Sei sulla pagina 1di 14

Composites: Part B 38 (2007) 584–597

www.elsevier.com/locate/compositesb

Optimal performance-based design of FRP jackets for seismic retrofit


of reinforced concrete frames
a,b a,*
X.K. Zou , J.G. Teng , L. De Lorenzis c, S.H. Xia d

a
Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hong Kong, China
b
Maunsell Structural Consultants Ltd, 18/F Grand Central Plaza, Tower 2, 138 Shatin Rural Committee Road, Shatin, NT, Hong Kong, China
c
Department of Innovation Engineering, University of Lecce, Lecce, Italy
d
School of Civil and Environmental Engineering, University of Adelaide, SA 5005, Australia

Received 10 May 2006; accepted 10 July 2006


Available online 23 December 2006

Abstract

External bonding of fiber-reinforced polymer (FRP) composites is now a well-established technique for the strengthening/retrofit of
reinforced concrete (RC) structures. In particular, confinement of RC columns with FRP jackets has proven to be very effective in
enhancing the strength and ductility of columns, and has become a key technique for the seismic retrofit of RC structures. Despite
the large amount of research on the behavior of RC columns confined with FRP, little research has been conducted on the behavior
of RC frames with FRP-confined columns. For the seismic retrofit of RC frames with FRP, apart from the structural response of a ret-
rofitted frame, an important issue is how to deploy the least amount of the FRP material to achieve the required upgrade in seismic
performance. With these two issues in mind, this paper presents an optimization technique for the performance-based seismic FRP ret-
rofit design of RC building frames. The thicknesses of FRP jackets used for the confinement of columns are taken as the design variables,
and minimizing the volume and hence the material cost of the FRP jackets is the design objective in the optimization procedure. The
pushover drift is expressed explicitly in terms of the FRP sizing variables using the principle of virtual work and the Taylor series approx-
imation. The optimality criteria (OC) approach is employed for finding the solution of the nonlinear seismic drift design problem. A
numerical example is presented and discussed to demonstrate the effectiveness of the proposed procedure.
 2007 Elsevier Ltd. All rights reserved.

Keywords: Confinement; Fiber-reinforced polymer (FRP); Performance-based design; Pushover analysis; Reinforced concrete; Seismic retrofit; Structural
optimization

1. Introduction strong beam type, which results in brittle soft-story or col-


umn sideway collapse mechanisms during strong ground
The seismic performance of existing reinforced concrete motions [2]. In order to reduce the risk of structural col-
(RC) framed structures designed for gravity loads or lapses during strong earthquakes, there is an urgent need
according to old codes has proven to be poor during recent to upgrade existing RC buildings to meet the requirements
earthquakes, due to insufficient lateral load-carrying capac- of current seismic design codes.
ity and limited ductility [1]. Such structures possess an The seismic retrofit of an RC building may involve tar-
inherently low resistance to horizontal loads, resulting in geted strengthening of deficient regions, to increase the
large inelastic deformations during earthquakes. More- strength, stiffness and/or ductility of the structure, or to
over, their structural behavior is of the weak column/ provide redundant load-carrying mechanisms. In general,
a combination of different techniques may be employed
in the seismic retrofit of a structure. The selection of a spe-
*
Corresponding author. Tel.: +852 2766 6012; fax: +852 2334 6389. cific retrofit strategy should be based on the retrofit objec-
E-mail address: cejgteng@polyu.edu.hk (J.G. Teng). tives as well as on economic considerations [1]. The retrofit

1359-8368/$ - see front matter  2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesb.2006.07.016
X.K. Zou et al. / Composites: Part B 38 (2007) 584–597 585

design should be based on appropriate performance criteria design codes, directly addresses inelastic deformations
to ensure that a defined level of damage is not exceeded or induced in structures by earthquakes [3,9,10]. In assessing
the collapse of the building is prevented during specified the nonlinear seismic behavior of framed structures, push-
ground motions [3]. In addition, the cost of implementa- over analysis has been increasingly accepted as part of the
tion is of great concern to both building owners and prac- performance-based design procedure. Pushover analysis is
ticing engineers [4]. a simplified, static, nonlinear procedure in which a prede-
The overall seismic retrofit strategy for an RC frame fined pattern of earthquake loads is applied incrementally
must consider a number of key issues in an integrated man- to the structure until a plastic collapse mechanism is
ner; these issues include the strengthening of beams, col- reached. This method of analysis generally adopts a
umns and beam-column joints to prevent brittle failure lumped-plasticity approach that tracks the spreading of
modes such as shear failure to become critical using exter- inelasticity through the formation of plastic hinges at the
nal FRP reinforcement or other appropriate methods. ends of the frame elements during the incremental loading
Once these brittle failure modes are suppressed, the seismic process.
retrofit design to enable the frame to satisfy specific The lateral drift performance of a multi-story building is
demands of an earthquake depends on the strength and an important indicator that measures the level of damage
ductility of the columns under combined axial compression to the structural and non-structural components of a build-
and bending. ing in current seismic design approaches and also in the
Retrofit of the columns is one of the most widely used newly developed performance-based design approach
seismic upgrading approaches for RC frames. Improving [1,3,9–11]. The economic design of structural elements for
the column behavior typically involves increasing its various levels of elastic and inelastic lateral drift perfor-
strength, ductility, stiffness or in most cases a combination mance under multiple levels of earthquake loads is gener-
of these parameters. Conventional retrofit measures for ally a rather difficult and challenging task [12]. Lateral
columns include RC overlays or steel jacketing. A more drift design is particularly challenging as it requires the
recent technique is the use of fiber-reinforced polymer consideration of an appropriate stiffness distribution of
(FRP) jackets to confine columns [5,6]. In such jackets, all structural elements and, in a severe seismic event, also
the fibers are oriented only or predominantly in the hoop the occurrence and redistribution of plasticity in the ele-
direction to confine the concrete so that both its compres- ments. Structural engineers are thus faced with the problem
sive strength and ultimate compressive strain are signifi- of efficiently distributing materials throughout the struc-
cantly enhanced [5–7]. Compared to conventional ture to optimize the elastic and inelastic drift responses of
techniques, FRP jacketing is easier and quicker to imple- structures. In absence of an automated optimization tech-
ment, adds virtually no weight to the existing structure, nique, sizes of members and amounts of steel reinforcement
has minimal aesthetic impact and is corrosion-resistant. are designed by trial-and-error methods based on intuition
As a result, FRP jacketing has been found to be a more and experience [12]. The need for an optimal design
cost-effective solution than conventional techniques in approach is thus clear, and structural optimization of
many situations and has thus been widely accepted [5,6,8]. dynamically excited structures has been an active research
For the seismic retrofit of RC frames employing FRP topic for the past few decades [12–18]. In recent years,
confinement of RC columns, apart from the structural much research has been devoted to the optimization of
response of a retrofitted frame, an important issue is how the emerging performance-based design approach. In par-
to deploy the least amount of the FRP material to achieve ticular, Chan and Zou [12], Zou [16] and Zou and Chan
the required upgrade in seismic performance. With these [17,18] proposed an optimization technique for elastic
two issues in mind, this paper presents an optimization and inelastic drift performance-based seismic design of
technique for the performance-based seismic FRP retrofit RC buildings. They showed that an automated optimiza-
design of RC building frames. The thicknesses of FRP jack- tion technique is capable of achieving the best seismic drift
ets in the columns are considered as the design variables, performance combined with the least expensive design.
while the least total material cost (i.e. costs associated with Specific research on the optimization of seismic retrofit
other aspects such as transportation are not included) of design of existing structures has been much more limited.
FRP and a uniform ductility demand are taken as design Martinez-Rodrigo and Romero [19] proposed a simple
objectives of the inelastic drift design optimization process. methodology leading to an optimal solution with nonlinear
viscous fluid dampers for the seismic retrofit of moment–
2. Existing work on optimal performanced-based seismic resisting frames. To the best of the authors’ knowledge,
design no research has been conducted on the optimization of seis-
mic retrofit design of RC structures when the retrofit strat-
Traditional design approaches for seismic retrofit, simi- egy is the confinement of columns with FRP jackets. At the
lar to traditional approaches for seismic design of new present, the performance-based retrofit design of RC struc-
structures, assume that structures respond elastically even tures with FRP confinement of columns can only be con-
to severe earthquakes [9]. Performance-based seismic ducted by trial-and-error methods based on subjective
design, which appears to be the future direction of seismic experience and much computational effort. The final design
586 X.K. Zou et al. / Composites: Part B 38 (2007) 584–597

