Sie sind auf Seite 1von 20

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2004; 33:687–706 (DOI: 10.1002/eqe.369)

Behavior of moment-resisting frame structures subjected to


near-fault ground motions

Babak Alavi1; ∗; † and Helmut Krawinkler2


1 Exponent Failure Analysis Associates; 149 Commonwealth Drive; Menlo Park; CA 94025-1133; U.S.A.
2 Department of Civil and Environmental Engineering; Stanford University; Stanford; CA 94305-4020; U.S.A.

SUMMARY
Near-fault ground motions impose large demands on structures compared to ‘ordinary’ ground motions.
Recordings suggest that near-fault ground motions with ‘forward’ directivity are characterized by a
large pulse, which is mostly orientated perpendicular to the fault. This study is intended to provide
quantitative knowledge on important response characteristics of elastic and inelastic frame structures
subjected to near-fault ground motions. Generic frame models are used to represent MDOF structures.
Near-fault ground motions are represented by equivalent pulses, which have a comparable eect on
structural response, but whose characteristics are dened by a small number of parameters. The results
demonstrate that structures with a period longer than the pulse period respond very dierently from
structures with a shorter period. For the former, early yielding occurs in higher stories but the high
ductility demands migrate to the bottom stories as the ground motion becomes more severe. For the
latter, the maximum demand always occurs in the bottom stories. Preliminary regression equations are
proposed that relate the parameters of the equivalent pulse to magnitude and distance. The equivalent
pulse concept is used to estimate the base shear strength required to limit story ductility demands to
specic target values. Copyright ? 2004 John Wiley & Sons, Ltd.

KEY WORDS: near-fault; near-eld; near-source, pulse; frame structures; seismic demands

INTRODUCTION

Ground motion recordings have provided increasing evidence that ground shaking near a
fault rupture may be characterized by a large, long-period pulse, capable of causing severe
structural damage. This holds true particularly in the ‘forward’ direction, where the fault

∗ Correspondence
to: Babak Alavi, Exponent Failure Analysis Associates, 149 Commonwealth Drive, Menlo Park,
CA 94025-1133, U.S.A.
† E-mail:
balavi@exponent.com

Contract=grant sponsor: CUREE=Kajima Research Program


Contract=grant sponsor: U.S. National Science Foundation, US–Japan Cooperative Research Program in Urban
Hazard Mitigation; contract=grant number: CMS-9812478
Contract=grant sponsor: California Department of Conservation, SMIP 1997 Data Interpretation Project;
contract=grant number: 1097-601
Received 11 February 2003
Revised 11 September 2003
Copyright ? 2004 John Wiley & Sons, Ltd. Accepted 3 October 2003
688 B. ALAVI AND H. KRAWINKLER

rupture propagates towards the site at a speed close to the shear wave velocity. As a result,
most of the seismic energy from the rupture arrives within a short time at the beginning of the
record [1]. The radiation pattern of shear dislocation around the fault causes the fault-normal
component to be typically more severe than the fault-parallel component [2]. This phenomenon
aects the response attributes of structures located in the near-fault region, which is assumed
to extend 10 to 15 km from the seismic source [3], and therefore requires consideration in the
design process.
Recent structural design codes, e.g. the 1997 Uniform Building Code [4], have recognized
near-fault eects by introducing source type and distance dependent near-fault factors to the
customary design spectrum. However, these factors are believed to be inadequate to provide
consistent protection because they pay little attention to the physical response characteristics
of near-fault ground motions. Moreover, emerging concepts of performance-based seismic
design require a quantitative understanding of response covering the range from nearly elastic
behavior to highly inelastic behavior. Much work is needed to identify and quantify the site-
dependent characteristics of near-fault ground motions and to address issues concerning the
response of dierent types of structures to these ground motions.
This paper summarizes the results of a study that aims to provide quantitative knowledge
on the response of frame structures in the near-fault region of active faults. The objective is
to identify salient response characteristics, to describe near-fault ground motions by simple
equivalent pulses, and to use the pulse response characteristics in order to represent behavior
attributes of structures subjected to near-fault ground motions. It is recognized that the near-
fault problem is very complex, and that much more research and data are needed before a
comprehensive understanding of all important aspects of the problem will be accomplished.
The work summarized here attempts to address the most important issues concerning near-
fault ground motions and their response attributes in order to form a foundation on which to
base future research and development of design guidelines.

GROUND MOTIONS USED IN THIS STUDY

A set of 15 near-fault ground motion records with forward directivity is used to evaluate
elastic and inelastic demands of SDOF (single degree of freedom) and MDOF (multi degree
of freedom) structures. These ground motions, which were assembled for the SAC Steel
Project [5] and the CDMG Strong Motion Instrumentation Program [2], are either recorded
on sti soil or have been modied to NEHRP [6] Soil Type D conditions. Table I lists the
designation and basic properties of the recorded motions. These motions cover a moment
magnitude range from 6.2 to 7.3 and a rupture distance (closest distance from site to fault
rupture plane) range from 0.0 to 8:9 km. To augment the relatively small record set, a set of
simulated ground motions is also used in the investigation, which covers systematic ranges of
moment magnitude (6.5, 7.0 and 7.5) and rupture distance (3, 5 and 10 km) for two stations
(f6 and f8) in the forward direction of a strike-slip fault. These records were generated for
a project sponsored by the CDMG Strong Motion Instrumentation Program [2]. Records with
backward directivity are typically less severe and do not exhibit pulse-type characteristics [1];
therefore, this study only focuses on ground motions with forward directivity.
Figure 1 illustrates ground acceleration, velocity and displacement time history traces for
the fault-normal component of a typical near-fault ground motion with forward directivity

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
BEHAVIOR OF MOMENT-RESISTING FRAME STRUCTURES 689

Table I. Designation and basic properties of recorded ground motions used in this study.
Designation Earthquake Station Mw R (km)

LP89lgpc Loma Prieta, 1989 Los Gatos 7.0 3.5


LP89lex Loma Prieta, 1989 Lexington 7.0 6.3
EZ92erzi Erzincan, 1992 Erzincan 6.7 2.0
LN92lucr Landers, 1992 Lucerne 7.3 1.1
NR94rrs Northridge, 1994 Rinaldi 6.7 7.5
NR94sylm Northridge, 1994 Olive View 6.7 6.4
KB95kobj Kobe, 1995 JMA 6.9 0.6
KB95tato Kobe, 1995 Takatori 6.9 1.5
IV79ar06 Imperial Valley, 1979 Array 6 6.5 1.2
IV79melo Imperial Valley, 1979 Meloland 6.5 0.0
KB95kpi1 Kobe, 1995 Port Island 6.9 3.7
MH84andd Morgan Hill, 1984 Anderson D 6.2 4.5
MH84cyld Morgan Hill, 1984 Coyote L D 6.2 0.1
NR94newh Northridge, 1994 Newhall 6.7 7.1
NR94spva Northridge, 1994 Sepulveda 6.7 8.9

