Professional Documents
Culture Documents
Abstract: This paper summarizes a number of key findings that affect the use of linear and nonlinear analysis procedures for the seismic
evaluation of steel frame buildings with supplemental damping devices and in particular, buckling-restrained braces (BRBs) and nonlinear
viscous dampers (NVDs). The assessment is based on a comparison of various engineering demand parameters (EDPs) with experimental
data obtained from a series of full-scale shaking table tests of a five-story steel building equipped with BRBs and NVDs. It is shown that:
(1) there is no clear advantage between three-dimensional (3D) and two-dimensional (2D) analyses in the prediction of the dynamic behavior
of regular plan view steel frame buildings regardless of the damper type; (2) incorporating the brace and nonlinear viscous damper axial
flexibility in the analytical model of the test structure with NVDs significantly improves the overall numerical predictions; and (3) the current
effective damping ratio recommended by ASCE 41-13 as part of linear static procedures for the evaluation of frame buildings with BRBs or
NVDs may not be conservative enough. A new performance-based design tool called performance-spectra (P-Spectra) is able to reliably
predict the EDPs of interest. The P-Spectra tool is also able to validate the efficiency of various damper solutions on the dynamic performance
of the test structure. It is demonstrated that P-Spectra can be employed to predict estimates of potential residual deformations that traditional
linear and nonlinear static procedures cannot. DOI: 10.1061/(ASCE)ST.1943-541X.0001474. © 2016 American Society of Civil Engineers.
Author keywords: Nonlinear viscous dampers; Buckling-restrained braces; Simplified assessment methods; Supplemental damping;
Performance-spectra; Full-scale shake table tests; Nonlinear response history analysis; Seismic effects.
Introduction widely used in the current seismic design practice (Buchanan et al.
2011; Kasai et al. 2013).
The concept of resilient cities is becoming essential (Cimellaro et al. Because of the increasing use of supplemental damping devices
2010). Recent earthquakes (e.g., Chile in 2010; Christchurch in for the design and/or the seismic retrofit of existing frame build-
2011) in developed countries demonstrate that although conven- ings, it is important to validate and refine various assessment pro-
tional buildings designed according to regional seismic provisions cedures that are currently used for the evaluation of these buildings.
had minimal structural damage afterwards they were still nonopera- Rigorous nonlinear response history analysis (NRHA) may be em-
tional because of extensive nonstructural damage (Dhakal et al. ployed for this purpose. However, this may not be practical for a
2011; Miranda et al. 2012). To this end, response modification structural engineer because several iterations may be required to
(e.g., base isolation, supplemental damping devices, weakening) achieve the desired design objectives with an optimal use of damp-
is widely advocated. Damage-free lateral load-resisting systems ing devices. Therefore, the computational cost of NRHA may be
(e.g., self-centering and rocking systems) have also been developed significant. The difficulty of selecting and scaling representative
that minimize structural damage in the aftermath of an earthquake ground motions that describe the regional seismicity may be an-
(Christopoulos and Filiatrault 2006). The use of such systems is other important consideration. To this end, several simplified as-
1
sessment procedures have been developed for the seismic design
Ph.D. Student, Dept. of Civil Engineering and Applied Mechanics, and the performance evaluation of frame buildings equipped with
McGill Univ., Montreal, QC, Canada H3A 2K6.
2 supplemental damping devices. These methods include linear and
Associate Professor, Dept. of Civil Engineering and Applied
Mechanics, McGill Univ., Montreal, QC, Canada H3A 2K6 (corresponding nonlinear static procedures in accordance with ASCE 41-13
author). E-mail: dimitrios.lignos@mcgill.ca (ASCE/SEI 2014) and FEMA 274, 356, and 440 (FEMA 1997,
3
Section Manager, Building Construction and Steel Structures Division, 2000, 2005) as well as performance curves (Kasai et al. 1998; Kasai
Nippon Steel and Sumikin Engineering Co. Ltd., Tokyo 141-8604, Japan; and Ito 2005; Kasai et al. 2007), which are primarily based on lin-
formerly, Researcher, E-Defense, National Research Institute for Earth earization and transformation of frame buildings with dampers into
Science and Disaster Prevention (NIED). equivalent linear single-degree-of-freedom (SDF) systems. More
4
Professor, Disaster Prevention Research Institute (DPRI), Kyoto Univ., recently, a performance spectra (so-called P-Spectra) method
Uji, Kyoto 611-0011, Japan.
(Guo and Christopoulos 2013a, b) was proposed for the simplified
Note. This manuscript was submitted on April 3, 2015; approved on
November 17, 2015; published online on February 4, 2016. Discussion design and retrofit of frame buildings with hysteretic and linear vis-
period open until July 4, 2016; separate discussions must be submitted cous/viscoelastic dampers. The P-Spectra method also provides an
for individual papers. This paper is part of the Journal of Structural En- estimate of the expected residual deformations of a frame building
gineering, © ASCE, ISSN 0733-9445. with dampers. The aforementioned methods have not been
d,1 d,2
Elastic Elastic
Connection portion Plastic portion portion Connection
(a)
Mortar
Steel
Downloaded from ascelibrary.org by New York University on 02/11/16. Copyright ASCE. For personal use only; all rights reserved.
tube
Core plate
Fig. 2. Schematic representation of a typical buckling restrained brace (BRB) including its main parts
Table 1. Characteristic Lengths and Cross Sectional Areas of Buckling-Restrained Braces Used as Part of the Testing Program (Data from Hikino 2012)
Lpl Apl Lel Ael Lcon Acon Core plate K di a Fdyi b
Frame Story (mm) (mm2 ) (mm) (mm2 ) (mm) (mm2 ) (mm) (kN=mm) (kN)
X direction (2 bays) 4 2,900 1,408 277.3 3,840 285 7,264 16 × 88 81 317
3 2,900 1,408 277.3 3,840 285 7,264 16 × 88 81 317
2 2,750 2,204 303.3 5,909 295 8,569 19 × 116 126 496
1 3,600 2,204 258.1 5,909 295 8,569 19 × 116 105 496
Y direction (1 bay) 4 2,600 2,794 393.3 7,568 280 9,856 22 × 127 161 629
3 2,600 2,794 393.3 7,568 280 9,856 22 × 127 161 629
2 2,450 4,400 469.5 12,276 230 12,276 22 × 200 249 990
1 3,300 4,400 434.4 12,276 230 12,276 22 × 200 209 990
a
Equivalent stiffness of BRBs used in SAP2000 model and simplified methods. The area between connections and working points is assumed equal to Acon .
b
Average yield strength of BRBs under quasi-static loading. The values are multiplied by 1.32 in simulation and simplified methods to account for the dynamic
amplification of strength due to rate effects.