may be overly conservative, resulting in an unnecessarily potential plastic hinge region in a column, the total thick-
expensive retrofit intervention and less than optimal seis- ness of the FRP jacket should be the sum of those deter-
mic performance. The optimization technique for the drift mined for the three failure modes, respectively [5]. This
performance-based seismic retrofit design of framed RC represents a conservative but realistic approach given the
buildings presented in this paper therefore fills a significant current stage of knowledge.
gap in existing research. The retrofit strategy is based on The design variables in the optimization process are
the FRP confinement of columns at the two ends, i.e. therefore the thicknesses, ti, of the FRP jackets required
in the regions of potential plastic hinge formation for confinement of the plastic hinges in each member.
[20–23]. The optimal design procedure is one that has been For a given type of FRP material, if the topology of the
appropriately modified from that previously developed by structure is predefined and each column is assumed to have
Chan and Zou [12], Zou [16] and Zou and Chan [17,18] the same FRP jacket thickness and the same length of the
for the seismic design of new structures. confined region at both ends, the total material cost of the
FRP composite used for column confinement is given by

3. Optimal seismic retrofit design problem X


Nc
FRP material cost ¼ wi ti ð1Þ
i¼1
3.1. Implicit design optimization problem
where wi is the cost coefficient for the FRP composite,
As shown in Fig. 1, FRP sheets for confinement of col- wi = 4Lci(Bi + Di)q; q is the cost per unit volume of the
umns are wrapped around columns with the fibers oriented FRP composite; and Lci is the length of the primary con-
in the hoop direction. The consequent increases in the axial finement region at each end of the ith column, which
compressive strength and the ultimate axial strain of the should be the largest of the plastic hinge length, 0.5D and
concrete core depend on several factors, including the 12.5% of member length [5,21]. In practical implementa-
thickness, tensile strength and elastic modulus of the con- tion, a secondary confinement region adjacent to the pri-
fining FRP jacket, unconfined concrete strength and mary confinement region should also be confined but
cross-sectional shape of the column [7]. For given material with the FRP jacket thickness reduced to half of that in
properties and cross-sectional dimensions, the thickness of the primary confinement region. The amount of FRP re-
the FRP jacket governs the strength and ductility of the quired for confining the secondary confinement region is
confined cross-section. not further considered in this paper.
This study considers an RC framed structure with Nc In performance-based design, it is necessary to check the
columns, Nb beams, and hence 2(Nc + Nb) potential plastic ‘‘capacity’’ of a retrofitted structure against the ‘‘demand’’
hinges (assuming one hinge at each end of each member). of an earthquake at the performance point, i.e. at the inter-
The ith column is assumed to have a rectangular cross-sec- section of the pushover capacity curve and the demand
tion, with width Bi and depth Di. Seismic retrofit is spectrum curve. It has been recognized that the damage
achieved with FRP confinement of the potential plastic to various components of the structure is a function of
hinge regions of each column, as shown in Fig. 1. the drift [1]. Thus, in this study, the inter-story drift of a
Only the thicknesses of the FRP jackets required for building, generated by a specified earthquake demand, is
confinement of the plastic hinges are considered as design checked against appropriate limits corresponding to a
variables in this study. This approach is realistic and also given performance level. Namely, for a multi-story building
reduces the design problem to a manageable size. The structure, the inter-story drift ratio caused by pushover
jacket thicknesses required for shear resistance and for con- loading should comply with the following requirement:
finement of lap splices are first calculated for each member
[5], but these thicknesses are not taken into account in the Duj uj  uj1
optimal design procedure presented in this paper. In prac- ¼ 6 dU
j ð2Þ
hj hj
tical implementation of the seismic retrofit strategy, for any
where Duj is the inter-story drift of the jth story; uj and uj1
are the story displacements of the two adjacent j and j  1
FRP
Lc floor levels, respectively; hj is the height of the jth story; and
confinement
dUj is the specified inter-story drift ratio limit, representing
the damage threshold for the jth story. This limit generally
RC column depends on the type of building and on its importance.
In addition to the design performance constraints on
story drifts, the FRP thickness may be given a minimum
FRP and a maximum limit based on practical, technological or
Lc other requirements:
confinement

Fig. 1. FRP-jacketed regions of column for seismic retrofit. tLi 6 ti 6 tU


i ð3Þ
X.K. Zou et al. / Composites: Part B 38 (2007) 584–597 587

2
where superscripts L and U denote the minimum and max- ðEc  E2 Þ 2
imum limits of the design variable ti. rc ¼ Ec ec  ec for 0 6 ec 6 et ð4aÞ
4fco0
In order to facilitate a solution of the drift design prob-
lem, it is necessary that the implicit inter-story drift con- and
straints of Eq. (2) be formulated explicitly in terms of the rc ¼ fco0 þ E2 ec for et 6 ec 6 ecu ð4bÞ
design variables ti. Before presenting the explicit formula-
tion of the design problem, the properties of a rectangular where
FRP-confined RC cross-section are first introduced briefly fcc0  fco0
as follows. E2 ¼ ð5Þ
ecu
2fco0
3.2. Lam and Teng’s model for FRP-confined concrete et ¼ ð6Þ
ðEc  E2 Þ