NR94rrs, Fault-Normal
1000
max = 873

500
ag (cm / sec2)

-500

-1000
200
max = 174

100
vg (cm / sec)

-100

-200
40
max = 38.3

20
ug (cm)

-20

-40
0 2 4 6 8 10 12 14
Time (sec)

Figure 1. Ground acceleration, velocity and displacement time histories for record NR94rrs.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
690 B. ALAVI AND H. KRAWINKLER

NR94rrs, ξ = 2%
3
Fault-Normal
2.5 Fault-Parallel
0.707(FN+FP)
2 0.707(FN-FP)

Sa (g)
1.5

0.5

0
400
Fault-Normal
Fault-Parallel
300 0.707(FN+FP)
Sv (cm / sec)

0.707(FN-FP)

200

100

0
100
Fault-Normal
Fault-Parallel
80 0.707(FN+FP)
0.707(FN-FP)
60
Sd (cm)

40

20

0
0 0.5 1 1.5 2 2.5 3 3.5 4
T (sec)

Figure 2. Acceleration (elastic strength demand), velocity and displacement spectra for record NR94rrs.

(NR94rrs = Northridge 94 Rinaldi Receiving Station). As indicated particularly by the velocity


and displacement traces, the record contains a large pulse within the time range from about
2 to 3 sec. Figure 2 presents acceleration (elastic strength demand), velocity and displacement
spectra for the fault-normal, fault-parallel and two 45-degree rotated components of the same
ground motion. The gure indicates that the fault-normal component is much more severe
than the fault-parallel component. When these two components are rotated by 45 degrees,
the dierence in the spectra becomes smaller, but one of the two rotated components still
will impose demands close to (and sometimes even higher than) those associated with the
fault-normal component. A more comprehensive study of the orientation eect conrmed this
observation [7]. Thus, when a structure composed of frames in two perpendicular directions is
subjected to a near-fault ground motion, frames in one of the two directions will be subjected
to excitations almost as severe as the fault-normal component. For this reason this study
focuses on the fault-normal component of near-fault ground motions.
To put the severity of near-fault ground motions in perspective with the severity of ground
motions represented by current codes, a reference set of 15 ‘ordinary’ records (recorded
outside the near-fault region) is used for comparison purposes. These records, which were
used in past studies [8], are scaled such that the spectrum of each individual record matches

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
BEHAVIOR OF MOMENT-RESISTING FRAME STRUCTURES 691

Elastic SDOF Strength Demand Spectra


15-D* Records and UBC 97 Soil Type SD, ξ = 5%
2
UBC 97

1.6 15-D* (mean)

1.2

Sa (g)
0.8

0.4

0
0 0.5 1 1.5 2 2.5 3 3.5 4
T (sec)

Figure 3. Mean acceleration (elastic strength demand) spectrum for reference record set, 15-D*.

Elastic SDOF Velocity Demands


15-D* vs. Recorded Near-Field, ξ = 2%
800
LP89lgpc LP89lex
EZ92erzi LN92lucr
NR94rrs NR94sylm
KB95kobj KB95tato
15-D* (mean)
600
Sv (cm / sec)

400

200

0
0 0.5 1 1.5 2 2.5 3 3.5 4
T (sec)

Figure 4. Velocity response spectra for near-fault and reference ground motions.

the UBC 97 soil type SD spectrum with a minimum error, using discrete periods in the range
of 0.6 to 4:0 sec (constant velocity range). The mean acceleration spectrum of the 15 scaled
records, referred to as 15-D* (mean), is shown in Figure 3 together with the UBC 97 spectrum
for Seismic Zone 4 (without a near-fault factor). Thus, on average the 15-D* time histories
reasonably represent the UBC design level.
Figure 4 illustrates the mean velocity spectrum of the 15-D* records superimposed on the
velocity spectra of several of the recorded near-fault time histories. This gure is presented for
two reasons: rst, to illustrate the great variability in near-fault response spectra and second,
to put the severity of near-fault ground motions in perspective with present design ground
motions. The gure demonstrates that near-fault records can impose very large demands that
should be taken into account in the design process.
An important observation from the near-fault spectra (Figures 2 and 4) is the existence
of a predominant peak in the velocity spectrum of most of the near-fault records—although
some of the records have more than one clear velocity peak. The predominant peak of the
velocity spectrum is used later to estimate the period of the pulse contained in the near-fault
record.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
692 B. ALAVI AND H. KRAWINKLER

MDOF SYSTEMS USED IN THIS STUDY

In order to quantify seismic demands, MDOF structures are represented by a generic


20-story, single-bay frame whose fundamental elastic period and base shear strength are var-
ied. The base shear strength is dened by the coecient  = Vy =W , where Vy is the base shear
strength and W is the weight of the structure. The story stinesses and shear strengths are
determined from a lateral load pattern that corresponds to story shear forces resulting from
the SRSS combination of modal responses for a constant velocity response spectrum, e.g.,
the UBC design spectrum for T ¿Ts [4]. This load pattern, which will be referred to as the
‘SRSS pattern’, is in line with current seismic design load patterns.
The variation of the moments of inertia of the frame members over the height is tuned
such that a linear static analysis of the frame subjected to the SRSS load pattern results in a
straight-line deected shape. The bending strengths of the elements are tuned such that under
the SRSS pattern all plastic hinges, which are allowed only at ends of the beams and at the
base of the columns, form simultaneously. Thus, the story shear strength pattern follows the
same distribution as the design SRSS pattern, which implies a constant overstrength factor
for each story. A bilinear non-degrading hysteresis model with a 3% strain-hardening ratio
is used at all plastic hinge locations. In the time history analyses, Rayleigh damping is used
to obtain a damping ratio of 2% at the rst mode period T and at 0:1T . The low damping
is used because the emphasis is on the behavior of steel frame structures, and because the
maximum response of structures subjected to pulse-type excitations is relatively insensitive to
the damping ratio [9].
This paper does not address P-delta eects. However, it must be emphasized that other
results of this study have demonstrated that P-delta may have a signicant eect on structural
response to near-fault ground motions and generally deserve serious consideration. The second-
order eects may be important even in cases in which current design provisions do not require
any measures to account for P-delta eects [7].

