CHAPTER 1

INTRODUCTION

1.1 Statement of the Problem

Establishment of relationships for predicting the lateral drift demands of near-fault

ground motions is one of the major challenges in earthquake engineering. Excessive
lateral drifts caused by earthquake ground motions are the major causes of structural
damage observed in structures. A large amount of research has been performed on the
estimation of global displacement demands, such as roof drift. In the last decade there
has been a growing interest in the estimation of local displacement demands. The
unexpected level of damage observed in well-engineered structures after the 1994
Northridge and 1995 Kobe earthquakes forced the engineering community to focus on
the effects of near-fault ground motions. Studies on the effects of near-fault ground
motions on structures showed that the strong ground motion pulses observed in some of
the near-fault ground motions might have resulted in unexpectedly high local
displacement demands. A number of methods for estimating the local displacement
demands of structures subjected to near-fault ground motions have been developed.
Iwan(1997) have introduced the drift spectrum, a measure of local displacement demand
imposed on structures. This drift spectrum was based on wave propagation analysis of a
shear beam model. Gülkan and Akkar (2002) have proposed a simple method for

1
estimating the elastic ground story drift, based on the first mode response of a shear
beam. Use of shear beam model may provide sufficiently good estimates for the ground
story drifts of frames having beams relatively stiffer than columns (shear frames).
However, for the frames having beams and slabs relatively more flexible than columns
the use of shear beam model results in inaccurate estimates of the ground story drift. To
overcome this inaccuracy, this study aims to introduce a set of modifying coefficients to
the approximate equation proposed by Gülkan and Akkar (2002).

This study also presents an initial survey for estimating the inelastic displacement
demands of structures subjected to near-fault ground motions. Estimation of inelastic
displacement demands, based on the elastic analysis of structures subjected to near-fault
ground motions would provide a useful tool for seismic engineering practice. Although
a number of methods exists in the literature for estimating the inelastic demands based
on elastic analysis, a method directly taking into account the intrinsic features of near-
fault ground motions has not been proposed. This study is an initial survey for such a
method. Furthermore, as it will be seen in Chapter 4, C1 coefficient employed in
FEMA356(2000), which is one of the most discussed technical documents on inelastic
analysis, may provide alarmingly inaccurate estimates of inelastic displacements for the
case of near-fault ground motions. A set of preliminary equations was established for
estimating the inelastic displacement demands imposed on inelastic non-degrading
SDOF systems subjected to near-fault ground motions based on elastic analysis of the
systems.

Integration of the two procedures stated above would provide a useful tool for
estimating the local inelastic displacement demands of moment resisting frames
subjected to near-fault ground motions. However, it should be noted that such an
integration process requires a comprehensive research on the effects of nonlinearity on
the response of frame structures subjected to near-fault ground motions. Such a
comprehensive research is beyond the scope of this study.

2
1.2 Review of Past Studies

Several researchers have performed experimental and analytical studies on the

estimation of lateral displacements of building type of structures since the 1960s. A
number of analytical procedures and computer programs have been developed for this
purpose. Also a number of static and dynamic tests have been conducted for evaluating
and further improving these methods. Some of the major studies on estimation of
displacement demands, effects of near-fault ground motions and characteristic properties
of near-fault ground motions are summarized below.

1.2.1 Past Studies on the Characteristics of Near-Fault Ground Motions

Somerville et al., (1997) studied the fundamental characteristics of near-fault

ground motions. They stated that, when the velocity of rupture is close to the shear
wave velocity of the rock near the source, a series of waves resulting from each
dislocation are accumulated in the direction of rupture. Accumulation of pulses in the
direction of rupture results in a single large pulse of strong ground motion at the
beginning of the record. This pulse represents the cumulative effect of almost all of the
seismic radiation from the fault. The resulting phenomenon, which is closely related to
the Doppler effect in physics, is called forward directivity. The large pulse of motion is
generally observed at the beginning of velocity or displacement traces of the near-fault
strong motion records affected by forward directivity. Somerville et al. (1997), stated
that forward rupture directivity effects are observed when the rupture front propagates
through the site and the direction of slip on the fault is aligned with the site. Somerville
et al. (1997) also stated that, according to the theory of seismic wave propagation, strike
normal components of near-fault ground motions have greater damage-causing potential
compared to other components. It should be noted that if most of the seismic energy is
released at the beginning or end of rupture, directivity effects would not be significant.
In other words, energy release through the rupture should be nearly uniform for
directivity effects to be noticeable (Steward et al., 2001). Somerville et al., (1997)

3
noticed that when the rupture propagates away from the site, arrivals of seismic waves
are distributed in time and named this phenomenon as backward directivity. They
observed that backward directivity resulted in long duration motions having low
amplitudes at long periods. Akkar and Gülkan (2002) examined the forward directivity
effects in the near-fault strong ground motion records taken during the 17 August 1999
Kocaeli and 12 November 1999 Bolu-Düzce earthquakes. They observed that the
ground motion component with highest displacement demands were not always in the
fault normal direction. Akkar and Gülkan (2002) stated that the ground motion
components in the maximum ground velocity direction does correlate better with the
larger drift demands, compared to the strike normal component for the near-fault ground
motion records taken from the 17 August 1999 Kocaeli and 12 November 1999 Bolu-
Düzce earthquakes.

1.2.2 Past Studies on the Estimation of Displacement Demands Imposed on Frame

Structures

Blume (1968) investigated the modal characteristics of a number of frames. He

examined the effects of the various types of degrees of freedom, joint rotation, axial
deformation of columns and base rocking, on the overall response of building type of
structures. He established a set of equations and graphs for estimating the dynamic
characteristics of frame type of buildings with joint rotation, overall flexure and base
motion, using simple rigid-floor shear frame analysis. Diaz et al. (1994) examined the
validity of alternate models (i.e., rigid frame model, equivalent shear-beam model and
equivalent shear-frame model) of building frames by comparing the ductility demands
found for two idealized buildings excited by five simulated ground motions. Results of
the study showed that a model that neglects the joint rotations may produce significant
errors in estimating the lateral drift demands of frame type of systems.

Heidebrecth and Stafford Smith (1973) studied the interaction of shear walls and
frames and derived closed form solutions for lateral displacements. Using these closed

4
form solutions Miranda (1999) set up an approximate procedure for estimating the
maximum lateral deformations, maximum roof displacement and maximum interstory
drifts of frame type of structures with uniform stiffness. Miranda and Reyes (2002)
further improved the approximate method proposed by Miranda (1999), by adding the
capability of estimating the lateral displacements of frame structures with non-uniform
stiffness distribution.

Anderson and Bertero in 1986 examined the significant displacement demands

imposed on frame type of structures by near-fault strong ground motions. However, it
was after the 1994 Northridge and 1995 Kobe earthquakes that engineering community
recognized the inadequacy of current seismic design and evaluation methodologies for
the case of near-fault excitations.

Chopra and Cruz (1986a) analyzed the accuracy of response spectrum analysis
using a set of 5 story generic frames having various fundamental periods of vibration
and beam-to-column stiffness ratios. Comparisons of the response quantities obtained
from response spectrum analysis with the elastic response history analysis for 8
simulated ground motions showed tolerable errors. Making use of those results they
proposed a method called Simplified Response Spectrum Analysis (Chopra and Cruz
1986b). This method involved a practical procedure to estimate the first two
fundamental modes, periods of vibration and modal participation factors.