In the last few years, many studies have been conducted In Eq. (4), rc and ec are the axial (compressive) stress and
on the stress–strain behavior of FRP-confined concrete and strain of confined concrete; Ec and fco0 are the initial elastic
various models have been proposed. Critical reviews and modulus and the compressive strength of unconfined con-
evaluations of existing models are available in a number crete, respectively; E2 is the slope of the straight line that
of papers (e.g. [7,24,25]). Nevertheless, existing work on intercepts the stress axis at f0 ¼ fco0 (Fig. 2); et is the axial
FRP-confined concrete in rectangular sections is still lim- strain of concrete at which the parabolic first portion meets
ited [7], as most work has been concerned with concrete the linear second portion with a smooth transition. In Eqs.
in FRP-confined circular sections where confinement is (5) and (6), fcc0 and ecu are the compressive strength and
uniform when the concrete is subject to axial compression ultimate compressive strain of FRP-confined concrete
alone. and can be obtained from
For FRP-confined rectangular sections, Lam and Teng’s fcc0 fl
stress–strain model for FRP-confined concrete [26] appears ¼ 1 þ 3:3k s1 0 ð7Þ
fco0 fco
to be the most suitable model for use in design, as it takes a   0:45
simple form but captures the main characteristics of the ecu fl eh;rup
¼ 1:75 þ 12k s2 0 ð8Þ
stress–strain behavior of concrete confined by different eco fco eco
types of FRP. A further advantage of the model is that
In Eqs. (7) and (8), eco is the axial strain at peak stress of
for unconfined concrete (i.e., ti = 0) this model reduces
unconfined concrete, taken as 0.002 [25]; eh,rup is the FRP
directly to the idealized stress–strain curve of concrete
hoop rupture strain; fl is the equivalent confining pressure,
given by several existing design codes, such as Eurocode
defined as follows; and ks1 and ks2 are the shape factors for
2 [25,26] (see Fig. 2). This model is a simple modification
strength and ultimate strain, respectively.
of Lam and Teng’s stress–strain model [25] for FRP-con-
In circular cross-sections, the confining pressure is uni-
fined concrete in circular sections.
form around the circumference. Conversely, in rectangular
In Lam and Teng’s model (Fig. 2), the stress–strain
sections, the confining pressure is non-uniform and is lar-
curve of FRP-confined concrete is approximated using a
ger in the corner regions, so that only a portion of the sec-
parabolic first portion and a linear second portion, as
tion is effectively confined (Fig. 3). The equivalent
described by the following equations:
confining pressure of the FRP jacket for a rectangular
cross-section of dimensions B and D (D P B) is defined
as the confining pressure provided by an FRP jacket of
the same thickness to a circular section with an equivalent
diameter [26]. This equivalent diameter is taken as the
f'cc
E2
Axial stress σc

1
fo = f'co

Effective
Unconfined concrete confinement
(ENV 1992) area B
FRP-confined concrete B2 + D2
Ec (Lam and Teng)
45º Rc

εco εt 0.0035 εcu


D
Axial strain ε c

Fig. 2. Lam and Teng’s stress–strain model for FRP-confined concrete. Fig. 3. Confinement action in rectangular columns.
588 X.K. Zou et al. / Composites: Part B 38 (2007) 584–597

length of the diagonal of the rectangular section, i.e.


pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sections remain plane; elastic-perfectly plastic steel rein-
D2 þ B2 . The equivalent confining pressure, fl, is then forcement in both tension and compression; the FRP
given by jacket provides only lateral confinement but no stiffness
in the longitudinal direction; the steel stirrups are so widely
2Efrp teh;rup
fl ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð9Þ spaced that their confinement effect can be ignored; com-
D 2 þ B2 pressive stresses and strains are taken to be positive; and
where Efrp is the elastic modulus of the FRP. Eq. (9) indi- the ultimate limit state is reached when the strain of the
cates that fl is a linear function of the FRP thickness t. extreme compression fiber of concrete attains the ultimate
The shape factors, ks1 and ks2, are to account for the compressive strain ecu. The direct use of stress–strain mod-
effect of the cross-sectional shape on the compressive els based on the behavior of FRP-confined concrete under
strength and ultimate compressive strain of confined con- axial compression in the analysis of columns subjected to
crete [26]. The shape factor for strength, ks1, is given by combined axial compression and bending is an approxima-
 2 tion that has been widely accepted [5], and has been shown
B Ae to lead to close predictions of test results [27]. Based on
k s1 ¼ ð10Þ
D Ac these assumptions, Mu and /u are expressed in terms of
the design variables ti as described below.
while the shape factor for strain, ks2, is given by
 0:5 General strain and stress distributions of a rectangular
D Ae section under combined compression and bending at the
k s2 ¼ ð11Þ
B Ac ultimate limit state are illustrated in Fig. 4, where the
equivalent stress block for concrete is characterized by
where Ae is the area of the effectively-confined concrete and
the two parameters a and c. Based on Lam and Teng’s
Ac is the total area of concrete enclosed by the FRP jacket.
stress–strain model [26], a and c for a confined concrete sec-
In Lam and Teng’s model [26], Ae is the area contained by
tion can be expressed as follows [27]:
four second degree parabolas as shown in Fig. 3. These
parabolas have the initial tangents coinciding with the E2 ecu 2fco0
a¼1þ  ð13Þ
adjacent diagonal lines, rather than having them at a fixed 2fco0 3ðEc  E2 Þecu
angle of 45 to the section sides. Consequently, the effective 2fco03 þ 3fco0 e2cu ðEc  E2 Þ2 þ 2E2 e3cu ðEc  E2 Þ2
c¼1
confinement area ratio, Ae/Ac, is expressed by 4fco02 ecu ðEc  E2 Þ þ 3E2 e3cu ðEc  E2 Þ2 þ 6fco0 e2cu ðEc  E2 Þ2
B 2 D
DðD2Rc Þ þ B ðB2Rc Þ
2
ð14Þ
Ae 1  3Ag
 qsc
¼ ð12Þ
Ac 1  qsc Through ecu and E2, these two parameters are a function of
the thickness of the FRP jacket. The neutral axis depth X is
where Ag is the gross area of the section; qsc is the cross-sec-
derived from the equilibrium of forces (Fig. 4):
tional area ratio of longitudinal steel; and Rc is the radius
of the rounded corners. Corner rounding is generally re- X
Ns

quired for the FRP jacketing of rectangular columns to re- P ¼ afco0 BX þ fsk Ask ð15Þ
k¼1
duce the detrimental effect of a sharp corner on the tensile
rupture strength of the FRP and to improve the effective- where fsk and Ask are the stress (positive if compressive)
ness of confinement [5]. and the cross-sectional area of the kth layer of steel rein-
It should be noted that in the original model of Lam and forcement, respectively, and Ns is the number of layers of
Teng [26], Eq. (7) is valid only if the FRP jacket provided steel reinforcement. The steel of the kth layer has a strain
ensures that k s1 f1 =fco0 P 0:07, otherwise no strength esk and a distance dsk of its centroid from the extreme com-
enhancement should be assumed. In the present study, this pression fiber (Fig. 4). Once the neutral axis depth X is ob-
limit is not included in the formulation of the optimization tained from Eq. (15), the ultimate moment Mu (about the
procedure (i.e. Eq. (7) is taken to be valid for all values of section mid-height) corresponding to the axial load P is
the FRP jacket thickness to simplify the formulation), but determined by the equilibrium of moments:
this limit can be imposed as the lower limit of the jacket   X Ns  
0 D D
thickness within or at the end of the optimization solution M u ¼ afco BX  cX þ fsk Ask  d si ð16Þ
process. 2 k¼1
2

and the ultimate curvature /u can be obtained from


3.3. Ultimate moment and curvature of FRP-confined
ecu
rectangular cross-sections /u ¼ ð17Þ
X
The simple stress–strain model of Lam Teng [26] As the thickness of the FRP jacket increases, the compres-
described above can be used to compute the ultimate sive strength fcc0 and the ultimate compressive strain ecu of
moment Mu and the ultimate curvature /u of an FRP-con- the confined concrete increase while the neutral axis depth
fined rectangular cross-section under combined compres- decreases. Thus, both the ultimate moment given by
sion and bending, under the following assumptions: plane Eq. (16) and the ultimate curvature given by Eq. (17)
X.K. Zou et al. / Composites: Part B 38 (2007) 584–597 589

