Chapter 2

Seismic Input for Displacement

Based Seismic Design

Engr. Muhammad Qaisar

2015-UET-PhD/CE-77
 Assignment

Submitted To:
Prof. Dr. Qaisar Uz Zaman Khan
Submitted By:
Engr. Muhammad Qaisar
2015-UET-PhD/CE-77
 2.1 Introduction
Characteristics of Accelerograms

Our understanding of the response of the structures to earthquakes, and our

design methodologies, either force-based or displacement-based, are
critically dependent on recordings of strong ground motion by
accelerographs. Accelerograms are recordings of strong ground motion
made by accelerographs during earthquakes, and the earliest records date
back to the 1930’s. Early accelerograms were recorded in analogue form on
photographic film, and required digitization to put them in a form where their
characteristics could be examined.
 2.1 Introduction
Characteristics of Accelerograms

Accuracy was limited, and the dynamic characteristics of the accelerographs

themselves meant that useful data for preparing response spectra could be
extracted only up to periods of about two to three seconds. In the past
twenty years, digital accelerographs, with much high resolution and longer
range of period integrity have become increasingly common, and the quality
of data from recorded earthquakes is steadily improving as more digital
records become available.
 2.1 Introduction
Characteristics of Accelerograms

This text will not attempt to present seismological information about source
mechanisms, physical and temporal distributions of earthquakes, attenuation
relationships and modern developments in source modelling. The interested
reader is encouraged to read any of the many specialized seismological texts
[e.g. S7, K8]. The treatment here will be limited to information of specific
relevance for displacement-based seismic design. Nevertheless, a brief
treatment of some of the common terms and a similarity brief discussion of
the characteristics of accelerograms is warranted.
 2.1 Introduction
Characteristics of Accelerograms

The vast majority of the earthquakes are initiated on or adjacent to tectonics

plate boundaries by the slow relative movement of the plates. These are
termed interplate earthquakes. Interplate earthquakes, occurring far away
from plate boundaries are less common, but nevertheless can be significant
for specific sites ( for examples the New Madrid region of the central USA,
Charleston, South Carolina, USA and various parts of Australia).

The two basics terms used to provide a measure of the importance of a

particular earthquake are the magnitude and intensity.
 2.1 Introduction
Characteristics of Accelerograms

The magnitude, normally related to the Richter scale[P1] is a measure of the

energy release at the fault zone, while the intensity is a measure of the local
significance of ground motion at a given site, as described by locally recorded
accelerograms, or by subjective scales, such as Modified Mercalli scale.

The magnitude depends on the length and transverse dimension of fault that
fractures during the earthquake and on the average stress – drop in the rock
immediately adjacent to the fault, resulting from rupture.
 2.1 Introduction
Characteristics of Accelerograms

A magnitude 5.0 – 5.5 earthquake may result from faulting over a length of

few km, while a magnitude 8 earthquake may involve fault slip over a length

as much as 400 km. The energy release is related to magnitude in

proportion to 101.5M, implying that energy released increases by a factor of

32 for each unit increase of magnitude. Earthquakes with low magnitude

occur frequently, those of large magnitude occur less frequently. Averaged

globally the relationship between magnitude and annual probability of

 2.1 Introduction
Characteristics of Accelerograms

occurrence agrees well with a Gumbel extreme type 1 distribution[P1]. However, this
relationship becomes less reliable as the area sampled reduces in size. The
concept of a stationary value for the annual probability of occurrence of small to
moderate earthquakes near a given site is generally reasonable, but for the
larger earthquakes, particularly where the site is predominantly affected by
the earthquake on single fault, this assumption may be less valid.
Immediately after a major earthquake and its related aftershocks have
ended, the probability of major fault movement of the section of fault is
 2.1 Introduction
Characteristics of Accelerograms

is significantly reduced, potentially reducing the major contribution to local

seismic hazard. An example is the stretch of plate boundary along the coast
of Chile, where major earthquakes trend to occur on specific segments of
the fault at rather regular time intervals, and with comparatively uniform
magnitudes. This of course will not be the case with smaller earthquakes,
and even for large earthquakes where fracture of one segment of a fault
may create additional stress on the adjacent section, increasing the
 2.1 Introduction
Characteristics of Accelerograms

the probability of fracture of this section in the near future. An example is

Anatolian fault in Turkey, and the subduction boundary between the Nazca
and South American tectonic plates where fault rupture tends to occur on
successive adjacent segments of the fault in a comparatively regular
sequence. Nevertheless, it is common in seismic hazard analysis to assume
that the local risk is time- invariant.
 2.1 Introduction
Characteristics of Accelerograms

Intensity is dependent on magnitude of the causative earthquake, distance

from the fault zone, mechanism and direction of rupture propagation, and
ground conditions at the site at which intensity is observed, and between the
fault zone and the site. There is no exact means of measuring intensity,
since it is generally assessed through the effect that the earthquake has at a
given site on the built environment. This has typically been defined in the
past through descriptive scales such as the Modified Mercalli scale[P1].
 2.1 Introduction
Characteristics of Accelerograms

