result in nonlinear behaviors. At present, the static the POA has become a popular tool to evaluate the safety of struc-
pushover analysis (POA) and dynamic time history analysis (THA) tures against earthquakes. However, the POA lacks a rigorous the-
are the two most popular methods that are employed to predict oretical foundation [3], which is based on the assumption that the
the structural nonlinear responses under seismic loads. Although nonlinear responses of a structure can be related to the response of
the THA is often regarded as more accurate compared with the an equivalent single degree of freedom (SDOF) system and neglects
POA, there are still some reservations about the dynamic analysis, ground motion duration and cyclic effects. Actually, the structure
which are mainly related to its computational cost and selection of progressively changes in the dynamic properties as it experiences
proper ground motion records for practical design applications loading, yielding and unloading during an earthquake. Therefore,
[1,2]. The POA is an alternative approximate analysis method for the accuracy of the POA method needs to be validated because of
estimating seismic response demands in structures as well as their these inherent limitations. Some efforts have been made to
local and global capacities. It involves applying a predefined lateral improve the accuracy of the POA [4–15]. However, these methods
load pattern which is distributed along the building height. The focused on extending the application of the POA method, e.g., con-
lateral loads are then monotonically increased until the target dis- sidering the multi-mode combination, while the ‘‘core step”, i.e.,
placement or the structure collapse is reached. The POA is a simple the increased monotonous push step with lateral loads for the fun-
option for estimating the resistant capacity in the post-elastic damental structural mode, is not changed.
The POA has been evaluated in several previous studies with
different emphases. Krawinkler and Seneviratna’s research [3]
indicated that the POA may be less accurate for structures in which
⇑ Corresponding author at: School of Civil Engineering, Harbin Institute of
the story shear force vs. story drift relationships are sensitive to the
Technology, Harbin 150090, China.
E-mail address: shuangli@hit.edu.cn (S. Li). applied load. The accuracy of POA is validated by comparing the

http://dx.doi.org/10.1016/j.engstruct.2017.01.033
0141-0296/Ó 2017 Elsevier Ltd. All rights reserved.
S. Li et al. / Engineering Structures 136 (2017) 430–440 431

results obtained from POA with those from THA [16–19]. Papaniko- the constant-ductility inelastic demand spectrum curve. Then, in
laou et al. [20,21] investigated the conventional and adaptive POA, Section 4, another five code-compatible ground motion records
the results showed that adaptive POA provides results generally are used, to give the comparison in mean results.
superior to those from conventional POA, while the consistency The diagram method [34] is used to obtain the performance
of the improvement is unreliable. Kazaz et al. [22] investigated point. As shown in Fig. 1, the bilinear capacity spectrum curve is
the accuracy of the POA procedure with a significantly stiff constructed and used to derive the ductility factor. A number of
shear-wall structure subjected to ground motions recorded on rock inelastic demand spectra with different ductility factors are drawn
sites. Their study showed that equivalent SDOF systems are not in the same figure together with the bilinear capacity spectrum.
adequate in representing the actual structural performance in the The performance point is determined by the intersection where
case of significant nonlinearity. Akkar and Metin [23] evaluated the ductility factor calculated from capacity curve is identical
the POA presented in the FEMA-440 with nondegrading three to with the ductility calculated from the demand curve. Then, the
nine-story RC frames. To assess the accuracy and applicability of structural top displacement response can be obtained through
POA, comparisons between the POA and the experimental results the inverse procedure to obtain the Sa - Sd format spectrum. The
have also been carried out. Lin et al. [24] performed the pseudo- inter-story drift ratio and curvature of column ends can be
dynamic tests, cyclic loading tests and pushover tests on three obtained by acquiring the data again for the structure while it
RC columns. Two scaled RC frame tests [25] and a series of in- deforms to the performance point.
situ RC frame tests [26,27] were conducted and compared with
the POA. 3. Shaking table test and analytical model
However, most of the studies (both the numerical simulations
and experimental tests) are confined to some limited loading The validity of the THA is evaluated by comparing the simula-
levels. The reliability of structural response prediction by POA, tion results with the experimental results of a RC frame structure
when the structure is close to or at collapse state, has seldom been shaking table test. It needs to be pointed out, although the struc-
investigated. The structural collapse tests are also lacking to verify tural displacements can be obtained from the shaking table test,
the accuracy of POA while assessing the structural collapse capac- it is difficult to obtain other responses data (e.g., inter-story shear
ity. Whether the accuracy of POA is still acceptable in structural force, curvature of column ends) from the test. Consequently, the
collapse states remains unclear. Thus, it is doubtful that the POA comparison of shear force or curvature between the POA results
could be used reliably to predict the collapse capacity of structures and shaking table test results is not feasible. Thus, the THA is a sup-
and estimate their margin of safety against a global collapse. Nev- plement to the test that can provide all the structural response
ertheless, the POA has already been used to assess the structural data. The POA and THA are then compared based on the THA vali-
collapse capacity in a series of studies [e.g., 28–33]. dated by the shaking table test results.
In light of the above discussions, the main aim of this paper is to
investigate the accuracy and applicability of the POA for a structure
3.1. Shaking table test
from the elastic state to the highly nonlinear state and finally to the
completely dynamic instability (i.e., collapse state). Based on the
As shown in Fig. 2, the test frame structure is a 1/5-scaled three-
THA validated by the results of a three-story RC frame structure
story, one-bay, three dimensional RC structure, whose prototype
shaking table test, the accuracy and applicability of the POA is dis-
structure is designed according to the Chinese building codes
cussed. The calculated structural responses by the POA and THA
[35,36]. Fig. 3 shows the member geometries and reinforcement
are compared, where the discrepancies are analyzed on the struc-
details of the test structure. The beams in all three stories share
tural top displacement, the inter-story drift ratio and the curvature
the same cross-sectional area, length, and reinforcement, and so
of column ends.
are the columns. Additional masses are added on the test structure
according to the similarity principle. The total additional masses on
the 1st, 2nd, and 3rd stories, are 1197.5 kg, 1197.5 kg, and
2. Basic of pushover analysis
726.5 kg, respectively. The test mechanical properties of concrete
include elastic modulus 22.34 GPa, maximum compressive
Several POA procedures were developed in last two decades. In
strength fc = 48.14 MPa at strain 2.82e-3 and 10% fc at strain
this study, the improved capacity spectrum method [34] is applied
0.012. For the #8 steel rebars (with a diameter of 3.98 mm), the
because it is a basic method in other more complex POA proce-
elastic modulus is 201.83 GPa and yield strength is 282.50 MPa;
dures, as well as it is still the most widely used method in practices
among those POA procedures. Three type of lateral load distribu-
tions (inverted triangular, uniform and adaptive) are adopted to
Sa
represent the distribution of inertia forces imposed on structures.
The relationship between the base shear and the top displacement Demand spectra
(i.e., capacity curve) is obtained by monotonically increasing the µ=1
lateral loads until the prescribed displacement is reached. The µ=2
capacity curve is then converted to the Acceleration Displacement µ=3 Capacity spectra
Response Spectrum (Sa - Sd) format through dividing the top dis- µ=4
placement by the modal participation factor multiply the natural
vibration mode value at the top of the structure and the base shear
dividing the effective modal mass. The capacity spectrum curve is
then mimicked by a bilinear form representation. The bilinear rep- Performance point
resentation of the capacity curve is repeatedly constructed and
updated during the iterative process for the trace of performance µ=1 µ=2 µ=3 µ=4
point. In order to make sure that the results from POA are compa-
rable with the results from shaking table test and dynamic analy- Sd
O
sis, the ground motion record (IMPVALL/I-ELC180), used in the
shaking table test and THA (see Section 3), is adopted to derive Fig. 1. The iterative sketch diagram of performance point.
432 S. Li et al. / Engineering Structures 136 (2017) 430–440