Realizing the fact that local displacements are of primary importance,

particularly under near-fault excitations, researchers attempted to develop some practical
tools for measuring local displacement demands. Iwan (1997) developed the drift
spectrum concept which was summarized in Section 1.1. Iwan (1997) demonstrated that
the single-degree-of-freedom (SODF) response spectrum is an inadequate tool for
measuring the interstory drift demands, particularly for near-fault ground motions.
Similarly, Heidebrecht and Naumoski (1997) developed an alternative method for
establishing a drift spectrum. They utilized the shear-beam theory together with

5
approximate equivalent shear stiffness formula for portal frames, derived by Heidebrecht
and Stafford Smith (1973). Heidebrecht and Rutenberg (2000) further improved this
method by employing the approximate fundamental period formulas for frame
structures. Gülkan and Akkar (2002) developed a simpler procedure, utilizing the shear-
beam model, for constructing the drift spectrum.

Chopra and Chintanapakdee (2001a) compared the ground story drift ratios
found from response spectrum analysis with that of drift spectrum analysis and stated
that the difference is due to effects of higher modes. They also pointed out that a
sufficient degree of accuracy may be obtained using response spectrum method, if at
least five modes are included in the analysis.

1.2.3 Past Studies on the Estimation of Inelastic Displacement Demands Imposed

on Single Degree of Freedom Systems

Estimation of inelastic displacement demands from elastic displacements is the

most popular research area in earthquake engineering. Among these methods some are
based on the modification of elastic spectral ordinates by functions of period (T) and
ductility (µ) (Newmark and Hall, 1982; Nassar and Krawinkler, 1991; Vidic et al., 1994;
Miranda, 2001). The displacement coefficients method employed in FEMA356 (2000)
is based on the amplification of elastic displacements for estimating the inelastic
displacement demands according to strength reduction factor, natural period and
characteristic site period. FEMA356 (2000) employs the inverted form of an equation
which was originally derived for constant ductility by Vidic et al. (1994). This inversion
of the equation proposed by Vidic et al. (1994) from constant ductility form to constant
strength form resulted in loss of accuracy. Aydınoğlu and Kaçmaz (2002) have
addressed this inaccuracy and performed regression analysis directly on the spectral
coefficients depending on R and T. It should be noted that all these studies were

6
primarily based on far-field ground motions whose characteristics are very different
from near-fault excitations.

Baez and Miranda (2000) studied the inelastic displacement demands imposed
on single degree of freedom systems by near-fault ground motions. They stated that
structures subjected to ground motions with large velocity pulses may experience
maximum inelastic deformations larger than those subjected to ground motions that do
not have these pulses, even if the linear elastic ordinates in the short period spectral
region are similar.

(Chopra and Chintanapakdee, 2001b) compared the responses of single-degree-

of-freedom (SDOF) systems to 15 near-fault and 15 far-fault strong ground motions,
they concluded that near-fault ground motions had much narrower velocity sensitive
regions. Chopra and Chintanapakdee (2001b) further indicated that, the design
equations proposed by Veletsos and Newmark (1960), which are mainly derived for far-
field ground motions, are valid for near-fault ground motions also, as long as spectral
regions are defined appropriately.

Another set of methods (Rosenblueth and Herera, 1964, Gülkan and Sözen,
1974, Iwan 1980) are based on equivalent linearization. In equivalent linearization
methods maximum response of the inelastic system is approximated by the elastic
response of a linear elastic system with increased damping and lowered stiffness.
Capacity Spectrum Method is a method developed by Freeman et al. (1975), and based
on the iterative use of equivalent linearization. Capacity spectrum method is employed
in ATC-40 (1996), which is one of the most debated technical documents on inelastic
analysis of structures.

All of these studies provided considerable information on the nature of near-fault

ground motions, response of frame structures and inelastic displacement demands
imposed on single degree of freedom systems by near-fault ground motions.

7
1.3 Object and Scope

This study examines the properties of near-fault ground motions and effects of
ground motions containing pulses on the response of structures. Important
characteristics of near-fault ground motions were studied for this purpose. Performances
of various attenuation relationships were evaluated. Effects of various parameters on the
characteristic properties of near-fault ground motions were examined.

Modal properties of frames having different overall lateral deformation patterns

were analyzed. A set of idealized moment resisting frames having various numbers of
stories and beam-to-column stiffness ratios was utilized for this purpose. A set of
equations, estimating the elastic ground story and maximum interstory drift ratios for
moment resisting frames, was established. Principles of modal analysis were utilized in
the derivation of these equations. The equations were based on the equation proposed
by Gülkan and Akkar (2002) for estimating ground story drift.

Another set of equations have been established for estimating the inelastic
deformation demands of non-degrading SDOF systems, based on the maximum elastic
displacement demand imposed on SDOF systems with the same natural period and
damping. Equations for estimating the inelastic displacements were established by
directly performing regression analysis on elastic to inelastic displacement ratio, natural
period and strength reduction factor. This study is an attempt for establishing a direct
link between the strength capacity of the structure and inelastic displacement demand.
Such a methodology may serve as a convenient and practical tool for performance
evaluation of existing structures excited by near-field earthquakes.

This thesis is composed of five main chapters and two appendices. Contents of
each chapter may be summarized as follows:

8
Chapter 1 Statement of the problem and review of past studies on the estimation of
the displacement demands imposed on structures, effects of near-fault
ground motions on the response of structures and fundamental
characteristics of near-fault ground motions.

Chapter 2 Inspection of some of the fundamental characteristics of near-fault

ground motions records used in the study. Evaluation of performances
of some of the attenuation relationships for the case of near-fault ground
motions.

Chapter 3 Inspection of the effects of near-fault ground motions on moment

resisting frames. Derivation of the equations for estimating the elastic
ground story and maximum interstory drifts of moment resisting frames.
Establishment of equations for estimating the maximum inelastic
displacement demands imposed on non-degrading SDOF systems
subjected to near-fault ground motions.

Chapter 4 Statistical study on the errors of the proposed equations. Comparison of

the error ranges with other studies available in the literature.

Chapter 5 Summary and conclusions.

Appendix A List of the near-fault strong ground motion records used.

Appendix B Coefficients of the attenuation relationships referred to in Chapter 2.

9
 CHAPTER 2

NEAR-FAULT GROUND MOTIONS

2.1 Ground Motion Records

A set of 148 near-fault ground motion records was utilized in this study. All the
records had been recorded on dense or stiff soils, with average shear wave velocities in
the upper 30 m ranging from 760m/s to 180m/s. Records utilized in this study were
downloaded from the PEER Strong Motion Data Archive Website
(http://peer.berkeley.edu/smcat/). The strong motion record set, comprised of horizontal
components taken from 10 different earthquakes, covered a moment magnitude (Mw)
range from 6 to 7.6. The nearest and farthest records used in the study were made at
distances of 0.24 km and 20 km to the rupture surface, respectively. In order to examine
the effects of strong long duration pulses on the elastic and inelastic response, the record
set was sub-divided into two parts. A total of 56 records having both significant velocity
pulses in their time history traces and corresponding sharp peaks in their pseudo-velocity
spectra formed the “Records with Pulse” group. The remaining 92 strong motion
records formed the “Records without Pulse” group. Basic properties of these ground
motion records are listed in Table A.1 in Appendix A.