RESPONSE TO NEAR-FAULT GROUND MOTIONS

The story ductility ratio, dened as the maximum story drift normalized by the story yield
drift, i.e. i = max; i =y; i , is used to quantify the response of MDOF structures to near-fault
ground motions. Figure 5 illustrates distributions of story ductility demands over the height of
the MDOF structure previously introduced subjected to the near-fault records whose velocity
response spectra are shown in Figure 4. The demands are computed for MDOF systems with
a fundamental period T = 2:0 sec. and base shear strength coecients of  = 0:4 and  = 0:15,
which correspond to a relatively strong and a relatively weak structure, respectively. The
relatively long period of 2:0 sec. is chosen because, as will be shown later, the period of
the pulse contained in most of the near-fault records is shorter than 2:0 sec. For comparison
purposes, the mean story ductility demands obtained from the reference record set 15-D* are
superimposed.
An important observation is that for most of the near-fault records, the maximum story
ductility demand occurs in the upper portion of the structure when the structure is strong
(large ). However, a migration of ductility demands toward the base takes place when
the structure becomes weaker (small ). The non-uniformity of the story ductility demands

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
BEHAVIOR OF MOMENT-RESISTING FRAME STRUCTURES 693

Story Ductility Demands Story Ductility Demands


15-D* vs. Recorded Near-Field, T = 2.0 sec, γ = 0.40 15-D* vs. Recorded Near-Field, T = 2.0 sec, γ = 0.15
1 1
LP89lgpc LP89lgpc
LP89lex LP89lex
EZ92erzi EZ92erzi
0.8 0.8 LN92lucr
LN92lucr
Relative Height

Relative Height
NR94rrs NR94rrs
NR94sylm NR94sylm
0.6 KB95kobj
0.6 KB95kobj
KB95tato KB95tato
15-D* (mean) 15-D* (mean)
0.4 0.4

0.2 0.2

0 0
0 1 2 3 4 5 0 2 4 6 8 10 12
(a) Story Ductility Ratio, µi = δmax,i / δ y,i (b) Story Ductility Ratio, µi = δmax,i / δ y,i

Figure 5. Story ductility demands for several near-fault and reference ground motions,
T = 2:0 sec: (a)  = 0:40; and (b)  = 0:15.

Story Ductility Demands Story Ductility Demands


NR94rrs, T = 0.5 sec, without P-∆ NR94rrs, T = 2.0 sec, without P-∆
1 1
γ = 2.00 γ = 0.80
γ = 1.50 γ = 0.50
γ = 1.00 γ = 0.40
0.8 γ = 0.80 0.8 γ = 0.25
γ = 0.60 γ = 0.20
Relative Height

Relative Height

γ = 0.50 γ = 0.15
0.6 γ = 0.40 0.6 γ = 0.10
γ = 0.30 γ = 0.07
γ = 0.25 γ = 0.05
0.4 0.4

0.2 0.2

0 0
0 4 8 12 16 20 0 4 8 12 16 20
(a) Story Ductility Ratio, µi = δ max,i / δ y,i (b) Story Ductility Ratio, µi = δmax,i / δ y,i

Figure 6. Dependence of distribution of story ductility demands on base shear strength for
record NR94rrs: (a) T = 0:5 sec; and (b) T = 2:0 sec.

observed for individual near-fault records holds true also in the mean, whereas the same
SRSS-based story shear strengths on average result in a relatively uniform ductility distribu-
tion for ordinary ground motions (15-D*). Figure 5 also indicates the severity of near-fault
ground motions in comparison with mean demands for the ordinary ground motions scaled to
the UBC spectrum.
The strength and period dependence of story ductility distributions is illustrated in Figure 6
for the record NR94rrs. The distributions are shown for relatively short (T = 0:5 sec:) and
relatively long (T = 2:0 sec:) period structures whose base shear strength coecient, , is
varied progressively. As with the previous observation from Figure 5, for the structures with
T = 2:0 sec. (a period that clearly exceeds the eective pulse period of 1:0 sec. for NR94rrs),
maximum ductility demands occur in the upper stories when the structure is strong. With
a reduction in the base shear strength, the ductility demands in the upper portion stabilize
and grow no more. Further strength reductions result in a migration of the maximum demand

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
694 B. ALAVI AND H. KRAWINKLER

Normalized Elastic Story Shear Demands


NR94rrs, without P-
P-∆, Time History
1

0.8

Relative Height
0.6 Design
T = 0.5 sec
T = 1.0 sec
0.4
T = 1.5 sec
T = 2.0 sec
0.2 T = 2.5 sec
T = 3.0 sec
T = 4.0 sec
0
0 0.2 0.4 0.6 0.8 1 1.2
Vi,max / Vbase,max

Figure 7. Normalized elastic story shear demands for record NR94rrs.

Maximum MDOF Story Ductility Demands


NR94rrs, SRSS Pattern, without P-∆
2
T = 0.38 sec
T = 0.50 sec

1.5 T = 0.75 sec


T = 1.00 sec
γ = Vy / W

T = 1.50 sec

1 T = 2.00 sec
T = 3.00 sec

0.5

0
0 4 8 12 16 20
Maximum Story Ductility, µmax

Figure 8. Base shear strength vs. maximum story ductility demands for record NR94rrs.

toward the base. This phenomenon is not observed for the structures with T = 0:5 sec:, a period
that is shorter than the eective pulse period. For these structures the maximum ductility
demands occur close to the base regardless of the base shear strength.
The reason for the early inelastic behavior in the top portion of long-period, strong struc-
tures lies in the distribution of elastic story shear demands. Figure 7 illustrates such distri-
butions over the height of structures with various periods subjected to the record NR94rrs.
Superimposed is also the SRSS story shear strength distribution used to design the MDOF sys-
tems (denoted as ‘Design’). All story shears are normalized by their corresponding base shear
values. While the shear distributions for short-period systems (T 61:0 sec:) are relatively
smooth and resemble the design distribution, their long-period counterparts exhibit the ef-
fect of a wave traveling up the structure, causing distributions that deviate considerably from
the design shear distribution [7]. The elastic story shear demands are particularly large in
the upper portion of long-period structures (T 1:0 sec:), where the shear demands reach the
corresponding story shear capacity rst, resulting in early yielding in the upper stories.
A comprehensive assessment of maximum story ductility demands (maximum over all sto-
ries) of the generic MDOF frames subjected to the near-fault record NR94rrs can be obtained
from the –max curves presented for various structure periods in Figure 8. It is noted that the