1300 1200

30

30
700

700
30

30
2100

2100
700

700
30

30
80 80

80

700

700
60

A B B C

300

300
Fig. 2. The elevation layouts of the test structure (Unit: mm).

1300
5 5
#14@30 #14@30
double layers
14@35
8(#8)

1200
80

80

5(#8)
concrete cover: 5

60 60

(a) Reinforcement of beam and column (b) Reinforcement of slab


Fig. 3. Geometry and reinforcement detail of the test structure (Unit: mm).

for the #14 steel rebars (with a diameter of 2.01 mm), the elastic technique [37–39] are used to record the structural displacement
modulus is 196.57 GPa and yield strength is 389.93 MPa. responses. The validation of accuracy of stereo vision technique
The input ground motion is an acceleration record (IMPVALL/I- and more information on this shaking table test were presented
ELC180) obtained during the El Centro earthquake in 1940. The in the literature [40].
input motion is unidirectional along the AB direction. The time
pffiffiffi
scale is compressed by the factor considering the 1= 5 scale of 3.2. Analytical model and verification
the test structure. In this way, the duration of the input motion
is 17.89s. The test structure is subjected to a sequence of input The structure is modeled using the nonlinear finite element
ground excitations. The intensity (Peak Ground Acceleration, analysis platform OpenSees [41]. In this finite element model, the
PGA) of ground motion is gradually increased. Fig. 4 shows the RC beams and columns are modeled by the displacement-based
ground motion acceleration time histories from Test 1 to Test 8. beam-column elements with four elements in one member and
It has to be pointed out that the frame structure is severely dam- fiber sections. Each fiber is assigned with a specific uniaxial
aged after Test 7 (PGA = 11 m/s2). Subsequently, the input ground stress-strain relationship representing either the unconfined con-
motion intensity is no longer increased, the same seismic excita- crete, confined concrete or steel rebars. The Kent-Scott-Park consti-
tion (PGA = 11 m/s2) is used again in Test 8. The white noise tutive model [42] is adopted to mimic the material behaviors of the
(PGA = 0.5 m/s2) is used as the input to measure the resonant fre- confined and unconfined concrete. The steel constitutive model
quencies of the frame structure before each test. The linear variable developed by Menegotto and Pinto [43], modified later by Filippou
differential transformers (LVDTs) and binocular stereo vision [44] to include the isotropic strain hardening effects, is used to

10
Acceleration(m/s )
2

5
0
-5
-10 Test 1 Test 2 Test 3 Test 4 Test 5 Test 6 Test 7 Test 8
2 2 2
PGA=0.8m/s PGA=1.3m/s2 PGA=2.2m/s PGA=3.7m/s2 PGA=6.0m/s PGA=9.0m/s2PGA=11m/s2 PGA=11m/s2
-15
-20
Load test

Fig. 4. Ground motion acceleration time histories from Test 1 to Test 8.


