You are on page 1of 12

NONLINEAR SEISMIC RESPONSE OF A

SEVEN-STORY STEEL REINFORCED CONCRETE CONDOMINIUM


RETROFITTED WITH LOW-YIELD-STRENGTH-STEEL
DAMPER COLUMNS

Kenji FUJII1, Kazuaki MIYAGAWA2

ABSTRACT

A simplified procedure is proposed for predicting the peak response of a building retrofitted with a low-yield-
strength steel damper panel considering the strain hardening behavior. To approximate the strain hardening
behavior, the use of two models, namely the lower-bound (LB) model and upper-bound (UB) model, is proposed.
The predicted peak response of each LB and UB model is predicted through nonlinear static (pushover) analysis
and the equivalent linearization technique using the equivalent single-degree-of-freedom model. The predicted
peak response of each component is determined from the envelope of the prediction results of LB and UB models.
In numerical examples, the peak seismic response of a seven-story steel reinforced concrete (SRC) building
retrofitted with steel damper columns is predicted employing the proposed simplified procedure and compared
with the results of time-history analysis. Results show the simplified procedure can satisfactorily estimate the peak
response of the retrofitted building model with damper columns considering the strain hardening effect.

Keywords: seismic retrofit; existing steel reinforced concrete condominium; steel damper column; equivalent
linearization technique; pushover analysis

1. INTRODUCTION

The seismic retrofitting of existing buildings is important in most seismically active regions. In Japan,
the seismic evaluation and retrofitting of existing reinforced concrete (RC) and steel reinforced concrete
(SRC) buildings, especially public buildings including school buildings and government offices, has
accelerated since the 1995 Hyogo-Ken Nanbu Earthquake. Although the seismic evaluation and
retrofitting of school buildings have almost been completed, there are still many condominiums that
have been designed according to the old seismic design code.
In recent years, there have been many studies on passive control systems for the seismic retrofitting of
existing RC buildings (e.g., Ishii et al. 2004, Takeuchi et al. 2006, Fujii and Kitamura 2007, Mazza and
Vulcano 2015) and the seismic design of new RC buildings (e.g., Izumi et al. 2004, Matsuura et al.
2008). A popular system is the buckling-restrained brace (BRB) system using low-yield-strength steel.
BRBs are undoubtedly effective in the seismic retrofitting of RC buildings because the structural system
is clear; i.e., the deformation of braces equals the shear deformation of the existing RC frame. However,
applying BRBs to the seismic retrofitting of an existing condominium may introduce architectural
problems; e.g., installing BRBs in front of a balcony may obstruct the view from a window of the living
space. The use of a damper column with a shear panel (Katayama et al. 2000) may therefore be an
attractive choice for the seismic retrofitting of existing condominiums. In the damper column, shown in
Figure 1(a), low-yield-strength steel is used for the shear panel, which absorbs the hysteresis energy.
One possible scheme for the seismic retrofitting of a concrete condominium using steel damper columns

1
Professor, Dr. Eng., Chiba Institute of Technology, Narashino, Chiba, Japan, kenji.fujii@it-chiba.ac.jp
2
General Manager, Dr.Eng., JFE Civil Engineering and Construction Corp., Tokyo, Japan, miyagawa@jfe-
civil.com
Figure 1. Seismic retrofit of an existing RC building using steel damper columns.

Figure 2. Detail of the connection of an additional steel frame to an existing RC structure

involves adding a new steel frame on the side of the balcony, as shown in Figure 1(b). Details of the
connection of the additional steel frame to the existing concrete structure are given in Figure 2.
For the seismic retrofit design of existing buildings using hysteresis dampers, several researchers have
already proposed simplified nonlinear analysis procedures that predict the peak seismic response (e.g.,
Mazza and Vulcano 2015, Fujii 2015). However, as far as the authors know, none of the procedures
consider the strain hardening behavior of low-yield-strength steel due to cyclic loading, which is
important in the case of a shear panel made from low-yield-strength steel (Nakashima 1995a).
The present paper proposes a simplified procedure for predicting the peak response of a retrofitted
building with a low-yield-strength-steel damper panel considering the strain hardening behavior. The
peak seismic response of a seven-story SRC building retrofitted with steel damper columns is then
predicted employing the proposed simplified procedure and compared with the results of time-history
analysis.