L
Ld Lb
d
Clevis & Brackets Damper Portion Brace Clevis & Brackets
(a)
(b)
Kdi Cdi ,
(c)
Fig. 3. Schematic representation of a typical nonlinear viscous damper including its main parts and mathematical model representation
c
Estimated stiffness values (K ci ) for damper portion due to lack of data of dampers at third and fouth story (Yu et al. 2013).
internal damper, the steel brace (K bi ), the clevis, brackets (K cli ), same figure the authors have superimposed the pseudoacceleration
and gusset plates are shown in Fig. 3 and summarized in Table 2. and velocity spectra that correspond to a maximum considered earth-
This table also provides the characteristic lengths and cross sec- quake (i.e., 2% probability of exceedance in 50 years) in urban
tional areas of the damper portion (Ld , Ad ) and the brace (Lb , California (location: 37.779°N, 122.419°W) for soil class D per
Ab ). The same figure illustrates the mechanical model for a ASCE (2010), whereas the 40% intensity of the JR Takatori record
NVD [Figs. 3(b and c)]. According to dynamic sinusoidal displace- is closer to a design level earthquake (i.e., 10% probability of exceed-
ment tests with varying amplitudes, the velocity exponent, α, of all ance in 50 years) in the same location.
the NVDs was found to be equal to 0.38 (Kasai et al. 2008). Finally, Figs. 4(c and d) show the relative displacement response
The predominant periods of the test structure in the X and Y- spectra of the unscaled JR Takatori record in X and Y-loading
loading directions were T 1x ¼ 0.650 s and T 1y ¼ 0.705 s, respec- directions for damping ratios ranging from β ¼ 2 to 30%. From
tively. The same periods of the test structure with BRBs were T 1x ¼ Fig. 4(c), the expected relative displacement demands in the X-
0.469 s and T 1y ¼ 0.487 s and those of the test structure with loading direction are not affected by much with respect to the damp-
NVDs were T 1x ¼ 0.536 s and T 1y ¼ 0.575 s. Note that the effect ing ratio for the range of periods of interest (i.e., 0.47–0.65 s). This is
of nonstructural components on the lateral stiffness of the test struc- not the case in the Y-loading direction [Fig. 4(d)]. A damping ratio
ture with various types of dampers was considered. The corre- increase from 2 to 30% results to a decrease of the relative displace-
sponding damping ratios for the test structure (i.e., bare frame) ment demands by about one-third for the same range of periods.
were 1.0%; while for the test structure with BRBs and NVDs
was 1.4% and 4.3%, respectively. Further information regarding
the dynamic characteristics of the test structure with various damp- Seismic Performance of the Test Structure with
ing configurations can be found in Kasai et al. (2011) and Ji et al. Buckling-Restrained Braces and Nonlinear Viscous
(2013). Notably, the nonstructural components of the building con- Dampers
tributed to approximately 30% of its total lateral stiffness (Kasai
et al. 2010). The experimental periods and damping ratios of the The test structure with BRBs and NVDs behaved nearly elastically
respective test structure were determined with the linear identifica- for the 15% scaled intensity of the JR Takatori record. Therefore,
tion method based on low-amplitude white noise tests and low- this section summarizes in brief the seismic performance of the test
amplitude ground motion shaking with the JR Takatori record from structure with BRBs and NVDs only for the 40 and 100% scaled
the 1995 Kobe earthquake as discussed in the next section (Kasai intensities of the JR Takatori record. Figs. 5 and 6 illustrate the peak
et al. 2010). During these tests the BRBs remained elastic. story drift ratios, peak story shear forces, and peak absolute floor
accelerations along the height of the test structure with BRBs and
NVDs, respectively, under the 40 and 100% of the JR Takatori re-
Testing Protocol cord. In these figures the experimental data is noted as Exp. The
seismic intensity of interest (i.e., 40 and 100%) is noted after.
The test structure with various damper types in place was subjected Superimposed in the same figure are the same engineering demand
to the three components of the JR Takatori record from the 1995 parameters of interest as predicted from rigorous NRHA with vari-
Kobe earthquake. This record was scaled incrementally at 15, 40, ous nonlinear model representations discussed in detail in the fol-
70, and 100% of its unscaled intensity. Figs. 4(a and b) illustrate the lowing section. From Figs. 5(a and d) and 6(a and d), the retrofitted
5% damped pseudoacceleration and pseudovelocity spectra of the test structure did not exceed 1% story drift ratio along its height
40 and 100% JR Takatori record, respectively, in the X and Y-loading regardless of the loading direction and the employed damper type.
directions of the test structure [Fig. 1(a)]. From this figure, at the This was the expected yield story drift ratio of the test structure
highlighted range of periods under consideration (0.47–0.71 s) without dampers (Kasai et al. 2008). From Fig. 5(b) peak story
the pseudovelocity of the 100% JR Takatori record in the Y-loading shear forces were relatively large at 40% of the JR Takatori record
direction is about 1.4 times higher than that in the X-loading direc- when BRBs were employed compared to the test structure with
tion [Fig. 4(b)]. Superimposed in Fig. 4 is the pseudoacceleration and NVDs [Fig. 6(b)]. This is attributed to the additional lateral stiff-
pseudovelocity spectra based on a Level-2 earthquake in Japan (BCJ ness that BRBs provided to the test structure compared to that of the
2008). To put things in perspective, the pseudovelocity of the un- NVDs. Similar observations hold true for the peak absolute floor
scaled ground motion in the Y-loading direction is about 1.1 times accelerations as shown in Figs. 5(c and f) and 6(c and f) for the test
larger than that of a Level-2 earthquake in Japan [Fig. 4(b)]. In the structure with BRBs and NVDs, respectively.