10
 In Figure 2.1, moment magnitudes (Mw) and distances of the records are
summarized. The horizontal crowding in the graph at Mw equal to 7.6 is due to
numerical dominance of the 1999 Chi-chi earthquake records in the dataset. This
earthquake, which was recorded by a very dense array of accelerographs, provided a
useful group of data for near-fault strong ground motion studies. Indeed, 82 of the near-
fault strong motion records used in this study have been recorded during the 1999 Chi-
chi earthquake.

7.8

7.6

7.4
Moment Magnitude, Mw

7.2

7.0

6.8

6.6

6.4

6.2

6.0

5.8
0 4 8 12 16 20

Closest Distance to Rupture Surface, rrup (km)

Figure 2.1 Moment Magnitude versus Distance Plot of the Record Set

2.2 Characteristics of Near-Fault Ground Motions

Understanding the nature of near-fault ground motions is crucial for establishing

proper design methodologies for buildings situated in such conditions. In Figure 2.2, a
near-fault and a far-fault ground motion records were plotted. The sample near-fault
strong motion record on the left of Figure 2.2 was taken at a distance of 2.94 km to the
rupture surface during 1999 Chi-chi, Taiwan Earthquake (Mw=7.6).

11
 TCU101 Station EW Component Taft Station 111 Component
Chi-Chi Taiwan 1999 Earthquake Kern County 1952 Earthquake
Peak Ground Acceleration = 0.2 g Peak Ground Acceleration = 0.18 g
0.2 0.2

Acceleration (g)
Acceleration (g) 0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

0 10 20 30 40 50 0 10 20 30 40 50
Time (s) Time (s)
75
Peak Ground Velocity = 68 cm/s Peak Ground Velocity = 17.5 cm/s
50 50
Velocity (cm/s)

Velocity (cm/s)
25 25

0 0

-25 -25

-50
-50

-75
0 10 20 30 40 50 0 10 20 30 40 50
Time (s) Time (s)
80 80
Peak Ground Displacement = 76 cm Peak Ground Displacement = 9 cm
60 Displacement (cm) 60
Displacement (cm)

40 40
20 20
0 0
-20 -20
-40 -40
-60 -60
-80 -80
0 10 20 30 40 50 0 10 20 30 40 50
Time (s) Time (s)

100
%5 damping %5 damping
100
Pseudo-velocity (cm/s)

Pseudo-velocity (cm/s)

10 10

1 1

0.02 0.1 1 10 50 0.02 0.1 1 10 50

T, Period (s) T, Period (s)

Figure 2. 2 A Sample Near-Fault Ground Motion Record with Pulse (1999 Chi-Chi
Earthquake) and Far-Fault Ground Motion Record (1952 Kern County Earthquake)

12
 A significant long duration pulse may easily be noticed in the velocity and
displacement traces of this near-fault record. The sample far-fault strong motion record
on the right side of Figure 2.2 was taken at a distance of 41 km to the rupture surface
during 1952 Kern County Earthquake. It may easily be noticed that a significant long
duration pulse, like the one seen in the velocity and displacement traces of the near-fault
record, is not observed in the velocity and displacement traces of the sample far-field
record. The pseudo-velocity spectra of the records seem to be different too. By
examining the pseudo-velocity spectra of the sample near-fault and far-fault records, one
may conclude that the sample near-fault has significantly higher spectral ordinates for
the periods greater than 1 s. Dominance of the strong pulse seen in the velocity trace of
the sample near-fault record is evident from the pseudo-spectral velocity diagram too.

2.2.1 Peak Ground Acceleration (PGA)

The most commonly used measure of amplitude of a particular ground motion is

peak ground acceleration. It is simply the largest (absolute) acceleration in the strong
motion record. Widespread use of PGA for characterizing strong ground motions comes
from its straight-forward relationship with inertial forces. Although it is strongly related
to the sampling rate of accelerogram and gives only an indication about the inertial
forces acting on very stiff structures, it has been used widely for characterizing strong
ground motions.

13
 Records with Pulse
Records without Pulse

PGA (g) 6.7 ≥ Mw PGA (g) 6.7 > Mw > 7.4 PGA (g) Mw ≥ 7.4
2.0 2.0 2.0

1.5 1.5 1.5

1.0 1.0 1.0

0.5 0.5 0.5

0.0 0.0 0.0

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Distance, rrup (km) Distance, rrup (km) Distamce, rrrup
Distance, (km)
rup(km)

Figure 2.3 Peak Ground Acceleration versus Distance to Rupture Surface

Figure 2.3 shows the Peak Ground Acceleration (PGA) versus distance scatter
plots of the record set grouped into Mw sets. Although a falling trend in PGA with
increasing distance may easily be noticed from the 6.7 > Mw > 7.4 and Mw ≥ 7.4 graphs,
such a trend is hardly noticed in the 6.7 ≥ Mw graph in Figure 2.3. Especially the strong
motion record taken at Tarzana Cedar Hill Nursery (CDMG 24436) during the 1994
Northridge (Mw=6.7) earthquake seems to be lying outside the rest of the measurements
with a PGA of 1.78g at a distance of 17.5 km. The same plot shows that there is no
significant relation between the velocity pulse content and PGA.

2.2.2 Peak Ground Velocity (PGV)

Velocity is the first integral of acceleration with respect to time. Therefore,

unlike PGA, PGV is less sensitive to higher frequency components of the ground
motion. Therefore, PGV is a more effective tool for characterizing ground motion’s
potential for damage to typical engineering structures. Wald et al. (1999) have
developed regression relationships between the peak ground velocity and Modified
Mercalli Intensity (Richter 1958). Modified Mercalli Intensity is the measure of damage
caused by an earthquake at a site. Wald et al. (1999) have stated that peak ground

14
velocity is strongly related to the Modified Mercalli Intensity for the earthquakes having
intensities larger than VII. Figure 2.4 shows the Peak Ground Velocity (PGV) versus
distance distributions of the record set. A narrowing trend in PGV values with
increasing distance from rupture surface may be noted in the Mw ≥ 7.4 graph in Figure
2.4. In other words, strong motion records taken at closer distances have significantly
wider ranges of PGV values compared to those recorded at more distant locations.
Nevertheless, it should be noted that there exists a number of records taken at closer
distances than 4 km and having relatively small PGV values. Additionally, PGV versus
distance graph does not show a notable difference between records with and without
significant pulses in their velocity traces.

Records with Pulse

Records without Pulse

PGV (cm/s) 6.7 ≥ Mw PGV (cm/s) 6.7 > Mw > 7.4 PGV (cm/s) Mw ≥ 7.4
250 250 250

200 200 200

150 150 150

100 100 100

50 50 50

0 0 0
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Distance, rrup (km) Distance, rrup (km) Distamce, rrup(km)
Distance, (km)

Figure 2.4 Peak Ground Velocity versus Distance to Rupture Surface

2.2.3 Arias Intensity

Despite the fact that they are popular ground motion characterization parameters,
PGA and PGV only provide simple information about the amplitude of strong ground
motion. For proper characterization of a strong ground motion for engineering purposes
amplitude, frequency content and duration of the record should be reflected in the
characterization parameter. Arias Intensity (Arias 1970), reflects the amplitude,

15
frequency content and duration of the ground motion. Arias Intensity proposed by Arias
(1970), relates the cumulative energy per unit weight absorbed by an infinite set of
single degree of freedom oscillators having fundamental frequencies uniformly
distributed in the range (0,∞). In the case of oscillators with zero damping, Arias
Intensity is defined as:

π ∞ (2.1)
( )
2

2 g ∫0 
Ia =  a x t 
 dt

where Ia is the the Arias Intensity (in units of meter per second), ax(t) is the acceleration
time history of the strong motion record (in units of meters per second square) and g is
the gravitational acceleration (9.81m/s2).