2. SIMPLIFIED PROCEDURE CONSIDERING STRAIN HARDENING BEHAVIOR

2.1 Lower-bound model and upper-bound model

Figure 3 compares seismic analysis methods considering the strain hardening behavior of the damper
made from low-yield-strength steel. The most rigorous method of considering this behavior is
2
Figure 3. Comparisons of seismic analysis methods.

undoubtedly conducting nonlinear dynamic (time-history) analysis, which considers directly the strain
hardening behavior due to cyclic loading using a hysteresis model (e.g., Nakashima et al. 1995b).
However, this approach is time consuming and requires huge computational effort to evaluate the
responses to all possible seismic intensities. Instead, simplified nonlinear analytical procedures that
combine the nonlinear static (pushover) analysis of a multi-degree-of-freedom model and the response
spectrum analysis of an equivalent single-degree-of-freedom (SDOF) model are available (e.g., Fajfar
2000). However, the strain hardening behavior under monotonic loading is not considered in these
simplified procedures.
To approximate the strain hardening behavior, the present paper proposes the use of two models, namely
the lower-bound (LB) model and upper-bound (UB) model. In the LB model, the yielding strength of
the damper panel is assumed to be the LB yielding strength, QyDL, which corresponds to the initial
yielding strength before strain hardening. In the UB model, meanwhile, the yielding strength of the
damper panel is assumed to be UB yielding strength, QyDU, which corresponds to the yielding strength
after appreciable cyclic loading. The predicted peak response of each component is determined from the
envelope of the prediction results of LB and UB models. This paper assumes that the initial nominal
yield stress of low-yield-strength steel is 100 N/mm2, while the nominal yield stress after appreciable
cyclic loading is assumed to be 200 N/mm2.

2.2 Outline of the proposed simplified procedure

STEP 1: Construct two nonlinear models of frame structures, namely the LB model and UB model.
STEP 2: Carry out pushover analysis of the LB model. This procedure employs displacement-based
mode-adaptive pushover analysis (DB-MAP), proposed previously by the first author (Fujii 2014). Let
N be the number of floors of the considered (planar) building model, M be the mass matrix and nd and
nfR respectively be the displacement vector and restoring force vector at loading step n:

 m1 0   n y1   n f R1 
M =  O  , d =  M , f =  M  . (1)
 n   n R  
 0 mN   y   f 
n N   n RN 

Here, mj is the mass of the jth floor while nyj and nfRj are respectively the displacement and restoring
forces of the jth floor at loading step n. The equivalent displacement and acceleration of the first mode
at loading step n, nD1* and nA1* respectively, are calculated assuming that nd is proportional to the first
mode vector at each loading step, n Γ 1 n φ 1 :

T
Γ1 n φ1 M n d
m
j
j n y j2
Γ1 n φ1T n f R
 y
j
n j n f Rj
*
n D1 =
n
= , n A1* = n
= , (2)
n M1
*
m j
j n yj n M1
*
m j
j n yj
2
   2 
n M 1* = n Γ12 n φ 1 T M n φ 1 =   m j n y j   mj n yj  . (3)
 j   j 
3
Here, nM1* is the equivalent modal mass of the first mode at loading step n. This paper refers to the
relationship between nA1* and nD1* as the capacity curve. The equivalent period of the first mode at
loading step n, nT1eq, is calculated as

T
n 1eq = 2π n D1* n A1* . (4)

The method of calculating the equivalent damping ratio of the first mode at loading step n, nh1eq, is given
in the Appendix.
STEP 3: Predict the peak equivalent displacement D1*max employing the equivalent linearization (Otani
2000). The spectral reduction factor relating to the equivalent damping ratio, F(nh1eq), is determined as