S. Li et al. / Engineering Structures 136 (2017) 430–440 433

describe the material behavior of steel rebar. The bond-slip behav- 4. Comparison between POA and THA
iors are not considered in the finite element model. The detailed
parameters of the materials (concrete and steel bar) can be found 4.1. Comparison of the structural responses
in literature [40]. Co-rotational method [41] is used to consider
the geometrical nonlinearity. The Rayleigh damping with 5% In this section, the structural responses (top displacement,
damping ratio is used in the analysis. inter-story drift ratio and curvature ductility of column ends)
Table 1 shows the comparison of the first and second frequen- obtained from POA are compared with those obtained from THA.
cies between the shaking table test and THA simulation. When In order to follow the shaking table test procedure, the accumu-
the PGA of the earthquake inputs increased gradually, the damage lated structural damage under multiple loading tests (from Test 1
developed progressively, and the natural frequencies decrease. The to Test 8) in the numerical simulation needs to be considered.
natural frequency of the first mode before Test 8 (0.67 Hz/0.74 Hz) The POA and THA are carried out by the following procedure: For
is reduced to about 1/5 of the initial frequency of the undamaged Test N = 1, since there is no previous loading test or structural dam-
structure (3.42 Hz/3.38 Hz). Besides the test described in this age, the POA and THA are carried out directly under the seismic
study, some other tests also show remarkable natural frequency intensity of Test 1. For Test N = 2, the ground motion records of
decreases of the structures [45–47]. Test 1 and Test 2 are applied successively to the finite element
Fig. 5 shows the inter-story drift ratio time histories of the 1st model in THA; while for the POA, the ground motion record of Test
and 3rd stories in Test 1, 3, 5 and 7. The damage developed progres- 1 is applied to finite element model firstly by THA to consider the
sively from Test 1 to Test 7. In Test 1 and 3, the frame was slightly damage caused by Test 1, then the POA is carried out under the
damaged and the peak inter-story drift ratios were all below 2%. seismic intensity of Test 2 using the damaged finite element model
The frame was severely damaged but yet not collapsed after Test after Test 1. For the other N (N = 3–8), the loading procedure of THA
7. The peak inter-story drift ratios in Test 7 of the 1st, 2nd and 3rd and POA are done in the similar manner. The comparison of struc-
stories were 12.51%, 5.20%, and 2.49%, respectively. It can be seen tural responses in different loading tests are shown in Figs. 9–13,
that the results of THA well agree with the results of the shaking respectively.
table test. As mentioned previously, the intensity of the input Fig. 9 shows the comparison of structural responses in Test 1
ground motions in Test 7 and Test 8 were same. However, the (PGA = 0.8 m/s2). Compared with THA, the POA presents a good
frame structure survived in Test 7 but collapsed in Test 8, indicat- estimation on the top displacement. For inter-story drift ratio
ing that merely the earthquake intensity may not be adequate to and curvature ductility, the POA also predicts good estimations
effectually represent the seismic damage potential, and other fac- at the 1st and 2nd stories but underestimates the responses at the
tors such as the loading history (i.e., initial damage degree) and 3rd story when using the inverted triangular and uniform load
loading duration may also influence the structural collapse capac- distributions.
ity. Fig. 6 shows the comparison of displacements (relative dis- As shown in Figs. 10 and 11 for Test 3 and Test 5, the POA pre-
placement to the shaking table) when the structure finally sents good estimations on the top displacement but less accurate
collapsed in Test 8. Fig. 7 shows the collapse process of the frame estimations on the inter-story drift ratio and curvature ductility.
structure in Test 8. It could be noticed that the test frame experi- For inter-story drift ratio and curvature ductility, the uniform load
enced large nonlinear deformation and underwent further damage, distribution overestimates the response at the 1st story but under-
which leaded to final loss of its stability. The collapse of the frame estimates the responses at the 2nd and 3rd stories. The inverted tri-
was a progressive process, starting from its 1st story and then prop- angular and adaptive load distributions also underestimates the
agating to the 3rd story. drift ratio and curvature ductility at the 2nd and 3rd stories.
The maximum curvature values of column ends of different sto- Fig. 12 shows the comparison of structural responses in Test 7
ries are calculated from the THA. Fig. 8 shows the development of (PGA = 11 m/s2). In this case, the structure is severely damaged
plastic hinges in Test 1, 3, 5 and 7. lmax is the maximum curvature and residual displacement appeared, which can be seen from
ductility (maximum curvature response divides the yield curva- Fig. 5. All the responses obtained from POA are greatly underesti-
ture) of the column ends of different stories. The hollow circle indi- mated when compared with those from THA.
cates that the columns have not yielded (lmax < 1). The solid circle Fig. 13 shows the structural response comparisons in Test 8. The
indicates that the columns have already yielded (lmax > 1). As top displacements obtained from POA are 94.0 mm (Inverted trian-
shown in Fig. 8, all the lmax are smaller than 1 in Test 1 and 3, gular), 94.5 mm (Uniform) and 94.1 mm (Adaptive). While in the
which means the curvatures of the column ends don’t reach the THA and shaking table test the structure has lost the seismic resis-
yield curvature. The lmax of the 1st and 2nd stories became larger tant capacity and collapse on to the ground, which has been shown
than 1 in Test 4. All the columns has yielded in Test 7, the lmax in Figs. 6 and 7. It turns out that the POA may be unconservative
of the 1st, 2nd and 3rd stories are 18.68, 1.92 and 1.80, respectively. and fail to predict the collapse of the structure compared with
The development of the plastic hinges shows that the damage is the THA or shaking table test. It is noticeable that the seismic
mainly lumped in the 1st story. Thus, the 1st story becomes a weak intensities in Test 7 (11 m/s2) and Test 8 (11 m/s2) are the same
story due to the local damage at the column ends and finally col- for the shaking table test. However, the structure survived in Test
lapse first. 7 but collapsed in Test 8. The different results of structural
response obtained from the POA and THA under the same seismic

Table 1
Comparison of natural frequencies between test and simulation.