α f co
εcu fcc′
εs1 ds1 S1 X
1 fs1
X D/2 α fco′BX
2 Neutral εs2 fs2 S2
axis P
D 3 εs3 fs3 ds4 S3
4 εs4 fs4 S4

B S k = f sk Ask

Section Strains Stresses Forces


Fig. 4. Section with strain, stress and force distributions.

a b

c d e

Fig. 5. Effects of the FRP thickness on various parameters.

increase. Fig. 5 illustrates the variations of several param- evaluate the pushover displacement. The pushover story
eters related to the cross-sectional behavior with the thick- displacement, uj, includes a component related to the vir-
ness of the FRP jacket. tual work due to deformations within the structural mem-
bers, uj,memb, and a component related to the virtual work
due to rotations in the plastic hinges of beams and col-
3.4. Explicit drift formulation umns, ubeam col
j;hinge and uj;hinge [12,16,17]:

As mentioned earlier, the story drift constraints of Eq. uj ¼ ðuj;memb þ ubeam col
j;hinge Þ þ uj;hinge ð18Þ
(2) should be explicitly expressed in terms of the design
variables, ti, in order to facilitate a solution of the design where
problem. Based on the internal element forces and
" #
moments as well as plastic hinge rotations of the structure, X
Nc X
2
obtained from a pushover analysis at the performance ucol
j;hinge ¼ m0pjh hph ð19Þ
point, the principle of virtual work can be employed to i¼1 h¼1 i
590 X.K. Zou et al. / Composites: Part B 38 (2007) 584–597

In Eq. (19), m0pjh is the virtual moment at the location of the neously fixed value of M, a second-order Taylor series
hth hinge of a member due to a unit virtual load applied to approximation of hp at an initial design point t0 is given as
the building at the jth story level; hph is the plastic rotation  
 ohp  1 o2 hp 
experienced by the hth plastic hinge, which is equal to zero 
hp ðtÞ ¼ hp t¼t0 þ 0
ðt  t Þ þ ðt  t0 Þ
2
ð22Þ
when no plastic hinge is found. During the process of ot t¼t0 2 ot2 t¼t0
inelastic drift optimization, the term in the bracket of Eq. Given the explicit expressions of hp of Eq. (20) and Mu of
(18) is kept unchanged since the section size and steel rein- Eq. (16) as functions of t, the first and second derivatives of
forcement ratio of each member are fixed and FRP confine- hp(t) can be analytically computed. By substituting the ex-
ment is only applied to columns. The only variable term is plicit plastic rotation, hp(t), given by Eq. (22) into Eqs. (18)
the displacement caused by the formation of plastic hinges and (19), the pushover displacement, uj, can also be explic-
in the retrofitted columns, ucol j;hinge , which depends on the itly expressed in terms of the design variable, ti, as
thicknesses of the FRP jackets.  
As shown in Fig. 6, the moment–rotation behavior of a  XNc
ouj  1X Nc
o2 uj 

uj ðti Þ ¼ uj ti ¼t0 þ 0
ðti  ti Þ þ ðti  t0i Þ2
plastic hinge is modeled by a bilinear curve, composed of i
i¼1
oti ti ¼t0 2 i¼1 ot2i ti ¼t0
i i
an elastic segment AB and a hardening segment BC [9].
ð23Þ
Based on this curve, the plastic rotation hp can be expressed
as follows:
3.5. Explicit design problem formulation
M  My U
hp ¼ h ; with hp 6 hU ð20Þ
Mu  My p p
Upon establishing the explicit inelastic drift formulation
given by Eq. (23), the optimization problem of minimizing
In Eq. (20), M is the applied moment at the location of the the material cost of FRP confinement for retrofitting a
plastic hinge; My is the bending moment at first yielding of multi-story RC building can be explicitly written in terms
the tension steel reinforcement; Mu is the ultimate moment of the design variables ti as follows:
(computed in the previous section) and hU p is the ultimate
plastic rotation expressed by Minimize :
X
Nc

hU
p ¼ ð/u  /y ÞLp ð21Þ F ðti Þ ¼ wi t i ð24Þ
i¼1

In Eq. (21), /y is the yield curvature of the cross-section Subject to :


"
and Lp is the length of the plastic hinge. In the present 1  XNc

study, Lp = D/2 was adopted, as recommended by ATC- gj ðti Þ ¼ Duj ti ¼t0 þ b1i ðti  t0i Þ
hj i
i¼1
40 [9]. #
The effects of the FRP jacket on the values of Mu and /u 1 XNc
0 2
are illustrated in Fig. 5. Its effects on the values of My and þ b ðti  ti Þ 6 d U j ; ðj ¼ 1; 2; 3; . . . ; N j Þ
2 i¼1 2i
/y are less important and are neglected herein for simplic-
ity. The expressions of My and /y for an un-strengthened ð25Þ
RC cross-section have been given by Zou and Chan [17]. tLi 6 ti 6 tU
i ; ði ¼ 1; 2; 3; . . . ; N c Þ ð26Þ
In pushover analyses, pure moment hinges and axial
force–moment hinges are generally assigned to the two where gj(ti) is the inter-story drift ratio at the jth story level
ends of beams and columns, respectively. As the columns and a function of the design variables ti, Nj is the total
are confined with FRP jackets of variable thickness, the number of stories, and
 
properties of the axial force–moment hinges vary with the oDuj  oD2 uj 
b1i ¼ ; b2i ¼ ð27a; bÞ
thickness of the FRP jacket. To account for the change oti ti ¼t0 ot2i ti ¼t0
of hp due to a change of t while maintaining an instanta- i i

Eq. (25) defines the set of seismic inter-story drift perfor-


mance constraints under specified earthquake ground mo-
M
tions. Eq. (26) defines the sizing constraints for the
Mu
C thickness of the FRP jacket, where tLi and tU
i are the lower
M
My and upper bounds specified for the FRP jacket thickness.
B

3.6. Optimization procedure

A θy θ θ Once the design optimization problem is explicitly


p U
θp expressed in terms of the design variable, the next task is
to adopt a suitable method to solve the problem. The
Fig. 6. Moment–rotation curve. optimality criteria (OC) approach is adopted in this study
X.K. Zou et al. / Composites: Part B 38 (2007) 584–597 591

due to its superior numerical efficiency for the design of 3.6.2. Gradient of the neutral axis depth
large-scale building structures. When using the OC tech- The neutral axis depth, X, obtained from Eq. (15),
nique, it is necessary to re-analyze the structure after each decreases nonlinearly with an increase of t, as shown in
design cycle and to reapply the continuous optimization Fig. 5c. The gradient of Xi for member i can be derived as
process until convergence to the minimum cost design is oX i 2fco0 Bi X 2i oai
obtained. Details of the OC approach are available in Chan ¼ P N s1;i PN sk;i
oti 2ai fco0 Bi X i  P i þ fy k¼1 Ask;i þ ecu;i Es i¼N A oti
and Zou [12], Zou [16] and Zou and Chan [17,18]. Briefly, s1;i þ1 sk;i

this approach is derived from the Kuhn–Tucker conditions ð34Þ


of optimality and can be stated as
PN j ogj where Nsl,i is the number of layers of steel reinforcement in
 j¼1 k
oti j
member i found to have reached yielding.
sij ¼ oF
¼ 1; ði ¼ 1; 2; . . . ; N c ; j ¼ 1; 2; . . . ; N j Þ
oti