Arias intensity is a simple and efficient index of damage for many structural and
geotechnical engineering problems, such as dynamic response of stiff structures,
earthquake induced liquefaction and seismic slope stability (Kayen and Mitchell 1997).

In Figure 2.5, variation of Arias Intensity with distance is presented. Similar

outliers, like those in PGA versus distance plot (Figure 2.3), are observed in Figure 2.5.
Two of these outliers in the 6.7 ≥ Mw graph in Figure 2.5, with distances of 17.5 km are
again the components of strong motion recorded at Tarzana Cedar Hill Nursery (CDMG
24436) during the 1994 Northridge earthquake (Mw = 6.7). The third outlier in the
Mw ≥ 7.4, with a distance of 10.4 km is the west component recorded by TCU084 station
during 1999 Chi-chi earthquake. Leaving aside these three outliers, a narrowing trend in
Arias Intensity with increasing distance may be seen. No significant difference may be
observed between Arias Intensities of records with and without pulses.

16
 Records with Pulse
Records without Pulse

Ia (m/s) 6.7 ≥ Mw Ia (m/s) 6.7 > Mw > 7.4 Ia (m/s)

Mw ≥ 7.4
25 25 25

20 20 20

15 15 15

10 10 10

5 5 5

0 0 0
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Distance, rrup (km) Distance, rrup (km) Distamce, rrup(km)
Distance, (km)

Figure 2.5 Arias Intensity versus Distance to Rupture Surface

2.2.4 Peak Ground Acceleration to Peak Ground Velocity Ratio (PGA/PGV)

Studies have indicated that the decrease of ground velocity with distance is
generally slower than that of acceleration. Therefore it is generally expected that, the
(PGA/PGV) ratio is relatively higher near the earthquake source and lower at larger
distances from the source of energy release. Because the peak velocities and peak
accelerations are usually associated with different frequencies, PGA/PGV should give
information about the frequency content of the earthquake (McGuire, 1978). Zhu et al.
(1988), using 36 strong motion records, studied the correlation between the inelastic
displacement demand and maximum of the ground acceleration to velocity ratio
computed for each time step of the strong motion record max(a/v). It should be noted
that although max(a/v) is not directly related to PGA/PGV. However, it provides a
similar information about the strong motion record. Zhu et al. (1988) found that strong
ground motion records with lower max(a/v) had higher inelastic displacement demands.
They also found that records with low max(a/v) ratios had significantly higher hysteretic
energy demands. Malhotra (1999), using 4 strong motion records, claimed that near-fault
records having forward directivity effects, which were summarized in Chapter 1, tended
to have lower PGA/PGV ratios.

17
 The PGA/PGV versus distance scatters for different magnitudes were plotted in
Figure 2.6. Through Figure 2.6, it is observed that for the ground motions taken from
earthquakes with Mw<7.4 records with pulse on the average have lower PGA/PGV ratios
than records without pulse.

Records with Pulse

Records without Pulse

PGA/PGV (s )
-1 6.7 ≥ Mw -1
PGA/PGV (s ) 6.7 > Mw > 7.4 PGA/PGV (s-1) Mw ≥ 7.4
25 25 25

20 20 20

15 15 15

10 10 10

5 5 5

0 0 0
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Distance, rrup (km) Distance, rrup (km) Distance,
Distamce, rrup(km)
(km)

Figure 2.6 Ratio of Peak Ground Acceleration to Peak Ground Velocity versus Distance
Plot

2.3 Near-fault Ground Motions and Attenuation Relationships

Estimation of the maximum level of ground acceleration that a structure will be

subjected to during its service life constitutes an important part of force-based
earthquake-resistant design. The ground motion parameters of interest, for common
engineering applications, are peak ground acceleration (PGA), pseudo-spectral
acceleration (PSA), pseudo-spectral velocity (PSV), and spectral displacement (SD).
There exists more than 120 equations proposed by various researchers (Douglas 2001)
for PGA alone. Another important parameter that has proven to be strongly correlated
with the potential destructiveness of an earthquake is the Arias Intensity (Travasarou et
al. 2003). Numerous studies have been made for capturing the effects of earthquake

18
magnitude, rupture mechanism, site characteristics and source to site distance on these
strong motion parameters. In this section, the performance of two studies for predicting
PGA (Boore et al. 1997, Sadigh et al. 1997) and one study for predicting Arias Intensity
(Travasarou et al. 2003) will be evaluated.

2.3.1 Attenuation Relationships Predicting Peak Ground Acceleration

Boore et al. (1997), using a record set mostly based on California earthquakes,
proposed the following empirical equation for estimating the PGA of the average
horizontal component:

Vs (2.2)
ln (Y ) = b1 + b2 ( M − 6 ) + b3 ( M − 6 ) + b5 ln ( r ) + bV ln
2

VA

where r = rjb 2 + h 2

where Y is the peak horizontal ground acceleration or pseudo-acceleration response

(measured in g), rjb is the closest distance to surface projection of rupture surface (the so-
called. ‘Joyner and Boore Distance’), M is the Moment Magnitude, VS (m/s) is the
average shear-wave velocity in the upper 30 m of the recording station, b1, b2, b5, bV, h
and VA are the coefficients to be found from weighted, two-stage, regression analysis.
Coefficients and standard deviations associated with the above equation may be found in
Appendix B.

Boore (2001) has compared predictions calculated with this equation with
records obtained from 1999 Chi-chi earthquake (Mw=7.6). In his study, Boore (2001)
concluded that, for periods less than 1 s, predictions on the average were about twice the
recorded ground motions. Additionally, Boore (2001) observed that long duration wave
trains present in the ground motions even at distances of 30-60 km produced spectral
accelerations as much as five times larger than predictions made by this equation, in the

19
range of periods from 2 to 20 seconds. Boore (2001) expressed the view that these
unexpected differences might be due to site and propagation effects.

Geometric means of the PGA’s found from two components of strong ground
motion records obtained from each station are plotted above the attenuation relationship
proposed by Boore et al. (1997) in Figure 2.7. The moment magnitudes used in the
generation of each attenuation curve have been indicated on the top right corner of the
graphs in Figures 2.7, 2.9 and 2.10.

In their original study Boore et al. (1997) have directly incorporated the site
effects using the average shear-wave velocity in the upper 30 m of the recording station.
However, in this study only the NEHRP site classes of the strong motion stations were
available. Hence, the Vs in Equation 2.2 was taken as equal to the recommended value
by Boore et al. (1997) for each NEHRP site class. Closest distance to surface projection
of rupture surface was not available for 6 stations in the record set, so, they are excluded
from these curves.

In the topmost two curves in Figure 2.7, the principal representation is for near-
fault strong motion records obtained from the 1999 Chi-chi earthquake (Mw=7.6). It is
evident from these two graphs that there is a considerable overestimation for both C and
D class sites. Additionally, it should be stated that a significant dispersion exists in the
data.

The second row of curves in Figure 2.7 contains measurements of the strong
motion records from 1992 Cape Mendocino (Mw = 7.1) and 1989 Loma Prieta (Mw =
6.9) earthquakes. Although there are too few points to draw a comment, it may be seen
that all the points lie within the 84 percentile (mean plus one standard deviation) range.