S A ( nT1eq , n h1eq ) S D ( nT1eq , n h1eq ) 1.5


F ( n h1eq ) =
p
= = , (5)
p S A ( nT1eq ,0.05 ) S D ( nT1eq ,0.05 ) 1 + 10 n h1eq

where pSA(nT1eq, nh1eq) and SD(nT1eq, nh1eq) are respectively the pseudo-spectral acceleration and spectral
displacement of the ground motion. The present study refers to the relationship between pSA(nT1eq, nh1eq)
and SD(nT1eq, nh1eq) as the demand curve. The predicted peak equivalent displacement D1*max is obtained
as the intersection of the capacity and demand curves.
STEP 4: From the results of STEPS 2 and 3, obtain the peak response for each component (e.g., the peak
story drift, peak shear deformation and force of the damper panel) by referring to the pushover analysis
result corresponding to D1*max. The maximum response experienced until the equivalent displacement
reaches D1*max is the predicted peak response of each component obtained from the LB model.
STEP 5: Repeat STEPS 2 to 4 to obtain the peak response of the UB model.
STEP 6: Determine the predicted peak response of each component from the envelope of the predicted
peak responses of the LB and UB models.

3. BUILDING AND GROUND MOTION DATA

3.1 Building Data

The present study investigated a seven-story RC/SRC condominium designed according to the pre-1981
seismic code of Japan. Figure 4 shows the plan of the model building while Figure 5 shows the elevation
of the existing RC/SRC frame of the model building. This building model is the simplified model studied
previously by the first author (Fujii and Kuramoto 2012). This building model (i.e., the existing building)
consists of two frames in the longitudinal direction. The structure below the middle of the third story is
the SRC structure, while that above the middle of the third story is the RC structure. The floor mass per
floor area is assumed to be 1.2 t/m2. A seismic slit and additional shear reinforcement by carbon fiber
are assumed to be properly provided to the existing column in frame Y1 to prevent shear failure. The

Figure 4. Plan of the model building. Figure 5. Elevations of the existing RC/SRC frame.
4
model without an additional damper frame is referred to as Model-07-O. The base shear coefficient of
Model-07-O is 0.282.
A new steel frame with a damper column is added to the balcony side of Model-07-O in the present
study. The additional steel frame is connected to the existing concrete structure using horizontal braces
as shown in Figure 2. Two retrofitted models are considered to verify the accuracy of the simplified
procedure. The elevations of additional steel frames are shown in Figure 6. Damper columns are
installed on the first to fifth stories of Model-07-D01-05 shown in Figure 6(a) and on the first to sixth
stories of Model-07-D01-06 shown in Figure 6(b). All damper columns are identical. The sections of H-
section columns and beams are determined to have enough stiffness and strength to behave in the elastic
range when the strength of the damper panel reaches the UB yielding strength. The ratio of QyDL to the
total weight of the whole structure is 0.098.

Figure 6. Elevation of the additional steel frame with a damper column.

3.2 Modelling of building structures

The building is modelled as a fishbone-type plane frame model in the present study. The floor slab in
existing buildings and horizontal braces that connects the existing structures and additional steel frame
is assumed to be rigid. Figure 7 shows the modelling of each frame. As shown in Figure 7(a), all columns
and beams in existing RC/SRC frames (frames Y0 and Y1) are modelled as one component with
nonlinear flexural and shear springs. The rigid zone is considered in the existing RC/SRC frame. In the
additional steel frame shown in Figure 7(b), the damper column is modelled as an elastic column with
a nonlinear shear spring in the middle. The beams in the additional steel frame are modelled as elastic
beams. No rigid zone is considered in the additional steel frame.

Figure 7. Modelling of each frame.

5
Figure 8. Detail of existing RC/SRC frames. Figure 9. Detail of the additional steel frame.

Figure 8 shows the detail of existing RC/SRC frames. In frame Y1, the nonstructural wall above the
beam is considered for the calculation of elastic stiffness of the beams. Figure 9 shows the detail of the
additional steel frame. The height of the damper panel is assumed to be 0.86 m.
Figure 10 shows the envelopes of the force–deformation relationship of members, while Figure 11
shows the hysteresis model for each nonlinear spring. For the flexural spring of existing RC/SRC
members, the Muto model (Muto et al. 1974) with one modification is used to model the flexural spring,

Figure 10. Envelopes of the force–deformation relationship of members.

Figure 11. Hysteresis model for each nonlinear spring.