Attempts to relate such scales that are dependent on observations of

damage to different structural types and materials, to such measurable
quantities as peak ground acceleration or velocity have not been particularly
successful, as different ground motion quantities have different significance
to different structural types. Thus peak ground acceleration may be
important to structures that have brittle failure modes, but may be of little
importance to a flexible well-confined structure. Duration of shaking may be
a key parameter for a flexible structure without adequate confinement.
 2.1 Introduction
Characteristics of Accelerograms

No two accelerograms are identical, even when the earthquakes originate in

the same part of a fault, with similar magnitudes, and the site where the
accelerograms are recorded is the same. Some of the differences and
similarities between accelerograms are illustrated by the three examples of
Fig.2.1.
 2.1 Introduction
Characteristics of Accelerograms

The first earthquake of these is from the moderate Whittier earthquake of

1987 (Mw = 6.0), recorded in analogue form at a distance of 15 km from the
rupture. The second is from the Mw = 6.7 Northridge earthquake of 1994,
recorded at the Sylmar site, at a distance of 6 Km from the fault rupture, and
the final record is from the Mw = 6.9 1995 Kobe earthquake, recorded
immediately adjacent to the fault. These are shown in Fig. 2.1 in next
 2.1 Introduction
Characteristics of Accelerograms

Fig. 2.1 Selected Time-Windows of different Accelerograms

 2.1 Introduction
Characteristics of Accelerograms

Fig. 2.1 Selected Time-Windows of different Accelerograms

 2.1 Introduction
Characteristics of Accelerograms

Fig. 2.1 Selected Time-Windows of different Accelerograms

 2.1 Introduction
Characteristics of Accelerograms

None of the records shows the complete duration of recorded motion, but
all include the section of greatest interest, including before, during and
immediately after the strong ground motion. All three records are plotted to
the same time and acceleration scales.

The record from the smaller Whittier earthquake has a comparatively short
period of strong ground motion compared with the other two records, the
peak ground acceleration (PGA) is lower, and it appears that high frequency
components are more dominant.
 2.1 Introduction
Characteristics of Accelerograms

On the other hand, all records show an initial period of comparatively high
frequency/low amplitude acceleration before the onset of the strong-motion
period of response, corresponding to the time period between arrival of the
P and S waves. Both the Sylmar and Kobe records show high amplitude/low
frequency pulses in the initial stages of the strong ground motion,
corresponding to a focusing effect related to the mechanism of energy
release and the local geology, known as a velocity pulse.
 2.1 Introduction
Characteristics of Accelerograms

ln the case of the Kobe record, this is primarily a result of forward directivity

where the fault fractures over a short period of time from one end to another,

focusing the energy in the downstream direction. In the Sylmar record, the

reasons are apparently more complex, involving basin edge effects[S5].

It is also of some interest to examine the time sequence of ground

displacement, found by double integration of the acceleration records.

These are shown for the same three records in Fig.2.2, and are plotted to
 2.1 Introduction
Characteristics of Accelerograms

the same time scales, but with a factor of 30 difference between the
displacement scales of the Whittier and the other two records. It should be
noted that integration of the acceleration records to obtain displacement
records is inevitably subject to some error. Small systematic errors in the
acceleration record can lead to large errors in the displacement record,
causing the apparent displacement to drift in one direction. Base-line
corrections are typically carried out to remove this drift, but the accuracy of
such corrections is uncertain.
 2.1 Introduction
Characteristics of Accelerograms

Fig. 2.2 Selected Time-Windows of Ground Displacement from the Accelerograms

Of Fig. 2.1
 2.1 Introduction
Characteristics of Accelerograms

Fig. 2.2 Selected Time-Windows of Ground Displacement from the Accelerograms

Of Fig. 2.1
 2.1 Introduction
Characteristics of Accelerograms

Fig. 2.2 Selected Time-Windows of Ground Displacement from the Accelerograms

Of Fig. 2.1
 2.1 Introduction
Characteristics of Accelerograms

It will be apparent from comparison of Figs.2.1 and 2.2 that the differences
between the displacement records are more pronounced than between the
acceleration records. Although the PGAs for the three records only vary by a
factor of about 2.5, the peak ground displacements (PGD) vary by a factor of
about 25. All three records exhibit much less high frequency content in
displacement terms than in acceleration terms, but the Whittier record is
significantly richer in high frequency components than the other two records.
The two more intense records appear to show dominant long period ground
 2.1 Introduction
Characteristics of Accelerograms

displacement response. In the case of the Sylmar record, this appears to

correspond to a period of about 3 to 4 seconds.
Intensity, for a given earthquake, decreases with distance from the fault.
Attenuation relationships are used to describe this reduction in intensity.
However, there is a large spatial variation in recorded ground motions
between different sites at equal distances from the epicentre of an
earthquake. Attenuation relationships are averages found from recorded
accelerograms, and typically do not account for many of the factors known to
 2.1 Introduction
Characteristics of Accelerograms

