Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: www.elsevier.com/locate/soildyn

An energy-based method for seismic retrofit of existing frames

using hysteretic dampers
Amadeo Benavent-Climent n
Department of Structural Mechanics, University of Granada, Edificio Politécnico Fuentenueva, Granada, Spain

a r t i c l e i n f o abstract

Article history: This paper proposes a method for the seismic retrofitting of existing frames by adding hysteretic energy
Received 12 December 2010 dissipating devices (EDDs). The procedure is based on the energy balance of the structure, and it is used
Received in revised form to determine the lateral strength, the lateral stiffness and the energy dissipation capacity of the EDDs
14 May 2011
needed in each story to achieve prescribed target performance levels for a given earthquake hazard. The
Accepted 16 May 2011
Available online 14 June 2011
performance levels are governed by the maximum lateral displacement. The earthquake hazard is
characterized in terms of input energy and several seismological parameters, and further takes into
account the proximity of the earthquake to the source. The proposed method deals with the effect of
the EDDs explicitly in terms of hysteretic energy, bypassing equivalent viscous damping approxima-
tions, and directly quantifies the cumulative damage induced in the EDDs. The validity of the method is
assessed numerically through nonlinear dynamic response analyses with near-fault and far-field
ground motions, as well as experimentally through dynamic shaking table tests.
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction requirements of the EDDs (strength, stiffness, ultimate energy

Energy dissipating devices (EDDs), also known as dampers, are
giving rise to technology capable of minimizing interstory drift
and increasing the earthquake resistance of buildings, to achieve
performance-based seismic design. EDDs have been used for a
decade, and continue to attract attention in the field of earth-
quake engineering. A main application, particularly in Europe, is
in the seismic rehabilitation of buildings. Mechanisms studied
and used for passive energy dissipation include metal yielding,
phase transformation of metals, friction sliding, fluid orificing,
and deformation of viscoelastic solids or liquids. The EDDs based
on the yielding of metals—commonly known as hysteretic
dampers—are among the most popular. The present study focuses
on the application of hysteretic EEDs to the seismic retrofit of
buildings. In order to develop cost-effective retrofit solutions, it is
of paramount importance to have practical and sufficiently
accurate design procedures, able to define the EDDs that satisfy
target building performance levels for a given earthquake hazard.
There are important differences between new construction
design and seismic retrofit design, using EDDs. When EDDs are
used in new structures, the main frame can be designed to
contribute to dissipating, through plastic deformations, a portion
of the total energy input by the earthquake. This can relax the
n
Tel./fax: þ34 918020974.
E-mail address: benavent@ugr.es
dissipation capacity) and lower their cost. In contrast, most
existing structures were built according to past seismic codes
and their energy dissipation capacity, if any, is very limited. The
seismic retrofit design of such structures can therefore not rely on
the contribution of the main frame to dissipate energy; the design
of the dampers must guarantee that, under the design earthquake,
the main structure remains ‘‘basically’’ within the elastic range. In
new construction design of frames with EDD, the members can be
easily dimensioned to accommodate the forces transferred by the
EDDs, while in seismic retrofit design the strength, the stiffness
and the location of the EDDs must be carefully studied in order to
control torsion effects and to avoid unexpected failures in the
main frame before the dampers attain their maximum strength.
The design of new lateral force-resisting elements consisting of
EDDs for the seismic retrofit of existing structures must consider
important issues such as: (i) the connection of the EDDs with the
existing frame; (ii) the compatibility of deformation with the
existing lateral force-resisting or gravity load-carrying system;
(iii) the extent to which the new system relieves the existing
structure of load or deformation at all levels; (iv) the significance
of the mass added by the EDDs, and (v) the need for extensive
new foundations.
The first guidelines for the application of hysteretic EDDs to
building rehabilitation are the standards developed by the Federal
Emergency Management Agency (FEMA) in Reports 273, 274 [1]
and 356 [2], which contain both linear and nonlinear static
procedures. In the linear static procedures, the energy dissipation