20
 Mw=7.6 Mw=7.6
1 1

PGA (g)

PGA (g)
0.1 0.1

th Site D
84 Percentile th
84 Percentile
Site C ( Reverse Slip )
Mw = 7.6 ( Reverse Slip )
Median Median
1999 Chi-Chi Earthquake (Mw=7.6)
1999 Chi-Chi Earthquake
0.01 0.01
0.1 1 10 100 0.1 1 10 100
Closest horizontal distance to vertical projection of rupture, rjb(km) Closest horizontal distance to vertical projection of rupture, rjb(km)

Mw=7.1 Mw=7.1
1 1
PGA (g)

PGA (g)
0.1 0.1

Site C (Reverse Slip )

th
84 Percentile 1992 Cape Mendocino (Mw=7.1) 84th Percentile Site D (Reverse Slip)
Median 1989 Loma Prieta (Mw=6.9) Median 1989 Loma Prieta (Mw=6.9)
0.01 0.01
0.1 1 10 100 0.1 1 10 100
Closest horizontal distance to vertical projection of rupture, rjb(km) Closest horizontal distance to vertical projection of rupture, rjb(km)

Mw=6.7 Mw=6.7
1 1
PGA (g)
PGA (g)

0.1 0.1

Site D ( Reverse Slip )

th
84th Percentile 84 Percentile 1994 Northridge (Mw = 6.7)
Site C (Revese Slip)
Median Median 1971 San Fernando (Mw = 6.6)
1994 Northridge (Mw = 6.7)
0.01 0.01
0.1 1 10 100 0.1 1 10 100
Closest horizontal distance to vertical projection of rupture, rjb(km) Closest horizontal distance to vertical projection of rupture, rjb(km)

Mw=6.5
1
PGA (g)

0.1

84th Percentile
Site D ( Stike Slip )
Median
1979 Imperial Valley (Mw = 6.5 )
0.01
0.1 1 10 100
Closest horizontal distance to vertical projection of rupture, rjb(km)

Figure 2. 7 Attenuation Relationship Proposed by Boore et al. (1997) and Scattergrams

of Strong Motion Records Used in This Study

21
 The third row of graphs in Figure 2.7 consists of strong motion records from
1994 Northridge (Mw = 6.7) and 1971 San Fernando (Mw = 6.6) earthquakes. Except for
a few outliers, it is observed that the attenuation equation follow the trend of scatter
points. Similar to the Mw = 7.1 graph, it seems that most of the data points are located
between the mean and 84 percentile. This may be an indication that the strong motion
records in the dataset have higher PGA’s than the average near-fault strong motion
records.

The single graph at the bottom of Figure 2.7 represents the strong motion records
from 1979 Imperial Valley (Mw = 6.5) earthquake. It is clearly observable from the
graph that, attenuation relationship by Boore et al. (1997) captures the individual values
and trend of data very well.

A similar equation for estimating PGA’s and PSA’s, resulting from earthquakes
with a moment magnitude range from 4 to 8, at distances up to 100 km, has been
proposed by Sadigh et al. (1997). In this study, unlike Boore et al.(1997) all the ground
motion records are grouped into two according to their site characteristics as “rock sites”
and “deep soil sites”. Sadigh et al. (1997) have stated that deep soil data has been
collected from sites having greater than 20 m of soil over the bedrock. In the light of
this statement, it may be assumed that record set used in this study was derived from
“deep soil sites”. The comparisons were made accordingly. Function and coefficients
proposed by Sadigh et al.(1997) for the prediction of PGA are given in Appendix B. An
important point to be noted is that, the attenuation relationship proposed by Sadigh et al.
(1997) and Boore et al. (1997) have different distance definitions. The two site to source
distances, namely closest distance to surface projection of the rupture surface (rjb) used
by Boore et al. (1997) and closest distance to rupture surface (rrup) used by Sadig et al.
(1997), are shown in the representative sketch in Figure 2.8.

22
 Surface projection of the
rupture surface

Recording station
rjb

rrup

Rupture surface

Figure 2.8 Sketch of rjb and rrup

Attenuation relationship proposed by Sadigh et al (1997) and measurement

points of the strong motion records used in this study may be inspected in Figure 2.9.
The graphs contain the same earthquakes as in Figure 2.7. Unlike Boore et al. (1997),
there is some underestimation for the Mw = 7.6 graph. This underestimation is even
more noticeable in the Mw = 6.7 graph. More importantly, a clear parallel trend can not
be observed between the points and attenuation relationship.

In conclusion, it may be stated that according to dataset used in this study

attenuation relationship proposed by Boore et al (1997) seems to capture the overall
trend better than Sadigh et al (1997). This may be due to the difference of the distance
definitions proposed for each relationship.

23
 Mw=7.6 Mw=7.1
1 1

PGA (g)

PGA (g)
0.1 0.1

( Reverse Slip )
84th Percentile ( Reverse Slip ) 84th Percentile
1992 Cape Mendocino (Mw=7.1)
Median 1999 Chi-Chi Earthquake (Mw=7.6) Median
1969 Loma Prieta (Mw=6.9)
0.01 0.01
0.1 1 10 100 0.1 1 10 100
Closest Distance to Rupture Surface, rrup (km) Closest Distance to Rupture Surface, rrup (km)

Mw=6.7 Mw=6.6
1 1
PGA (g)

PGA (g)
0.1 0.1

( Reverse Slip ) ( Stike Slip )

84th Percentile 1994 Northridge(Mw=6.7) 84th Percentile
1987 Superstition Hills (Mw=6.6)
Median 1971 San Fernando (Mw=6.6) Median
1979 Imperial Valley (Mw=6.5)
0.01 0.01
0.1 1 10 100 0.1 1 10 100
Closest Distance to Rupture Surface, rrup (km) Closest Distance to Rupture Surface, rrup (km)

Figure 2. 9 Attenuation Relationship Proposed by Sadigh et al. (1997) and Scattergram

of Strong Motion Records Used in This Study

2.3.2 Attenuation Relationships Predicting Arias Intensity

Emphasizing the fact that Arias Intensity is a useful measure for characterizing
strong ground motions, Travasarou et al. (2003) proposed an attenuation relationship for
estimating Arias Intensity. This relationship, which was derived from point-source
model and the coefficients found from regression analysis on 1208 ground motion
records, is given in Appendix B. It is stated that this relationship is applicable for
earthquakes with moment magnitudes in a range from 4.7 to 7.6 and distances from 0.1
km to 250 km (Travasarou et al. 2003). It should be noted that the equation proposed by
Travasarou et al. (2003) estimates the average of the Arias Intensities of the two
components.

24
 In Figure 2.10 this attenuation equation is plotted over the data obtained from the
near-fault strong motion record set used in this study. Travasarou et al. (2003) have
used the site classification scheme proposed by Rodrigez-Marek et al. (2001). This site
classification scheme, unlike NEHRP(2000) site classification which is only based on
the average shear-wave velocity of the soil profile in the upper 30 m, takes both the
shear-wave velocity and depth of soil layer above the bedrock as parameters. Since the
classification data of all of the stations according to this scheme was not available, plots
have been prepared assuming all the sites were BRM -D class (which results in higher Ia
values). It is seen from Figure 2.10 that the proposed attenuation relationship captures
most of the measurement points very well. The same over estimation in Mw = 7.6 set
and outliers in Mw=6.7 are seen in these graphs too.