influence intensity. Somerville et al[S5] have noted that the variations in

ground motion, which are particularly apparent at periods greater than one
second (and hence are of particular importance to structural design -
whether force-based or displacement-based) “can usually be attributed to
features of the earthquake source such as the orientation of the fault plane,
the style of faulting (strike-slip or dip-slip), and the evolution and distribution
of slip on the fault plane”[S5]. Thus it is reasonable to assume that the current
uncertainty associated with site intensity predicted by probabilistic
 2.1 Introduction
Characteristics of Accelerograms

seismic hazard analyses (PSHA) will be reduced as site and source

modelling improves.
A number of the factors affecting spatial variation of ground motion from a
given earthquake are discussed in [S5] in relation to the Los Angeles Basin.
The following notes provide a summary of the discussion in [S5]:

Near fault rupture directivity pulse:

Near fault recordings from recent earthquakes indicate that ground motion is
dominated by a large long-period narrow-band pulse in the fault normal motion,
 2.1 Introduction
Characteristics of Accelerograms

whose period increases with magnitude[S6]. This pulse may have a dominant
period of about 1 sec. for earthquakes of magnitude Mw = 6.7 - 7.0, and as
high as 4 sec. for earthquakes of magnitude Mw = 7.2 - 7.6.

Reverse faulting earthquakes:

Ground motions from reverse faulting earthquakes are systematically

stronger than ground motions from strike-slip earthquakes. The influence
may be as much as 20-40%.
 2.1 Introduction
Characteristics of Accelerograms

Buried faulting earthquakes:

Ground motions from shallow earthquakes that do not break the ground
surface are systematically stronger than from earthquakes that result in
surface faulting. Again the influence may be in the order of 20-40%. The
1989 Loma Prieta, and the 1994 Northridge earthquakes are examples of
shallow earthquakes without surface faulting.
 2.1 Introduction
Characteristics of Accelerograms

Ground motion from large surface faulting earthquakes:

Ground motions from earthquakes that produce large surface faulting (eg.
the Chi-Chi earthquake in Taiwan) tend to be significantly lower than
predicted by current ground motion models, and substantially lower than
ground motions from buried faulting earthquakes.
 2.1 Introduction
Characteristics of Accelerograms

Basin effects:

Current codes modify design ground motions on the basis of the shear-wave
velocity in the upper 30m of soil. This is only appropriate for rather short-
period motion, as at periods greater than one second, seismic wave lengths
are much longer than 30m, and response is likely to be influenced by soil
properties at depths of hundreds, and perhaps thousands of metres. Basin
edge effects can also be significant, with constructive interference between
waves entering from the edge and from the basin below, particularly when
 2.1 Introduction
Characteristics of Accelerograms

the basin has steep fault-controlled margins.

Recently developed hybrid simulation procedures[G4] are capable of

incorporating all of the above features in calculating broadband ground
motion time-histories for prescribed earthquake scenarios. PSHAs based on
these techniques are already more reliable than those based on attenuation
relationships, and it can confidently be expected that improved
characterization of seismicity of specific sites will continue.
 2.1 Introduction
Characteristics of Accelerograms

It is likely that future developments will be less towards improved accuracy

of code spectra, than towards improved mapping of local characteristics
defining seismicity, with spectral shape as well as spectral ordinates being a
mapped variable.
 2.2 RESPONSE SPECTRA
2.2.1 Response Spectra from Accelerograms

The fundamental information extracted from accelerograms or PSHA’s for

use in design is typically expressed in the form of response spectra, which
represent the peak response of single degree of freedom oscillators of
different periods of vibration to the accelerogram. The quantities most
commonly represented in response spectra are absolute acceleration (with
respect to “at rest” conditions), and relative displacement (with respect to
instantaneous ground displacement), though relative velocity response
spectra are also sometimes computed.
 2.2.1 Response Spectra from Accelerograms

The procedure is represented in Fig.2.3 where five different SDOF

oscillators are depicted in Fig.2.3(a) subjected to the earthquake ground
motion ἃg. The peak absolute acceleration and relative displacements
recorded during response to the accelerograms are plotted against the
period of the structure in Fig.2.3(b).

Normally response spectra provide information on the peak elastic response

for a specified elastic damping ratio (typically 5%), and are plotted against
the elastic period.
Fig. 2.3 Formation of Response Spectra
 2.2.1 Response Spectra from Accelerograms

lt is, however, also possible to plot inelastic spectra related to specified

displacement ductility levels. In this case the period may represent the initial
elastic period, or the effective period at peak displacement demand, related
to the effective stiffness.