0267-7261/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2011.05.015
1386 A. Benavent-Climent / Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396
afforded by the EDDs is accounted for indirectly—by reducing the
design earthquake loads or the response spectrum through a
factor that depends on an equivalent effective viscous damping.
In the nonlinear static procedures (pushover analysis), the effect
of the hysteretic EDDs is either recognized indirectly, by the
increase in building stiffness afforded by such devices, or directly,
by means of an equivalent effective viscous damping similar to
that used in linear static procedures. The linear static procedures
oversimplify the highly nonlinear random response of the entire
building-device structure, however, and their use is subjected to
severe restrictions (e.g. the effective damping afforded by the EDD
cannot exceed 30% of critical damping in the fundamental mode).
The nonlinear static pushover procedures likewise have deficien-
cies and theoretical inconsistencies [3,4]. One is that they replace
the hysteretic damping provided by the EDDs with an equivalent
viscous damping, although there is no physical principle that
justifies the existence of a stable relationship between hysteretic
energy dissipation of the maximum displacement excursion
and equivalent viscous damping, particularly for highly inelastic
systems. Moreover, the current design philosophy—basically
force-based and displacement-based procedures—considers cyc-
lic degradation only in an implicit manner; it is difficult to include
the effect of duration-related cumulative damage [5], which can
lead to unsatisfactory performance at levels associated with
repairable damage and collapse prevention [6].
In the literature, many experimental and theoretical studies
have focused on EDDs and buildings equipped with such devices.
Ruiz et al. [7] suggested determining the strength of the EDDs
using dynamic step-by-step analyses in a trial and error proce-
dure to attain minimum damage to the structure and maximum
nonlinear behavior in the EDDs. Gluck et al. [8] proposed a
method of designing supplemental EDDs based on optimal linear
control theory. Rivakov and Gluck [9] proposed replacing the
hysteretic EDDs with equivalent viscous dampers that dissipate
the same energy per cycle, and applying optimal control theory to
determine the levels of viscous damping in each story. Uetani
et al. [10] developed a computer program based on optimization
techniques to determine the required story stiffness of a building
with EDDs so that it would exhibit a specified distribution of
interstory drifts. Karami et al. [11] proposed a method based on
iterative time-history response analyses. Choi and Kim [12]
developed an energy-based design method for a particular type
of EDD—the buckling-restrained brace—that involves performing
series of time-history analyses. Lin et al. [13] developed a direct-
displacement-based procedure for nonlinear viscous dampers.
Other interesting proposals have been made in recent years with
respect to design methodologies for metallic hysteretic dampers
[14–16], several of which entail deformation-based design pro-
cedures [17–19].
Over the past few decades, the seismic design methodology
based on energy first proposed by Housner [20] has been largely
accepted [21–24] and has been recently introduced in the Japanese
seismic code [25]. This paper proposes an energy-based procedure
for the seismic upgrading of existing structures by adding hystere-
tic EDDs. The procedure is grounded on recent research on the
cumulative demand of earthquake ground motions [26]. The
intention is to develop a practical design method to be used as a
tool for conventional low- to medium-rise structures, without
resorting to nonlinear time history analyses. The method deter-
mines the lateral stiffness, the lateral strength and the energy
dissipation capacity that must be provided by the dampers
installed in each story, so that the main existing frame does not
exceed a limit of drift beyond which plastic deformations (i.e.
structural damage) would take place. Once the required strength,
stiffness and energy dissipation capacity of the dampers are
determined, an appropriate EDD and a solution for connecting
the dampers to the main frame that satisfies these requirements
must be selected. This second part must take into account the
performance of the connection, which is an important issue, but
lies outside the scope of this paper. The connection of the damper
is a big problem in the seismic retrofit design. When RC frame
structures are strengthened by using the brace-type dampers, the
stress concentration in the connection between the RC beam-
column joints and the damper devices must be carefully studied
to avoid damage on the concrete. Problems can also arise from the
need of tension-type connections between the damper and the
frame that require drilling for anchorage of tension/shear connec-
tions. The connection details must be designed to satisfy the
requirement of rigid connection; that is, neither uplift, sliding nor
damage occurs under a design working force larger than the
maximum force expected on the dampers. It is very important to
avoid relative displacements between the main frame and the end
of the dampers at the connection, especially in stiff structures. This
is a point of paramount importance so that the damper can
dissipate energy at small lateral drifts of the main frame, and thus
protect efficiently the construction from structural and non-struc-
tural damage.
2. Background on energy-based design
The equation of motion of an inelastic single-degree-of-free-
dom system (SDOF) subjected to a unidirectional horizontal
ground motion can be written as follows:
My€ þC y_ þ Q ðyÞ ¼ M z€ g , ð1Þ
where M is the mass, C is the damping coefficient, Q(y) is the
restoring force, y is the relative displacement, y_ and y€ its first and
second derivates with respect to time, and z€ g is the ground
acceleration. Multiplying (1) by dy ¼ ydt _ and integrating over
the entire duration of the earthquake, i.e. from t ¼0 to t ¼to, the
energy balance equation becomes
Wk þ Wx þ Ws ¼ E: ð2Þ
R
_ ydt
In the left-hand-side term, Wk ¼ yM € is the kinetic energy,
R R
Wx ¼ C y_ 2 dt is the damping energy, and Ws ¼ Q ðyÞydt_ is the
absorbed energy, which is composed of the recoverable elastic
strain energy, Wse, and the irrecoverable plastic energy, Wp, i.e.
R
Ws ¼Wse þWp. The right-hand-side term, E ¼ Mz€ g ydt _ is, by
definition, the input energy, which can be expressed in the form
of an equivalent velocity VE as
rffiffiffiffiffiffi
2E
VE ¼ : ð3Þ
M
Since Wk þWse is the elastic vibrational energy, We, the Eq. (2)
can be rewritten as
We þ Wp ¼ EWx : ð4Þ
Further, We þ Wp can also be expressed in the form of an
equivalent velocity VD so that
MVD2
We þ Wp ¼ : ð5Þ
2
Eq. (4) holds also for a multi-degree-of-freedom system (MDOF)
subjected to a unidirectional horizontal ground motion if the
above expression for Wk, Wx, Ws and E is replaced by Wk ¼
R T R R R
y_ Mydt,
€ Wx ¼ y_ T Cydt,
_ Ws ¼ y_ T Q dt and E ¼  y_ T Mrz€ g dt,
respectively. Here, M is the mass matrix, C the damping matrix
€
and Q(t) the restoring force vector; yðtÞ €
and yðtÞ are the accelera-
tion and velocity vectors relative to the ground respectively; and r
represents the displacement vector yðtÞ resulting from a unit
support displacement.
 A. Benavent-Climent / Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396
On the basis of numerous response analyses, Akiyama [22]
concluded that, in general MDOF damped inelastic systems, the
total input energy supplied by the earthquake E—and conse-
quently VE—coincides approximately with that of an equivalent
elastic SDOF system with mass M equal to the total mass of the
MDOF system, and period T equal to that of its first vibration
mode. This conclusion has been validated experimentally [23]. In
the energy-based seismic design approach, the energy input
spectrum in the form of equivalent velocity VE T characterizes
the loading effect of the earthquake. Design input energy spectra
VE  T [22,27,21] have been proposed in past studies.
The cumulative damage of the structure is strongly related to
the plastic strain energy Wp. The sum of Wp and the elastic
vibrational energy We is what Housner [20] called the energy that
damages a structure subjected to seismic action. From Eqs.(2)–(5)
it follows that for undamped systems VD ¼VE; otherwise, the
difference between VD and VE is the energy dissipated by the
inherent damping of the structure. Several empirical expressions
have been proposed that allow us to obtain VD from VE
[22,27–30]. Moreover, attenuation relationships have been estab-
lished for use in energy-based seismic design [5] that directly
provide Ws—the absorbed energy—for a given earthquake mag-
nitude, source-to-site distance, site class and ductility factor, in
terms of an equivalent velocity Va defined by
rffiffiffiffiffiffiffiffiffiffi
2WS
Va ¼ : ð6Þ
M
Although VD T spectra have been recently introduced in some
building codes [25], characterization of the design earthquake in
terms of energy input spectra VD  T is not as common for
professionals as in the form of absolute acceleration spectra Sa T.
Energy input spectra in terms of VD  T can be obtained from the
Sa  T spectra on the basis of the following considerations: (i) over
the range of damping ratios exhibited by actual structures—say
less than 0.10—the spectral absolute acceleration Sa of damped
elastic SDOF systems is related to the pseudo-velocity spectral
response, Spv, by the well known expression Sa ¼ oSpv, where o is
the circular frequency. (ii) Hudson [31,32] demonstrated that,
except in the case of very long period oscillators, the spectral
relative velocity, Sv, differs very little from Spv. (iii) Akiyama [22]
showed that Sv provides a good approximation of VD, and
validated Housner’s [20] assumption that VD can be taken as
equal to Sv for the purposes of earthquake-resistant design.
3. Modeling and design criteria
The proposed method requires knowledge of the mass, mi, the
lateral yield strength, fQyi, the initial stiffness fki of each story i, and
the fundamental period T1 of the existing building (referred to as
main structure hereafter). The seismic retrofit strategy proposed
consists of adding hysteretic EDDs in each story. The existing
structure and the added EDDs are arranged so as to form a dual
system consisting of two inelastic springs connected in parallel as
shown in Fig. 1. The lateral load–displacement relationship, fQi–di,
of a given i-th story of the main structure—without EDDs—under
f Qi Pushover analysis
Idealized curve
f Qyi
0.6f Qyi
tan-1fki
i
f yi
Fig. 1. Idealized interstory drift-shear force curve of each story. (a) Existing (main) structure. (b) Energy dissipating device. (c) Dual system
1387
monotonic loading is represented by the elastic-perfectly plastic
model shown with bold lines in Fig. 1a. The lateral load–
displacement relationship, sQi–di, of a given i-th story accounting
only for the EDDs is also assumed to be of the elastic-perfectly-
plastic type and is defined by the lateral yield strength sQyi and the
initial stiffness ski as shown in Fig. 1b. The lateral shear force-
interstory drift relationship of the entire building-device structure
of a i-th story under monotonic loading, Qi  di, is obtained by
summing up the forces sustained by each element at a given
displacement level di, as shown in Fig. 1c.
The goals of the proposed procedure are: (i) to determine the
sQyi and ski of the EDDs needed in each story to achieve the
required building performance levels for a given earthquake
hazard; and (ii) to evaluate the energy dissipation demand on
the EDDs. Usually, one of the objectives of the EDDs is to avoid
inelastic deformations in the structural elements outside the
EDDs, and many researchers consider this objective as one of
the basic requirements for a system with EDDs [12,33]. Further,
many existing structures were designed according to past seismic
codes and their ductility, if any, is very limited. In the case of
reinforced concrete frames, it is worth noting that although they
do not have a large elastic range, especially if compared with steel
frames, there is a range of lateral drift within which the behavior
can be assumed to be ‘‘basically’’ elastic and the damage on the
main frame would be null or negligible. Accordingly, it is imposed
that
dmax i r f dyi , ð7Þ
where dmax i is the maximum interstory drift of the entire
building-device structure and fdyi ( ¼ fQyi/fki) is the yielding inters-
tory drift of the main structure.
The condition imposed by Eq. (7) is intended to avoid the
severe degradation effects in the response of the main structure
that may arise from stiffness degradation and deterioration as
well as shear effects in some types of structures. In structures for
which such degradation cannot be avoided, the proposed method
is not applicable. Such is the case, for example, of precast
structures with slender columns and/or existing reinforced con-
crete frame systems designed primarily for gravity loads and with
high shear degradation. Yet the proposed formulation is indeed
applicable to symmetric structures or systems with prevalent
translation modes of vibration. If the main existing structure has
important irregularities that may place extraordinary displace-
ment demands on elements due to torsional response, measures
must be adopted to mitigate torsional effects. One such measure
is to locate the EDD so that they can balance the stiffness and
make the mass and stiffness centers very close.
The lateral yield strength of the entire building-device struc-
ture at the i-th story, Qyi, is simply
Qyi ¼ s Qyi þ s dyif ki , ð8Þ
where sdyi ( ¼ sQyi /ski) is the yield deformation of the EDDs
installed in that story. The maximum lateral force sustained by
the main structure, fQmax i, is fQmax i ¼ dmax i  fki. For the building-
device structure surviving the earthquake, the plastic strain
energy accumulated in the i-th story, Wpi, must not exceed the
Qi
s Qi
Qyi
sQyi
tan-1ski i i
syi syi yi
1388 A. Benavent-Climent / Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396