6
as shown in Figure 11(a). Specifically, the unloading stiffness after yielding decreases in proportion to
μ−0.5 (where μ is the ductility ratio of the flexural spring) to represent the degradation of the unloading
stiffness after yielding of the RC members, as in the model of Otani (Otani, 1981). The origin-oriented
model (Figure 11(b)) is used to model the shear spring of columns with a wing wall. The shear behavior
of other existing columns and beams is assumed to be elastic. The I-K model proposed by Nakashima
et al. (Nakashima et al. 1995b) is used to model the shear spring of the damper panel in the nonlinear
time-history analysis: the two parameters, α and β, are respectively assumed to be 0.005 and 0.85 as
shown in Figure 11(c) and as proposed by Nakashima et al. (Nakashima et al. 1995b). Note that for the
nonlinear time-history analysis, the initial yield strength of the damper panel shown in Figure 11(c) is
set to be QyDL while the UB yield strength is set to be QyDU.
The damping matrix at time t, C(t), is assumed to be proportional to the instant stiffness matrix without
a damper column. The damping ratio of the elastic first mode of the model without a damper column is
assumed to be 0.05.

3.3 Ground motion data

Two groups of artificial ground motions are used for the nonlinear time-history analysis. The target
elastic spectrum with 5% critical damping, pSA(T, 0.05), determined from the Building Standard Law of
Japan (BCJ 2016) considering the type-1 (rock) soil condition, is calculated as

4.8 + 45T m s 2 : T < 0.16s



p S A ( T ,0.05 ) = 12.0 : 0.16s ≤ T < 0.576s , (6)
12.0 ( 0.576 T ) : T > 0.576s

where T is the natural period of the SDOF system. Two records are used to determine the phase angle
of artificial ground motions: one is the horizontal major component of Sendai Government Office
building #2 recorded during the 2011 earthquake off the Pacific coast of Tohoku (Kashima et al. 2012)
while the other is the horizontal major component of the 1995 JMA Kobe record (respectively referred
to as the SND record and JKB record). To generate artificial ground motion with the same phase
difference but different time histories, the phase angle is shifted by the constant Δϕ0. The detail of the
generation of artificial ground motion is as follows.

(i) Perform a Fourier transform of the original record ag(t) to obtain the amplitude Ai and phase angle
ϕi for the circular frequency ωi.
(ii) Shift each phase angle ϕi by adding the constant Δϕ0. The new (shifted) phase angle for the circular
frequency ωi, ϕi*, is expressed as

φi* = φi + Δφ0 . (7)

(iii) Calculate the new wave ag*(t) using the amplitude Ai and shifted phase angle ϕi* according to

ag * ( t ) =  Ai cos (ωit − φi* ) . (8)


i

(iv) Calculate the pseudo acceleration response spectrum of the wave ag*(t). Then calculate the ratio of
the calculated response spectrum and target response spectrum.
(v) Modify each amplitude Ai according to the ratio of the response spectrum. Then calculate the new
wave ag*(t) using the modified amplitude and ϕi*. Repeat this process until the calculated response
spectrum is close enough to the target response spectrum.

Twelve artificial ground motions are generated for each record by varying the constant Δϕ0 in intervals
of π/12 from 0 to 11π/12; the Art-L series (wave Art-L00 to Art-L11) are the artificial waves generated
from the SND record while the Art-S series (wave Art-S00 to Art-S11) are the artificial waves generated

7
Figure 12. Elastic pseudo acceleration spectrum of artificial ground motions

Figure 13. Example wave of the generated artificial ground motions.

from the JKB record. Figure 12 shows the pseudo acceleration spectra of the generated artificial ground
motions while Figure 13 shows an example wave of the generated artificial ground motions. From the
comparison of Art-L00 and Art-S00, it is expected that the strain hardening behavior of the damper
column may be notable if the input ground motion is the Art-L series rather than the Art-S series,
although the same target spectrum is considered for the two series. Figure 13 (c) compares the 12 waves
of the Art-S series over a short time. It is seen that the details of all waves are different owing to the
shift of the phase angle.