Examples of elastic acceleration and displacement spectra for the three

accelerograms represented in Figs.2.1 and 2.2 are shown in Fig.2.4. The
spectra are shown for four levels of elastic damping, expressed as ratios to
the critical damping.
Fig. 2.4 Acceleration and displacement Elastic Response Spectra for
Accelerograms of Figs. 2.1 and 2.2
Fig. 2.4 Acceleration and displacement Elastic Response Spectra for
Accelerograms of Figs. 2.1 and 2.2
Fig. 2.4 Acceleration and displacement Elastic Response Spectra for
Accelerograms of Figs. 2.1 and 2.2
 2.2.1 Response Spectra from Accelerograms

Some interesting conclusions can be drawn from examination of these

figures. The Whittier aecelerogram has a PGA of about O.4g (see Fig.2.1
(a)), and a peak response acceleration of more than O.8g at a period of
about 0.25 sec. This might be considered to represent reasonably strong
ground motion, since design PGA’s and peak response accelerations for
high- seismicity regions are often in the range 0.4g and 1.0g respectively.
However, when we examine the displacement spectra from the same
accelerogram, we find that the peak response displacement is less than
 2.2.1 Response Spectra from Accelerograms

20mm (O.79in), for a damping level of 5% of critical damping. Thus, if a

given structure is capable of sustaining this very minor peak response
displacement within the elastic range of response, no damage would be
expected, despite the high peak response acceleration. To put this in
perspective, information provided in Section 4.4.6 indicates that for
reinforced concrete Frame buildings of typical proportions, effective yield
displacements for two- and four-storey buildings might be approximately
45mm (1.8 in) and 90mm (3.6in) respectively – significantly larger than the
 2.2.1 Response Spectra from Accelerograms

Peak displacement response for this accelerogram. It is apparent that only

very stiff and brittle structures would be expected to be at risk from an
accelerograms similar to the Whittier record. This is in agreement with the
recorded damage in the Whittier earthquake.
Another point of interest is apparent from Fig, 2.4(a). Information from the
acceleration response spectra cannot be extracted for periods of T > 1.5 sec
since the response accelerations are so low.
 2.2.1 Response Spectra from Accelerograms

The displacement spectra provide much more readily accessible information

for the medium to long period range, but indicate surprisingly regular
displacements at periods greater than about 2 seconds. In fact this is false
data, since the accelerogram was recorded by an analogue, rather than
digital accelerograph, and a filter at 3 seconds was used to determine the
displacement response. Bommer et al[A8] have shown that the roll-off
associated with filtering makes the response spectra unreliable for periods
greater than about 2/3rds of the filter period.
 2.2.1 Response Spectra from Accelerograms

Thus the data in the displacement spectra of Fig.2.4(a) are meaningless for
periods greater than about 2 sec.

The Northridge Sylmar acceleration spectra of Fig.2.4(b) show peak

acceleration response for 5% damping of about 2.7g - about three times the
response for the Whittier earthquake. The displacement spectra, which
result from a digital accelerograph, and are reliable up to significantly longer
periods, indicate peak displacement response of about 800mm (31.5in) -
more than 40 times that of the Whittier record.
 2.2.1 Response Spectra from Accelerograms

Clearly this record would be expected to have much greater potential for
damage than the Whittier record. Note that after reaching a peak response
at about 3 sec., displacement response decreases at higher periods.

The Kobe record of Fig,2.4(c) also has high peak displacement response,
and somewhat similar characteristics to the Sylmar record, though the peak
displacement response appears to occur at a reduced period.
 2.2.2 Design Elastic Spectra

(a) Elastic Acceleration Spectra:

Until recently, design spectra for seismic design of structures were typically
specified in design codes as a spectral shape related to soil conditions,
modified by a design PGA, reflecting the assessed seismicity of the region
where the structure was to be built. Typically only acceleration spectra were
provided and mapping of the variation of PGA with location was coarse. This
is still the case with many seismic design codes.
 2.2.2 Design Elastic Spectra

Recently, more detailed information has been provided in different design

codes, such as the IBC[X4] of the USA, and the new Italian seismic design
code[X14], where spectral acceleration ordinates at two or three key periods
are provided for a given site for different return periods. Typically this is
provided through a computerized data base, enabling design data to be
extracted based on site longitude and latitude. However, this is typically
provided only for acceleration response spectra; peak displacement
response is not yet available in many design codes.
 2.2.2 Design Elastic Spectra

The typical form of elastic acceleration response spectra is illustrated in

Fig.2.5(a). The shape is smoothed, reflecting the average of many
accelerograms, and is based on probabilistic estimates of the contribution to
seismic risk of a larger number of smaller earthquakes, and a reduced
number of larger earthquakes. The result is a spectrum where the
acceleration ordinates have uniform probability of occurrence for a given
return period (see Section 2.2.2(c) below).
Fig. 2.5 Design Elastic Acceleration and Displacement Spectra
 2.2.2 Design Elastic Spectra

The spectrum rises from the PGA at T = 0 to a maximum value at a period

TA (typically about 0.15 seconds). For soft soils, codes typically amplify the
PGA above the value applicable for firm ground, or rock. The plateau
typically has a response acceleration of about 2.5 to 2.75 times the PGA.
The acceleration plateau continues to a period of TB, the value of which
depends on the ground conditions in the near-surface layers, with larger
values applying to soft soils, as indicated in Fig.2.5(a). The value of TB also
typically depends on the magnitude of the earthquake, as is apparent from
 2.2.2 Design Elastic Spectra