Mw=7.6 Mw=7.1
10 10
Arias Intensity (m/s)
Arias Intensity (m/s)

1 1
BRM Site Class D
(Revese Slip)
1999 Chi-Chi (Mw=7.6) BRM Site Class D
(Revese Slip)
84th Percentile 84th Percentile 1992 Cape Mendocino (Mw=7.1)
Median Median 1969 Loma Prieta (Mw=6.6)
0.1 0.1
0.1 1 10 100 0.1 1 10 100
Closest Distance to Rupture Surface, rrup (km) Closest Distance to Rupture Surface, rrup (km)

BRM Site Class D

Mw=6.7 (Strike Slip) Mw=6.6
1987 Superstition Hills (Mw=6.6)
10 10
1979 Imperial Valley (Mw=6.5)
Arias Intensity (m/s)

Arias Intensity (m/s)

1 1

BRM Site Class D

(Revese Slip)
84th Percentile 1994 Northridge (M =6.7)
w 84th Percentile
Median 1971 San Fernando (Mw=6.6)
Median
0.1 0.1
0.1 1 10 100 0.1 1 10 100
Closest Distance to Rupture Surface, rrup (km) Closest Distance to Rupture Surface, rrup (km)

Figure 2.10 Arias Intensity Attenuation Relationship Proposed by Travasarou et al.

(2003) and Plotted against Strong Motion Records Used in This Study

25
2.4 Correlation between the Pulse Period and the Predominant Period from
Pseudo-Velocity Response Spectra

Two important parameters, characterizing the frequency content of near-fault

records are pulse period (Tν) and predominant period from pseudo-velocity response
spectrum (Tp-v). Pulse Period (Tν) is defined as the duration of the largest amplitude
pulse in the velocity trace of the strong motion record. Starting and ending times of the
pulse are either defined as zero crossing time or the time at which velocity is equal to 10
percent of the peak velocity (Stewart et al. 2001). Likewise, Somerville et al. (1999)
define pulse period as the largest cycle of motion in the strong motion record.
Predominant period in the pseudo-velocity response spectrum (Tp-v), as the name
implies, is measured as the period corresponding to the peak at the pseudo-velocity
response spectrum (Stewart et al., 2001). These two measures are capable of providing
information about the frequency content and spectral shapes of near-fault strong
motions, particularly for those affected by forward directivity.

Pulse period (Tν)’s of all of the near-fault records were measured by careful
examination of velocity traces of near-fault strong motion records that had “significant”
pulses in their velocity traces. Predominant period of the pseudo-velocity response
spectrum (Tp-v) of each record were read from the response spectra of all strong motion
records. A sample velocity trace and pseudo-velocity response spectrum for 5 percent
damping, together with corresponding Tν and Tp-v measurements are given in Figure
2.11.

26
 300
100 %5 damping
250

Pseudo-Velocity (cm/s)
50
Velocity (cm/s) 200

0 150

100
-50

50
-100 Tν ≅ 3.87s Tp-v≅ 3.40s
0
0 5 10 15 20 25 0.1 1 10
Time (s) Period, T (s)

Figure 2.11 Determination of Tν and Tp-v from 1979 Imperial Valley (Mw=6.5)
Earthquake Strong Motion Record (Station: El Centro Array #6, Component 230)

Stewart et al. (2001) have stated that coincidence of Tν and Tp-v for a strong
ground motion indicates that velocity pulse contains energy in a narrow period band. In
order to examine the correlation in between Tν and Tp-v, the plot given in Figure 2.12
was prepared, and coefficient of correlation was calculated. It is apparent from the
figure and calculated correlation coefficient that there is a strong connection between Tν
and Tp-v, especially for the periods shorter than 4 seconds. Mean and standard deviation
of the ratio between Tν and Tp-v were found to be 0.96 and 0.18, respectively. In a
similar study Rodrigez-Marek (2000) found a mean ratio of 0.84 and a standard
deviation of 0.28.

27
 12

Correlation Coefficient
10 ρx,y = 0.951

Pulse Period, Tν (s)

8

0
0 2 4 6 8 10 12

Predominant Period from Pseudo-Velocity Response Spectrum, Tp-v (s)

Figure 2. 12 Correlation between the Pulse Period and Predominant Period from Pseudo-
Velocity Response Spectrum for the Records Used in This Study

Sommerville et al. (1999), utilizing large set of ground motion records from 15
earthquakes, proposed the following relationship for Tν:

log10 Tν = −3.0 + 0.5M w (2.3)

where Tν is the pulse period (period of strongest pulse in velocity trace) in seconds and
Mw is the moment magnitude of ground motion.

A similar study, both relating Tν and Tp-v to magnitude, was presented by

Rodrigez-Marek (2000). Using 48 strong motion records from 11 earthquakes, he
proposed the following relationships:
ln Tv = −5.81 + 0.97 M w (2.4)

ln Tp-v = −5.66 + 0.91M w (2.5)

28
where Tν is the pulse period (period of strongest pulse in velocity trace) in seconds and
Mw is the magnitude of ground motion. Standard errors (σtotal) associated with Equations
(2.4) and (2.5) were given as 0.51 and 0.61, respectively.

Alavi and Krawinkler (2001) proposed the following relation for Tp-v:
log10 Tp-v = −1.76 + 0.31M w (2.6)

where Mw is the moment magnitude.

Tν versus Mw and Tp-v versus Mw relationships given above are plotted over the
individual points found using the record set in Figures 2.13 and 2.14, respectively. It is
evident from the two figures that although all relationships follow the trend observed in
the data sets, they are not fully capable of predicting the applicable ranges for related
variables. Additionally, the scatter in Tν-Mw graph (Figure 2.13) seems to be smaller
compared to Tp-v-Mw graph (Figure 2.14). Large scatter seen in the figures may be due
to other factors which are not taken into account in the proposed relationships.

18
Records used in this study
16 Somerville et al (1999)
Rodrigez-Marek (2000) - Median
14 Rodrigez-Marek (2000) - 84th Percentile
Pulse Period, Tν (s)

12

10

0
6.0 6.4 6.8 7.2 7.6 8.0
Moment Magnitude, Mw

Figure 2. 13 Scatter Plot of the Records Used and Tν vs. Mw Relationships

29
 12
Records used in this study
Alavi and Krawinkler (2001)

(s)
p-v (s)
10 Rodrigez-Marek (2000) Median

Predominant Period in PSV, TTv-p

Rodrigez-Marek (2000) 84th Percentile

0
6.0 6.4 6.8 7.2 7.6 8.0
Moment Magnitude, Mw

Figure 2.14 Scatter Plot of the Records Used and Tp-v vs. Mw Relationships

In light of the observations stated above, it is reasonable to examine the variation

in Tν with other parameters such as site to source distance and faulting mechanism.
Variation of pulse period with distance may be inspected from the graph in Figure
2.15(a). A narrowing trend with increasing distance is hardly noticeable in this graph.
Magnitude clusters significantly show the variation with magnitude which was selected
as the governing parameter by other researchers. The graph in Figure 2.15(b) was
prepared for making a rough comparison among pulse periods observed from events
which have similar magnitudes but different faulting mechanisms. A rough conclusion
that may be derived from this graph is that, for the data set used in this study, strike slip
events tend to produce longer velocity pulses compared to reverse slip events.