Fig. 4. Response spectra of the 40 and 100% scaled intensities of the JR Takatori record
Nonlinear Response History Analysis of the Test material properties of the steel cross sections as reported by Kasai
Structure with Buckling-Restrained Braces and et al. (2010) were explicitly assigned to the respective beam and
Nonlinear Viscous Dampers column elements of the various numerical models of the test struc-
ture. In the SAP2000 model, steel beams and columns are modeled
The test structure is modeled in two different simulation platforms. as elastic elements. The composite action attributable to the pres-
The first one is SAP2000, which is a commercial software widely ence of the slab is explicitly considered in all model representa-
used by structural engineering offices in North America. The sec- tions. In particular, in the OpenSees numerical model of the test
ond one is the Open System for Earthquake Engineering Simula- structure. This is done with a fiber cross section on top of the bare
tion (OpenSees) platform (McKenna 1997). A three-dimensional steel wide-flange section. The reinforced concrete slab is modeled
(3D) model of the test structure is developed in both platforms. with a concrete (Concrete02) material (Yassin 1994), which ac-
For comparison purposes a two-dimensional (2D) model is also counts for the concrete linear tension softening. The effective width
developed in OpenSees. For the 2D model representation, all three of the concrete slab is calculated based on Section I3.1a of ANSI/
moment resisting frames per loading direction [Figs. 1(c and d)] are AISC 360-10 (AISC 2010). The contribution of nonstructural com-
modeled in series and linked at each floor level with axially rigid ponents to the lateral resistance of the test structure is not consid-
truss elements in order to simulate the rigid diaphragm action. The ered. In a prior study (Lignos 2012), the autoclaved lightweight
2D models consider the panel zone shear deformation based on the concrete (ALC) panels that represent part of the nonstructural com-
Krawinkler (1978) model. For both the 2D and 3D OpenSees mod- ponents of the test structure were simulated with the SAW hyster-
els the steel beams and columns of the test structure are modeled etic model (Folz and Filiatrault 2001) through diagonal struts that
with single force-based distributed plasticity beam-column ele- were installed around the perimeter of the test structure. Based on
ments. Five integration points are considered in order to trace Lignos (2012), the effects of nonstructural components on the
the onset of yielding and cyclic hardening of steel beams and col- lateral resistance of the test structure with dampers did not seem
umns. The steel cross sections of the wide flange beams and HSS to be important. Geometric nonlinearities are considered with the
columns are discretized with a 5 × 3 fiber element grid along P-Delta transformation.
the width and thickness of the flange and web, respectively. The In order to model numerically a BRB in OpenSees, such element
Giuffré-Menegotto-Pinto (Menegotto and Pinto 1973) steel is divided into three parts. A force-based distributed plasticity
material model is assigned at each fiber element. The measured beam-column element is employed to model the core (center)
Fig. 5. Peak response of test structure with BRBs for 40 and 100% of the JR Takatori record: (a–c) X-loading direction; (d–f) Y-loading direction
portion of the BRB in which extensive axial yielding is expected; with elastic beam-column elements. The core portion of the BRB is
the rectangular section is discretized into 5 × 10 fibers along its modeled with the Giuffré-Menegotto-Pinto (Menegotto and Pinto
width and the length, respectively. The portions that are expected 1973) material model. The input parameters to define an engineering
to remain elastic, and the end connections of the BRBs are modeled stress-strain curve for the BRB core are obtained from calibrations
Fig. 6. Peak response of test structure with NVDs for 40 and 100% of the JR Takatori record: (a–c) X-loading direction; (d–f) Y-loading direction
Fig. 7. Comparison between experimental data and analytical predictions (data from E-Defense 2008)
with experimental data available from the E-Defense Blind Analy- element available in SAP2000. Fig. 7(b) shows the hysteretic
sis Contest (E-Defense 2008) as shown in Fig. 7(a). In this figure, behavior of a NVD similar to the one used as part of the shaking
the hysteretic response of the BRB includes the deformations from table test series (i.e., velocity exponent α ¼ 0.38). In this figure, the
both the elastic and plastic steel brace segments (Fig. 2). The BRB measured displacement of the NVD corresponds to the deformation
models were also calibrated with component test results obtained of the damper portion only (Fig. 3). The NVD is subjected to sinus-
from dynamic loading with a 2 Hz frequency by adjusting the oidal dynamic loading with 2 Hz frequency. In the same figure the
cyclic hardening parameters of the Giuffré-Menegotto-Pinto authors have superimposed the simulated behavior of the NVD as
material model to inherently capture the strain-rate effects on predicted by the numerical model discussed earlier. From this fig-
the axial strength of the BRBs. The 2 Hz frequency approximately ure, it is evident that the proposed model represents reasonably well
corresponds to the natural fundamental frequency of the building the hysteretic behavior of a NVD.
with BRBs (Hikino 2012). The required amplification factor to ob- In all cases, the Rayleigh damping model is considered with
tain the dynamic axial yield stress from the static one was found to 2% damping ratio assigned to the first and third translational
be equal to 1.32. This study employed different isotropic hardening modes of vibration of the respective numerical model of the test
parameters for axial tension and compression in order to take into structure with dampers. The Newmark average acceleration method
account the higher axial yield strength in compression than that in (Newmark 1959) is employed for time integration of the equations
tension because of the restrainer-core interaction within a BRB. of motion. The integration step for the NRHA is assumed to be
Similarly, in the SAP2000 3D model representation of the test struc- 0.01 sec. This was found to be adequate to guarantee both the
ture with BRBs, these were modeled as diagonal link elements by numerical stability and accuracy of the simulated results with re-
using the plastic Wen model (Wen 1976) with 1% postyield stiff- spect to the experimental data.