4. ANALYSIS RESULTS

4.1 Prediction of the peak response in a simplified procedure

Figure 14 shows the prediction of the peak equivalent displacement D1*max for all models. The elastic
and equivalent periods of the first modes are also shown. Comparisons in the figure reveal that the
predicted peak response of the model building is effectively reduced by the addition of dampers as
expected: the predicted D1*max of Model-07-O is 0.173 m while values for models with dampers are
within the range of 0.0978–0.107 m. Figure 14(b) shows that the predicted D1*max of the UB model is
almost the same as that of the LB model for Model-07-D01-05, even though the equivalent period of
the UB model at D1*max is shorter than that of the LB model. A similar result is obtained for Model-07-
D01-06, as shown in Figure 14(c). Figure 14(b) shows that in case of Model-07-D01-05, D1*max of the
LB model is close to the loading step when the story drift R reaches 1/100. In the case of the UB model,
meanwhile, D1*max is close to the loading step when the story drift R reaches 1/50. This is because in the
UB model of Model-07-D01-05, the partial collapse mechanism is on the sixth and seventh stories.
To clarify the differences between LB and UB models, Figure 15 compares the relationships of the
equivalent damping nh1eq and equivalent displacement nD1*. The figure suggests that nh1eq at predicted
peak displacement D1*max of the LB model is larger than that of the UB model. This is because the yield
deformation of the damper panel assumed in the UB model is twice that assumed in the LB model.
Therefore, one reason why the predicted peak equivalent displacement D1*max of the UB model is almost

8
Figure 14. Peak equivalent displacement predicted using the equivalent linearization.

Figure 15. Comparisons of the relationships of equivalent damping and equivalent displacement.

the same as that of the LB model is that the equivalent damping of the LB model at D1*max is larger than
that of the UB model while the equivalent period of the UB model at D1*max is shorter than that of the
LB model.

4.2 Comparison of the results of nonlinear time-history analysis

Figure 16 compares the peak response (story drift Rmax, shear deformation of damper panel γDmax, shear
force of damper panel normalized by the LB yield strength QDmax / QyDL, and moment of the damper
column at the end normalized by the yield moment of the H-section column MDmax / MyDe) obtained from
the nonlinear time-history analyses and simplified procedure. The figure shows the mean and standard
deviation of peak responses obtained from each Art-L/S series.
Figure 16(a) shows that the predicted Rmax of Model-07-O is close to the mean of Rmax for the Art-L
series while the mean of Rmax for the Art-S series exceeds the predicted peak. In contrast, the mean of
Rmax for the Art-L series is larger than that of the Art-S series on most stories in the cases of Model-07-
D01-05 and Model-07-D01-06 (Figure 16(b) and (c)). The predicted values of Rmax of Model-07-D01-
05 and Model-07-D01-06 agree well with the time-history analyses results on all stories except the upper
stories of Model-07-D01-05. The predicted γDmax, QDmax/QyDL, and MDmax/MyDe are conservative,

9
Figure 16. Comparisons of peak responses obtained in time-history analyses and the simplified procedure.

compared with the results of time-history analyses. Figure 16(b) and (c) shows that the mean values of
QDmax/QyDL and MDmax/MyDe obtained for the Art-L series are larger than those obtained for the Art-S
series, as expected. However, the differences are not appreciable.
Note that on most stories, the predicted Rmax and γDmax are determined from the results of the LB model,
while the predicted QDmax and MDmax are determined from the results of the UB model. In conclusion,
the use of two models in the prediction of peak responses and their envelope is reasonable for the seismic
evaluation of buildings retrofitted using steel damper columns considering strain hardening behavior.

5. CONCLUSIONS

A simplified procedure was proposed for predicting the peak response of a retrofitted building with a
low-yield-strength steel damper panel considering the strain hardening behavior. The peak seismic
response of a retrofitted seven-story SRC building with steel damper columns was then predicted using
the proposed simplified procedure and compared with the results of time-history analysis. The results
show the simplified procedure can satisfactorily estimate the peak response of a retrofitted building
model with damper columns considering the strain hardening effect.

REFERENCES

BCJ (2016) The building standard law of Japan on CD-ROM. The Building Center of Japan, Tokyo.