Fig.2.4, with smaller values being appropriate for earthquakes of lower

magnitude. For periods greater than TB the response acceleration reduces,

typically in proportion to T; implying a constant-velocity response. In many

codes this constant-velocity part of the spectrum continues indefinitely. More

advanced codes specify an upper limit of T = TC for the constant-velocity

range, above which the acceleration decreases in proportion to T2 . A

completely opposite trend is apparent in some less advanced codes, where

 2.2.2 Design Elastic Spectra

a constant plateau corresponding to a minimum specified response

acceleration is sometimes defined. This is shown by the dash-dot line in
Fig.2.5(a). The intent of such a provision is to ensure that the lateral strength
of a structure is not less than a code-specified minimum value. However,
this is better controlled by limits on P-Δ moments (see Section 3.8). As
discussed below, when the logic of the minimum acceleration plateau of the
acceleration spectrum is translated to equivalent displacements, impossible
trends result.
 2.2.2 Design Elastic Spectra

The general form of the elastic acceleration spectrum can be defined by

the following equations:
 
 PGA.1  C A  1
T
0  T  TA : S AT   ( 2.1a )
 TA 
0  T  TA : S AT   C A . PGA ( 2.1b)

TB
TB  T  TC : S AT   C A . PGA. (2.1c)
T
 2.2.2 Design Elastic Spectra

TB .TC
T  TC : S AT   C A . PGA. ( 2.1d )
T2
In Eqs.(2.1) SA is the spectral acceleration, and CA is the multiplier (e.g. 2.5)
applied to the PGA to obtain the peak response acceleration. Figure 2.5(a)
has been developed from these equations with the following parameters:

firm soil: PGA = 0.4g, TA = 0.15 sec, TB = 0.5 sec, TC = 4.0 sec, CA = 2.5
soft Soil: PGA = 0.48g, TA = 0.15 sec, TB = 0.8 sec, TC = 4.0 sec, CA = 2.5
 2.2.2 Design Elastic Spectra

(b) Elastic Displacement Spectra:

Although many codes still do not define design displacement spectra, they
are becoming more common [see, e.g. X2, X3, X4, X8]. Ideally these should
be developed separately, though using the same data, from acceleration
spectra. However, most code-based design displacement spectra are
generated from the acceleration spectra assuming that the peak response is
governed by the equations of steady-state sinusoidal response. Thus the
relationship between displacement and acceleration can be expressed as
 2.2.2 Design Elastic Spectra

T2
 (T )  .S AT  g ( 2.2)
4 2

where g is the acceleration due to gravity, and SA(T) is expressed as a multiply

of g as in Fig.2.5(a). Equation (2.2) has been used to generate the elastic
displacement spectra of Fig.2.5(b) from the acceleration spectra of Fig.2.5(a).
Some characteristics are immediately obvious:

• The displacement spectra are essentially linear with period up to the period .
TC For obvious reasons this is termed the corner period. The non-linearity at
 2.2.2 Design Elastic Spectra

low periods, corresponding to T < TB, will be found to have little relevance to

most displacement-based designs, as will become apparent in later chapters

of this text.

• The general shape agrees reasonably well with the response spectra
generated from specific accelerograms in Fig.2.4. The peak displacements
are also compatible with the more intense accelerograms of Fig.2.4(b) and
(C) considering differences in PGA.
 2.2.2 Design Elastic Spectra

• The displacement spectrum corresponding to a minimum design

acceleration as shown in Fig.2.5(a) has completely unrealistic
displacement demands for long period structures. This curve, again shown
by the dash-dot line in Fig.2.5(b) has only been shown up to a period of 5
seconds. At 10 seconds the response displacement would be 6.2 m (20ft).
This illustrates the illogical nature of some design codes that specify
minimum design acceleration levels, and require displacement demand to
be determined from Eq.(2.2).
 2.2.2 Design Elastic Spectra

A more general form of the elastic displacement response spectrum is

defined by Eurocode EC8[X9] and is shown in Fig.2.6. This shows the linear
displacement increase up to the corner period TC, with a subsequent plateau
of displacement up to a period TD, followed by a decrease in displacement
up to a period TE, at which stage the response displacement has decreased
to the peak-ground displacement (PGD). Information on the period at which
the response displacement starts to decrease is less reliable than data on
other key periods in Figs 2.5 and 2.6, and is of little interest to the designer
Fig. 2.6 General Characteristics of Elastic Displacement Response Spectra
 2.2.2 Design Elastic Spectra

other than extremely long-period structures such as suspension bridges and

large-diameter fluid storage tanks, where convective modes of vibration may
be very long. lt will be conservative to assume that the plateau displacement
remains constant with period above TC. However, it is worth noting that all
three of the displacement spectra shown in Fig.2.4 conform to the general
shape defined by Fig.2.6.
 2.2.2 Design Elastic Spectra

Recent work by Faccioli et al[F6] analyzing a large number of recent high-

quality digital records, has provided new information on the factors
influencing the shape of displacement spectra. The records investigated
included the very large data set from the large 1999 Chi-chi (Taiwan)
earthquake, (magnitude MW = 7.6), and a number of moderate European
and japanese earthquakes in the magnitude range 5.4 < MW <6.9.