ultimate energy dissipation capacity of the EDDs installed in that distribution of sai be that given by Eq. (13) (Appendix B):
story, Wui. In turn, Wpi and Wui can be expressed in the form of
Ki ðK1 þ1Þ
two non-dimensional coefficients, Zi and Zui, defined by a ¼ ai s a1
s i : ð14Þ
K1 ðKi þ1Þ
Wpi Wui
Zi ¼ ; Zui ¼ , ð9Þ Of course, instead of Eq. (13), a more refined lateral strength
s Qyi s dyi s Qyi s dyi
distribution could be used from dynamic analysis by applying a trial
thus, the above condition can be written as and error iterative method. This would not alter the proposed
method at all. However, it implies selecting a set of earthquake
Zi r Zui : ð10Þ
records, conducting many nonlinear dynamic response analyses, and
averaging the distribution derived for each record, which can be a
4. Formulation of the method very cumbersome process. Such a process should be used for systems
with a degrading type of response under earthquake loading.
For the sake of convenience, dmax i, Qyi, fQmax i, sQyi and fk1 will
be also expressed herein in non-dimensional form by the plastic
deformation ratio mi, the shear-force coefficients ai, famax i, sai, and 4.2. Lateral strength to be provided by the EDDs of the first story
the stiffness ratio w1, respectively, which are defined as follows:
Once the stiffness ratios Ki and the lateral shear-force coeffi-
ðdmaxi s dyi Þ Qyi fQmaxi
mi ¼ ; ai ¼ PN ; f amaxi ¼ PN ; cient distribution ai are determined, the lateral shear-force
s dyi k¼i mk g k¼i mk g coefficient to be provided by the EDDs of the first story, sa1, must
s Qyi
be calculated in order to obtain the required lateral shear force
f k1
a ¼ PN
s i ; w1 ¼ : ð11Þ coefficient of the EDDs in the other stories sai with Eq. (14). The
k ¼ i mk g
keq
equations that govern the sa1 required for a given seismic hazard
Here N is the total number of stories, g is the acceleration of the and building performance level are derived next by establishing
gravity, and keq is the stiffness of an equivalent SDOF system of the energy balance of the structure.
P
mass M ¼ mj and period equal to the fundamental period of the When using EDDs, the yield displacement sdyi of the EDDs is made
main structure, T1, i.e. keq ¼4p2M/T21. smaller than that of the main structure, fdyi, so that the EDDs begin
dissipating energy before the main structure might yield. Moreover,
sQyi is commonly smaller than fQyi. As a result, the elastic strain
4.1. Stiffness and strength distribution of the EDDs among the stories
energy stored by the EDDs is commonly negligible in comparison to
that of the main structure, and the elastic vibrational energy of the
The ratio between the lateral stiffness provided by the EDDs
whole building, We, can be approximated from the maximum shear
and the lateral stiffness of the main structure in each story is
force sustained by the main structure on the first story as follows
referred to as
(refer to Eq. (A.25) and the last paragraph of Appendix A)
s ki
Ki ¼ ð12Þ 2
Mg 2 T12 f amax1
f ki We ¼ : ð15Þ
4p2 2
There is no need to make Ki equal in all stories, although this
criterion has been used on occasion in the past. The lateral From Eq. (9) and taking into account the coefficients defined in
strength distribution of the entire building-device structure, Eq. (11), the plastic strain energy accumulated in the i-th story
Qyi/Qy1, can be expressed in terms of shear-force coefficients by Wpi can be expressed as follows:
ai ¼ ai =a1 . The criterion adopted in the proposed method to 2
!2
s Qyi
X
N
1
determine the ai distribution is to attain an even distribution of Wpi ¼ Zis Qyis dyi ¼ Zi ¼ Zis a2i mk g : ð16Þ
damage among the EDDs. The damage in the EDDs installed in a s ki k¼i s ki

given story i can be characterized by the non-dimensional para-

Provided that the strength distribution given by Eq. (13) is
meter Zi defined by Eq. (9). Past studies [22] showed that the
adopted, the normalized plastic strain energy Zi can be assumed
strength distribution ai that makes Zi equal in all stories (Zi ¼ Z) in
equal in all stories, i.e. Zi ¼ Z. Thus, taking into account Eq. (12)
a low-to- medium rise multi-story building subjected to seismic
and using the non-dimensional parameters sai and ai defined
loads coincides approximately with the maximum shear-force
above, the total plastic strain energy dissipated by the EDDs of the
distribution in an equivalent elastic undamped shear strut with
whole structure, Wp, can be expressed in terms of the plastic
similar lateral stiffness distribution along its height. The deriva-
strain energy dissipated by the EDDs of the first story, Wp1, by
tion of the ‘‘exact’’ shear-force coefficient distribution ai for an
introducing a new ratio g1 ¼Wp/Wp1, which is obtained as follows:
elastic undamped shear strut subjected to a design earthquake
2 !2 3
characterized by a bilinear VD  T spectrum is explained in detail
PN PN
in Appendix A. The ‘‘exact’’ solution cannot be expressed with 4Z a 2
mk g =s ki 5
is i
Wp i¼1 k¼i
simple equations, but it can be approximated for design purposes
g1 ¼ ¼
with the following expression: Wp1 Z1s a21 M2 g 2 =s k1
" ! ! # (  )
a k T k T XN X mj  ðK1 þ1Þ2 f k1 Ki
ai ¼ i ¼ exp 10:02 f 1 0:16 1 x 0:50:05 f 1 0:3 1 x2 : ¼ ai , ð17Þ
a1 f kN TG f kN TG M ðKi þ 1Þ f ki K1
i¼1
ð13Þ
thus
Here x ¼ ði1Þ=N, fkN is the lateral stiffness of the uppermost N-th
story of the main structure, whereas TG defines the change of 2
g1s Qy1 Z
slope of the VD  T bilinear spectra. TG may be called the pre- Wp ¼ g1 Wp1 ¼ g1s Qy1s dy1 Z ¼
s k1
dominant period of the ground motion [29].
Further, from the definition of sai and Ki given by Eqs. (11) g1s a21 M2 g 2 Z g a2 M2 g 2 Z g1s a21 Mg 2 ZT12
¼ ¼ 1s 1 ¼ : ð18Þ
and (12) the following relation must be satisfied so that the K1 f k1 K1 w1 keq 4p2 K1 w1
 A. Benavent-Climent / Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396 1389

Substituting Eqs. (15) and (18) in Eq. (5) gives substituting in Eq. (25) gives
" 2 #
Mg 2 T12 f amax1 g1 MVD2 m þ 1Þ
s dy1f k1 ð m þ 1Þ
s dy1s k1 ð m þ 1Þ
s Qy1 ð a m þ1Þ
s 1ð
þ 2
Za ¼ : ð19Þ f amax1 ¼ ¼ ¼ ¼ :
4p2 2 K1 w1 s 1 2 Mg K1 Mg K1 Mg K1
ð26Þ
Now, a new parameter ae is introduced. ae is defined as the
Substituting Eq. (26) in Eq. (21), recalling that m ¼ mi ¼ Z/neq and
base shear-force coefficient that the main structure should have
solving for m gives
in order to absorb by itself—i.e. without EDDs—the amount of (sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi )
 
input energy MVD2 =2 supplied by the earthquake. The expression neq g1 2 2neq g1 a2 neq g1
for ae can be obtained placing sa1 ¼0 in Eq. (19)—this implies m ¼ K1 þ þ e2  1: ð27Þ
w1 K1 w1 s a1 w1
ignoring the EDDs—and solving for famax1( ¼ ae), which gives
Noting that for the other stories m ¼(dmax i  sdyi)/sdyi, using Eqs.
2pVD
ae ¼ : ð20Þ (14) and (27), and solving for dmax i gives the equation that
gT1 predicts the maximum displacement of a given story i
Using Eqs. (20), Eq. (19) can be rewritten as follows: P
ai a ðK1 þ 1Þð Nj¼ i mj gÞ
dmaxi ¼ s 1
f a2max1 g1 a2 f ki ðKi þ 1Þ
þ Zs a21 ¼ e : ð21Þ (s ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi )
2 K1 w1 2 neq g1 2 2neq g1 a2e neq g1
 þ þ 2 : ð28Þ
The relation between Zi and mi is a key parameter in seismic w1 K1 w1 s a1 w1
design methodologies based on the energy concept, and it has
been addressed in different ways in the past [23,34,35]. Based on
the results of regression analyses performed with 128 near-fault 5. The procedure
and 122 far-field earthquake records, Manfredi et al. [26] pro-
posed the following formulae for estimating the equivalent In the proposed method, the characteristics of the existing main
number of plastic cycles neq at the maximum value of plastic structure—without EDDs—to be retrofitted (i.e. mi, fki, fdyi and T1)
excursion that a SDOF system of mass m, elastic period T and are assumed to be known data. mi, fki, fdyi and T1 can be estimated
yielding force Fy must develop in order to dissipate the total by using approximate formulae [22], or by creating a finite element
amount of hysteretic energy input by the earthquake based model and performing a pushover analysis using a triangular
rffiffiffiffiffiffiffiffi lateral load distribution. In the latter option, a fQi–di curve is
TNH obtained for each story—the dotted line in Fig. 1a—from which
neq ¼ 1 þ c1 Id ðR1Þc2 : ð22Þ
T
fki and fdyi can be estimated by using the secant stiffness at 60% of