Before Test First frequency Second frequency First frequency Second frequency Error of simulation (first Error of simulation (third
No. (Test) (Test) (simulation) (simulation) frequency) frequency)
Test 1 3.42 11.23 3.38 10.87 1.17% 3.21%
Test 3 3.17 10.00 3.16 10.01 3.15% 0.10%
Test 5 2.44 8.61 2.38 7.54 2.46% 12.4%
Test 7 1.40 6.19 1.43 4.39 2.14% 29.1%
Test 8 0.67 2.53 0.74 2.21 10.45% 12.6%
434 S. Li et al. / Engineering Structures 136 (2017) 430–440

0.004
0.004 Test 1 (0.8m/s2)
Test 1 (0.8m/s2) 0.002 Simulation (0.8m/s2)
Drift ratio
0.002 Simulation (0.8m/s2)
0.000 0.000

-0.002 -0.002
-0.004 -0.004
0 5 10 15 20 25 0 5 10 15 20 25
0.02 0.015 2
Test 3 (2.2m/s )
Test 3 (2.2m/s2) 0.010
Simulation (2.2m/s2)
Drift ratio

0.01
Simulation (2.2m/s2) 0.005
0.00 0.000
-0.005
-0.01 -0.010
-0.02 -0.015
0 5 10 15 20 25 0 5 10 15 20 25
0.02
0.04 Test 5 (6.0m/s2)
Test 5 (6.0m/s2) 0.01
Drift ratio

0.02 Simulation (6.0m/s2)


Simulation (6.0m/s2)
0.00 0.00
-0.02 -0.01
-0.04
-0.02
0 5 10 15 20 25 0 5 10 15 20 25

0.15 Test 7(11m/s )2


0.04
0.10 Simulation (11m/s2)
Drift ratio

0.05 0.02
0.00
Test 7(11m/s2) 0.00
-0.05
-0.10 Simulation (11m/s2) -0.02
-0.15
0 5 10 15 20 25 0 5 10 15 20 25
Time(s) Time(s)
(a) The 1st story (b) The 3rd story
Fig. 5. The drift ratio time histories at the 1st and the 3rd stories.

0.4 4.2. Error analysis of POA


Displacement(m)

0.3 Test
Simulation The error of POA is discussed in this section. The error index
0.2 used in the study is presented as
0.1 Result POA  Result THA
1st story Error ¼ ð1Þ
0.0 ResultTHA

0 1 2 3 4 where Result POA is the structural response obtained by POA, and


Time(s) ResultTHA is the structural response obtained by THA.

0.4 4.2.1. Single ground motion used in shaking table test


Displacement(m)

Test Fig. 14 shows the error of top displacement in different loading


0.3
Simulation tests. As can be seen from the figure, in Test 1, 3 and 5 for all the
0.2 three load distributions, the errors are under 10%, while in Test 7
0.1 the POA greatly underestimates the top displacement, the average
error of three load distributions is 25.1%.
0.0 3rd story
Figs. 15 and 16 show the errors of inter-story drift ratio and cur-
0 1 2 3 4 vature ductility of column ends in different tests. As shown in
Time(s) Fig. 15a, the uniform load distribution overestimates the
inter-story drift ratio by 8.26%, 10.2% and 25.7% in Test 1 3 and
Fig. 6. The relative displacements of the 1st and the 3rd stories in Test 8. 5, respectively. All the three load distributions underestimate the
inter-story drift ratio in Test 7, where the errors reach 28.1%,
14.9% and 32.7% respectively and the average error is 25.2%.
Fig. 15b shows the error of inter-story drift ratio for the 3rd story.
intensity indicates that the effect of structural deterioration in All the three load distributions underestimate the inter-story drift
strength and stiffness may be more sensitive to the cyclic dynamic ratio in all tests, with the exception of adaptive load distribution
loading. The monotonic static POA may underestimate the struc- slightly overestimates the response in Test 1. The errors increase
tural responses and overestimate its collapse capacity. as the structural damage develops, and the errors reach 49.6%,
S. Li et al. / Engineering Structures 136 (2017) 430–440 435

t = 0s t = 1.73s t = 2.80s

t = 3.04s t = 3.26s t = 3.72s

Fig. 7. Collapse process of the frame structure in Test 8.

max=0.06 max=0.43 max=0.79 max=1.80

max=0.11 max=0.53 max=1.20 max=1.92

max=0.22 max=0.97 max=4.11 max=18.68

Fig. 8. The development of plastic hinges in Test 1, 3, 5 and 7.

0.4 7 3500
Inverted triangular Inverted triangular
6 3000
Uniform Uniform
Story shear force

0.3
Top disp (mm)

5 Adaptive 2500 Adaptive


4 Dynamic 2000 Dynamic
Sa (g)

0.2 Inverted triangular Test


µ=1 3 1500
Uniform
0.1 Adaptive 2 1000
Demand spectrum
1 500
0.0 0 0
0.00 0.02 0.04 0.06 0.08 0.10 0.000 0.001 0.002 0.003
Sd (m) Drift ratio
(a) Capacity and demand spectra (b) The top displacement (c) Drift ratio of the 1st story

2500 1600 0.25


Inverted triangular Inverted triangular Inverted triangular
1400
2000 Uniform Uniform 0.20 Uniform
Story shear force

Story shear force

1200 Adaptive
Adaptive Adaptive
1500 Dynamic 1000 Dynamic 0.15 Dynamic
800
µ

1000 600 0.10


400
500 0.05
200
0 0 0.00
0.000 0.001 0.002 0.003 0.004 0.005 0.000 0.001 0.002 0.003 1 2 3 4
Drift ratio Drift ratio Story