Their findings include the following points:

 2.2.2 Design Elastic Spectra

• The 5% damped displacement spectra tend to increase essentially linearly

with period up to a “corner period". Beyond this, the response displacement

either stays essentially constant (for large earthquakes), or tends to decrease

(for moderate earthquakes). It is thus conservative to assume a constant

spectral displacement for periods higher than the corner period.

• The 10-second spectral displacement for 5% damping can be considered to

be a measure of the peak ground displacement.

 2.2.2 Design Elastic Spectra

• Soil amplification of displacement occurs throughout the period range, up

to 10 seconds. There is a slight tendency for the corner period to increase

for soft soils with large earthquakes, but this is less obvious for moderate

earthquakes.

• Soft soil amplification is more pronounced at longer distances (30-50 km)

for both moderate and large earthquakes.
 2.2.2 Design Elastic Spectra

• The corner period appears to increase almost linearly with magninicle. For
earthquakes with moment magnitude greater than MW = 5.7, the following
relationship seems conservative:

TC  1.0  2.5( M W  5.7) seconds (2.3)

• Peak response displacement, δmax, depends on the magnitude, the
epicentral distance r km (or nearest distance to the fault plane for a large
earthquake), and the stress drop during rupture (generally in the range 1 -
10 MPa).
 2.2.2 Design Elastic Spectra

Based on analytical considerations, Faccioli et al derive the following

relationship for peak response displacement, δmax, in cm, for firm ground
conditions:
log10  max  4.46  0.331log10   M W  log10 r (2.4)
where Δδ the stress drop, in MPa, MW is the moment magnitude, and r is the
epicentral (or fault plane) distance in km. Substituting an average value of
Δδ = 6 MPa, and reformatting as a power expression, with δmax in mm:
 2.2.2 Design Elastic Spectra

10 ( M W 3.2 )
 max  CS . ( mm) (2.5)
r

where CS =1.0 for firm ground. The response displacements resulting from

Eq.(2.5) should be modified for other than firm ground. Tentative

suggestions are as follows:

Rock: Cs = 0.7

Firm Ground Cs = 1.0

Intermediate SoiL Cs = 1.4 Very soft Soil Cs = 1.8

 2.2.2 Design Elastic Spectra

These tentative coefficients for Cs have been interpreted from typical

acceleration modifiers for PGA and period TB (see Fig.2.5(a)). More refined
estimates can be expected as further analyses of digital records become
available.

Equation (2.5), with Cs = 1.0 is plotted in Fig.2.7 together with Eq.(2.3) for
moment magnitudes of 6.0, 6.5, 7.0 and 7.5. The very strong influence of
magnitude and epicentral distance on spectral shape and maximum
displacement is clearly evident in Fig,2.7. It should be noted, however, that
 2.2.2 Design Elastic Spectra

It is probable that the peak displacements for high seismic intensity and low
(r = 10 km) epicentral distance are overestimated by Eq.(2.5). Within 10 km
of the fault rupture, spectral parameters tend to saturate, and the values for
10 km can be assumed to apply for distances less than 10 km.

Note also that the peak displacements predicted for earthquakes on the low
end of the moderate range (i.e. 5.5 < MW <6.0) are rather small, even at
epicentral distances of 10km.
 2.2.2 Design Elastic Spectra

In these cases, most frame buildings of three or more storeys in height, in

either reinforced concrete or steel construction, would have effective yield
displacements exceeding the 5% damped peak displacement δmax and
hence these buildings would be expected to respond elastically to the
design-level earthquake. At this early stage in the development of spectral
displacement characteristics, the above recommendations should be viewed
as preliminary. The consequences in terms of the shape of uniform-risk
spectral displacement spectra are also less obvious than for specific
earthquake intensities.
 2.2.2 Design Elastic Spectra

The current Eurocode EC8[X9] regulations specify that the corner period TC
(Fig.2.6) is taken as 1.2 sec, for causative earthquakes with MW < 5.5, and
2.0 sec. for MW ≥ 5.5. This would imply elastic response for most buildings
with more than about 8 storeys. It has been suggested by Boore and
Bommer[B7] that this low corner period is a result of interpretation of data
from analogue records which have been processed with low-order filters set
at periods that make the interpretation unreliable at periods above 2
seconds.
 2.2.2 Design Elastic Spectra

NEHRP recommendations[X8], based primarily on seismology theory include

the following relationship between corner period and moment magnitude:

log10 TC  1.25  0.3M W (2.6)