ð28Þ 3.6.3. Gradients of the ultimate moment and the ultimate


plastic hinge rotation
where kj is the Lagrange multiplier associated with the jth The ultimate moment Mu in Eq. (16), the ultimate cur-
constraint. As the optimal design is attained, sij approaches vature, /u, in Eq. (17) and the ultimate plastic hinge rota-
unity. A recursive relationship using Eq. (28) was derived tion, hU
p , in Eq. (21) are very sensitive to the change of t, as
by Chan and Zou [12], Zou [16] and Zou and Chan [17,18]. evident in Figs. 5d and e. The gradient of Mu,i can be for-
mulated in terms of ecu,i, ai, Xi, oai/oti and oXi/o ti, all of
3.6.1. Gradients of a and c which can be expressed as functions of ti for member i:
The gradient of the confining pressure, fl,i, for member i  
is easily derived from Eq. (9): oM u;i Di oai
¼ fco0 Bi X i  ci X 2i
oti 2 oti
ofl;i 2Efrp eh;rup
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ a1;i ð29Þ fco0 Bi Di
oti B2 þ D2 þ ai  2ci ai fco0 Bi X i
i i 2
PN sk;i !
In Eq. (29), ofl,i/oti is constant and thus denoted by a1,i for ecu;i Es i¼N d A ðDi =2  d sk;i Þ oX i
s1;i þ1 sk;i sk;i
þ ð35Þ
convenience. The confined concrete strength fcc0 and the X 2i oti
ultimate stain ecu,i, given by Eqs. (7) and (8), are linear
functions of fl,i and hence of ti. Their gradients are as Similarly, the gradient of hU p;i can be formulated in terms of
follows: ti as
"   0:45 #
0
ofcc;i ohUp;i 12k s2;i a1;i eco eh;rup ecu;i oX i
¼ 3:3k s1;i a1;i ð30Þ ¼  2 lp;i ð36Þ
oti oti Xi fco0 eco X i oti
  0:45
oecu;i eco eh;rup
¼ 12k s2;i 0 a1;i ð31Þ
oti fco eco
3.6.4. Gradient of the inelastic displacement
Using Eqs. (5), (30), (31), the gradient of E2,i is derived as According to Eqs. (18)–(20), the first and second deriv-
"   0:45 # atives of the drift uj are given by
0
oE2;i 3:3k s1;i 12k s2;i ðfcc;i  fco0 Þ eco eh;rup
¼  a1;i
oti ecu;i e2cu;i fco0 eco ouj X 2
¼ mjh;i ðM h;i  M y;i Þ
oti
ð32Þ h¼1
"
where fcc0 and ecu,i can be expressed through Eqs. (7) and (8) hU p;i oM u;i
  2
as functions of ti. ðM u;i  M y;i Þ oti
The factors a and c are nonlinear functions of t. As evi- #
1 ohU
p;i
dent in Fig. 5b, c is relatively insensitive to variations of t þ ð37Þ
and hence its gradient is assumed to be zero for simplicity ðM u;i  M y;i Þ oti
(i.e., oc/ot  0), while the gradient of ai for member i can be o2 uj X 2
derived from Eq. (13) as ¼ mjh;i ðM h;i  M y;i Þ
! ot2i h¼1
oai ecu;i 2fco0 1 oE2;i "  2
¼  ð33Þ 2hU
p;i oM u;i
oti 0
2fco 3ecu;i ðEc;i  E2;i Þ 2
oti  3
ðM u;i  M y;i Þ oti
#
Since ecu,i, E2,i and oE2,i/oti are found from Eqs. (8), (9), U
2 oM u;i ohp;i
(5), (7), and (32) to be functions of ti, oai/oti in Eq. (33)  2
ð38Þ
can be expressed in terms of ti. ðM u;i  M y;i Þ oti oti
592 X.K. Zou et al. / Composites: Part B 38 (2007) 584–597

As a result, b1 in Eq. (27a), b2 in Eq. (27b) and ogj/oti in 50kN/m


Eq. (28) can be determined on the basis of Eqs. (37) and
C6 C5

3.6*3=10.8m
(38).
C4 C3
3.7. Overall design optimization procedure
C2 C1
The design optimization procedure is summarized as
6m 6m 6m
follows:
Fig. 7. A three-story three-bay RC frame.
1. Establish an initial design with a given set of member
sizes, steel reinforcement ratios and initial thicknesses
of the FRP jackets for columns. Possibly set lower and 1.1% and a bottom reinforcement ratio of 0.9%. Exterior
upper bound limits of the design variables ti. Determine columns (denoted by C1, C3, and C5) have a 300-mm ·
a design spectrum as the earthquake demand and con- 300-mm cross-section with a 1.25% reinforcement ratio,
duct a static virtual load analysis to obtain the virtual while interior columns (denoted by C2, C4 and C6) have
internal forces in all members; a 400-mm · 400-mm cross-section with a 1% reinforcement
2. Based on Eqs. (5)–(17), (21), calculate the Mu to My ratio. The steel reinforcement in each column is evenly dis-
ratio, the ultimate axial force–moment hinge rotation, tributed on its four sides. The concrete cover to the longi-
hUp , and the axial force-bending moment interaction tudinal steel reinforcement is 25 mm in all beams and
curve for each column cross-section. These values will 30 mm in all columns. The frame needs to be retrofitted
be used for defining the column hinge properties in the to withstand a severe earthquake with a peak acceleration
nonlinear pushover analysis; of 0.5 g. Using the Chinese seismic design code [29], the
3. Carry out a nonlinear pushover analysis (using for shear capacity of the frame was found to be higher than
instance a commercially available software package) to the maximum shear demand. This indicates that flexural
determine the inelastic drift response of the retrofitted failure governs the capacity of this particular frame.
structure at the performance point and to obtain the Wet layup CFRP jackets with fibers oriented in the
actual internal forces of all members; hoop direction only were proposed to retrofit the columns.
4. Track down the locations of the plastic hinges for all The CFRP tensile strength and elastic modulus Efrp are
members, compute the gradients using Eqs. (29)–(38) 3550 MPa and 230 GPa, respectively, based on the fiber
and then determine the first-order and second-order thickness. The FRP jacket was assumed to reach rupture
derivatives of the drift response using Eqs. (27a) and at a hoop tensile stress of 2100 MPa with a corresponding
(27b) for those columns containing plastic hinges; hoop rupture strain eh,rup = 0.00913, based on the well
5. Establish the explicit inter-story drift constraints using a established fact that the hoop rupture strain of an FRP
second-order Taylor series approximation and formu- jacket is significantly lower than that from a flat coupon
late the explicit design problem (Eqs. (24)–(26)); tensile test [30]. The radius of the rounded corners of col-
6. Apply the recursive OC optimization algorithm to mod- umn sections Rc is 50 mm.
ify the FRP thicknesses, and to resize the thicknesses of No lower and upper bounds for the FRP jacket thick-
the FRP jackets for columns and determine the optimal nesses were assumed. Hence, the initial value of the FRP
total material cost of FRP; jacket thickness in the design optimization was set to zero
7. Check convergence of the FRP material cost and of the for all columns. Inter-story drifts were considered, with an
inelastic drift performance of the structure. Terminate assumed drift ratio limit of 1%. The design process was
with the optimal design if convergence of the solution deemed to have converged when the difference in the
is found; otherwise return to Step 2. FRP material cost between two successive design cycles
was within 0.5%.
The pushover analysis of the structure was conducted
4. Numerical example with the SAP2000 program [31] for both the original and
the retrofitted structures. Results of the pushover analysis
4.1. Problem statement can be used to estimate the potential ductility of the struc-
ture, to evaluate its lateral load capacity and to identify the
The proposed optimal design method was applied to a failure mechanism. The loads considered in the pushover
low-rise three-storey RC frame in an office building, shown analysis were lateral seismic loads and vertical gravity
in Fig. 7. The frame was designed using GBJ68-84 [28] only loads. While the lateral loads were applied incrementally,
for gravity loads and was assumed to be located in intensity the gravity loads were maintained unchanged during the
VII seismic zone [29]. The yield strength of the steel rein- nonlinear pushover analysis. The initial lateral loads were
forcement is 300 MPa and the unconfined concrete com- proportional to the product of the story mass and the first
pressive strength is 21 MPa. All beams have a 250-mm · mode shape of the structure. The P-delta effect was not
600-mm cross-section with a top reinforcement ratio of taken into account. Flexural moment hinges and axial
X.K. Zou et al. / Composites: Part B 38 (2007) 584–597 593