(d) Drift ratio of the 2nd story (e) Drift ratio of the 3rd story (f) Curvature ductility
Fig. 9. Comparison of the structural responses (Test 1: PGA = 0.8 m/s2).
436 S. Li et al. / Engineering Structures 136 (2017) 430–440

1.0 30 6000
Inverted triangular
25 5000 Uniform
0.8 Inverted triangular

Story shear force


Adaptive

Top disp (mm)


Uniform 20 4000 Dynamic
Inverted triangular
0.6 Adaptive
Sa (g)

Uniform
Demand spectrum 15 3000
Adaptive
0.4 µ = 1.1 10 Dynamic 2000
Test
0.2 5 1000

0.0 0 0
0.00 0.02 0.04 0.06 0.08 0.10 0.000 0.004 0.008 0.012 0.016
Sd (m) Drift ratio

(a) Capacity and demand spectra (b) The top displacement (c) Drift ratio of the 1st story

4000 3000
Inverted triangular Inverted triangular 1.2
2500 Inverted triangular
Uniform Uniform 1.0
Story shear force

Story shear force

3000 Uniform
Adaptive 2000 Adaptive
0.8 Adaptive
Dynamic Dynamic
Dynamic
2000 1500

µ
0.6
1000 0.4
1000
500 0.2
0 0 0.0
0.000 0.004 0.008 0.012 0.000 0.004 0.008 1 2 3 4
Drift ratio Drift ratio Story

(d) Drift ratio of the 2nd story (e) Drift ratio of the 3rd story (f) Curvature ductility
2
Fig. 10. Comparison of the structural responses (Test 3: PGA = 2.2 m/s ).

1.0 60 5000

0.8 50 4000
Inverted triangular
Story shear force
Top disp (mm)

Uniform 40 Inverted triangular


0.6 Adaptive 3000
Sa (g)

Uniform
Demand spectrum 30
Adaptive Inverted triangular
0.4 2000
µ = 2.6 20 Dynamic Uniform
Test Adaptive
0.2 10 1000
Dynamic
0.0 0 0
0.00 0.02 0.04 0.06 0.08 0.10 0.00 0.01 0.02 0.03 0.04 0.05
Sd (m) Drift ratio

(a) Capacity and demand spectra (b) The top displacement (c) Drift ratio of the 1st story

5000 3000 6
Inverted triangular Inverted triangular
4000 Uniform 2500 Uniform 5
Story shear force

Story shear force

Adaptive Adaptive Inverted triangular


2000 4
Uniform
3000 Dynamic Dynamic
1500 3 Adaptive
µ

2000 Dynamic
1000 2
1000 500 1

0 0 0
0.00 0.01 0.02 0.03 0.000 0.004 0.008 0.012 0.016 1 2 3 4

Drift ratio Drift ratio Story

(d) Drift ratio of the 2nd story (e) Drift ratio of the 3rd story (f) Curvature ductility
Fig. 11. Comparison of the structural responses (Test 5: PGA = 6.0 m/s2).

66.0% and 41.2% in Test 7. Fig. 16 shows the error of curvature 4.2.2. Multiple ground motions selected with design spectra
ductility. The error diagrams of curvature ductility is similar to In order to validate the observations derived from one single
those of inter-story drift ratio, but the error is larger. ground motion record (IMPVALL/I-ELC180), another five ground
S. Li et al. / Engineering Structures 136 (2017) 430–440 437

1.6 160 5000


1.4 140
Inverted triangular 4000

Story shear force


1.2 120

Top disp (mm)


Uniform
1.0 100 Inverted triangular
Adaptive 3000
Sa (g)

Uniform
0.8 Demand spectrum 80
Adaptive
0.6 µ = 4.1 60 Dynamic
2000 Inverted triangular
0.4 40 Test Uniform
1000 Adaptive
0.2 20
Dynamic
0.0 0 0
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.00 0.04 0.08 0.12
Sd (m)
Drift ratio

(a) Capacity and demand spectra (b) The top displacement (c) Drift ratio of the 1st story

4000 3000 20
Inverted triangular
2500 Inverted triangular
Uniform
Uniform
Story shear force

Story shear force

3000 Adaptive 15
2000 Adaptive
Dynamic
Dynamic
2000 1500 10

µ
Inverted triangular
Uniform 1000
1000 Adaptive 5
Dynamic 500

0 0 0
0.00 0.01 0.02 0.03 0.04 0.05 0.00 0.01 0.02 0.03 1 2 3 4

Drift ratio Drift ratio Story


nd rd
(d) Drift ratio of the 2 story (e) Drift ratio of the 3 story (f) Curvature ductility
Fig. 12. Comparison of the structural responses (Test 7: PGA = 11 m/s2).

1.2 600

1.0 Inverted triangular 500


Top disp (mm)

Uniform infinite
0.8 400 dynamic instable
Adaptive
Sa (g)

Demand spectrum
0.6 300 Inverted triangular
µ = 4.5
Uniform
0.4 200
Adaptive
0.2 100 Dynamic
Test
0.0 0
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
Sd (m)
(a) Capacity and demand spectra (b) The top displacement
Fig. 13. Comparison of the structural responses (Test 8: PGA = 11 m/s2).