Here TNH is the initial period of medium period region in the
Newmark and Hall [36] spectral representation. R is the reduction
factor defined as R ¼mSa/Fy where Sa is the elastic spectral
acceleration. Id is a seismological parameter [34] defined by
R to 2
€
0 z g dt
Id ¼ ð23Þ
PGA  PGV
where PGA and PGV are the peak ground acceleration and velocity,
respectively. In Eq. (22), Manfredi et al. [26] proposed to take
c1 ¼0.23, c2 ¼ 0.4 for near-fault earthquakes; and c1 ¼0.18, c2 ¼ 0.6
for far-field earthquakes. In order to apply the Manfredi et al.
equation to the proposed method, the multi-story structure is
assimilated to an equivalent SDOF system with elastic period
T¼T1, mass m¼M and Fy ¼ sQy1 þ fk1  sdy1. Based on this equiva-
lence, while taking into account that Sa is approximately equal to
(2p/T)Sv, and that the elastic spectral velocity Sv coincides
approximately with VD [5,22], Eq. (22) is rewritten as
sffiffiffiffiffiffiffiffi c2
TNH K1 ae
neq ¼ 1 þ c1 Id 1 : ð24Þ
T1 ðK1 þ 1Þs a1
For the EDDs with elastic-perfectly-plastic characteristics
dealt with in this study, neq is, by definition [26]: neq ¼Wpi/
[sQyi(dmax i  sdyi)], which coincides with Zi/mi. In the proposed
method, the same neq( ¼ Zi/mi)—given by Eq. (24)—is adopted for
all stories. Since Zi was also assumed as constant, Zi ¼ Z, the
maximum plastic deformation ratio mi has the same value mi ¼ m
( ¼ Z/neq) in all stories. On the other hand, since the condition
given by Eq. (7) was adopted, the maximum base shear-force
coefficient of the main structure famax1 is
dmax1 f k1
the yield strength, as suggested by FEMA 273 [1], and T1 is
obtained from an eigenvalue analysis. The goal of the proposed
method is to determine the lateral stiffness ski, lateral strength sQyi,
and the normalized energy dissipation demand Z of the EDDs to be
installed in each story, so that the entire building-device structure
satisfies predetermined performance levels defined by the max-
imum allowed interstory drift, dallow i, for a given earthquake
hazard characterized by VD, TG, TNH, Id and the proximity to the
source. The basic steps involved in the procedure are as follows.
Step 1: Characterize the design earthquake in terms of a
bilinear VD  T spectra defined by the maximum demand VDmax
and the predominant period TG—i.e. VD ¼TVDmax/TG for ToTG,
VD ¼VDmax for TZTG—and the values of the seismological
parameters Id, TNH, c1 and c2.
Step 2: Prescribe the maximum interstory drift allowed in each
story, dallow i, in accordance with the acceptance criteria for
building components at the target building rehabilitation
performance level.
Step 3: Calculate ai for each story i with Eq. (13), ae with
Eq. (20) and w1 with Eq. (11).
Step 4: Choose a set of values for Ki, and compute g1 with
Eq. (17). From i¼1 to i ¼N proceed for each story as follows.
Starting with sa1 ¼0, iterate in Eq. (28)—with neq given by
Eq. (24)—increasing the values of sa1 until the predicted dmax i
gets close to dallow i within an acceptable tolerance (for
example, 5% of dallow i). In these iterations, sa1 shall not be
larger than the value given by the following expression, so that
sdyi r fdyi (see Appendix C)
f dyif ki K1 ðKi þ1Þ
s 1r
a : ð29Þ
P
N

f amax1 ¼ : ð25Þ ai ðK1 þ 1Þ mk g

Mg k¼i

From the definition of mi ( ¼ m)—Eq. (11)—particularized for If in a given story i it is not possible to find a sa1 that makes
the first story, it is obtained that dmax1 ¼ sdy1(m þ1), and dmax i close enough to dallow i, restart step 4 with larger values
1390 A. Benavent-Climent / Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396

for Ki. Once the appropriate sa1 is obtained, keep this value as

1st design earthquake 2nd design earthquake

sa1i ¼ sa1 and proceed with the next story. The parameter sa1i

0.04
0.04
0.03
0.03
0.03
0.02
0.02
0.02
0.02
s i
represents the shear-force coefficient required for the EDDs of

a
the first story so that the maximum interstory drift at the i-th

K¼ 4, Z ¼ 35
story does not exceed the allowed limit dallow i.

(kN/cm)
(far-field)
Step 5: Select the maximum of the sa1i, i.e. sa1max ¼max{sa1i},
which gives the required lateral strength for the EDDs of the

324
276
284
372
372
372

376
424
404
s ki
first story. Obtain the lateral strength required in the other
stories, sai, by making sa1 ¼ sa1max in Eq. (14). Calculate the

0.18
0.14
0.12
0.10
0.08
0.08
0.07
0.07
0.06
Design of the EDDs
lateral stiffness ski and the lateral strength sQyi required for the

a s i
EDDs of each story taking into account that ski ¼Ki fki and
P
s Qyi ¼ s ai

(near-fault)
mk g.

(kN/cm)

K ¼7, Z ¼5
Once the lateral stiffness ski and the lateral strength sQyi of the

567
483
497
651
651
651

658
742
707
s ki
EDDs to be installed in each story are determined, an appropriate

d (cm) (%)
type of hysteretic damper must be chosen. To this end, it is

(0.62)

(0.68)

(0.71)
(0.91)
(0.89)
(0.52)
(0.73)

(0.75)
(0.73)
necessary to check that the normalized ultimate energy dissipa-
tion capacity of the damper Zui is larger than the demand Zi ( ¼ Z)

1.72

1.88

1.95
2.51
2.46
1.95
2.01

2.05
2.01
f yi
(without EDDs)
9-Story frame
as indicated by Eq.(10). Z is simply calculated by making

Main frame

(kN/cm)
sa1 ¼ sa1max in Eq. (24) to obtain neq, substituting this neq and

T1 ¼ 1.02 s
sa1 ¼ sa1max in Eq. (27) to calculate m ( ¼ mi), and recalling that

101

106
Z ¼neqm. The estimation of Zui for a given type of hysteretic

81
69
71
93
93
93

94
f ki
damper is beyond the scope of this paper; yet a procedure is

2nd design earthquake

proposed by Benavent-Climent [37].

0.06
0.05
0.04
0.04
0.03
0.03
a s i
Finally, it is worth emphasizing that under earthquake excita-
tions, structures undergo torsional as well as lateral motions if

K¼3, Z ¼ 22
(kN/cm)
their mass and stiffness centers do not coincide. The coupling

(far-field)
between lateral and torsional motions can significantly increase
the lateral displacements as compared to those experienced by

246

222
288
276
210

270
ski
the same structure with a symmetric plan. Existing structures
Design of the EDDs

0.17
0.14
0.11
0.09
0.08
0.08
typically exhibit irregularities that may place extraordinary dis-

a s i
placement demands on elements due to torsional response. In the

(near-fault)
earthquake

(kN/cm)

K¼ 8, Z ¼ 8
1st design
proposed method, it is assumed that the EDD are appropriately
placed to balance the stiffness, so that the mass and stiffness

656

592
768
736
560

720
centers of each story are very close and torsion effects are

s ki
mitigated.
d (cm) (%)

(0.62)

(0.68)
(0.81)
(0.80)
(0.54)
(0.73)
1.71

1.87
2.24
2.21
2.04
2.00
f yi
(without EDDs)

6. Examples of application and numerical validation

6-Story frame

Main frame

(kN cm)

T1 ¼ 0.69 s
To illustrate the application of the proposed method, three
reinforced concrete (RC) frames with wide beam-column connec-
82

74
96
92
70

90
fki

tions of 3, 6 and 9 stories were retrofitted with hysteretic

dampers. The structural configuration and detailing of the main
2nd design earthquake

0.06
0.04
structures corresponded to typical residential buildings located in 0.03
a s i

the moderate-seismicity southern part of Spain, and they were

designed according to earlier seismic codes, which required
K¼ 4, Z ¼ 61
(kN/cm)
(far-field)

relatively reduced lateral strength and did not prescribe any

seismic detailing. Table 1 shows the fki, fdyi obtained from a
0.24 388

0.14 248
0.18 308

pushover analysis, and the T1 calculated by an eigenvalue analysis

s ki
Properties of the main frames and hysteretic dampers.