As a consequence of the current interest in displacement-based design, new

data are continually being developed by seismologists, related to

displacement spectra. A current study of ltalian seismicity with the aim of

 2.2.2 Design Elastic Spectra

developing design information for displacement based design in a form

similar to the maps of spectral accelerations at key defined periods seems
likely to end up with mapped corner periods and peak response
displacements in a data base related to GPS coordinates. Preliminary
results from this study indicate that both the corner period and the peak
elastic response period given by Eqs.(2.3) and (2.5) respectively may be
revised upwards by approximately 20%.
Fig. 2.7 Influence of Magnitude and Distance on 5% Damped Displacement
Spectra for Firm Ground using Eqs. (2.3) and (2.5)[after F6]
Fig. 2.7 Influence of Magnitude and Distance on 5% Damped Displacement
Spectra for Firm Ground using Eqs. (2.3) and (2.5)[after F6]
Fig. 2.7 Influence of Magnitude and Distance on 5% Damped Displacement
Spectra for Firm Ground using Eqs. (2.3) and (2.5)[after F6]
 2.2.2 Design Elastic Spectra

Figure 2.8 compares the different equations for corner period as a function
of magnitude. lt is clear that the current EC8 equation is severely non-
conservative, and there are significant differences between the NEHRP
equation, and the equation determined from the work by Faccioli et al[F6]. It
appears that the work in progress in Italy, using a world-wide data-base of
some 1700 digital records, and shown tentatively in Fig.2.8 by three dots
interpreted from data supplied by Faccioli2 may lie somewhere between
Eq.(2.3) and (2.6).
Fig. 2.8 Relationship between Corner Period, TC of Displacement Spectra
and Moment Magnitude MW
 2.2.2 Design Elastic Spectra

Figure 2.9 shows average displacement spectra interpreted by Faccioli et al

from the world-wide data base of 1700 digital records for earthquakes of
magnitude 6.4<MW<6.6 at different distances from the fault plane. The records
have been grouped in distance bands (10-30 km; 30-50 km etc), so the
averages can be interpreted as appropriate for the average distance (20km,
40km etc). Also shown are bi-linear displacement spectra fitted to average
spectra to obtain the best fit in the plateau range, while still satisfying normal
acceleration/displacement relationships in the shorter period range,
Fig. 2.9 Comparison of Averaged Displacement Spectra from World Wide
Data Base of Digital Accelerogram with Bi-Linear Approximations
 2.2.2 Design Elastic Spectra

lt will be noted that the corner period shows only slight increase with distance,
and that the plateau displacements decrease with distance at a rate that is
very close to, but slightly more rapid than that predicted by Eq.(2.5).

The consequences of the differences in corner period estimation are

potentially serious for long-period structures. Consider a frame building 60m
tall (i.e. about 20storeys). The structure is founded on firm soil, and the design
PGA is 0.4g, with the majority of seismic risk resulting from an MW =7
earthquake on a nearby fault.
 2.2.2 Design Elastic Spectra

As noted in Section 1.3.2 (Eq.(1.12)) a reasonable approximation of the

structural period is 0.1 x 60 = 6.0 sec. Combining Eqs.(2.1c) and (2.2), the

displacement at the corner period (and hence the peak elastic response

displacement) can be expressed as

 C  C A .PGA.TB .TC .g/(4 ) 2

(2.7)
Taking TB = 0.5 sec, and g = 9.805m/s2, Eq. (2.7) results in corner-period

displacement estimates of 250mm (9.8in), 530mm (20.9in) and 880mm (34.6in)

 2.2.2 Design Elastic Spectra

from the current EC8 value of TC = 2 sec, the value from Eq.(2.3) (TC = 4.25
sec), and the NEHRP value from Eq.(2.6) (TC = 7.1 sec) respectively. For
both the ECB and the Faccioli approach the structural period would exceed
the corner period, and hence the response displacement equals the peak
response displacement (assuming validity for the “equal-displacement”
approximation). Since the NEHRP equation for corner period (Eq.(2.6))
exceeds the estimated elastic period, the response displacement is
estimated from Eq.(2.7) using the elastic period of 6 seconds, instead of TC,
 2.2.2 Design Elastic Spectra

resulting in a displacement of 740mm (29.1in). The ltalian work in progress

would appear to suggest a value of about 640mm (25 in) corresponding to a
period of TC = 5.1 sec. It is felt that the correct answer may lie somewhere
between Eqs.(2.3) and (2.6). The yield displacement of the frame building
can be estimated at approximately 400mm (15.7in) using procedures
developed in Section 4.4. Thus the EC8 approach would predict that the
response would be elastic, with a maximum displacement of about 60% of
yield, while the other two approaches would predict inelastic response with
 2.2.2 Design Elastic Spectra

displacement ductility demands of about 1.3 and 1.8 respectively.