the elastic design response spectrum in the Chinese building codes


0.2
[36]. As shown in Fig. 17, the response spectrum of the selected
0.1 records have effective comparisons and calibrations with the
Top disp error

design spectrum. All the PGA of the five ground motions are scaled
0.0
to four different values (PGA = 0.8 m/s2, 2.2 m/s2, 6.0 m/s2 and
-0.1 11 m/s2). The accuracy of POA is evaluated by calculating the errors
Inverted triangular via Eq. (1).
-0.2
Uniform Figs. 18–20 show the mean errors of structural responses for the
-0.3 Adaptive five ground motion records. Fig. 18 shows the mean error of top
displacement and the error increases with the increase of ground
-0.4
1 3 5 7 motion intensity. The mean errors of the three load distributions
are all smaller than 10%, which are 1.7%, 4.2% and 8.6% in cases
Test
PGA = 0.8 m/s2, 2.2 m/s2 and 6.0 m/s2, respectively, while the
Fig. 14. The error of top displacement in different loading tests. errors reach 21.7% in case PGA = 11 m/s2. Figs. 19 and 20 show
the errors of inter-story drift ratio and curvature ductility of
column ends corresponding to different intensities. The error
motion records (see Table 2) are selected from the PEER ground diagrams of inter-story drift ratio and curvature ductility share
motion database. The ground motion records are selected to fit similar trends. For the 1st story, the uniform load distribution
438 S. Li et al. / Engineering Structures 136 (2017) 430–440

0.6
Inverted triangular 0.0
0.4 Uniform -0.2

Drift ratio error


Drift ratio error
Adaptive
0.2 -0.4
-0.6
0.0
-0.8
-0.2 -1.0 Inverted triangular
-0.4 -1.2 Uniform
-1.4 Adaptive
-0.6
1 3 5 7 1 3 5 7
Test Test
(a) The 1st story (b) The 3rd story
Fig. 15. The error of drift ratio in different loading tests.

0.6
Inverted triangular 0.0
0.4 Uniform
-0.2

Curvature error
Curvature error

Adaptive
0.2 -0.4
-0.6
0.0
-0.8
-0.2 -1.0 Inverted triangular
-0.4 -1.2 Uniform
-1.4 Adaptive
-0.6
1 3 5 7 1 3 5 7
Test Test
(a) The 1st story (b) The 3rd story
Fig. 16. The error of curvature ductility in different loading tests.

Table 2
Information of the ground motions.

ID No. Ground motion Earthquake name Year Recording station Predominant period (s) Duration (s)
1 BORREGO_A-SON303 Borrego Mtn 1968 San Onofre So Cal Edison 0.30 45.17
2 SFERN_PEL090 San Fernando 1971 LA-Hollywood Stor FF 0.26 79.38
3 PARKF_C12050 Parkfield 1966 Cholame Shandon Arry #12 0.18 44.23
4 FRIULI.B_B-COD000 Friuli, Italy-02 1976 Codroipo 0.28 33.53
5 SFERN_PVE155 San Fernando 1971 2516 Via Tejon PV 0.36 70.16

0.8 0.3
Target spectrum Inverted triangular
Mean Top Disp Error

BORREGO_A-SON303 0.2 Uniform


0.6 SFERN_PEL090 Adaptive
PARKF_C12050 0.1
FRIULI.B_B-COD000
Sa (g)

0.0
0.4 SFERN_PVE155
-0.1

0.2 -0.2

-0.3
80 220 600 1100
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Ground motion intensity(gal)
T (s)
Fig. 18. The mean error of top displacement.

Fig. 17. The ground motion acceleration spectra and target spectra.

errors increase with the increase of ground motion intensity. The


overestimates the inter-story drift ratio and curvature ductility, uniform load distribution gives the worst estimation on the results.
while all the three load distributions underestimate the curvature The average error of inter-story drift ratio is 42.1% and the aver-
ductility at PGA = 11 m/s2. The average error of inter-story drift age error of curvature ductility is 65.1% at PGA = 11 m/s2.
ratio is 20.4% and the average error of curvature ductility is As a general trend, the POA can provide acceptable estimation
30.0%. For the 3rd story, all the load distributions of POA underes- on the structural responses when the structure is not severely
timate the inter-story drift ratio and curvature ductility, and the damaged or the degradation of structural strength and stiffness is
S. Li et al. / Engineering Structures 136 (2017) 430–440 439

0.6
Inverted triangular 0.0

Mean drift ratio error


Mean drift ratio error 0.4 Uniform -0.2
Adaptive
0.2 -0.4
-0.6
0.0
-0.8
Inverted triangular
-0.2 -1.0 Uniform
-0.4 -1.2 Adaptive
-1.4
-0.6
80 220 600 1100 80 220 600 1100
Ground motion intensity(gal) Ground motion intensity(gal)
(a) The 1st story (b) The 3rd story
Fig. 19. The mean error of story drift ratios.

0.6
Inverted triangular 0.0

Mean curvature error


Mean curvature error

0.4 Uniform
-0.2
Adaptive
0.2 -0.4
-0.6
0.0
-0.8
-0.2 -1.0 Inverted triangular
-0.4 -1.2 Uniform
-1.4 Adaptive
-0.6
80 220 600 1100 80 220 600 1100
Ground motion intensity(gal) Ground motion intensity(gal)

(a) The 1st story (b) The 3rd story


Fig. 20. The mean error of curvatures ductility.