30
 12 12 55

Mw
Mw == 7.6
10 10 6.7 << Mw <7.6
6.7
44
Mw ≤≤ 6.7
Mw

nν (s)
Pulse Period, Tn (s)
Tnν (s)
(s) 8 8

(s)
Period, TTTn (s)
33
Period, T

PulsePeriod,
PulsePeriod,

Period,
6 6

22

Pulse
Pulse
Pulse

4 4

1
2 1
2 MMw ==6.5
6.5Strike
StrikeSlip
Slip
w
MMw ==6.7
6.7Reverse
ReverseSlipSlip
0 w
0 0
0 3 6 9 12 15 18 21 00 2 4 6 8 10 12 14
0 3 6 9 12 15 18 21 0 2 4 6 8 10 12 14
Closest Distance to Rupture Surface, rrup (km) Closest Distance to Rupture Surface, rrup (km)
Closest Distance to Rupture Surface, rrup (km) Closest Distance to Rupture Surface, rrup (km)

(a) (b)

Figure 2. 15 Variation of Pulse Period with Parameters Other than Moment Magnitude

2.5 Spectral Displacements of Ground Motion Components

Spectral displacements are the displacement demands imposed on the single

degree of freedom structures by ground motions. Therefore, they are of primary
importance for earthquake resistant design. Earthquake ground motion is a three
dimensional motion. However, in common engineering applications only the two
horizontal components are utilized. Using these two components, an infinite number of
components in different directions, may be generated. Eventually, all these generated
components would have different amplitude and frequency characteristics. As a result,
they would have various different displacement spectra. Since it is not practical to
analyze the structure under all these generated components, a severest component should
be selected. In general, there is no such thing as “severest” component of a ground
motion record if individual periods for maximum base shear are considered. Usually,
for one period highest spectral displacement is observed in one direction, while for
another period it is observed in another direction.

Somerville et al. (1997) state that, for the strong motion records showing effects
of forward directivity, maximum spectral displacements are in the fault normal

31
component. However, for some cases the fault normal direction may not be readily
available. Sometimes, a consensus on the layout of the fault can not be established or
such information may not be reachable for a design engineer. Therefore, there is a need
for a more practical method for finding the ‘relatively’ severest component of a near-
fault ground motion.

In Figure 2.16, displacement spectra found by simple methods such as taking the
arithmetic and geometric means of individual spectral displacements found from two
recorded components are plotted. It is evident from the top four frames that there is no
significant difference between taking the means and geometric means. Another
important observation is that, records with pulse have noticeably higher spectral
displacements compared to records without pulse. Akkar and Gülkan (2002) have
studied the fault normal and maximum velocity direction components of near-fault
strong motion records from 1999 Kocaeli (Mw=7.4) and 1999 Bolu-Düzce (Mw=7.2).
Akkar and Gülkan (2002) pointed out that, the ground motion components in the
maximum velocity directions, in general, have higher spectral drifts than other
components. They have also stated that the maximum velocity direction does not
always coincide with the fault-normal direction. In light of this study, spectral
displacements in the maximum velocity direction are plotted in the two frames at the
bottom of Figure 2.16. Noticeably, higher spectral displacements are seen in the
maximum velocity direction components of strong motion records with pulse. It should
also be noted that, such a difference is not observed in records without pulse.

In conclusion, when fault normal directions of the strong motion records are not
readily available, a relatively severe component may be found in the maximum velocity
direction.

32
 Strong Motion Records with Pulse Strong Motion Records without Pulse
Mean of Components Mean of Components

Individual Records 120 120

Individual Records
Spectral Displacement, Sd (cm) Median Median

Spectral Displacement, Sd (cm)

Median +/- σlnY 100 Median +/- σlnY 100

80 80

60 60

40 40

20 20

0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Period, T (s) Period, T (s)

Geometric Mean of Components Geometric Mean of Components

Individual Records 120 Individual Records 120

Median Median

Spectral Displacement, Sd (cm)

Median +/- σlnY Median +/- σlnY
Spectral Displacement, Sd (cm)

100 100

80 80

60 60

40 40

20 20

0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Period, T (s) Period, T (s)

Maximum Velocity Direction Component Maximum Velocity Direction Component

Individual Records 120 Individual Records 120

Median Median
Spectral Displacement, Sd (cm)

Median +/- σlnY

Spectral Displacement, Sd (cm)

100 Median +/- σlnY 100

80 80

60 60

40 40

20 20

0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Period, T (s) Period, T (s)

Figure 2.16 Comparisons among Spectral Displacements of Records with and without
Pulse

33
 Taking into account all of these observations it may be concluded that, near-fault
ground motions have characteristics that are not observed in far-field ground motions.
An in-depth understanding of these properties would eventually lead to development of
reliable models for predicting demands imposed by near-fault ground motions on
engineering structures. Such models are crucial both for designing earthquake resistant
structures near-active faults and for reliable evaluation of existing structures affected by
near-fault strong motions.

34
 CHAPTER 3

BASIC THEORY AND DERIVATION OF EQUATIONS

3.1 Models and Analysis Procedures

In this chapter, elastic response of multi-degree-of-freedom (MDOF) systems

and inelastic response of singe-degree-of-freedom (SDOF) systems subjected to near-
fault ground motions will be analyzed. Despite the fact that, moment resisting frames
are usually designed to yield under strong ground motion excitations, it is generally
believed that analysis of the elastic response of these frames under near-fault excitations
provides information on their response characteristics. During elastic MDOF response
history analyses local displacement demands will be the main items of concern. In
addition to examining the elastic response of conventional moment resisting frames, one
of the main aims of this study is the enhancement of the simple drift estimation equation
proposed by Gülkan and Akkar (2002). A set of coefficients will be introduced for
further improving the predictive power of the equation by adding the capability of
estimating the ground story drift ratios and maximum interstory drift ratios of frames
having beams and slabs relatively more flexible than columns.

Response of a series of SDOF systems to the near-fault ground motions in the

record set will be used to examine the fundamental characteristics of the response of

35
inelastic systems to near-fault ground motions. Using the results of these analyses,
differences between the response of systems to records with and without pulse will be
compared.

3.1.1 MDOF Models

3.1.1.1 Properties of the Idealized Frames

In order to capture the effects of different structural properties on the response, a

set of idealized moment resisting frames were generated. Idealized frames were single-
bay, rectangular plane frames, with a constant story height equal to 3 m and, bay width
equal to 6 m. Structural configuration of a sample frame may be viewed in Figure 3.1.
Only flexural deformations were considered in the structural members. P-delta effects
were not taken into account in the analyses. The frames had uniform stiffness through
their height. Moments of inertia of beams (Ib) were varied to investigate response
characteristics of different frame systems, while moments of inertia of all the columns
(Ic) were kept constant. The damping ratio for all modes of vibration was assumed to be
equal to 5 percent of critical damping.

2h
Ib
0.75m

h Ic Ic

Ib
m

h Ic Ic

Ib
m

h Ic Ic

Figure 3.1 Representative Sketch of the Structural Configuration of Idealized Frames

36
 The story masses were assumed to be concentrated at the floor levels and
rotational inertias of stories were neglected. Mass of the roof story of each idealized
frame was set to be 75% of the other stories. Story masses were adjusted so that the
fundamental period Tn of the frames were equal to the period found from the
approximate relation below:
Tn = 0.1N ( 3. 1 )
where N is the number of stories and Tn is the fundamental period in seconds.