2.5 Table 1
Initial and optimal thicknesses of the FRP jackets
2.046
Element Story Member Member sizes Initial Optimal
2 type level group (mm) thickness thickness
FRP Volume(10 m )
3

(mm) (mm)
Width Depth
-2

1.5 Columns 3rd C5 300 300 0.000 0.000


C6 400 400 0.000 0.000
2nd C3 300 300 0.000 0.282
1 C4 400 400 0.000 0.956
1st C1 300 300 0.000 0.773
C2 400 400 0.000 1.792
0.5

0 4.2. Numerical results and discussions


0 1 2 3 4 5 6
Design Cycle The history of the optimization process is illustrated in
Fig. 8 in terms of the total volume of FRP versus the design
Fig. 8. History of the design optimization process.
cycle. It is seen that the optimal design process with the
multiple drift constraints converges after five design cycles,
force–moment hinges were assigned to the end locations of with an increase in the FRP volume from 0.00 to the final
the beams and the columns, respectively. The ultimate 2.046 · 102 m3. The main reason for such rapid conver-
rotation, hU
p , of the flexural moment hinge was assumed
gence is the low sensitivity of the internal forces and
to be equal to 0.02 radian [9], while hU
p of the axial force–
moments in all members to the thicknesses of the FRP
moment hinge was not constant in the design optimization jackets applied to columns. The FRP jacket thicknesses
and varied with the thickness of the FRP jacket as dis- were checked and found to be large enough to ensure that
cussed previously in the paper. k s1 fl =fco0 P 0:07.

a 100 b 8000
C1 for the original frame
Bending Moment M (kN-m)

80 C1 for the retrofitted frame


6000
Axial Force (kN)

60
4000

40
2000
C1 for the original frame
20 C1 for the retrofitted frame
0
0 100 200 300 400
0
0 0.1 0.2 0.3 0.4 -2000
Curvature (1/m) Bending Moment (KN-m)

c 100 d 8000
C3 for the original frame
Bending Moment M (kN-m)

80 6000 C3 for the retrofitted frame


Axial Force (kN)

60
4000

40
2000
C3 for the original frame
20
C3 for the retrofitted frame
0
0 100 200 300 400
0
0 0 .1 0.2 0 .3 0.4 -2000
Curvature (1/m) Bending Moment (KN-m)

Fig. 9. Moment–curvature and moment-axial load interaction curves for exterior columns: (a) Moment–curvature curve of C1 (at the axial load level of
the performance point); (b) moment-axial load interaction curve of C1; (c) moment–curvature curve of C3 (at the axial load level of the performance
point); (d) moment-axial load interaction curve of C3.
594 X.K. Zou et al. / Composites: Part B 38 (2007) 584–597

Table 1 presents the optimal FRP jacket thicknesses for reduced effectiveness of FRP jackets in confining rectangu-
all columns of the three stories. The original frame with no lar/square columns when compared with circular columns.
FRP jackets was found not to comply with the assumed When desirable and feasible, rectangular/square columns
inter-story drift limit. At the end of the optimization pro- may be modified in shape to enhance the effectiveness of
cess, the thicknesses of the FRP jackets on the columns FRP confinement. For example, by modifying a rectangu-
are rather large at the lower levels of the structure, partic- lar section into an elliptical section, the confinement effec-
ularly in the first level, with the FRP jackets on the interior tiveness can be substantially enhanced [32].
columns being thicker than those on the exterior columns Figs. 9 and 10 show the moment–curvature curves (at
due to the larger hinge rotations required. Although the the axial load level corresponding to the performance
overturning action of the lateral loading causes uplifting point) and moment-axial load interaction curves of col-
tension on the left side columns and down-pressing com- umns C1–C4 before and after FRP jacketing. From these
pression on the right side columns, the FRP jacket thick- figures, it is evident that FRP confinement leads to
nesses for the columns to the left of the vertical axis of increases in the strength and the ductility of the columns.
symmetry must be the same as those for the columns to In particular, Figs. 9a and c, 10a and c show that the ulti-
the right of this axis, to account for the reversal of seismic mate moment and the ultimate curvature of the confined
loads. No FRP jackets are needed for the columns at the cross-section are substantially higher than those of the
third level. This indicates that the performance of these col- unconfined cross-sections. The increase in ultimate curva-
umns is mainly controlled by strength requirements under ture results in an enhancement of the hinge rotational
gravity loads, while the performance of the columns at capacity and the energy dissipation capacity, and hence
the first and second levels is governed by inelastic inter- may prevent early collapse and reduce the level of damage.
story drift requirements which can be enhanced by FRP Figs. 9b and d, 10b and d indicate that FRP confinement
confinement. enhances the ultimate axial load and the ultimate moment
The required FRP jacket thicknesses are rather large for of the cross-sections of the columns. In particular, the ulti-
some of the columns (Table 1) and this is a result of the mate moment increases of the interior columns are larger

a b 8000 C2 for the original frame


250
C2 for the retrofitted frame
Bending Moment M (kN-m)

200 6000
Axial Force (kN)

150 4000

100
2000
C2 for the original frame
50 C2 for the retrofitted frame
0
0 100 200 300 400
0
0 0 .1 0.2 0 .3 0.4
-2000
Curvature (1/m) Bending Moment (KN-m)

c d 8000 C4 for the original frame


250
C4 for the retrofitted frame
Bending Moment M (kN-m)

200 6000
Axial Force (kN)