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
BEHAVIOR OF MOMENT-RESISTING FRAME STRUCTURES 695

1.5 1.5

1 1

0.5 0.5
ag / ag,max

ag / ag,max
0 0

-0.5 -0.5

-1 -1

-1.5 -1.5
1.5 1.5
vg,max = ag,max.Tp / 4 vg,max = ag,max.Tp / 4
1 1

0.5 0.5
vg / vg,max

vg / vg,max
0 0

-0.5 -0.5

-1 -1

-1.5 -1.5
1.5 1.5

1 ug,max = ag,max.Tp2 / 16 1
ug,max = ag,max.Tp2 / 32

0.5 0.5
ug / ug,max

ug / ug,max

0 0

-0.5 -0.5

-1 -1

-1.5 -1.5
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
(a) t / Tp (b) t / Tp

Figure 9. Acceleration, velocity and displacement time histories of basic pulses:


(a) pulse P2; and (b) pulse P3.

–max curves for systems with T = 1:5 and 2:0 sec. include a range with a very steep slope at
=∼ 4, which corresponds to the migration of ductility demands from the upper stories to the
bottom story (see Figure 6(b)). The steep slope implies that for a signicant range of base
shear strength (represented by ) the maximum story ductility demand remains about the same.

RESPONSE TO PULSE INPUT


As pointed out in the introduction, ground shaking in the forward directivity region of a fault
rupture is characterized by a large pulse early in the time history. If simple pulse models can
be introduced to represent near-fault ground motions with reasonable accuracy, the process
of design and response evaluation will be greatly facilitated because of common patterns for
structures whose fundamental period T has a specic relation to the pulse period Tp .
This study focuses on three basic pulses for response evaluations: half pulse (P1), full
pulse (P2) and multiple pulse (P3). Ground acceleration, velocity and displacement time
histories for pulses P2 and P3 are shown in Figure 9. Pulse P1, whose time histories are
not shown, consists of only the rst half of pulse P2. These pulses are fully dened by
a pulse shape and two parameters: the pulse period Tp (duration of a full velocity cycle)
and an intensity measure, which can be either the maximum pulse acceleration, ag; max , or the

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
696 B. ALAVI AND H. KRAWINKLER

8
P1
P2
6 P3

g,max
S /a
4

a
2

0
5
P1
P2
4 P3
g,max

3
S /V

2
v

0
6
P1
P2
5
P3

4
g,max

3
S /u

2
d

0
0 0.5 1 1.5 2 2.5 3
T / Tp

Figure 10. Acceleration (elastic strength demand), velocity and displacement spectra for basic pulses.

maximum pulse velocity, vg; max (= ag; max Tp =4). The authors have shown that these square-wave
acceleration pulses are adequately capable of representing other pulse shapes such as triangular
acceleration pulses [7]. Other researchers have also investigated various pulse shapes such as
sine-wave acceleration pulses to represent both fault-normal and fault-parallel components of
ground motions with forward directivity [10–12]. The great advantage of the square pulse
is the unambiguous denition of maximum acceleration and the simple relationship between
maximum acceleration and maximum velocity.
The elastic response spectra for the three basic pulses are presented in Figure 10. The
period axis is normalized by the pulse period Tp , and the spectral ordinates are normalized
by their corresponding peak time history value. The velocity spectra for P2 and P3 show a
peak at T=Tp = 1:0, indicating that the maximum spectral velocity for these pulses occurs at
a period that matches well with the pulse period. This property can be taken advantage of
to estimate the equivalent pulse period for near-fault records. The spectra also show larger
spectral peaks for pulse P3 than for the other two pulses, particularly in the displacement
domain. But in this domain it must be considered that for a given pulse intensity (e.g. ag; max ),
the peak ground displacement for P3, used to normalize the displacement spectrum, is only
half of that for P2 (see Figure 9). Nevertheless, it is evident that the multiple pulse P3 is
more damaging than a single pulse.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
BEHAVIOR OF MOMENT-RESISTING FRAME STRUCTURES 697

Story Ductility Demands Story Ductility Demands


Pulse P2, SRSS Pattern, T / Tp = 0.5, without P-∆ Pulse P2, SRSS Pattern, T / Tp = 2.0, without P-∆
1 1
η= 2.50 η = 1.25
η= 2.00 η = 1.00
0.8 η= 1.50 0.8 η = 0.75
η= 1.25 η = 0.50
Relative Height

Relative Height
η= 1.00 η = 0.25
0.6 η= 0.75 0.6 η = 0.15
η= 0.60 η = 0.10
η= 0.50 η = 0.07
η= 0.40 η = 0.05
0.4 0.4

0.2 0.2

0 0
0 4 8 12 16 20 0 4 8 12 16 20
(a) Story Ductility Ratio, µi = δ max,i / δ y,i (b) Story Ductility Ratio, µi = δ max,i / δ y,i

Figure 11. Dependence of distribution of story ductility demands on base shear


strength for pulse P2: (a) T=Tp = 0:5; and (b) T=Tp = 2:0.

The generic MDOF frame previously introduced is used to study the distributions of story
ductility demands when the structure is subjected to the basic pulses. The base shear strength
in this case is dened by the coecient  = Vy =(m:ag; max ), which expresses the base shear
strength relative to the pulse intensity. Figure 11 illustrates story ductility distributions for
pulse P2 and period ratios T=Tp = 0:5 and 2.0, representing structures with fundamental periods
clearly shorter and longer than the pulse period. The distributions can be directly compared
to those presented in Figure 6 for the near-fault record NR94rrs, whose Tp is 1:0 sec. This
comparison reveals notable similarities between the response to the record NR94rrs and the
response to pulse P2. These consistent similarities between MDOF response properties of
near-fault ground motions with forward directivity and those of the basic pulses support the
argument that near-fault records can indeed be represented by equivalent pulses.
A comprehensive assessment of maximum story ductility demands (maximum over all sto-
ries) can be obtained from the –max diagrams shown for pulse P2 and various T=Tp ratios
in Figure 12. For T=Tp = 1:5 to 3.0, there is a range in which  (base shear strength) can
be reduced in half without leading to an increase in the maximum ductility demands. This
close-to-vertical range, which is located in the ductility range of 3 to 4, corresponds to the
migration of the maximum ductility demand from upper stories to the bottom story. Once
the maximum ductility demand has reached the bottom story, it increases rather rapidly with
a further reduction in strength. This pattern is in agreement with the observations made for
near-fault ground motions (see Figure 8).
In design it is often more useful to rearrange the information presented in Figure 12 in order
to evaluate the base shear strength, , required to limit the maximum story ductility demand,
max , to specic target values. Vertical cuts through the –max diagram (supplemented by
a linear interpolation scheme) provide values for the MDOF strength demands shown in
Figure 13. This gure represents MDOF base shear strength demand spectra for pulse P2 and
target maximum story ductility ratios of max = 1, 2, 3, 4, 6 and 8. Inherent in these spectra
is the assumption that the story shear strength distribution over the height follows the SRSS
story shear distribution. Provided that the basic pulses introduced in this study can represent
near-fault ground motions, these MDOF spectra can be used for design against near-fault
eects, as will be shown later.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
698 B. ALAVI AND H. KRAWINKLER