Beam-to-column stiffness ratio (ρ), originally named as the joint rotation index
by Blume (1968), is a parameter for quantifying the distribution pattern of lateral
deformations in frames. ρ is calculated as follows:

Ib
∑ L
ρ = beams b ( 3. 2 )
Ic
∑
columns Lc

where Ib and Lb are the moments of inertia and lengths of the beams in the story closest
to mid-height of the building, respectively, and Ic and Lc are the moments of inertia and
heights of the columns, in the story closest to mid-height of the building, respectively. It
should be noted that units of moments of inertia and lengths should be consistent. It is
assumed that for a building with nearly uniform distributed stiffness, a representative
value of ρ may be found by using the member properties of the elements in the mid-
height of the building. For the case of idealized frames used in this study Equation 3.2
reduces to:
Ib
ρ= ( 3. 3 )
4I c

The parameter ρ is a measure of relative beam-to-column stiffness. It is directly

related to the degree of frame action. When ρ equals to zero, indicating Ib equals to

37
zero, and beams do not impose any restraint to rotation of joints. As a result, system
shows an overall lateral deformation pattern, like a vertical cantilever bending beam.
Solid structural walls, with no considerable framing or effective spandrels, would
portray such deformations. When, the ρ equals infinity, theoretically beams are
infinitely stiff relative to columns and joint rotations are completely restrained. This
frame model is named as shear building or shear frame in the literature (Ayre 1956, Paz
1985, Chopra 2000). Paz (1985) has stated that the name shear building has originated
from the overall lateral deformation pattern of these frames which is similar to a shear
beam deflected by shear forces. It should be noted that, for most of the time, the
deformations taken into account in the analysis of shear buildings are only the flexural
deformations of structural components. It should also be noted that in this study the
deformations taken into account are only the flexural deformations of structural
members. In the shear buildings, since all joint rotations are completely restrained, all
the columns would bend in double curvature in the displaced configuration.
Fundamental mode shapes of idealized four story frames, with various ρ values, are
shown in Figure 3.2.

ρ=0 ρ = 0.125 ρ=4 ρ=∞

Overall lateral Overall lateral
deformation pattern deformation pattern
like a bending beam like a shear beam

Figure 3.2 Fundamental Mode Shapes of Idealized Frames with Different ρ Values

38
 Blume (1968) calculated ρ values, changing between 0.1 and 1.55, for 27 frames
designed by several engineers. In the light of this observation, for the MDOF analyses
in this study, moment resisting frames with ρ equal to 0.125, 0.25, 0.5, 0.75, 1, 1.5, 2, 3
and 4 were used. A separate case, utilizing shear-frame model with completely
restrained joint rotations was included in the analysis ( ρ = ∞).

A total of 190 moment resisting frames (corresponding to 19 different number of

stories from 2 to 20 and 10 different ρ values) were used in the MDOF analyses.
Studying such a wide set of moment resisting frames led to better understanding of
various factors affecting the response of frames.

3.1.1.2 Analysis Procedure for MDOF Systems

For the elastic response history analysis and inspection of modal properties, a
special MATLAB program was prepared. The program has the capability of generating
any frame with the given number of story, fundamental period and ρ. The seismic
response of frames was found by superposing the individual modal responses. Response
in each mode was computed by exact solution of the response of linear system under the
ground motion excitation, interpolated over each time step. This numerical procedure
has proven to be useful, especially when the excitation is defined at closely spaced time
intervals, such as the ground motion records used in this study (Chopra 2000).

Organization of the output file was arranged so that, it could be directly imported
to common spreadsheet programs. This compatibility led to easier examination and
processing of the output data.

39
3.1.2 SDOF Systems

3.1.2.1 Properties of the SDOF Systems

A set of inelastic SDOF systems was used for analyzing the basic response
characteristics of yielding systems under near-fault ground motions. Natural periods of
SDOF systems covered the range from 0.1 s to 3.0 s, with increments of 0.05s. In order
to examine the significance of pulse period, which was defined in Chapter 2, on the
inelastic displacement demands imposed on structures a series SDOF systems having
natural period to pulse period ratios (Tn /Tν) between from 0.1 to 3 were generated for
each strong motion record with pulse. Damping was set to be equal to 5 percent of
critical. Elastoplastic load deformation model was used for simulating the inelastic
behavior of SDOF systems. This load deformation model is commonly used for
modeling the inelastic behavior of structural elements that display small strength or
stiffness degradation. This is one of most widely used hysteretic models. Figure 3.3
shows the load deformation curve of a typical elastoplastic system and corresponding
linear elastic system.

F
Corresponding
fo linear system

fy
Elastoplastic
system
k k
1 1

uy uo um u
k
1

-fy

Figure 3.3 Elastoplastic Load Deformation Model and Corresponding Linear

Elastic System

40
 In Figure 3.3, F is the force applied on the system, k is the stiffness of the system
in the elastic range, uy is the yield displacement, fy is the yield strength, fo is the
maximum earthquake force on the corresponding linear system, um is the maximum
displacement of inelastic system and uo is the maximum displacement of the
corresponding linear system.

Strength reduction factor (R), is the ratio of the strength demand imposed on the
linear elastic system to the strength capacity of corresponding inelastic system with the
same natural period. It is formulated as follows:
f o uo
R= = ( 3. 4 )
f y uy

. R is a convenient parameter for normalizing the capacity of a system and it has been
used as the main capacity parameter in this study. Inelastic response history analyses
were performed for SDOF systems having R’s equal to 1.5, 2, 3, 4, 5, 6, 7 and 8.

3.1.2.2 Analysis Procedure for SDOF Systems

A special FORTRAN program was prepared for performing elastic and inelastic
response history analysis of SDOF systems. The program was based on Newmark’s
(1959) Linear Acceleration Method for Nonlinear Systems.

The special FORTRAN program had the capability of calculating the

displacement response spectra for a given list of records and a given set of Ry values
automatically for any range of periods.

41
3.2 Effects of Near-Fault Ground Motions on the Response of Elastic Moment
Resisting Frames

Results obtained from the elastic response history analysis of the idealized
frames subjected to near-fault ground motions provided a comprehensive data. In this
section, observations made on the effects of strong ground motion pulses on the
response of elastic moment resisting frames are presented. The importance of local
displacement demands, particularly for the structures excited by near-fault ground
motions, has been addressed by several studies (Iwan 1997, Alavi and Krawinkler 2001,
Gülkan and Akkar, 2002). Studies by Alavi and Krawinkler (2001) have shown that
high amplitude pulses observed in some of the near-fault ground motions result in
significantly high local displacement demands. In light of this observation, ground story
drift ratio (GSDR) and maximum interstory drift ratio (MIDR) were selected as the main
displacement demand parameters in this study. Ground story drift ratio (GSDR) is
defined as the lateral drift of the ground story normalized by the height of the ground
story columns and maximum interstory drift ratio (MIDR) is defined as the maximum
interstory displacement divided by the story height. GSDRALL and MIDRALL are short
for ground story drift ratios and maximum interstory drift ratios found by considering
the contributions of all modes, respectively. Similarly, GSDR1 and MIDR1 are the
ground story drift ratios and maximum interstory drift ratios found by considering only
the contribution of the first mode, respectively.

3.2.1 Effects of Pulse Period on Local Displacement Demands

In order to examine the effects of strong velocity pulses on the local

displacement demands, ratio of fundamental period to pulse period (Tn/Tν) were
computed for all the individual response history analyses of frames. GSDRALL versus
Tn/Tν points are plotted in Figure 3.4. It may be observed from these plots that, GSDR’s
are noticeably higher near Tn/Tν=1. In other words, for the frames with fundamental
periods close to pulse period GSDR’s are significantly higher, compared to frames with

42