Design of the EDDs

using the initial elastic stiffness of the members. fdyi is expressed

a s i

in centimeters and as a percentage of the story height (into

K ¼11, Z ¼10

brackets). All the stories had the same mass mi ¼0.57 kNs2/cm.
(near-fault)
earthquake

(kN/cm)
1st design

The story height was also equal in all stories (275 cm) excepting
1067

the first story (375 cm). This exceptionally small story height of
847
682
s ki

275 cm is typical of residential constructions built in Spain during

d (cm) (%)

the 70s, 80s and 90s of the 20th century.

1.57 (0.57)
2.22 (0.81)
2.24 (0.60)

Two design earthquakes were considered, and for each of

them the required EDDs were obtained. The first design earth-
f yi
(without EDDs)
Story 3-Story frame

quake represented a near-fault ground motion—thus c1 ¼0.23,

Main frame

(kN/cm)

c2 ¼0.4—and it was defined by VDmax ¼45 cm/s, TG ¼0.52 s,

T1 ¼ 0.37 s

TNH ¼0.65 s and Id ¼7.5. The second design earthquake repre-

sented a far-field ground motion —thus c1 ¼0.18, c2 ¼0.6— and
97
77
62
f ki

it was defined by VDmax ¼45 cm/s, TG ¼1 s, TNH ¼1 s and Id ¼16.

Table 1

The value VDmax ¼45 cm/s was determined from input energy
9
8
7
6
5
4
3
2
1

spectra proposed for moderate-seismicity regions [27], and it is

 A. Benavent-Climent / Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396 1391

associated to a PGA of 0.13g and a 5% damping fraction. This PGA
is a typical value prescribed by the current Spanish seismic code
for the southern part of Spain. It corresponds to a design earth-
quake having a probability of exceedance of 10% in 50 years and a
return period of 475 years (i.e. 10/50/475). According to FEMA
356 [2], for an enhanced rehabilitation objective, the required
building performance level for this earthquake hazard 10/50/475
is Immediate Occupancy (IO). Only light damage is allowed in the
IO performance level, and according to FEMA-356 [2] in RC
structures this involves limiting the interstory drifts below 1%
of the story height. The lateral yield displacement fdyi of the bare
frames used in this study ranged between 0.52% and 0.6% in the
first story—375 cm height—and 0.57–0.91% in the other
stories—275 cm height. In all cases fdyi was below the 1% set for
the IO performance level. Accordingly, so as to achieve the
aforementioned enhanced rehabilitation objective, the EDDs were
designed so that dmax i r fdyi.
The same Ki ¼K was adopted in all stories. The resulting lateral
stiffness ski, lateral strength sQyi expressed in terms of sai, and
energy dissipation capacity in terms of Z required for the EDDs for
each design earthquake are indicated in Table 1. It is worth noting
that although all structures were seismic-upgraded for the same VD,
the near-fault design earthquake demanded EDDs with much larger
K and sai (i.e. greater strength) than the far-field design earthquake.
To validate the proposed design method, a series of nonlinear
dynamic response analyses were carried out with the three
structures equipped with EDDs described above. Two sets of
acceleration records were considered. The first set included the
near-fault records JMA-Kobe (Hyogo-ken nambu earthquake,
1995), El Centro (Imperial Valley earthquake, 1979), Rocca (Ancona
earthquake, 1972), Tolmezzo (Friuli earthquake, 1976) and
Korinthos (Alkion earthquake, 1981); their Id index ranged bet-
ween 6.9 and 7.9 with the average Id ¼7.5; the values of TG and TNH
for the corresponding VD  T and Newmark–Hall envelope spectra
were TG ¼ 0.52 s and TNH ¼0.65 s. The second set included the
far-field records Calitri (Campano Lucano earthquake, 1980), SCT
(Mexico, 1985), Petrovac (Montenegro earthquake, 1979), Taft
(Kern county, 1952) and Montebello (Northridge earthquake,
1994); their Id index ranged between 14 and 17 and the average
was Id ¼16; the values of TG and TNH for the corresponding VD T
and Newmark–Hall envelope spectra were TG ¼1 s and TNH ¼1 s. All
the records were scaled so that the total energy input contributable
to damage expressed in terms of equivalent velocity was
VDmax ¼45 cm/s. The frames with EDDs designed for the first and
second design earthquakes were subjected to the respective set of
acceleration records. The results of the analyses are shown in Fig. 2.
The solid and the dashed lines show the maximum interstory drift
dmax i —in percentage of the story height— predicted with Eq. (28)
for the sa1max that governed the design of the EDDs. The solid lines
correspond to the near-fault design earthquake, while the dashed
lines correspond to the far-field one. The symbols with thin lines
show the results of the time-history response analyses. In all cases
the maximum interstory drift obtained from the response analyses
remains below the maximum allowed drift in the critical story
(i.e. the first story). Also, Fig. 2 shows that Eq. (28) provides a
satisfactory upper-bound prediction of the maximum drifts exhib-
ited by the entire building-device structure, and that this predic-
tion is close to the response under the severest earthquake.
7. Experimental validation
To experimentally support the proposed method, dynamic
tests were conducted on a half-scale model with the 3 m  3 m
shaking table of the University of Granada. Fig. 3 gives an overall
view. The test structure consisted of a reinforced concrete slab of
125 mm-depth supported on four steel columns made with
80 mm  80 mm  4 mm box-sections (referred to as main struc-
ture hereafter), and two brace dampers whose source of energy
dissipation was the plastic deformation of steel plates installed
between two U-shaped steel sections. The main structure was
designed to behave as a strong slab-weak column structure and
its total weight was 72.4 kN. The purpose of the tests was to
assess the accuracy of the proposed method, which idealizes the
existing frame as elastic. In order to use a specimen whose
behavior was as close as possible to the model used in the
idealization, RHS steel columns were used instead of reinforced
concrete columns. Before performing the dynamic tests, the main
structure—without the brace dampers—was tested under lateral
forces and it was found that: (i) its lateral stiffness was
fk¼37.2 kN/cm and (ii) it could sustain lateral displacements
beyond 1.8 cm within the elastic range. The lateral stiffness and
strength provided by the two brace dampers was obtained from
static tests conducted prior to the dynamic experiments with

9 Story 9 Story 9 Story

8 Prediction (far-field)
8 8
Prediction (near-fault)
Montebello
7 Petrovac
Far-field
7 7
SCT
Calitri
6 Taft 6 6
Tolmezzo
Korinthos
5 Rocca Near-fault 5 5
El Centro
JMA-Kobe
4 4 4

3 3 3

2 2 2

1 δi(%) 1 δi(%) 1 δi(%)