ness ratio and a yield exponent equal to unit. The equivalent BRB In Figs. 5 and 6 the numerical results from the 2D and 3D
axial stiffness was computed as discussed in Yu et al. (2013) and numerical model representations of the test structure are superim-
summarized in Table 1. posed with BRBs and NVDs, respectively, in terms of peak story
In order to simulate the hysteretic behavior of NVDs the Max- drift ratios, peak story shear forces, and peak absolute floor accel-
well model (Makris and Constantinou 1991) was employed, which erations for the two seismic intensities of interest of the JR Takatori
consists of a nonlinear dashpot and an elastic spring element in record (i.e., 40 and 100%). In these figures, the numerical simula-
series. The input parameters for the NVDs (Table 2) are the velocity tion results based on the OpenSees models are simply referred as
coefficient Cdi , velocity exponent, α, and the equivalent damper OS-2D and OS-3D. Results from the SAP2000 3D model are noted
stiffness K di , which represents the equivalent axial stiffness of the as SAP-3D. The responses of the 3D model were recorded at the
damper including its brace portion. In order to solve the ordinary geometric centers of each floor. Minor torsion was observed
differential equation (ODE) that expresses the force equilibrium during the numerical simulation. However, even at largest seismic
within a NVD a new material model was implemented into the intensities (e.g., 100% of the unscaled JR Takatori record), the
OpenSees platform (i.e., ViscousDamper) that uses the Dormand- maximum story drift ratios computed at perimeter frame corner no-
Prince method (Dormand and Prince 1980) with an adaptive step des of the test structures with BRBs and NVDs were found to be 10
size. The details of this implementation are summarized in and 5% larger than those computed at the geometric centers of each
Akcelyan et al. (2015). The absolute relative error between fifth floor, respectively. Figs. 5 and 6 indicate that for regular plan view
and fourth order solutions of the first order ODE is employed to low-rise steel frame buildings, there is no clear advantage between
control the substepping within the material model. This implemen- 3D and 2D analyses, regardless of the type of the passive control
tation does not require small integration steps during the NRHA device (i.e., BRB or NVD). From the same figures, there is practi-
when NVDs with small values of α or large values of K di are em- cally no difference between the simulated results from the 3D
ployed. In SAP2000 the NVDs in the 3D model of the test structure OpenSees and SAP2000 numerical models. This indicates that steel
are modeled with the Maxwell model assigned to a damper link frame buildings equipped with BRBs or NVDs with virtually no
Fig. 9. Damper hysteretic response in Y-loading direction at 100% JR Takatori record; comparison of experimental data with analytical predictions
from the 2D OpenSees model of the five-story test structure
0 0
-20 -20
-40 -40
-60 -60
Downloaded from ascelibrary.org by New York University on 02/11/16. Copyright ASCE. For personal use only; all rights reserved.
P1 SP2 hod 1 hod 2 SDF ctra RHA P1 SP2 hod 1 hod 2 SDF ectra RHA
LS L et t . e N LS L et et uiv. -Sp
Me uiv P-Sp N
S P-M SP- -Eq S P-M SP-M -Eq P
N N P N N S P
NS N
(a) (b)
20
0 0
-20 -20
-40 -40
-60 -60
Fig. 10. Relative error in prediction of peak roof displacement for the five-story test structure equipped with BRBs and NVDs
the experimental data from the full-scale tests conducted at E- an equivalent linear single-degree-of-freedom (SDF) system. Two
Defense. The efficiency of each method is justified based on the LSP methods have been utilized herein. The first one (noted as
relative error of the predicted values with respect to the experimen- LSP1) is based on ASCE-41-13 (ASCE/SEI 2014) and FEMA 274,
tal data at 40 and 100% seismic intensities of the JR Takatori record 356 (FEMA 1997, 2000). The second one (noted as LSP2) is based
per loading direction. In the linearization methods discussed in this on the current Japanese practice for the design of frame buildings
paper the authors used the spectral displacement directly from the with supplemental damping devices (Kasai and Ito 2005; Kasai
damped displacement spectra, Sd ðT eff ; β eff Þ of the respective et al. 2007). This method is also known as performance-curve.
ground motion record rather than the spectral damping modifica- The fundamental difference between LSP1 and LSP2 arises in
tion factors (i.e., B-factors in ASCE 41-13). This was done with the the calculation of the effective damping ratio, β eff of the equivalent
intent of conducting a fair comparison between the effective damp- SDF system. In particular, based on LSP1 if it is assumed that the
ing recommendations suggested by different linearization methods. excitation frequency of a harmonic motion is the same as the cyclic
For instance, from Figs. 4(c and d), if spectral damping modifica- frequency of the equivalent linear SDF system, its β eff can be com-
tion factors were to be used for a single ground motion record this puted as follows:
would result in significant errors in the predicted damped spectral
displacement values. However, in the design practice, the authors 2ðμd − 1Þð1 − pÞ 1
β eff ¼ β 0 þ ; p¼ ;
typically employ the linearization methods by using an idealized πμd ð1 − p þ pμd Þ 1 þ K d =K f
design spectrum. In this case, damping modification factors are PN
i¼1 ðμdi F dyi cos θi ϕri Þ
valid. μd ¼ P ð1Þ
i¼1 ðFdyi cos θi ϕri Þ
N
0 0
-20 -20
-40 -40
-60 -60
Downloaded from ascelibrary.org by New York University on 02/11/16. Copyright ASCE. For personal use only; all rights reserved.