Maximum MDOF Story Ductility Demands


Pulse P2, SRSS Pattern, without P-∆
5
T/Tp = 0.375
T/Tp = 0.50
4 T/Tp = 0.75

η = Vy / (m.ag,max)
T/Tp = 1.00
T/Tp = 1.50
3
T/Tp = 2.00
T/Tp = 3.00
2

0
0 4 8 12 16 20
Maximum Story Ductility, µmax

Figure 12. Base shear strength vs. maximum story ductility demands for various T=Tp values, pulse P2.

MDOF Strength Demands for Constant Ductility


Pulse P2, SRSS Pattern, without P-∆
5
µ=1
µ=2
4 µ=3
µ=4
η = Vy / (m.ag,max)

µ=6
3 µ=8

0
0 0.5 1 1.5 2 2.5 3
T / Tp

Figure 13. MDOF base shear strength demand spectra for target maximum story ductility ratios.

EQUIVALENT PULSES FOR NEAR-FAULT GROUND MOTIONS

Since near-fault ground motions come in great variations (as evidenced by the responses shown
in Figures 4 and 5), any attempt to evaluate or predict structural response will be complex
unless near-fault ground motions can be represented by simplied motions that reasonably
replicate important near-fault response characteristics. An inspection of near-fault time history
traces, as well as the study of similarities between the response of structures subjected to
near-fault records and simple pulses, provides evidence that—within limitations—near-fault
records with forward directivity may be represented by equivalent pulses of the type introduced
previously.
It is, however, unreasonable to expect that this equivalence can be established accurately
for the full period range of interest. In the very short period range, a near-fault record is likely
to be contaminated by high-frequency components that have little to do with characteristics of
the long-period high-energy pulse generated by the propagation of fault rupture. In the very
long period range, it is also likely that other phenomena (e.g., basin eects and instrument
errors) contaminate the record. In this study it is postulated that an equivalence between a
near-fault ground motion and a pulse can be reasonably established in a range of T=Tp from

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
BEHAVIOR OF MOMENT-RESISTING FRAME STRUCTURES 699

0.375 to 3.0. This range is based on a comparison between spectral shapes for near-fault
records and the basic pulses, giving consideration to the range in which Tp varies for typical
near-fault ground motions. It should be emphasized that the equivalent pulse is by no means a
precise representation of near-fault ground motions. Rigorous procedures for the identication
of equivalent pulse parameters have been established, but in some cases judgmental decisions
need to be employed in order to arrive at nal values for these parameters [7].
In order to establish an equivalent pulse for a record, three parameters need to be evaluated:
the pulse type (P1, P2 or P3), pulse period Tp , and pulse intensity ag; max (see Figure 9). An
inspection of time history traces and a comparison between ground motion and pulse spectral
shapes are employed to decide on the pulse type. The pulse period Tp is identied as the
period corresponding to a global peak in the velocity response spectrum. The basis for this
approach is that both P2 (full cycle) and P3 (multiple cycle), which represent the near-fault
ground motions used in this study, show a hump at T=Tp = 1 (see Figure 10). While in many
cases a narrow range for Tp can be identied, there are near-fault records whose velocity
spectra include a wide peak or multiple peaks (see Figure 4). In such cases judgment has to
be employed to decide on a nal value. The authors have performed a sensitivity study to
evaluate these judgmental decisions and have come to the conclusion that, within limitations,
the structural response is not very sensitive to the choice of Tp , provided that the intensity
matching follows the procedure outlined in the next paragraph [7].
To estimate the intensity of the equivalent pulse, a rigorous process is employed whose
objective is to minimize the dierences between the maximum story ductility demands obtained
from the near-fault record and the equivalent pulse. In the following discussion the maximum
acceleration, ag; max , and velocity, vg; max , of the equivalent pulse are referred to as the eective
acceleration, ae , and velocity, ve . In summary, the procedure includes the following steps:
1. Compute –max curves (see Figure 12) for the appropriate pulse type and T=Tp = 0:375,
0.5, 0.75, 1.0, 1.5, 2.0 and 3.0.
2. Compute –max curves for the near-fault record (see Figure 8) and T=Tp = 0:375, 0.5,
0.75, 1.0, 1.5, 2.0 and 3.0.
3. For each T=Tp value, convert the pulse –max curve into a –max curve [ = Vy =
(m:g) = (ae =g)] and nd best-t values for ae by minimizing the relative dierences
between the pulse and ground motion −max curves for 16max 610 (from nearly elas-
tic to highly inelastic behavior). The dierences between the two curves are minimized
using the least-squares method.
4. Obtain nal values for ae by averaging the best-t values for the seven period ratios.
Tables II and III summarize the results of this procedure for the recorded and simu-
lated near-fault ground motions, respectively. The Landers record (LN92lucr) is omitted from
Table II because its pulse period is longer than 4 sec. (see Figure 4), and the matching proce-
dure for this record would involve computing demands for structures with unreasonably long
periods (e.g., T = 3Tp ¿12:0 sec:). The tables list the pulse type, pulse period, and maximum
acceleration and velocity of the equivalent pulse. The peak velocity of the pulse is computed
from the equation ve = ae Tp =4.
Comparing the so-computed ve with the recorded peak ground velocity (PGV) listed in the
last column of Tables II and III indicates that in most of the cases the peak velocity of the
equivalent pulse is very close to the PGV of the near-fault record. Only for the MH84cyld
record does the dierence exceed 20%. Thus, it appears to be feasible to use the PGV of the

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
700 B. ALAVI AND H. KRAWINKLER

Table II. Equivalent pulses for recorded ground motions.