10
Fajfar P. (2000) A nonlinear analysis method for performance-based seismic design, Earthquake Spectra, 16(3),
pp.573-592.
Fujii K, Kitamura H. (2007) Seismic performance evaluation of existing R/C buildings with hysteresis dampers
based on energy balanced response, Third International Conference on Urban Earthquake Engineering, Tokyo
Institute of Technology, 5-6, March, Tokyo, Japan.
Fujii K, Kuramoto H. (2012) Earthquake response evaluation of an existing steel reinforced concrete building
using equivalent SDOF system, Journal of Structural and construction Engineering, AIJ, 77(678), pp.1227-1236.
(in Japanese).
Fujii K. (2014) Prediction of the largest peak nonlinear seismic response of asymmetric buildings under bi-
directional excitation using pushover analyses, Bulletin of Earthquake Engineering, 12, pp. 909-938.
Fujii K. (2015) Application of the pushover-based procedure to predict the largest peak response of asymmetric
buildings with buckling-restrained braces, 5th International conference on computational methods in structural
dynamics and earthquake engineering (COMPDYN 2015), 25-27, May, Crete Island, Greece.
Ishii T, Mukai T, Kitamura H, Shimizu T, Fujisawa K, and Ishida Y. (2004) Seismic retrofit for existing R/C
building using energy dissipative braces, Proceedings of the 13th World Conference on Earthquake Engineering,
1-6, August, Vancouver, Canada.
Izumi M, Chiba O, Takahashi K, Iizuka S (2004) Earthquake resistant performance of reinforced concrete frame
with energy dissipation devices, Proceedings of the 13th World Conference on Earthquake Engineering, 1-6,
August, Vancouver, Canada.
Kashima T, Koyama S, Okawa I. (2012) Strong motion records in buildings from the 2011 off the Pacific coast of
Tohoku earthquake, Building Research Institute (in Japanese).
Katayama T, Ito S, Kamura H, Ueki T, and Okamoto H. (2000) Experimental study on hysteretic damper with low
yield strength steel under dynamic loading, Proceedings of the 12th World Conference on Earthquake Engineering,
30 January – 4 February, Auckland, New Zealand.
Matsuura T, Inai E, Fujimoto T. (2008) A study on structural performance considering connecting condition of
pillar type hysteretic dampers for R/C buildings, Proceedings of the 14th World Conference on Earthquake
Engineering, 12-17, October, Beijing, China.
Mazza F, Vulcano A. (2015) Displacement-based design procedure of damped braces for the seismic retrofitting
of r.c. framed buildings, Bulletin of Earthquake Engineering, 13(7), pp 2121–2143.
Muto K, Hisada T, Tsugawa T, Bessho S. (1974) Earthquake resistant design of a 20 story reinforced concrete
buildings, Proceedings of the fifth world conference on earthquake engineering. 25-29, June, Rome, Italy.
Nakashima M. (1995a) Strain-hardening behaviour of shear panels made of low-yield steel 1: Test, Journal of
Structural Engineering, ASCE, 121(12), pp.1742-1749.
Nakashima M, Akazawa T., Tsuji B. (1995b) Strain-hardening behaviour of shear panels made of low-yield steel
2: Model, Journal of Structural Engineering, ASCE, 121(12), pp.1750-1757.
Otani S. (2000) New seismic design provision in Japan, Proceeding of the Second U.S.-Japan Workshop on
Performance-Based Earthquake Engineering Methodology for Reinforced Concrete Structures. 11-13, September,
Sapporo, Japan.
Otani S. (1981) Hysteresis models of reinforced concrete for earthquake response analysis. J Faculty of
Engineering, University of Tokyo, 36(2), pp. 125-156.
Takeuchi T, Yasuda K, Yuasa K, Iwata M. (2006) Seismic retrofit of existing building with hysteretic dampers,
Third International Conference on Urban Earthquake Engineering, Tokyo Institute of Technology, 6-7, March,
Tokyo, Japan.

APPENDIX. METHOD OF CALCULATING EQUIVALENT DAMPING

The method of calculating the equivalent damping nh1eq is summarized in Figure A1. The initial damping
ratio, h0, is assumed to be 0.05, which is consistent with the damping ratio used in the nonlinear time-
history analysis. The strain energy of the axial deformation of columns is neglected because the
buildings are modelled as fishbone-type plane frame models.
11
Figure A1. Method of calculating equivalent damping of the whole structure.

12

You might also like