P1 SP2 hod 1 hod 2 SDF ectra RHA P1 SP2 hod 1 hod 2 SDF ectra RHA
LS L et t . LS L et et uiv. -Sp
M Me uiv P-Sp N N
P- SP-M -Eq
M
S P- SP- -Eq S P
N N P N N P
NS NS
(a) (b)
20 20
Relative Error, RE [%]
-20 -20
-40 -40
-60 -60
P1 P2 a1 ra2 A1 A2 P1 P2 a1 ra2 A1 A2
LS LS pectr pect NRH NRH LS LS pectr pect NRH NRH
S S S S
P- P- P- P-
(c) (d)
Fig. 11. Relative error in prediction of peak base shear for the five-story test structure equipped with BRBs and NVDs
stiffnesses of the BRBs and bare frame within the equivalent SDF instead of computing it directly at a peak displacement of the equiv-
system, respectively; and μdi = displacement-ductility ratio of alent SDF system [Eq. (1)]. For both methods the effective period of
BRBs at level i in the horizontal direction. This is computed based the equivalent linear SDF system, T eff can be obtained as follows:
on the displacement of a relative mode shape, ϕri , by assuming a rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uniform drift profile. Fdyi and θi are the tensile yield force and in- pμd
T eff ¼ T f ð3Þ
clination angle of BRBs at level i, respectively. In particular, from 1 − p þ pμd
Eq. (1), it is understood that LSP1 is an iterative procedure because
the effective damping and period formulations employed from where T f is the predominant period of the bare frame. For frame
ASCE 41-13 (ASCE/SEI 2014) are displacement-dependent. buildings with NVDs, unlike LSP1, the LSP2 method incorporates
The iterative procedure to obtain μd and β eff was repeated until the axial flexibility of NVDs; therefore, the NVDs are explicitly
the initially assumed displacement demand matched the computed represented as a Maxwell model. For frame buildings with NVDs
one. It should be pointed out that the absolute acceleration response that concentrate their inelastic behavior into NVDs, the effective
spectrum of the JR Takatori record was employed to conduct such damping ratio and period formulations of the equivalent linear
iterations. SDF system can be obtained as follows based on the LSP2 method
However, according to LSP2 (Kasai and Ito 2005), the effective
damping ratio formulation, β eff can be computed as follows: ψK d0 0 u2c Tf
β eff ¼ β 0 þ ; T eff ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ;
2πðK a0 þ K f Þu2 1 þ K a0 =K f
2 1 − p þ pμd Cd ωαeff
β eff ¼ β 0 þ ln ð2Þ K d0 0 ¼ ð4Þ
μd πp μpd u1−α
c
The damping formulation in Eq. (2) is the average damping ratio where ψ = function of the velocity exponent, α, of the NVD, which
during an earthquake that accounts for the variation of ductility can be approximately computed as 4e−0.24α ; K d0 0 and Cd = loss
%
%
20 1 50 1
eff
eff
Teff
β
β
16 0.9 40 0.9
Effective Damping
Effective Damping
T eff / Tf
Teff / Tf
12 eff 0.8 30 0.8
8 0.7 20 0.7
eff
(a) (b)
Fig. 12. Effective damping ratio and period variation of equivalent linear SDF system based on LSP1 and LSP2 methods in Y-loading
direction
stiffness and damping coefficient of the nonlinear dashpot, respec- [i.e., Eq. (1)] may not be representative and should account for
tively; K a0 = storage stiffness of the Maxwell model; K f = lateral the displacement-based ductility variation of BRBs as reflected
stiffness of the bare frame; and u = maximum displacement of by Eq. (2).
the frame building with dampers. Further details regarding the com- For the test structure with NVDs, for moderate earthquake
putation of the β eff based on the LSP2 method can be found in Kasai events (i.e., 40% of the JR Takatori record), the LSP1 method under-
et al. (2007). If the axial flexibility of NVDs is neglected, then the estimates both EDPs of interest by more than 60% in the Y-loading
maximum displacement of the nonlinear dashpot, uc , becomes equal direction [Figs. 10(c) and 11(c)]. However, in the X-loading direction
to the maximum displacement u of the equivalent linear SDF system the relative error of the predicted peak roof displacements and base
of a frame building with NVDs. Thus, K a0 becomes zero and the shear of the test structure with NVDs is in the range of 10% or less.
effective angular frequency of the frame building with NVDs This is attributed to the fact that the reduction of the displacement
(ωeff ) becomes equal to that of the bare frame (ωf ). Consequently, spectrum of the JR Takatori record in the X-loading direction in the
β eff becomes the same as the one proposed by Ramirez et al. (2001) range of effective periods under consideration is less sensitive to sup-
for frame buildings with fluid viscous devices plemental damping compared to that in the Y-loading direction. This
is shown in Figs. 4(c and d).
ψK d0 0
β eff ¼ β 0 þ ð5Þ For higher ground motion intensities (i.e., 100% of the JR
2πK f Takatori record), the relative error of the computed EDPs of interest
based on the LSP1 method is in the range of 20–40% [Figs. 10(d)
Eq. (5), which refers to LSP1, is identical to Eq. (14.33) from
and 11(d)]. However, the same error drops to about 10% or less
ASCE 41-13 (Section 14.3.4.1.2, ASCE/SEI 2014) for frame struc-
when the LSP2 method is employed. This is attributed to the con-
tures with fluid viscous devices.
sideration of the axial flexibility of NVDs as part of the LSP2
Figs. 10 and 11 show the relative error of LSP1 and LSP2 meth-
method. This issue can be further explained based on Fig. 12(b)
ods in predicting the peak roof displacements and base shear, re-
that shows the effective damping ratio and period of the equivalent
spectively, of the test structure with BRBs and NVDs for the 40 and
linear SDF oscillator based on the LSP1 and LSP2 methods
100% intensities of the JR Takatori record. From these figures and
for the test structure with BRBs, LSP1 underestimates the two with respect to the K d0 0 =K f ratio in the Y-loading direction. The
EDPs of interest compared to the experimental data by about curves representing the LSP2 method are constructed for the case
20–30% regardless of the loading direction and ground motion in- of p ¼ 0.54, where p is computed as defined in Eq. (1). However,
tensity [Figs. 10(a and b) and 11(a and b)]. in this case K d is the lateral stiffness of NVDs within the equivalent
When the LSP2 method is employed the relative error of the SDF system. Based on the results of LSP2 method in Y-loading
predicted EDPs is in the order of 10% or less (Figs. 10 and 11). direction, note that K d0 0 =K f ¼ 0.84 and 0.41 at 40 and 100% inten-
This notable difference is primarily attributed to the effective damp- sities of the JR Takatori record, respectively. The larger the roof
ing ratio formulation [Eqs. (1) and (2)] and to a lesser extent to displacement of the test structure the smaller the expected
higher mode effects that are not captured by either one of the lin- K d0 0 =K f ratio. This is attributed to the fact that K d0 0 =K f is inversely
earization methods discussed in this paper. In order to illustrate this proportional to the damper axial displacement in case that the
issue more clearly, Fig. 12(a) shows the effective damping ratio, velocity exponent α of the NVD becomes smaller than unit. From
β eff and period T eff variation with respect to displacement-based Fig. 12(b), it is clear that based on the LSP1 method β eff is over-
ductility, μd , as computed by LSP1 and LSP2 methods. From this estimated by more than two times compared to that based on the
figure, for the corresponding displacement-based ductility levels LSP2 method. According to ASCE-41-13 (Section 14.3.4) the
for the test structure with BRBs in the Y-loading direction at LSP1 method should not be used if the computed β eff > 30%. This
40% (i.e., μd ¼ 1.9) and 100% (i.e., μd ¼ 3.6) intensities of the is the case for both loading directions of the test structure with
JR Takatori record when the LSP1 method is employed β eff is over- NVDs. Furthermore, from Fig. 12(b), the LSP2 method is able
predicted by at least a factor of two compared to that from the LSP2 to capture the softening of the T eff of the equivalent linear SDF
method. This indicates that ASCE-41-13 (ASCE/SEI 2014) and system with respect to the bare frame only while the ground motion
FEMA 356 (FEMA 2000) effective damping ratio formulation intensity increases. The LSP1 method neglects this issue.