Designation Mw R (km) Pulse type Tp (sec) ae (g) ve (cm/sec) PGV (cm/sec)

LP89lgpc 7.0 3.5 P3 3.0 0.23 169 173


LP89lex 7.0 6.3 P2 1.0 0.70 172 179
EZ92erzi 6.7 2.0 P2 2.3 0.17 96 119
NR94rrs 6.7 7.5 P2 1.0 0.72 177 174
NR94sylm 6.7 6.4 P3 2.4 0.18 106 122
KB95kobj 6.9 0.6 P3 0.9 0.86 190 160
KB95tato 6.9 1.5 P3 2.0 0.40 196 174
IV79ar06 6.5 1.2 P2 3.4 0.13 108 110
IV79melo 6.5 0.0 P2 2.8 0.15 103 117
KB95kpi1 6.9 3.7 P2 1.8 0.25 110 100
MH84andd 6.2 4.5 P2 0.8 0.16 31 27
MH84cyld 6.2 0.1 P3 0.8 0.47 92 65
NR94newh 6.7 7.1 P2 1.3 0.37 118 119
NR94spva 6.7 8.9 P3 2.7 0.09 60 63

Table III. Equivalent pulses for simulated ground motions.


Station Mw R (km) Pulse type Tp (sec) ae (g) ve (cm/sec) PGV (cm/sec)

f6 6.5 3 P2 1.7 0.17 71 68


f8 6.5 3 P2 1.2 0.41 121 116
f6 6.5 5 P2 2.0 0.10 49 46
f8 6.5 5 P2 2.1 0.19 98 70
f6 6.5 10 P2 2.6 0.03 19 25
f8 6.5 10 P2 3.0 0.05 37 32

f6 7.0 3 P2 3.2 0.13 102 98


f8 7.0 3 P2 3.4 0.21 175 146
f6 7.0 5 P2 3.5 0.11 94 89
f8 7.0 5 P2 3.6 0.16 141 124
f6 7.0 10 P2 5.0 0.04 49 51
f8 7.0 10 P2 3.3 0.10 81 76

f6 7.5 3 P3 3.2 0.21 165 148


f8 7.5 3 P3 3.2 0.27 212 210
f6 7.5 5 P3 3.2 0.19 149 138
f8 7.5 5 P3 3.2 0.24 188 201
f6 7.5 10 P3 3.2 0.14 110 84
f8 7.5 10 P3 3.2 0.18 141 131

near-fault record to estimate the pulse intensity parameter (i.e., ae = 4PGV=Tp ) rather than
following the elaborate procedure outlined previously.
Examples of story ductility distributions over the height for the NR94rrs record and its
equivalent pulse are shown in Figure 14 for cases of strong and weak structures. Although
some dierences exist, the equivalent pulse appears to reasonably capture the important re-
sponse characteristics of the near-fault record, particularly the migration of ductility demands
from the top to the bottom portion of the long-period structure (T=Tp = 2:0) if the strength of
the structure is decreased.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
BEHAVIOR OF MOMENT-RESISTING FRAME STRUCTURES 701

Story Ductility Demands Story Ductility Demands


Low Ductility, T / Tp = 2.0, γ = 0.4, without P-∆ High Ductility, T / Tp = 2.0, γ = 0.1, without P-∆
1 1
NR94rrs NR94rrs

0.8 0.8 Pulse P2


Pulse P2
Relative Height

Relative Height
0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 1 2 3 4 0 2 4 6 8 10 12

(a) Story Ductility Ratio, µi = δ dyn,i / δ y,i (b) Story Ductility Ratio, µi = δ dyn,i / δ y,i

Figure 14. Comparison of story ductility demands obtained from record NR94rrs and its equivalent
pulse, T=Tp = 2:0: (a) high strength, low ductility; and (b) low strength, high ductility.

MAGNITUDE AND DISTANCE DEPENDENCE OF EQUIVALENT


PULSE PROPERTIES

As Table II and, more systematically, Table III indicate, the parameters of the equivalent
pulse for a near-fault record, i.e. Tp and ve (or ae ), appear to be related to the magnitude
of the event (Mw ) and the closest distance from the rupture plane (R). Somerville showed
that Tp is mostly aected by slip rise time, dened as the duration of slip at a given point on
the fault, which in turn is related to Mw [2]. He proposed a preliminary model that expresses
log10 Tp as a linear function of Mw , independent of distance. Employing his functional form
for Tp and performing a linear regression analysis on the equivalent pulse periods obtained for
the combination of recorded and simulated ground motions (Tables II and III), the following
regression equation is obtained:

log10 Tp =− 1:76 + 0:31Mw (1)

Figure 15 illustrates this equation together with the data points (solid circles) to which the
line is tted. Some of the circles represent multiple data points with identical Mw and Tp
values. Superimposed on each graph are prediction bands [13], which represent the level of
condence on the model. For example, for a given Mw value, there is a 90% likelihood that
the Tp value falls within the 90% prediction band that surrounds the regressed line. The large
scatter observed in the data points in part can be attributed to the fact that the ground motions
come from various events with various faulting mechanisms and geological conditions. The
scatter translates into wide prediction bands, which indicate large uncertainties in predicting
Tp from the regression equation.
The magnitude of the earthquakes used in developing Equation (1) does not exceed 7.5
(Mw = 7:5 for the six simulated events, and Mw = 7:3 for the Landers earthquake). Somerville
studied the magnitude 7.4 Izmit (Turkey, 1999) and magnitude 7.6 Chi-Chi (Taiwan, 1999)
earthquakes [14]. He downplayed the prevalence of directivity eects in the Chi-Chi earth-
quake and attributed the relatively small size of the forward directivity region to the
shallow hypocenter, the small amount of up-dip slip, and the thrust-fault mechanism of this

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
702 B. ALAVI AND H. KRAWINKLER

Pulse Period - Magnitude Relationship


Combined Ground Motions with Forward Directivity
0.8

0.6

log10Tp = - 1.76 + 0.31 Mw


0.4

log10Tp
0.2 data points
regressed, mean
90% prediction band
0 80% prediction band
70% prediction band
Taiwan, Tsaotun
Turkey, Yarimca
-0.2
6 6.2 6.4 6.6 6.8 7 7.2 7.4 7.6 7.8
Magnitude, Mw