slight and medium. However, the POA tends to underestimate the made in terms of the top displacement, the inter-story drift ratio
structural responses when the structure is severely damaged, and the curvature of member and the discrepancies are analyzed,
regardless of the lateral load distributions considered. The errors the following conclusions are drawn:
increase with the increase of ground motion intensity. The maxi-
mum error reaches more than 60% for the shaking table tested (1) The POA can provide reasonably accurate estimation of the
structure. It seems that the errors due to the effects of cyclic load- structural responses when the structure is not severely dam-
ing and duration under real earthquake excitations may be ampli- aged. For the structural responses (top displacement, inter-
fied when the structure approaches the collapse state. As a story drift ratio and curvature of member), the error of the
consequence, the POA overestimates the structural capacity and top displacement is the lowest while the error of the curva-
underestimates the structural responses in earthquakes, especially ture is the largest. The error of POA increases as the struc-
when the deterioration of structural strength and stiffness is tural damage develops. While the structure is severely
significant. damaged, the POA greatly underestimated the structural
responses, regardless of the lateral load distributions
5. Summary and conclusions (inverted triangular, uniform and adaptive) considered. The
maximum error of structural responses reach more than
60%.
A three-story RC frame structure was tested on the shaking
(2) The structure collapsed in Test 8, the different responses
table. The damage of the structure developed progressively with
from POA and THA of the structure under the same seismic
the increase of seismic intensity, from an elastic state to a highly
intensity indicates that the dynamic loading history and
nonlinear state, then experienced dynamic instability, and finally
duration may play roles in assessing the structural collapse
collapsed to the ground. The responses of THA well agree with
capacity. The monotonic POA capacity curve, which doesn’t
the results of the shaking table test. The improved capacity spec-
consider such factors, fails to predict the occurrence of struc-
trum method, one of the most widely used POA method in practice
tural collapse. Thus, when the structure is close to collapse
and the basis for most other complex POA procedures, is adopted
state, i.e., the degradation of stiffness and strength is serious,
to predict the structural responses. It is well known that the POA
the POA may not be reliable to access the collapse capacity
is adequate in predicting the structural responses for such a
of the structure. The POA, when is used as a tool of seismic
three-story ductile RC frame (non-shear failure dominated). How-
collapse-resistance design and evaluation, needs further
ever, this conclusion may be true when the structure is not
developments in the cases when severe deterioration of
severely damaged, it is unclear that the accuracy of POA is still
structure may occur.
acceptable when using this method to estimate the structural col-
(3) This study carried out a preliminary research about the val-
lapse. Thus, based on the THA validated by the shaking table test
idation of POA when the structures are near collapse state.
results, the accuracy and applicability of POA is discussed. The
The conclusions are obtained from the comparison of the
POA and THA are compared from the elastic state to nonlinear state
POA with THA validated by a RC frame collapse shaking table
and collapse state. The comparison between POA and THA has been
440 S. Li et al. / Engineering Structures 136 (2017) 430–440