(i.e., 0.9% roof drift). Because the peak roof displacement of displacement estimate is as follows:
the test structure with dampers installed at various excitations
was always less than 140 mm for all practical purposes the test β eff ¼ β 0 þ Aðμ − 1Þ2 þ Bðμ − 1Þ3 ;
structure did not yield during the entire testing program (Kasai et al.
T eff ¼ T e ½Gðμ − 1Þ2 þ Hðμ − 1Þ3 þ 1 ð6Þ
2008). Therefore, a NSP is conducted for the test structure with
BRBs only. In Fig. 13 the authors have superimposed the base where T e = elastic period of the idealized bilinear SDF oscillator;
shear-roof displacement relation of the test structure with BRBs μ = displacement-based ductility demand computed as the ratio
based on its first mode lateral load pattern per loading direction. of the maximum-to-yield displacement of the idealized bilinear
From this figure, the BRBs yield well before the test structure does capacity curve; and A, B, G, H = constants that depend on the hys-
(i.e., around 40 mm roof displacement); thus, unlike with what teretic type. Because the postyield stiffness ratio of the idealized
ASCE-41-13 suggests (i.e., bilinear approximation), the pushover bilinear curve of the equivalent SDF system is 0.20, based on
curve of a steel frame building with BRBs should have been ideal- FEMA 440 (Tables 6.1 and 6.2), A ¼ 4.6, B ¼ −0.99, G ¼ 0.10
ized as a trilinear curve. This agrees with recommendations by and H ¼ −0.015.
Ramirez et al. (2001). In this paper this assumption is considered Approach 3: Equivalent SDF. The third option for computing a
as an alternative evaluation procedure (see Approach 3). In order to target roof displacement of a frame building with BRBs is a trilin-
compute the target roof displacement to conduct a NSP three differ- ear equivalent SDF system that is subjected to the ground motion of
ent approaches are evaluated. These approaches are summarized as interest. Nonlinear response history analysis is conducted in this
follows: case. To facilitate these computations a versatile nonlinear SDF
Approach 1: ASCE-41 Method 1. This approach refers to the analysis tool called IIIDAP (Lignos 2010) is employed. For the case
coefficient method for calculating the target displacement of a of the test structure with BRBs such trilinear idealization is shown
frame building in accordance with Section C3.3.3.3 of FEMA in Fig. 13.
274 (FEMA 1997) and FEMA 356 (FEMA 2000) based on a bi- Figs. 10(a and b) and 11(a and b) illustrate the relative error in
linear idealization of the test structure with BRBs. It is assumed that predicting the peak roof displacement and peak base shear, respec-
BRBs are treated as displacement-dependent devices. Because the tively, of the test structure with BRBs per loading direction and
use of BRBs results in substantially small displacement demands ground motion intensity based on the three approaches previously
Fig. 13. Base shear-roof displacement relation of the test structure with/without BRBs based on its first mode lateral load pattern per loading direction
Fig. 14. Equivalent SDF system for frame buildings with supplemental damping devices
K d and Fdy values, a triangular mode shape is assumed instead of a to Figs. 15(a and b), μd ¼ 2.9 refers to the displacement-based duc-
shear mode of the bare frame as suggested by Guo and Christopou- tility obtained for the test structure with BRBs in the Y-loading
los (2013a). In order to construct a P-Spectra, μd is computed as direction. For comparison purposes with other simplified assess-
the peak-to-yield displacement ratio obtained from NRHA of the ment methods, the relative error of the predicted peak displacement
equivalent SDF with BRBs. The yield displacement, udy of the and base shear compared to the experimental data is also shown in
equivalent SDF system is equal to Fdy =K d depending on the load- Figs. 10(a and b) and 11(a and b) for the 40 and 100% intensities of
ing direction of interest. For the test structure with BRBs the fol- the JR Takatori record, respectively. From these figures it is evident
lowing parameters can be identified per loading directions: K d;x ¼ that a P-Spectra approach generally yields to smaller relative errors
38.1, K d;y ¼ 37.8 KN=mm, Fdy;x ¼ 1,062, Fdy;y ¼ 1,059 KN, in terms of relevant EDPs of interest compared to most of the other
respectively. simplified assessment approaches regardless of the ground motion
The P-Spectra can be constructed for both loading directions of intensity and loading direction of interest. With the use of P-Spectra
the test structure with BRBs for the 100% JR Takatori record as a structural engineer is able to obtain potential residual deforma-
shown in Figs. 15(a and b), respectively. The damper stiffness, tions in any loading direction of the test structure. As expected
K d was kept constant while the BRB yield force, Fdy was varied these are nearly zero as shown in Figs. 15(a and b). For μd > 15,
in order to check how the test structure’s performance changes for the expected residual roof displacements of the test structure with
Fig. 15. Performance spectra for test structure with BRBs and NVDs for the 100% JR Takatori record
without dampers (i.e., K f , Fy , rf ); and (2) a nonlinear Maxwell and d) illustrate the predicted response of these two EDPs of the test
model that represents the NVDs in parallel with the elastoplastic structure with NVDs for the 40% and 100% JR Takatori record in
spring. From Fig. 14(b), the damper coefficient Cd of the equivalent both loading directions (noted as P-Spectra1). In all cases, the effect
SDF system can be derived as follows: of axial flexibility of the NVDs on the predicted EDP of interest can
be critical if the response of the building under consideration is
X
N
sensitive to the increase of the corresponding NVD damping coef-
Cd ¼ Γ1þα
1 Cdi ðϕri cos θi Þ1þα ð8Þ
ficient. For example, in the Y-loading direction, for low ground
i¼1