Figure 15. Dependence of equivalent pulse period on magnitude.

earthquake [15]. Two of the pulse-type records used in his study are Yarimca (Turkey) and
Tsaotun (Taiwan). The pulse period for the fault-normal component of these ground motions
(from the location of a global peak in the velocity spectrum) is estimated to be 3:7 sec. and
4:2 sec:, respectively. Figure 15 compares these values with predictions from Equation (1),
and indicates a good agreement.
Adopting the functional form used in Equation (1), a similar study was conducted by
Rodriguez-Marek [12], in which he investigated a larger set of recorded ground motions,
spanning wider ranges of magnitude and distance. He classied the ground motions into two
categories: (1) rock, including competent rock, weathered rock and shallow sti soil; and (2)
soil, including deep sti clay and soft clay. This soil denition is broader than the one used
in this study. He concluded that the existence of soil at the site elongates the pulse period,
but the period elongation diminishes for large magnitudes.
The conclusion from these studies is that there is a pattern of increase in pulse period with
magnitude, but the scatter of the data is large, and regression results will depend on soil type,
source mechanism, record selection, and on the procedure employed to dene the pulse period.
What is needed most to improve predictions of Tp –Mw relationships is the improvement of
simulation techniques and the availability of more records. More research should take care of
the former and time will take care of the latter.
Somerville developed a model that expresses the PGV of the record as a function of moment
magnitude and distance [2]. His proposed functional form is used here in a linear two-variable
regression analysis that relates the equivalent pulse intensity parameter, ve , to Mw and R.
A combination of ground motions, also used by Somerville, is used that includes the records
listed in Table II and a subset of the simulated ground motions listed in Table III. This
subset, which consists of simulations for Mw = 6:5 and 7.0 with strong forward directivity
eects (station f8), is most compatible with the recorded time histories [2]. Records with R
values smaller than 3 km are not used in the derivation of the regression equation on account
of the logarithmic form of this equation, which results in unreasonably large values for ve
at small R values. The following equation results from the regression analysis:

log10 ve =− 2:03 + 0:65Mw − 0:47 log10 R (2)

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
BEHAVIOR OF MOMENT-RESISTING FRAME STRUCTURES 703

Effective Pulse Velocity


Data Points
Surface Points

200

150

Veff (cm / sec)


100

50

3
0
4
6 5
6.2 6
6.4 7
6.6 8
6.8 9
7 10 R (km)
Mw

Figure 16. Dependence of equivalent pulse velocity on magnitude and distance.

Figure 16 illustrates this equation together with the data points to which the surface is tted.
The solid circles identify the data points used in the regression analysis, and the empty circles
represent points on the regression surface. The quality of the t appears to be relatively good.
However, the data contain a relatively small number of recorded motions and a relatively large
number of simulations, which is likely to bias the outcome of the regression analysis. The
relatively steep slope of the surface at Mw = 7 translates into excessively large predictions for
ve for events with Mw ¿7. PGV data from recent earthquakes (the 1999 Turkey and Taiwan
earthquakes) provide a clear indication that PGV saturates at large magnitudes and that the
predictions from Equation (2) are much too high for magnitudes greater than 7.
When compared to Equation (2), the regression equation developed by Rodriguez-Marek
[12] indicates a much weaker variation of PGV with magnitude, and for large magnitudes
results in smaller pulse intensity. This is again an indication that Equation (2) cannot be
considered a reliable measure for the pulse intensity of near-fault ground motions in very
large magnitude earthquakes.
In conclusion, it is claimed that the eective velocity as dened in this paper (or in a
simplied way, the PGV) together with the pulse period are good descriptors of the equivalent
pulse. The functional forms of the regression equations relating Tp and ve to Mw and R are
believed to be suitable, but further investigations and much more data are needed to develop
more reliable regression coecients and uncertainty measures. Nevertheless, Equations (1) and
(2) are used in the following discussion to estimate the period and intensity of the equivalent
pulse for specic Mw and R values. For the range of Mw and R values used in the discussion,
the results are believed to be reasonable.

MAGNITUDE AND DISTANCE DEPENDENCE OF BASE SHEAR


STRENGTH DEMANDS

Given the moment magnitude, Mw , and the closest distance from the rupture plane, R, the
equivalent pulse parameters Tp and ag; max (= ae = 4ve =Tp ) can be estimated from Equations

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
704 B. ALAVI AND H. KRAWINKLER

Figure 17. Preliminary results on magnitude and distance dependence of MDOF base shear strength
demand spectra for constant ductilities, equivalent pulse P2: (a) Mw = 6, R = 10 km; (b) Mw = 6,
R = 3 km; (c) Mw = 7, R = 10 km; and (d) Mw = 7, R = 3 km.

(1) and (2), and the pulse –T=Tp curves presented in Figure 13 can be converted into
–T curves [ = Vy =W = (ag; max =g)]. The so obtained –T curves represent base shear strength
demand spectra for specied target ductilities, suitable for the types of frames represented by
the generic frame used in this study. Examples of such strength demand spectra are presented
in Figure 17 for various combinations of magnitude and distance, assuming that the ground
motion is represented by pulse P2. Inherent is the assumption of the SRSS story shear strength
distribution over the height of the structure. The graphs illustrate the magnitude and distance
dependence of the base shear strength demands obtained from the equivalent pulse approach.
To put these demands in perspective with present design practice, superimposed on each graph
are two design strength curves related to the UBC 97 soil Type SD design spectra for Seismic
Zone 4, one with and one without the code specied near-fault factor (denoted as ‘UBC/4’)
[4]. The UBC near-fault factor is computed using Source Type A for Mw = 7 and Source
Type C for Mw = 6. The design strength curves are obtained by scaling down the code base
shear by a factor of 4, which is arrived at by choosing a response modication factor of 8
(close to that for special moment frames) and assuming a period independent overstrength
factor of 2. These values are purely for comparison, as the real strength of a structure will
vary on a case-by-case basis.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
BEHAVIOR OF MOMENT-RESISTING FRAME STRUCTURES 705