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

Fig. 2. Comparison between prediction and results of dynamic analyses. (a) 3 story frame, (b) 6 story frame and (c) 9 story frame.
1392 A. Benavent-Climent / Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396
Added weigth (steel blocks)
Gauges
Slab
Accelerometers U-section
Brace damper Column
Steel plates that plastify (inside)
Displacement transducer
Gauges
Dynamic
actuator
3×3m shaking table
Fig. 3. Overall view of the shaking table test.
similar brace dampers, giving sk¼ 80 kN/cm and sQy ¼46 kN,
respectively. These values give a shear force coefficient for the
EDDs of sa ¼ sQy/Mg¼ 0.64 and a stiffness ratio of K ¼ sk/fk¼2.2.
The 1980 Campano Lucano earthquake recorded in Calitri was
used as the input motion for the shaking table. The dynamic
shaking table test of the main structure with the brace dampers
was carried out scaling the earthquake record to PGA¼0.55 g.
The maximum lateral displacement of the slab—relative to the
table—recorded by the displacement transducers was 1.55 cm. The
parameters that characterized the input earthquake were also
determined from the measurements provided by the instrumenta-
tion during the test. Since the main structure remained elastic, all
plastic strain energy consumed by the system, Wp, was dissipated
by the brace dampers. Also, at the end of the tests the velocity was
nearly 0 and the elastic strain energy stored in the system was very
small in comparison to Wp, thus We was negligible. Consequently
VD was calculated from the total energy dissipated by the brace
dampers, which was estimated from the readings provided by the
displacement transducers and the strain gages at both ends of each
brace damper, giving VD ¼120 cm/s. The parameters Id, TG and TNH
were obtained from the actual acceleration record measured by an
accelerometer located on the shaking table during the test, giving
pffiffiffiffiffiffiffiffiffi
Id ¼18.5, TG ¼0.75 s and TNH ¼0.9 s—the time was scaled by 1=2
to fulfill similarity laws between prototype and test model. By
substituting these values in Eq. (28) and adopting for c1 and c2 the
values corresponding to a far-field earthquake, the maximum
lateral displacement predicted is 1.62 cm, which is very close to
the experimental result (1.55 cm).
The numerical model used for idealizing the structure tested
on the shaking table was a single degree of freedom system,
whose restoring force was controlled by two elements connected
in parallel: one is the ‘‘flexible’’ element that represents the main
structure without dampers, and the other is the ’’stiff’’ element
that represents the EDDs. The flexible element was an elastic
spring of stiffness fk¼37.2 kN/cm. The stiff element was assumed
to follow an elastic-perfectly-plastic behavior (i.e. without strain
hardening effects) and it was defined by a lateral yield strength of
sQy ¼ 46 kN and an initial stiffness of sk¼80 kN/cm. One of the
factors that can justify the higher lateral deformability obtained
with the numerical simulation (1.62 cm) as compared to the
shaking table tests (1.55 cm) is that in the former the strain
hardening effects actually exhibited by the EDDs were neglected.
8. Conclusions
This paper presents an energy-based design procedure for
seismic retrofit of multi-story buildings by installing hysteretic
dampers. The method provides the lateral strength, the lateral
stiffness and energy dissipation capacity required for the dampers
to be installed in each story of the existing structure to achieve a
desired building performance level for a given earthquake hazard.
The maximum allowed interstory drift controls the target perfor-
mance level. The earthquake hazard is characterized in terms of
energy input, several seismological parameters used in the
literature—i.e. the Cosenza and Manfredi adimensional index Id,
the predominant period of the earthquake TG, and the initial
period of medium period region in the Newmark and Hall
spectrum, TNH—and it takes into account the proximity to the
source. Under the proposed method, the effect of the hysteretic
dampers is recognized directly in terms of hysteretic energy,
without having to resort to equivalent viscous damping approx-
imations; further, the cumulative damage induced in the dampers
is explicitly evaluated. The proposed method is intended to be
applied to low- to medium-rise structures. The application of the
method is illustrated with three structures of 3, 6 and 9 stories,
and its validity is assessed numerically through nonlinear
dynamic response analyses with near-fault and far-field ground
motions; and experimentally, by means of dynamic shaking
table tests.
Acknowledgments
This research was funded by the local government of Spain,
Consejerı́a de Innovación, Ciencia y Tecnologı́a (project P07-TEP-
02610) and by the European Union (Fonds Européen de Dévelop-
ment Régional).
Appendix A
The elastic response of a low- to medium-rise multi-story
building can be analyzed replacing the structure by a shear strut
of height H and using the mode superposition method. The
undamped response under a ground motion acceleration z€ g ðtÞis
governed by
 
@2 y @ @y
m  G ¼ mz€ g , ðA:1Þ
@t 2 @x @x
where x is the height at an arbitrary point; t is the time; y(x,t) is
the horizontal displacement; m(x) is the mass, and G(x) is the
shear rigidity. Setting the right-hand side of Eq. (A.1) to zero,
gives the equation of the free vibration
 
@2 y @ @y
m  G ¼ 0, ðA:2Þ
@t 2 @x @x
whose solution takes the form of
X
y¼ qj fj : ðA:3Þ
Here qj(t) and fj(x) are the time function and the mode function
for the j-th mode of vibration, and the summatory S is extended
from j ¼1 to N. qj(t) is expressed by
qj ðtÞ ¼ aj sinðoj t þ aj Þ, ðA:4Þ
where the amplitude aj, frequency oj and phase-angle aj are
constant for each mode j. The shape function fj(x) is an
 A. Benavent-Climent / Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396 1393

orthogonal function, that is, it satisfies Multiplying both sides of Eq.(A.15) by mfs and integrating
Z H over O–H gives
mfi fj dx ¼ 0 if ia j: ðA:5Þ RH
o mfs dx
Ds ¼ o : ðA:16Þ
The function fj(x) can be normalized so that M
Z H
Raising both sides of Eq. (A.15) to the second power, multi-
2 plying by m, integrating over the height 0–H and recalling
mfi dx ¼ M, ðA:6Þ
o Eqs.(A.5),(A.6) gives
where M is the total mass of the system. Substituting Eqs.(A.3) X
D2j ¼ 1: ðA:17Þ
and (A.4) in Eq.(A.2) gives
X  d

dfj
 Substituting Eq. (A.16) in Eq. (A.14), the total energy input E
qj mo2j fj þ G ¼ 0, ðA:7Þ can be rewritten as follows:
dx dx X
E¼ E1j D2j : ðA:18Þ
and to satisfy this equation for arbitrary values of qj the following
equation is obtained: The above defined E1j can be expressed in form of an equiva-
  lent velocity V1j(oj) given by
d dfj
G ¼ mo2j fj : ðA:8Þ 2
dx dx MV1j
E1j ¼ , ðA:19Þ
The solution of Eq. (A.1) takes also the form of Eq. (A.3). 2
Substituting Eq. (A.3) in Eq. (A.1) and taking into account Eq. (A.8) and then the total input energy E given by Eq. (A.18) can be
gives expressed by
X X P
m fj q€ j þ m o2j fj qj ¼ mz€ g : ðA:9Þ M ðV1j Dj Þ2
E¼ ðA:20Þ
2
Multiplying both sides of Eq. (A.9) by a given mode function
For t Zt0, the system is oscillating in an undamped free
fs, integrating over the height H, and recalling Eqs. (A.5) and
oscillation, thus the term V1j Dj represents the maximum velocity
(A.6), the equation that governs the vibration in the s-th mode is:
of the system vibrating in mode j. Since the maximum absolute
Z H
acceleration can be obtained multiplying the maximum velocity
M q€ s þM o2s qs ¼ z€ g mfs dx: ðA:10Þ
o by the frequency oj, the maximum force at height x is
Multiplying the second member of Eq. (A.9) by ydt _ ¼
P
fj q_ j dt ojV1jDjf(x)m(x) and the corresponding shear force Q^ j is given by
Z H
and integrating over O–H with respect to x and over 0–t0 with
Q^ j ðxÞ ¼ oj V1j Dj mfj dx: ðA:21Þ
respect to t, the total energy input E is expressed by x
Z H Z t0 X  X Z t0 Z H 
Combining the Q^ j of each vibration mode with the rule of the
E¼ mz€ g jj q_ j dt dx ¼  z€ g q_ j mfj dx dt:
o o o o square root of sum of the squares, the maximum total shear force
ðA:11Þ at height x normalized by the weight over height x and expressed
in terms of a shear-force coefficient a^ ðxÞ, is expressed by
Eq. (A.10) has been already multiplied by fs. Multiplying now qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P ^2 P RH
the second member of Eq. (A.10) by q_ s dt and integrating over 0–t0 Q j ðxÞ ðoj V1j Dj x mfj dxÞ2
with respect to t, the energy input in a single mode of vibration s, a^ ðxÞ ¼ R H ¼ RH , ðA:22Þ
g x mdx g x mdx
Es, is obtained by
Z t0 Z H  making x ¼0 in Eq. (A.22) and recalling Eq. (A.16), the base shear
Es ¼  z€ g q_ s mfs dx dt: ðA:12Þ force coefficient is given by
o o qP
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
From Eqs.(A.11) and (A.12) it follows that ðoj V1j D2j Þ2
X a^ ð0Þ ¼ : ðA:23Þ
g
E¼ Ej : ðA:13Þ
Further, E and a^ ð0Þ can be expressed as a fraction of the
Now, a new parameter E1j is defined as the energy input in a corresponding values in a fictitious SDOF system of mass M and
fictitious undamped SDOF of mass M and frequency oj subjected frequency equal to that of the first mode o1 of the real MDOF
to the ground motion acceleration z€ g ðtÞ. The equation of motion of system, subjected to the ground motion z€ g ðtÞ by two new
this fictitious SDOF system would be M y€ þ Mo2j y ¼ M z€ g (here y coefficients e and a as follows:
indicates a fictitious displacement), and the corresponding input MV12
Rt o1 V1
energy would be E1j ¼  oo M z€ g ydt.
_ For j ¼s, the force represented a^ ð0Þ ¼ a ; E¼e , ðA:24Þ
g 2
RH
by the second member of Eq. (A.10) is ð o mfs dxÞ=M times larger
where V1 is the input energy, in form of equivalent velocity, in the
than that acting on the fictitious SDOF system (i.e. M z€ g ); there- fictitious SDOF system of frequency o1. In an undamped elastic
hR i2
H
fore, the energy Es given by Eq. (A.12) is ð o mfs dxÞ=M times system, E coincides with the elastic vibration energy We for t Zt0.
Thus, since o1 ¼2p/T1 and using E and a^ ð0Þ given by Eq. (A.24),
larger than E1j. Accordingly, recalling Eq. (A.13) this means that
We ( ¼E) is expressed by
RH !2
X mfj dx M a^ ð0Þ2 g 2 Mg 2 T12  e  a^ ð0Þ2
E¼ E1j o : ðA:14Þ We ¼ e ¼ : ðA:25Þ
M 2a o12 2 4p2 a2 2