motion intensities (e.g., 40%) the relative error in the prediction
where Cdi are the damper coefficients of the NVD installed in story of both the peak roof displacement and base shear drops from about
i. The equivalent stiffness K d of the Maxwell model can be sim- 65–70% to about 9–17% [Figs. 10(c) and 11(c)]. For higher ground
ilarly found based on Eq. (7). Note that Cd in Eq. (8) represents the motion intensities (e.g., 100%) the relative errors in the prediction
damping coefficient of the equivalent SDF system. Therefore, sim- of the same EDPs become smaller but they are still in the range of
ilar to the BRB case, the roof displacement in the P-Spectra plots 50% different [Figs. 10(d) and 11(d)]. According to ASCE 41-13
are computed by multiplying the peak displacement of the equiv- (ASCE/SEI 2014) Section C1. 14.3.3.2.3, linear fluid viscous
alent SDF system obtained from NRHA with Γ1 . For frame build- dampers exhibiting stiffness in the frequency range 0.5f 1 to
ings with NVDs the fundamental mode shape of the bare frame is 2.0f1 (f1 : fundamental frequency of the rehabilitated building)
considered as suggested by (Guo and Christopoulos 2013a). For the shall be modeled as fluid viscoelastic devices. This may be valid
test structure with NVDs, Cd is found to be 79.6 KN=ðmm=sÞ0.38 for linear viscous dampers. However, if nonlinear fluid viscous
regardless of the loading direction. Similarly, K d;x ¼ 35.6 kN=mm dampers with K d =K f ratios smaller than 10 are modeled as fluid
and K d;y ¼ 30.5 kN=mm per loading direction. There is a percep- viscous devices (i.e., K d =K f ¼ ∞) the damper displacement de-
tion that the axial flexibility of NVDs may be ignored when the mands may be significantly underestimated (Kasai et al. 2007).
predicted response of frame buildings equipped with NVDs is com- The numerical models that have been developed for the test
puted with simplified assessment techniques. In order to evaluate structure (i.e., bare frame structure without dampers) as part of
this assumption, two SDF oscillators are developed. One considers the present work have been validated with shake table experiments
the axial flexibility of NVDs and the other does not. of the test structure (i.e., no dampers installed) as well as other steel
Figs. 15(c and d) show the P-Spectra for the test structure with frame structures tested at the E-Defense facility in previous years
NVDs in the X and Y-loading directions, respectively, under the (e.g., Lignos 2012; Lignos et al. 2013). Therefore, the authors have
100% JR Takatori record. In order to construct the P-Spectra in this confidence on the performance of the developed numerical models
case the equivalent SDF oscillator with the inclusion of the axial discussed in this paper once the test structure becomes nonlinear. In
flexibility of the NVD is considered. The Cd value of the NVD is the same manner, in case that the main frame behaves in its non-
varied unlike the axial stiffness of the NVD, K d that is considered linear range then the effective damping and period formulations in
to be constant (i.e., K d;x ¼ 35.6, K d;y ¼ 30.5 kN=mm). From linearization methods should be adjusted accordingly. Prior studies
Figs. 15(c and d), the response of the test structure with NVDs (e.g., Ramirez et al. 2001; Guo and Christopoulos 2013a, b) pro-
under the 100% JR Takatori record in the Y-loading direction is vide guidance related to this scenario for buildings with BRBs and
more sensitive to the increase of the corresponding NVD damping NVDs. However, it should be pointed out that in the aforemen-
coefficient than that in the X-loading direction. From Figs. 15(c and tioned studies the flexibility of NVDs was neglected. In cases that
d), if a Cd ¼ 79.6 kN=ðmm=sÞ0.38 is selected the expected mini- inelastic deformations concentrate in a single story of the frame
mum peak base shear and peak roof displacement of the test struc- building then simplified methods are not able to capture the ex-
ture with NVDs can be computed for both loading directions. The pected building performance when traditional mode shape assump-
relative error of such predictions compared to the experimental data tions are employed.
for the 40 and 100% of the JR Takatori record is shown in Figs. 10(c
and d) and 11(c and d) for the peak roof displacement and base
shear, respectively (noted as P-Spectra2). From Figs. 10(c and d), Conclusions
for the selected Cd value, the relative error of the predicted peak
roof displacement compared to the experimental data is comparable This paper summarizes a comprehensive evaluation of simplified
with that of the LSP2 and in average smaller than that of the other and state-of-the-art methods for the performance evaluation of steel
simplified assessment techniques. Both the LSP2 and P-Spectra2 frame buildings equipped with buckling-restrained braces (BRBs)
consider the axial flexibility of NVDs. The same observation holds and nonlinear viscous dampers (NVDs). The assessment is based
true for the peak base shear predictions based on the P-Spectra2 on unique experimental data from a full-scale shake table test of a
approach for the 40% JR Takatori record [Fig. 11(c)]. For the five-story steel frame building equipped with BRBs and NVDs
concentrate their inelastic behavior into the BRBs or NVDs do 360-10, Chicago.
not require the utilization of sophisticated nonlinear modeling Akcelyan, S., Lignos, D. G., and Hikino, T. (2015). “Adaptive numerical
approaches of the lateral load resisting system for computing method algorithms for nonlinear viscous and bilinear oil damper models
various EDPs of interest for structural and nonstructural damage under random vibrations.” J. Eng. Mech., in press.
control. Emphasis should be placed on the nonlinear modeling ASCE. (2010). “Minimum design loads for buildings and other structures.”
of the employed supplemental damping device. ASCE/SEI 7-10, Reston, VA.