A comparison of the near-fault strength demand spectra with the UBC=4 curves reveals
that for given magnitude and distance values, a structure designed according to present code
provisions will experience quite dierent levels of inelasticity depending on the fundamental
period. For instance, Figure 17(d) indicates that for Mw = 7, R = 3 km, and a ground motion
of the type represented by pulse P2, structures designed according to the UBC 97 provisions
(including the near-fault factor) may experience story ductility demands less than 4 when the
period is longer than 3:4 sec:, but may experience ductility demands larger than 8 if the period
is between 1.0 and 2:2 sec.
Assuming that the  = 4 spectrum for Mw = 7 and R = 3 km is representative of design
conditions, it is observed that this spectrum has a much wider plateau (up to a period of about
2 sec.) than the UBC=4 curve. This plateau is followed by a descending curve that falls below
the UBC=4 curve around 3:4 sec. This indicates that long-period and short-period structures are
adequately protected by the UBC 97 provisions, but structures of intermediate period (from
about 0.8 to 3:4 sec.) may experience ductility demands signicantly larger than 4.0.
The introduction of the UBC near-fault factor is an improvement that gives recognition
to the existence of the problem. However, the results of this study indicate that the code
criteria do not provide a consistent level of protection against near-fault eects, and in certain
period ranges may provide inadequate protection. The problem cannot be solved by introducing
additional factors to conventional design spectra, and a more rigorous approach appears to be
necessary. The equivalent pulse concept can provide the basis for such an approach.

CONCLUSIONS
The study summarized in this paper addresses important response characteristics of SDOF and
MDOF structures subjected to near-fault ground motions and basic pulses. Equivalent pulses,
which are fully dened by a small number of parameters, are introduced to represent near-
fault ground motions. Relationships are proposed for the magnitude and distance dependence
of these parameters. A procedure for estimating the required design base shear strength is
introduced based on the response characteristics of equivalent pulses. It is emphasized that
the conclusions and results presented in this paper are applicable only within the context of
the assumptions made. The conclusions of this study are summarized as follows:
• Spectral values for individual near-fault ground motions can be several times the values
given by the UBC 97 design spectrum over a wide range of periods.
• For structures with T ¿Tp , large elastic story shear forces are generated in the upper
stories of the structure.
• These large elastic shear forces result in early yielding of upper stories when the structure
is relatively strong. When the strength is reduced, story ductility demands stabilize in
the upper portion and the maximum demand migrates to the base, where it grows rather
continuously with a further reduction in strength.
• For short-period structures (T 6Tp ) the maximum story ductility demands occur in the
bottom portion regardless of strength.
• There are clear similarities between the response of frame structures to near-fault ground
motions and the response to pulse-type excitations. Within the approximate period range
of 0:375¡T=Tp ¡3:0, important response characteristics of near-fault ground motions can
be represented by simple equivalent pulses.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706
706 B. ALAVI AND H. KRAWINKLER

• Preliminary regression equations are developed that relate equivalent pulse parameters
to moment magnitude and closest distance to the fault rupture plane. Equations of this
type, together with pulse MDOF strength demand spectra, can be used to estimate the
required base shear strength for target story ductility ratios.
• For Mw = 7 and R = 3 km, a comparison between the base shear strength demand spec-
tra obtained from the regression equations and UBC 97 design estimates indicates that
code-compliant structures with an intermediate period may experience excessive ductility
demands.

ACKNOWLEDGEMENTS
The research summarized in this paper was supported by a grant from the CUREE/Kajima Research
Program, by the National Science Foundation through Grant CMS-9812478 of the US–Japan Cooperative
Research Program in Urban Hazard Mitigation, and by the California Department of Conservation as
a SMIP 1997 Data Interpretation Project (Department of Conservation Contract No. 1097-601). This
support is gratefully acknowledged. The constructive collaboration of Dr Paul Somerville on the ground
motion aspects of this research is much appreciated.

REFERENCES
1. Somerville P, Smith N, Graves R, Abrahamson N. Modication of empirical strong ground motion attenuation
relations to include the amplitude and duration eects of rupture directivity. Seismological Research Letters
1997; 68:180 –203.
2. Somerville P. Development of an improved ground motion representation for near fault ground motions.
Proceedings of SMIP98 Seminar on Utilization of Strong-Motion Data, Oakland, CA, September 1998.
3. Structural Engineers Association of California. Recommended Lateral Force Requirements and Commentary.
7th Edition, SEAOC: Sacramento, California, 1999.
4. International Conference of Building Ocials. Uniform Building Code. Whittier, California, 1997.
5. Somerville P, Smith N, Punyamurthula S, Sun J. Development of ground motion time histories for phase 2 of
the FEMA/SAC steel project. Report No. SAC/BD-97-04, 1997.
6. Federal Emergency Management Agency. NEHRP Recommended Provisions for Seismic Regulations for New
Buildings and Other Structures. 2000 Edition. FEMA 368. FEMA: Washington, DC, 2001.
7. Alavi B, Krawinkler H. Eects of near-fault ground motions on frame structures. John A. Blume Earthquake
Engineering Center Report No. 138, Stanford University, February 2001.
8. Seneviratna G, Krawinkler H. Evaluation of inelastic MDOF eects for seismic design. John A. Blume
Earthquake Engineering Center Report No. 120, Stanford University, June 1997.
9. Chopra A. Dynamics of Structures: Theory and Applications to Earthquake Engineering. Prentice Hall: Upper
Saddle River, New Jersey, 1995.
10. Stewart J, Chiou S, Bray J, Graves R, Somerville P, Abrahamson N. Ground motion evaluation procedures
for performance-based design. PEER-2001/09, Pacic Earthquake Engineering Research Center, University of
California, Berkeley, September 2001.
11. Sasani M, Bertero V. Importance of severe pulse-type ground motions in performance-based engineering:
historical and critical review. 12th World Conference on Earthquake Engineering, New Zealand Society for
Earthquake Engineering, Upper Hutt, New Zealand, 2000.
12. Rodriguez-Marek A. Near-fault site response. Ph.D. dissertation, Department of Civil and Environmental
Engineering, University of California, Berkeley, Fall 2000.
13. Ramsey F, Schafer D. The Statistical Sleuth: a Course in Methods of Data Analysis. Duxbury Press: Belmont,
California, 1996.
14. Somerville P. Characterization of near-fault ground motions. Proceedings of the US–Japan Workshop on Eects
of Near-Field Earthquake Shaking, San Francisco, CA, March 2000.
15. Somerville P. Magnitude scaling of near fault ground motions. Proceedings of the International Workshop
on Annual Commemoration of Chi-Chi Earthquake, National Center for Research on Earthquake Engineering,
Taipei, Taiwan, September 2000.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:687–706

Das könnte Ihnen auch gefallen