test. More tests and researches of different type of structures [24] Lin YY, Chang KC, Wang YL. Comparison of displacement coefficient method
and capacity spectrum method with experimental results of RC columns.
will be needed to investigate the validity of POA in the
Earthquake Eng Struct Dynam 2004;33(1):35–48.
future. [25] Cardone D. Nonlinear static methods vs. experimental shaking table test
results. J Earthquake Eng 2007;11(6):847–75.
[26] Sharma A, Reddy GR, Vaze KK, Eligehausen R. Pushover experiment and
analysis of a full scale non-seismically detailed RC structure. Eng Struct
Acknowledgements 2013;46:218–33.
[27] Weng YT, Lin BZ, Hwang SJ. Experimental and analytical performance
This research project was supported by the National Natural assessment of in-situ pushover tests of school buildings in Taiwan. In:
Proceedings of the 4th international conference of earthquake engineering
Science Foundation of China (Grant Nos. 51578202, 51278150). (4ICEE), Taipei, Taiwan; 2006.
The financial supports are greatly appreciated by the authors. [28] Zacharenaki AE, Fragiadakis M, Papadrakakis M. Reliability-based optimum
seismic design of structures using simplified performance estimation
methods. Eng Struct 2013;52:707–17.
References [29] Fragiadakis M, Vamvatsikos D. Fast performance uncertainty estimation via
pushover and approximate IDA. Earthquake Eng Struct Dynam 2010;39
[1] Elnashai AS, Di Sarno L. Fundamentals of earthquake engineering. UK: Wiley (6):683–703.
and Sons; 2008. [30] Borzi B, Pinho R, Crowley H. Simplified pushover-based vulnerability analysis
[2] Villaverde R. Methods to assess the seismic collapse capacity of building for large-scale assessment of RC buildings. Eng Struct 2008;30(3):804–20.
structures: State of the art. J Struct Eng-ASCE 2007;133(1):57–66. [31] Huang W, Gould PL. 3-D pushover analysis of a collapsed reinforced concrete
[3] Krawinkler H, Seneviratna GDPK. Pros and cons of a pushover analysis of chimney. Finite Elem Anal Des 2007;43(11–12):879–87.
seismic performance evaluation. Eng Struct 1998;20(4–6):452–64. [32] Shafei B, Zareian F, Lignos DG. A simplified method for collapse capacity
[4] Chopra AK, Goel RK. A modal pushover analysis procedure for estimating assessment of moment-resisting frame and shear wall structural systems. Eng
seismic demands for buildings. Earthquake Eng Struct Dynam 2002;31 Struct 2011;33(4):1107–16.
(3):561–82. [33] Fiore A, Spagnoletti G, Greco R. On the prediction of shear brittle collapse
[5] Chintanapakdee C, Chopra AK. Evaluation of modal pushover analysis using mechanisms due to the infill-frame interaction in RC buildings under pushover
generic frames. Earthquake Eng Struct Dynam 2003;32(3):417–42. analysis. Eng Struct 2016;121:147–59.
[6] Antoniou S, Pinho R. Advantages and limitations of adaptive and non-adaptive [34] Chopra AK, Goel RK. Capacity-demand-diagram methods based on inelastic
force-based pushover procedures. J Earthquake Eng 2004;8(4):497–522. design spectrum. Earthquake Spect 1999;15(4):637–56.
[7] Goel RK, Chopra AK. Evaluation of modal and FEMA pushover analyses: SAC [35] Ministry of Housing and Urban-Rural Development of P. R. China. Code for
buildings. Earthquake Spect 2004;20(1):225–54. design of concrete buildings, GB 50010–2010. Beijing, China: China
[8] Antoniou S, Pinho R. Development and verification of a displacement-based Architecture and Building Press; 2010.
adaptive pushover procedure. J Earthquake Eng 2004;8(5):643–61. [36] Ministry of Housing and Urban-Rural Development of P. R. China. Code for
[9] Kalkan E, Kunnath SK. Adaptive modal combination procedure for nonlinear seismic design of buildings, GB 50010–2010. Beijing, China: China
static analysis of building structures. J Struct Eng 2006;132(11):1721–32. Architecture and Building Press, 2010.
[10] Poursha M, Khoshnoudian F, Moghadam AS. A consecutive modal pushover [37] Chang CC, Xiao XH. An integrated visual-inertial technique for structural
procedure for estimating the seismic demands of tall buildings. Eng Struct displacement and velocity measurement. Smart Struct Syst 2010;6
2009;31(2):591–9. (9):1025–39.
[11] Shakeri K, Shayanfar MA, Kabeyasawa T. A story shear-based adaptive [38] Busca G, Cigada A, Mazzoleni P, Zappa E. Vibration monitoring of multiple
pushover procedure for estimating seismic demands of buildings. Eng Struct Bridge points by means of a unique vision-based measuring system. Exp Mech
2010;32(1):174–83. 2014;54(2):255–71.
[12] Reyes JC, Chopra AK. Three-dimensional modal pushover analysis of buildings [39] Wang H. Stereovision based digital image correlation method for monitoring
subjected to two components of ground motion, including its evaluation for three dimensional deformation. Master Dissertation, Harbin Institute of
tall buildings. Earthquake Eng Struct Dynam 2011;40(7):789–806. Technology; 2013.
[13] Abbasnia R, Davoudi AT, Maddah MM. An adaptive pushover procedure based [40] Li S, Zuo ZX, Zhai CH, Xu SY, Xie LL. Shaking table test on the collapse process of
on effective modal mass combination rule. Eng Struct 2013;52:654–66. a three-story reinforced concrete frame structure. Eng Struct
[14] Panyakapo P. Cyclic Pushover Analysis procedure to estimate seismic demands 2016;118:156–66.
for buildings. Eng Struct 2014;66:10–23. [41] Mazzoni S, McKenna F, Scott MH, Fenves GL. Open system for earthquake
[15] Poursha M, Samarin ET. The modified and extended upper-bound (UB) engineering simulation: User command-language manual. Pacific Earthquake
pushover method for the multi-mode pushover analysis of unsymmetric- Engineering Research Center. Berkeley: University of California; 2009.
plan tall buildings. Soil Dyn Earthquake Eng 2015;71:114–27. [42] Spacone E. Flexibility-based finite element models for the nonlinear static and
[16] Mwafy AM, Elnashai AS. Static pushover versus dynamic collapse analysis of dynamic analysis of concrete frame structures. PhD Dissertation, University of
RC buildings. Eng Struct 2001;23(5):407–24. California, Berkeley; 1994.
[17] Kalkan E, Kunnath SK. Assessment of current nonlinear static procedures for [43] Menegotto M, Pinto PE. Method of analysis for cyclically loaded reinforced
seismic evaluation of buildings. Eng Struct 2007;29(3):305–16. concrete plan frames including changes in geometry and non-elastic behavior
[18] Ferracuti B, Pinho R, Savoia M, Francia R. Verification of displacement-based of elements under combined normal force and bending. Proceedings of IABSE
adaptive pushover through multi-ground motion incremental dynamic symposium on resistance and ultimate deformability of structures acted on by
analyses. Eng Struct 2009;31(8):1789–99. well-defined repeated loads 1973:15–22.
[19] Valente M, Milani G. Non-linear dynamic and static analysis on eight historical [44] Filippou FC, Popov EP, Bertero VV. Effects of bond deterioration on hysteretic
masonry towers in the North-East of Italy. Eng Struct 2016;114:241–70. behavior of reinforced concrete joints. Report EERC83-19, Earthquake
[20] Papanikolaou VK, Elnashai AS. Evaluation of conventional and adaptive Engineering Research Center, University of California, Berkeley; 1983.
pushover analysis I: methodology. J Earthquake Eng 2005;9(6):923–41. [45] Xiao JZ, Wang CQ, Li J, Tawana MM. Shake-table model tests on recycled
[21] Papanikolaou VK, Elnashai AS, Pareja JF. Evaluation of conventional and aggregate concrete frame structure. ACI Struct J 2012;109(6):777–86.
adaptive pushover analysis II: comparative results. J Earthquake Eng 2006;10 [46] Ghannoum WM, Moehle JP. Shake-table tests of a concrete frame sustaining
(1):127–51. column axial failures. ACI Struct J 2012;109(3):393–402.
[22] Kazaz I, Yakut A, Gulkan P. Seismic response assessment of a stiff structure. [47] Kim Y, Kabeyasawa T, Igarashi S. Dynamic collapse test on eccentric reinforced
Earthquake Eng Struct Dynam 2006;35(6):737–59. concrete structures with and without seismic retrofit. Eng Struct
[23] Akkar S, Metin A. Assessment of improved nonlinear static procedures in 2012;34:95–110.
FEMA-440. J Struct Eng 2007;133(9):1237–46.

You might also like