• The current effective damping ratio formulation that is recom- ASCE/SEI (Structural Engineering Institute). (2014). “Seismic evaluation
mended by FEMA 356 (FEMA 2000) and ASCE-41-13 (ASCE/ and retrofit of existing buildings.” ASCE/SEI 41-13, Reston, VA.
SEI 2014) to be used in linear static procedures (LSPs) for BCJ (Building Center of Japan). (2008). “Seismic provisions for design of
evaluating EDPs in frame buildings with BRBs and NVDs building structures.” Tokyo.
may not be conservative enough. It is advisable that such for- Buchanan, A., Bull, D., Dhakal, R., MacRae, G., Palermo, A., and
mulation should consider the displacement-based ductility de- Pampanin, S. (2011). “Base isolation and damage-resistant technologies
mand variation of BRBs in the case of frame buildings with for improved seismic performance of buildings: A report written for
the Royal Commission of Inquiry into building failure caused by the
BRBs. The equivalent formulation for frame buildings with
Canterbury earthquakes.” Research Rep. 2011-02, Dept. of Civil and
NVDs should consider the axial flexibility of the dampers. Natural Resources Engineering, Univ. of Canterbury, Christchurch, U.K.
This is consistent with linear static procedures that are cur- Chopra, A. K., and Goel, R. K. (2001). “A modal pushover analysis
rently being used in the Japanese design practice (Kasai and procedure for estimating seismic demands for buildings.” Pacific
Ito 2005; Kasai et al. 2007). Earthquake Engineering Research Center, Berkeley, CA.
• Currently employed nonlinear static procedures (NSP) as Christopoulos, C., and Filiatrault, A. (2006). Principles of supplemental
discussed in ASCE-41-13 (ASCE/SEI 2014) and FEMA 356/ damping and seismic isolation, IUSS Press, Milan, Italy.
FEMA 440 (FEMA 2000) may overestimate the predicted Cimellaro, G. P., Reinhorn, A. M., and Bruneau, M. (2010). “Framework
roof displacement demands in frame buildings with BRBs for for analytical quantification of disaster resilience.” Eng. Struct., 32(11),
moderate and extreme ground motion intensities associated 3639–3649.
with design level and maximum considered earthquakes in Constantinou, M. C., and Symans, M. D. (1993). “Seismic response of the
North America. This is in part attributed (1) to the idealized structures with supplemental damping.” Struct. Des. Tall Build., 2(2),
bilinear approximation of the base shear-roof displacement 77–92.
relation of frame buildings with BRBs and (2) the effective Dao, N. D., Ryan, K. L., Sato, E., and Sasaki, T. (2013). “Predicting the
displacement of triple pendulum™ bearings in a full-scale shaking
damping formulations proposed in FEMA 273/356/440 (FEMA
experiment using a three-dimensional element.” Earthquake Eng. Struct.
1997, 2000, 2005). A simple trilinear equivalent SDF oscillator
Dyn., 42(11), 1677–1695.
with a versatile nonlinear SDF analysis tool retains the simplicity Dhakal, R. P., MacRae, G. A., and Hogg, K. (2011). “Performance of ceil-
of NSP procedures for evaluating frame buildings with BRBs; in ings in the February 2011 Christchurch earthquake.” Bull. N. Z. Soc.
this case, based on the available full-scale experimental data the Earthquake Eng., 44(4), 379–389.
relative errors in the respective EDPs of interest are in the range Dormand, J. R., and Prince, P. J. (1980). “A family of embedded Runge-
of 10% or less regardless of the ground motion intensity and Kutta formulae.” J. Comput. Appl. Math., 6(1), 19–26.
loading direction of interest. E-Defense. (2008). “2008 blind analysis contest by E-Defense.” Hyogo
• When a performance spectra (i.e., P-Spectra) approach is em- Earthquake Engineering Research Center, 〈http://www.bosai.go.jp/
ployed to validate the efficiency of various damper solutions hyogo/blind-analysis/2008/eng/outline/outline.html〉 (Dec. 5, 2008).
on the dynamic response of the test structure, the relative errors Fahnestock, L., Ricles, J., and Sause, R. (2007). “Experimental evaluation
in predicting both the peak roof displacement and base shear of of a large-scale buckling-restrained braced frame.” J. Struct. Eng.,
the test structure with BRBs are in general smaller than those 10.1061/(ASCE)0733-9445(2007)133:9(1205), 1205–1214.
from currently employed nonlinear static approaches. The same FEMA. (1997). “NEHRP provisions for the seismic rehabilitation of build-
conclusion holds true for the test structure with NVDs when the ings.” FEMA 273 (Guidelines) and 274 (Commentary), Washington, DC.
FEMA. (2000). “Prestandard and commentary for the seismic rehabilitation
axial flexibility of such dampers is considered for the construc-
of buildings.” FEMA-356, Washington, DC.
tion of P-Spectra. This is also confirmed with rigorous nonlinear
FEMA. (2005). “Improvement of nonlinear static seismic analysis proce-
response history analysis with a 3-dimensional representation of dures.” FEMA-440, Washington, DC.
the test structure with NVDs. FEMA. (2012). “Seismic performance assessment of buildings.” FEMA
• A P-Spectra approach is able to provide a more realistic repre- P-58-1, Washington, DC.
sentation of the structural system performance including esti- Folz, B., and Filiatrault, A. (2001). “SAWS—Version 1.0, A computer
mates of residual deformations. Traditional static approaches program for the seismic analysis of woodframe structures.” Structural
are not able to predict such EDP, which is essential for structural Systems Research Project: SSRP-2001/09, Dept. of Structural Engi-
and nonstructural damage control. neering, Univ. of California, San Diego.