150
4000

100
2000
C4 for the original frame
50 C4 for the retrofitted frame
0
0 0 100 200 300 400
0 0.1 0 .2 0 .3 0 .4
-2000
Curvature (1/m)
Bending Moment (KN-m)

Fig. 10. Moment–curvature and moment-axial load interaction curves for interior columns: (a) Moment–curvature curve of C2 (at the axial load level of
the performance point); (b) moment-axial load interaction curve of C2; (c) moment–curvature curve of C4 (at the axial load level of the performance
point); (d) moment-axial load interaction curve of C4.
X.K. Zou et al. / Composites: Part B 38 (2007) 584–597 595

than those of the exterior columns, due to the axial force drift of the retrofitted frame is smaller than that of the ori-
being larger in the interior than in the exterior columns. ginal frame for any given value of lateral load. For exam-
The larger the axial load, the larger the increase in moment ple, at a lateral load of 280 kN, the third story drift is equal
capacity due to FRP confinement. to 0.064 m for the original frame and 0.052 m for the retro-
Fig. 11a shows the base shear-top displacement relation- fitted frame. Finally, the top displacement of the retrofitted
ship for the original frame (labeled as ‘‘initial’’) and the ret- frame at ultimate (0.15 m) is larger than that of the original
rofitted frame (labeled as ‘‘final’’). The original frame frame (0.10 m), which indicates that retrofit produces a lar-
sustains a lateral load (base shear) at first yielding of steel ger displacement ductility of the frame as a result of the lar-
of 82.89 kN and an ultimate lateral load of 285.36 kN. The ger curvature ductility of the confined cross-sections.
ultimate lateral load of the retrofitted frame is about 9% In Fig. 11b, the performance points of the original and
higher (being equal to 312.04 kN). This indicates that the the retrofitted frames are indicated. The A–B–C–D curve
retrofit design obtained through the proposed optimization represents the performance of the initial structure, with a
procedure results in a higher lateral load-carrying capacity. spectral acceleration at first yielding of 0.035 g (at point
The two curves for the original and the retrofitted frames B). Point ‘‘D’’ (corresponding to a spectral acceleration
are coincident up to the first yielding of steel, and this is of 0.097 g and to a spectral displacement of 0.076 m) repre-
the direct result of the assumed moment–rotation relation- sents the performance point of the original structure. The
ships (e.g. FRP-confinement has no effect on the slope of A–B–E–F curve is the capacity curve of the optimized ret-
the moment–rotation curve before first yielding). This rofitted structure. Retrofit leads to no appreciable increase
assumption is a good approximation of the real behavior in the spectral acceleration at first yielding (point B). The
that confinement of the columns produces little stiffness final performance point, ‘‘F’’, corresponds to a spectral
increase of the structure in the pre-yielding phase. This is acceleration of 0.111 g and a spectral displacement of
a distinct advantage of retrofit by FRP confinement of col- 0.072 m. The increase in the spectral acceleration capacity
umns, as an increase in stiffness of the structure would from 0.097 g to 0.111 g indicates that the optimized retro-
attract higher seismic forces and hence decrease the effec- fitted structure attracts an increase in the seismic load
tiveness of retrofit. Beyond the first yielding of steel, the action. The decrease in the spectral displacement from
0.076 m to 0.072 m results from the optimization of the
inelastic drift response through the optimal sizing of the
FRP jackets.
a 350
Final curve E F Fig. 12 shows the inter-story drift ratios of the original
300 (0.15, 312.04)
C D
and the retrofitted frames at the performance points. The
original first-floor inter-story drift ratio is seen to violate
Base Shear (kN)

250 (0.10, 285.36)


substantially the limit of 1% due to the initial weak, uncon-
200
Initial curve fined columns, resulting in a non-ductile soft-story or
150 column side-sway collapse mechanism at the initial
100
performance point. However, the inter-story drift ratios
B of the retrofitted frame at the first and second floors are
50 (0.01, 82.89)
close to the upper limit values, indicating that a uniform
0
A
inter-story drift distribution has been achieved as a result
0.00 0.05 0.10 0.15 0.20 of optimization and the occurrence of a weak story mech-
Top Displacement (m) anism has been prevented. Confinement of the columns
0.2
b
0.18
Final demand
4
0.16 Drift limit ratio = 1/100
Spectral Acceleration (g)

Final performance point Initial drift ratio


Initial demand
0.14 (0.072,0.111g)
Final drift ratio
0.12 Final capacity F
E 3
0.1
Story Level

C D
0.08
Initial capacity Initial performance point 2
0.06
(0.076,0.097g)
0.04
B
0.02 (0.01,0.035g) 1
0
A
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Spectral Displacement (m) 0
0 0 . 005 0. 01 0. 015
Fig. 11. Results of pushover analysis for the original and the retrofitted Interstory Drift Ratio
frames: (a) Base shear-top displacement relationship; (b) acceleration–
displacement spectrum. Fig. 12. Inter-story drift ratios at the performance points.
596 X.K. Zou et al. / Composites: Part B 38 (2007) 584–597

frame, most hinges are located in the columns. Almost all


the columns of the first and second floors exhibit hinges
at both ends and their hinge rotations are large, particu-
larly at the base of the first story. The rotations of two plas-
tic hinges at the base columns exceed the specified
threshold of plastic rotation, hU p , as shown in Fig. 13a.
These large hinge rotations lead to a brittle column side-
sway mechanism at the first floor level. Conversely, the
frame with the optimal retrofit design shows an improved
plastic hinge distribution with more hinges in beams and
fewer hinges in columns, and with no indication of soft-
story behavior. All plastic hinge rotations are smaller than
hUp and only one beam is close to its rotation capacity, as
seen in Fig. 13b. Hence, the seismic performance of a frame
governed by a soft-story mechanism has been enhanced by
FRP confinement of some of the columns. The rotation
capacities of the hinges in the columns now exceed the
demands and the failure mode has shifted from a column
side-sway mechanism to a mechanism with acceptable
story deformation levels and weak beam-strong column
behavior.

5. Summary and conclusions

This paper has presented an optimization technique for


the performance-based seismic FRP retrofit design of RC
building frames. The thicknesses of FRP jackets used for
the confinement of columns are taken as the design vari-
ables, and minimizing the volume and hence the material
cost of the FRP jackets is the design objective in the opti-
mization procedure. A simple yet accurate stress–strain
model for FRP-confined concrete is employed to formulate
the ultimate moment and rotation expressions of the axial
force–moment hinge assigned at both ends of each FRP-
confined column explicitly in terms of the design variables.
Then, by tracking the formation of the plastic hinges gen-
erated in the pushover analysis, the inelastic drift is also
Fig. 13. Plastic hinge distributions of the original and the retrofitted
frames: (a) Original frame; (b) retrofitted frame; (c) definition of hinge
expressed explicitly in terms of the design variables using
status. the principle of virtual work and the Taylor series approx-
imation. The OC approach is employed for finding the
solution of the nonlinear seismic drift design problem.
results in a lower value for the story drift ratio at the first The effectiveness of the proposed optimal design method
level at the performance point. The inter-story drift ratio of has been demonstrated by a numerical example of a planar
the third floor level is seen to be smaller than the 1% limit, frame. The example also demonstrated that the seismic
resulting from the fact that the drift is governed by member resistance of an RC frame designed for gravity loads only
sizes and steel reinforcement ratios. Confinement is clearly can be substantially enhanced through confinement of col-
shown to improve the behavior of the frame, and the opti- umns using FRP jackets. FRP confinement enhances the
mization method proves to have been able to automatically strength of columns but has little effect on their stiffness,
size the FRP jacket thicknesses of all columns to attain a which is an important advantage in seismic retrofit as lar-
uniform ductility demand. ger stiffnesses lead to higher seismic forces. The optimiza-
Figs. 13a and b show the plastic hinge distributions tion technique is able to design the FRP thicknesses in
under the pushover loading at the performance points of the columns so as to efficiently distribute and control the
the structure before and after retrofit, whilst Fig. 13c shows damage throughout the structure, changing the failure
the definition of hinge status in color.1 In the original mode from a column side-sway mechanism to a mechanism
with acceptable story deformation levels and weak beam-
1
For interpretation of color in Figs. 5, 8–13 the reader is referred to the strong column behavior. The procedure is thus capable of
web version of this article. producing retrofit designs that minimize the total material
X.K. Zou et al. / Composites: Part B 38 (2007) 584–597 597