For convenience, E is now expressed in terms of new coeffi- Based on the above equations, the response of a system of
cients Dj, which satisfy uniform mass m(x)¼ m and linearly changing shear rigidity
X G(x)¼Go þ(GT Go)(x/H)—here Go and GT are values of G at the
Dj fj ¼ 1: ðA:15Þ
bottom and top part of the building—subjected to an earthquake is
1394 A. Benavent-Climent / Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396

x/H T1 = 1 2 3 T1 1
5 x/H = 2 3 5
1.0 TG 1.0 TG

0.8 Calculated 0.8

Approximate
0.6 0.6
Go
=1 Go
0.4 GT 0.4 =2
GT
0.2 0.2

0.0 0.0
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

T1 1 3 T1 1
x/H = 2 5 x/H = 2 3 5
1.0 TG 1.0 TG

0.8 0.8

0.6 0.6
Go Go
=4 = 10
0.4 GT 0.4 GT

0.2 0.2

0.0 0.0
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

Fig. A1. Calculated and approximated distributions of shear-force coefficient for: (a) Go/GT ¼ 1; (b) Go/GT ¼2; (c) Go/GT ¼4; (d) Go/GT ¼10.

obtained next. The earthquake is characterized by a bilinear input Appendix B

energy spectra expressed in terms of equivalent velocity VD  T by
Taking into account Eq. (12) and the ratios defined in Eq. (11),
for T oTG : VD ¼ VDmax T=TG , ðA:26aÞ it follows from Eqs. (8)
 
k 1
for T ZTG : VD ¼ VDmax , ðA:26bÞ Qyi ¼ s Qyi þ s dyif ki ¼ s Qyi þ s dyi s i ¼ s Qyi 1 þ : ðB:1Þ
Ki Ki
PN
where VDmax is the maximum VD in the spectra and TG is Dividing Eq. (B.1) by k ¼ i mk g and recalling the ratios defined
the predominant period of the ground motion. Substituting in Eq. (11), the first and last terms of Eq. (B.1) give
G(x)¼ Go þ(GT Go)(x/H) in Eq. (A.2) and performing a variable  
1
transformation X¼Go þ(GT  Go)(x/H), Eq .(A.2) relates to Bessel’s ai ¼ s ai 1 þ , ðB:2Þ
Ki
equation and the oi and fi(x) can be obtained through numerical
calculations. Substituting these fi(x) in Eq. (A.16) allows one to which, particularized for the first story, yields
calculate the Dj’s. With the frequencies oi obtained above, the  
1
corresponding V1j are calculated from the VD  T spectra defined by a1 ¼ s a1 1 þ : ðB:3Þ
K1
Eq. (A.26a) and (A.26b). Substituting oj, fj(x) and these Dj’s, V1j’s in
Eqs. (A.20), (A.22) and (A.23), the values of E, a^ ðxÞ and a^ ð0Þ are Dividing Eq. (B.2) by Eq. (B.3), the distribution of the shear
derived. From these a^ ðxÞ and a^ ð0Þ, the exact shear-force coefficient force coefficients is obtained
distribution aðxÞ ¼ a^ ðxÞ=a^ ð0Þ is obtained. The distribution aðxÞ ai a K1 ðKi þ 1Þ
calculated in this way is shown in Fig. A1 with solid lines for ¼ s i : ðB:4Þ
a1 s a1 Ki ðK1 þ1Þ
different values of T1/TG and Go/GT. For simplicity, in this study
these solid lines have been approximated with the dashed lines
given by Eq. (13). In the proposed Eq. (13), Go/GT has been replaced In the proposed method, the distribution ai/a1 is made equal to
by fk1/fkN for lumped-mass systems. ai defined given by Eq. (13), therefore making ai =a1 ¼ ai in
Finally, the values of parameters a and e in Eq. (24), which Eq. (B.4) and solving for sai gives Eq. (14).
relate a^ ð0Þ and E with the corresponding values in a fictitious
SDOF of mass M and period T1, were computed for realistic values
of T1/TG from 1 to 5 and Go/GT from 1 to 10. It was obtained that e Appendix C
remains very close to 1—the mean is 0.87 and the standard
deviation 0.059. Recalling Eq. (A.24), this indicates that the total From the coefficients defined in Eq. (11) and recalling Eq. (14),
energy input in the system E is governed by the first mode of d is expressed by
s yi

vibration. Also, e/a2 was found to be slightly above 1—the mean is P

N PN
1.33 and the standard deviation 0.045. Taking e/a2 ¼1 for simpli- a
s i mk g mk g
s Qyi k¼i Ki ðK1 þ 1Þ k ¼ i
city in Eq. (A.25) yields Eq. (15), where a^ ð0Þ has been renamed as s dyi ¼ ¼ ¼ ai s a1 , ðC:1Þ
s ki Ki f ki K1 ðKi þ 1Þ Ki f ki
famax1 for a lumped-mass model. Eq. (15) provides a safe-side—
and not over-conservative—estimation of We from the base shear substituting this expression in sdyi r fdyi and solving for sa1 gives
force coefficient of the building. Eq. (29).
 A. Benavent-Climent / Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396 1395

Appendix D. List of symbols Qyi yield shear force at the i-th story of the entire building-
device structure
a coefficient that expresses the base shear force coefficient R strength reduction factor
in a MDOF system of mass M and fundamental fre- r displacement vector resulting from a unit support
quency o1 as a fraction of the corresponding value in a displacement.
fictitious SDOF of mass M and frequency o1 a
s 1i shear-force coefficient required on the dampers of the
C damping coefficient first story so that the maximum interstory drift at the
C damping matrix i-th story does not exceed dallow i.
c1 seismological parameter a
s i shear-force coefficient representing sQyi normalized by
c2 seismological parameter the total weight of the upper stories
Dj coefficient associated with a given vibration mode j used d
s yi yielding interstory drift of the dampers of the i-th story
to relate E and E1j Sa spectral absolute response acceleration
E total input energy ski initial lateral stiffness of the dampers of the i-th story
e coefficient that expresses the input energy in a MDOF Spv spectral relative pseudo-velocity
system of mass M and fundamental frequency o1 as a Sv spectral relative velocity
fraction of the corresponding value in a fictitious SDOF sQi restoring force exerted by the dampers on the i-th story
of mass M and frequency o1 sQyi yield shear force of the dampers of the i-th story
E1j energy input in a fictitious undamped SDOF of mass M t time
and frequency oj T period of vibration
Ej energy input in a single mode of vibration j T1 fundamental period of the main structure (without
d
f yi yielding interstory drift of the i-th story of the main dampers)
structure (without dampers) TG period corresponding to the change of slope of the VD  T
fki lateral stiffness of the i-th story of the main structure bilinear spectra (predominant period of the ground
(without dampers) motion).
fkN lateral stiffness of the uppermost N-th story of the main TNH initial period of medium period region in the Newmark
structure (without dampers) and Hall spectral representation
fQi restoring shear force at the i-th story of the main to instant when the ground motion fades away
structure (without dampers) V1j equivalent velocity converted from E1j
fQmax i maximum shear force at the i-th story of the main Va equivalent velocity converted from the absorbed energy
structure (without dampers) VD equivalent velocity converted from the energy that
fQyi yield shear force at the i-th story of the main structure contributes to damage ( ¼Wp þWe)
(without dampers) VDmax maximum value of VD in the VD  T design spectrum
Q^ j contribution of the j-th vibration mode to the shear VE equivalent velocity converted from the total input energy
force acting at a given height of a shear strut. Wx damping energy
Fy yielding force of an equivalent SDOF system We elastic vibrational energy
G(x) shear rigidity per unit length of a shear strut that Wk kinetic energy
represents the building Wp irrecoverable plastic strain energy
g acceleration of gravity Wpi plastic strain energy accumulated at the i-th story
Go shear rigidity at the bottom of the building Ws absorbed energy
GT shear rigidity at the top of the building Wse recoverable elastic strain energy
H total height of the building Wui ultimate energy dissipation capacity of the dampers
Id seismological parameter installed in the i-th story
K ratio between the lateral stiffness of the dampers and x height with respect to the ground at an arbitrary point
the main structure at the i-th story (equal for all stories) of a shear strut that represents a building
keq stiffness of an equivalent SDOF system of mass and x coefficient that represents the position of the story
period equal to the total mass and fundamental period y relative horizontal displacement
of the MDOF system that represents the main structure y_ relative horizontal velocity
Ki ratio between the lateral stiffness of the dampers and the y€ relative horizontal acceleration
lateral stiffness of the main structure at the i-th story _
yðtÞ relative velocity vector
M mass matrix €
yðtÞ relative acceleration vector
m(x) mass per unit length of a shear strut that represents the z€ g ground acceleration
building ae base shear-force coefficient required on the main struc-
M total mass ture so that it can absorb by itself (i.e. without the
mi mass of the i-th story dampers) the amount of energy that contributes to
N number of stories of the building damage input by the earthquake
neq equivalent number of plastic cycles at the maximum a^ ðxÞ shear-force coefficient representing the maximum total
value of plastic excursion that an equivalent SDOF shear force at a given height x normalized by the weight
system must develop to dissipate the total amount of over the height x
hysteretic energy input by the earthquake ai shear-force coefficient representing Qyi normalized by
PGA peak ground acceleration the total weight of the upper stories
PGV peak ground velocity a
f max i shear-force coefficient representing fQmax i normalized
Q(t) restoring force vector by the total weight of the upper stories
Q(y) restoring force ai lateral strength distribution of the entire building-
Qi restoring force at the i-th story of the entire building- device structure expressed as ai/a1
device structure w1 stiffness ratio representing fk1 normalized by keq
qj(t) time function for the j-th mode of vibration di interstory displacement of the story i
1396 A. Benavent-Climent / Soil Dynamics and Earthquake Engineering 31 (2011) 1385–1396
dallow i maximum allowed interstory drift at the i-th story
dmax i maximum interstory drift of the entire building-device
structure at the i-th story
fj(x) mode function of the j-th mode of vibration
g1 ratio between Wp and Wp1
Z cumulated plastic strain energy (equal for all stories)
Zi normalized plastic strain energy accumulated at the
i-th story
Zui normalized ultimate energy dissipation capacity of the
dampers installed at the i-th story
m plastic deformation ratio (equal for all stories)
mi ratio representing the normalized plastic deformation of
the dampers of the i-th story
o circular frequency of a SDOF system
oj circular frequency of the j-th mode of vibration
References
[1] FEMA. Guidelines and commentary for the seismic rehabilitation of buildings
(FEMA 273, 274). Federal Emergency Management Agency, Washington DC;
1998.
[2] FEMA. Prestandard and commentary for the seismic rehabilitation of build-
ings. Federal Emergency Management Agency (FEMA 356), Washington DC;
2000.
[3] Krawinkler H. Pros and cons of a pushover analysis of seismic performance
evaluation. Engineering Structures 1988;20:452–64.
[4] Kim S, D’Amore E. Push-over analysis procedure in earthquake engineering.
Earthquake Spectra 1999;15:417–34.
[5] Chou CC, Uang CM. Establishing absorbed energy spectra—an attenuation
approach. Earthquake Engineering and Structural Dynamics 2000;29:
1441–55.
[6] Kunnanth SK, Chai H. Cumulative damage-based inelastic cyclic demand
spectrum. Earthquake Engineering and Structural Dynamics 2004;33:
499–520.
[7] Ruiz SE, Urrego E, Silva FL. Influence of the spatial distribution of energy-
dissipating bracing elements on the seismic response of multistory frames.
Earthquake Engineering and Structural Dynamics 1995;24:1511–25.
[8] Gluck N, Reinhorn AM, Gluck J, Levy R. Design of supplemental dampers for
control of structures. Journal of Structural Engineering 1996;122:1394–9.
[9] Rivakov Y, Gluck J. Optimal design of ADAS damped MDOF structures.
Earthquake Spectra 1999;15:317–30.
[10] Uetani H, Tsuji M, Takewaki I. Application of an optimum design method to
practical building frames with viscous dampers and hysteretic dampers.
Engineering Structures 2003;25:579–92.
[11] Karami MR, El Naggar MH, Moghaddam H. Optimum strength distribution for
seismic resistant shear buildings. International Journal of Solids and Struc-
tures 2004;41:6597–612.
[12] Choi H, Kim J. Energy-based seismic design of buckling-restrained braced
frames using hysteretic energy spectrum. Engineering Structures 2006;28:
304–11.
[13] Lin YY, Chang KC, Chen CY. Direct displacement-based design for seismic
retrofit of existing buildings using nonlinear viscous dampers. Bulletin of
Earthquake Engineering 2008;6:535–52.
[14] Moreschi LM, Singh MP. Design of yielding metallic and friction dampers for
optimal seismic performance. Earthquake Engineering and Structural
Dynamics 2003;32:1291–311.
[15] Dargush GF, Sant RS. Evolutionary aseismic design and retrofit of structures
with passive energy dissipation. Earthquake Engineering and Structural
Dynamics 2005;34:1601–26.
[16] Apostolakis G, Dargush GF. Optimal seismic design of moment-resisting steel
frames with hysteretic passive devices. Earthquake Engineering and Struc-
tural Dynamics 2010;39:355–76.
[17] Lin YY, Tsai MH, Hwang JS, Chang KC. Direct displacement-based design for
buildings with passive energy dissipation systems. Engineering Structures
2003;25:25–37.
[18] Kim JK, Seo YL. Seismic design of low-rise steel frames with buckling-
restrained braces. Engineering Structures 2004;26:543–51.
[19] Vargas R, Bruneau M. Seismic design of multi-story buildings with metallic
structural fuses. In: Proceedings of the eighth U.S. national conference on
earthquake engineering, San Francisco, California; 2006.
[20] Housner GW. Limit design of structures to resist earthquakes. In: Proceedings
of the first world conference on earthquake engineering, Berkeley CA; 1956.
[21] Zahrah TF, Hall WJ. Earthquake energy absorption in SDOF systems. Journal
of Structural Engineering 1984;110:1757–72.
[22] Akiyama H. Earthquake-resistant limit state design for buildings. Tokyo:
University of Tokyo Press; 1985.
[23] Uang, CM, Bertero, VV. Use of energy as a design criterion in earthquake-
resistant design. Report No. UBC/EERC-88/18, University of California at
Berkeley, CA; 1990.
[24] Fajfar P, Vidic T, Fischinger M. On energy demand and supply in SDOF
systems. In: Fajfar P, Krawinkler H, editors. Nonlinear seismic analysis and
design of reinforced concrete buildings. Amsterdam: Elsevier; 1992.
p. 41–61.
[25] BSL. The building standard law of Japan. Tokyo: The Building Center of Japan;
2009. [English version on CD available at /http://118.82.115.195/en/ser
vices/publication.htmlS].
[26] Manfredi G, Polese M, Cosenza E. Cumulative demand of the earthquake
ground motions in the near source. Earthquake Engineering and Structural
Dynamics 2003;32:1853–65.
[27] Benavent-Climent A, Pujades LG, Lopez-Almansa F. Design energy input
spectra for moderate seismicity regions. Earthquake Engineering and Struc-
tural Dynamics 2002;31:1151–72.
[28] Benavent-Climent A, Lopez-Almansa F, Bravo-Gonzalez DA. Design energy
input spectra for moderate-to-high seismicity regions based on Colombian
earthquakes. Soil Dynamics and Earthquake Engineering 2010;30:1129–48.
[29] Kuwamura H, Galambos TV. Earthquake load for structural reliability. Journal
of Structural Engineering 1989;115:1446–62.
[30] Fajfar P, Vidic T. Consistent inelastic design spectra: hysteretic and input
energy. Earthquake Engineering and Structural Dynamics 1994;23:523–37.
[31] Hudson, DE. Response spectrum techniques in engineering seismology. In:
Proceedings of first world conference on earthquake engineering, Berkeley
CA; 1956.
[32] Hudson DE. Some problems in the application of spectrum techniques to
strong motion earthquake analysis. Bulletin of the Seismological Society of
America 1962;52-2:231–52.
[33] Dubina D, Stratan A, Dinu F. Dual high-strength steel eccentrically braced
frames with removable links. Earthquake Engineering and Structural
Dynamics 2007;37:1703–20.
[34] Cosenza E, Manfredi G. The improvement of the seismic-resistant design for
existing and new structures using damage criteria. In: Fajfar P, Krawinkler H,
editors. Seismic Design Methodologies for the Next Generation of Codes.
Rotterdam: Balkema; 1997. p. 119–30.
[35] Manfredi G. Evaluation of seismic energy demand. Earthquake Engineering
and Structural Dynamics 2001;30:485–99.
[36] Newmark NM, Hall WJ. Earthquake spectra and design. Berkeley, California:
Earthquake Engineering Research Institute; 1982.
[37] Benavent-Climent A. An energy-based damage model for seismic response of
steel structures. Earthquake Engineering and Structural Dynamics 2007;36:
1049–64.