cost of FRP while optimizing the inelastic drift perfor- [14] Arora JS. Optimization of structures subjected to dynamic loads. In:
mance of the structure. Leondes CT, editor. Structural dynamic systems computational
techniques and optimization. Gordon and Breach Science Publ.;
1999. p. 1–73.
Acknowledgements [15] Foley CM. Optimized performance-based design for buildings. In:
Burns SA, editor. Recent advances in optimal structural design.
The authors are grateful for the financial support re- American Society of Civil Engineers; 2002. p. 169–240.
[16] Zou XK. Optimal seismic performance-based design of reinforced
ceived from the Research Grants Council of the Hong
concrete buildings, PhD Dissertation, Hong Kong, China: Hong
Kong SAR (Project No: PolyU 5059/02E) and The Hong Kong University of Science and Technology; 2002.
Kong Polytechnic University provided through its Area [17] Zou XK, Chan CM. Optimal seismic performance-based design of
of Strategic Development (ASD) Scheme for the ASD in reinforced concrete buildings using nonlinear pushover analysis. Eng
Urban Hazard Mitigation. Struct 2005;27:1289–302.
[18] Zou XK, Chan CM. An optimal resizing technique for seismic drift
design of concrete buildings subjected to response spectrum and time
References history loadings. Comput Struct 2005;83:1689–704.
[19] Martinez-Rodrigo M, Romero ML. An optimum retrofit strategy for
[1] Ghobarah A, El-Attar M, Aly NM. Evaluation of retrofit strategies moment resisting frames with nonlinear viscous dampers for seismic
for reinforced concrete columns: a case study. Eng Struct 2000;22(5): applications. Eng Struct 2003;25(7):913–25.
490–501. [20] Saadatmanesh H, Ehsani MR, Jin LL. Repair of earthquake-
[2] Bracci JM, Reinhorn AM, Mander JB. Seismic retrofit of reinforced damaged RC columns with FRP wraps. ACI Struct J 1997;94(2):
concrete buildings designed for gravity loads: Performance of 206–15.
structural model. ACI Struct J 1995;92(6):711–23. [21] Seible F, Priestley MJN, Hegemier GA, Innamorato D. Seismic
[3] SEAOC, Vision 2000, Performance Based Seismic Engineering of retrofit of RC columns with continuous carbon fiber jackets. J
Buildings, part 2: conceptual framework, structural engineers asso- Compos Construct ASCE 1997;1(2):52–62.
ciation of California, Sacramento, CA; 1995. [22] Kobatake Y. A seismic retrofitting method for existing rein-
[4] Kunnath SK, Hoffmann G, Reinhorn AM, Mander JB. Gravity load- forced concrete structures using CFRP. Adv Compos Mater
designed reinforced-concrete buildings – 2. Evaluation of detailing 1998;7(1):1–22.
enhancements. ACI Struct J 1995;92(4):470–8. [23] Saiidi MS, Martinovic F, McElhaney B, Sanders D, Gordaninejad F.
[5] Teng JG, Chen JF, Smith ST, Lam L. FRP-strengthened RC Assessment of steel and fiber reinforced plastic jackets for seismic
structures. New York: Wiley; 2002. retrofit of reinforced concrete columns with structural flares. J Struct
[6] Teng JG, Chen JF, Smith ST, Lam L. Behavior and strength of FRP- Eng ASCE 2004;130(4):609–17.
strengthened RC structures: a state-of-the-art review. Proc Inst Civil [24] De Lorenzis L, Tepfers R. A comparative study of models on
Eng-Struct Build 2003;156(1):51–62. confinement of concrete cylinders with FRP composites. J Compos
[7] Teng JG, Lam L. Behavior and modeling of fiber reinforced polymer- Construct ASCE 2003;7(3):219–37.
confined concrete. J Struct Eng ASCE 2004;130(11):1713–23. [25] Lam L, Teng JG. Design-oriented stress–strain model for FRP-
[8] Xiao Y. Application of FRP composites in concrete columns. Adv confined concrete. Construct Build Mater 2003;17:471–89.
Struct Eng 2004;7(4):335–43. [26] Lam L, Teng JG. Design-oriented stress–strain model for FRP-
[9] ATC-40. Seismic evaluation and retrofit of concrete buildings, vol. 1 confined concrete in rectangular columns. J Reinf Plast Compos
(ATC-40), Report No. SSC 96-01, Redwood City, CA: Applied 2003;22(13):1149–84.
Technology Council; 1996. [27] Yuan YF, Lam L, Xia SH, Smith. Analysis and behavior of FRP-
[10] FEMA 274. NEHRP guidelines for the seismic rehabilitation of confined short concrete columns subject to eccentric loading. To be
buildings (FEMA 273). Washington, DC, USA: Building Seismic published.
Safety Council; 1997. [28] GBJ68-84. Chinese Code for Design of Building Structures. Beijing,
[11] Moehle JP, Mahin SA. Observations on the behavior of reinforced China; 1984.
concrete buildings during earthquakes. In: Ghosh SK, editor. [29] GBJ11-89. Chinese Code for Seismic Design of Buildings. Beijing,
Earthquake-resistant concrete structures – inelastic response and China; 1989.
design. American Concrete Institute SP-127; 1991. [30] Lam L, Teng JG. Ultimate condition of FRP-confined concrete. J
[12] Chan CM, Zou XK. Elastic and inelastic drift performance optimi- Compos Construct ASCE 2004;8(6):539–48.
zation for reinforced concrete building under earthquake loads. [31] CSI. SAP2000/NL-PUSH Software, Version 7.40, Computer and
Earthquake Eng Struct Dyn 2004;33(8):929–50. Structures, Inc., Berkeley, California, USA; 2000.
[13] Beck JL, Papadimitriou C, Chan E, Irfanoglu A. A performance- [32] Teng JG, Lam L. Compressive behaviour of carbon fiber reinforced
based optimal structural design methodology. Report No. EERL 97- polymer-confined concrete in elliptical columns. J Struct Eng ASCE
03, CA, USA; 1998. 2002;128(12):1535–43.

Potrebbero piacerti anche