In this text we shall in some cases use code-specified displacement spectra,

and in others, generate displacement spectra based on Eqs.(2.3) and (2.5)
to illustrate the general procedure, while recognizing that the corner period
may well be revised upwards in the future. We will discount the current EC8
value of 2 seconds for the corner period as being unrealistically low and
unsafe.
 2.2.2 Design Elastic Spectra

It should be noted that the difference between current procedures affects the
design and response only of long-period structures, and also does not affect
the validity of the displacement-based seismic design approach developed in
the subsequent chapters of this text. It should also be noted that criticism of
displacement-based design approaches on the basis that they use the
lengthened effective period corresponding to peak displacement response,
where greater uncertainty on spectral ordinates exists as noted above, rather
 2.2.2 Design Elastic Spectra

than the initial period, where spectral ordinates may be better defined, is
invalid. Force-based seismic design relies on displacement-equivalence
rules relating the elastic displacement to the displacement of the effective
ductile structure (see Section 1.3.8). Since the effective degraded period of
the ductile structure is greater than the elastic period, the displacement-
equivalence rules are subject to exactly the same uncertainty as is present
in designs based on the effective stiffness.
 2.2.2 Design Elastic Spectra

(c) Design Return Periods:

It is customary to assign different return periods, or more correctly annual

probabilities of exceedence, for design to different limit states. For example,
Eurocode EC8[X3] defines three levels of design earthquake corresponding to
different limit states:
 2.2.2 Design Elastic Spectra

(i) Level 1 Seismicity – Serviceability limit state:

The level 1 design-level seismicity has a probability of exceedence of 50%

in 50 years approximately corresponding to an average return period of 72
years. Under an earthquake with this probability of exceedence, there
should be no damage requiring repair, and normal operations of the
structure should not be significantly affected.
 2.2.2 Design Elastic Spectra

(ii) Level 2 Seismicity – Damage control limit state:

The level 2 design-level seismicity has a probability of exceedence of 10%

in 50 years, approximately corresponding to an average return period of 475
years. Repairable damage is permitted under an earthquake with this
probability of exceedence, and normal operations may be suspended while
repairs are carried out.
 2.2.2 Design Elastic Spectra

(iii) Level 3 Seismicity – Life – safety Limit State:

The level 3 design – level seismicity has a probability of exceedance of 2%

in 50 years, approximately corresponding to an average return period of
2500 years. The structure should not collapse under ground shaking with
this probability of occurrence, but it may not be possible to economically
repair the structure following the earthquake.
 2.2.2 Design Elastic Spectra

It is common for only the first two limit states to be considered for structures
of other than special importance. In Europe the life-safety limit state is not
used for new designs, but is considered in assessment of existing
structures.

There are some problems with a probabilistic approach such as that defined
in EC8 and other similar definitions of seismicity, particularly when only the
first two limit states are used for design. In regions of high seismicity, such
an approach tends to produce rational increments of seismic intensity for the
 2.2.2 Design Elastic Spectra

different limit states. However in regions of lower seismicity the seismic

intensity corresponding to the first two limit states may be very low, as the
seismicity is defined by very rare, but very intense ground shaking. Thus the
intensity corresponding to the Life-safety earthquake may be much greater
than that corresponding to the damage control earthquake.

Consider a rather artificial case where seismic hazard is dominated by a

fault which fractures every 1000 years, on average inducing PGA’s at the
 2.2.2 Design Elastic Spectra

site being considered 0.8g. The remainder of seismic risk is attributed to

background seismicity capable of inducing PGA’s of only 0.2g.

A probabilistic analysis might then determine the 475 year risk as being
characterized by a PGA of 0.4g, since the probability of occurrence of the
design earthquake occurring within the 475 year period is approximately
50%. However, the earthquake either occurs during that period, or it doesn’t.
If the structure is designed for 0.4g and the earthquake occurs, the structure
 2.2.2 Design Elastic Spectra

fails, since the intensity is much greater than the design strength. lf the
earthquake doesn’t occur, the structure is over-designed.

In the USA a slightly different approach has been used in recent times[X4, X8].
A “maximum considered earthquake” (MCE) is defined with a return period
of 2500 years- essentially the same as the Level 3 seismicity defined above.
The structure is then designed for an intensity of O.67xMCE to the
performance criteria applicable to the damage-control limit state.
 2.2.2 Design Elastic Spectra

For regions of high seismicity there is little difference from the 10% in 50
years seismicity defined above for the Level 2 seismicity, but for other
regions, particularly those with infrequent large earthquakes, the difference
is large. The justification for the 2/3rds factor is that structures are expected
to have a 50% excess displacement capacity above that corresponding to
design to the damage-control limit state performance criteria.
 2.2.2 Design Elastic Spectra

In many existing codes the design acceleration spectra for Level 1 or Level
3 seismicity are found by multiplying the spectrum for the Level 2 seismicity
by specified multipliers. These can differ quite markedly between different
codes. For example, in Europe and New Zealand the multipliers for the
serviceability-level seismicity are 0.4 and 0.167 respectively. This approach
does not seem reasonable, as high annual probabilities of exceedence
(Level 1 seismicity) are dominated by earthquakes of smaller magnitude,
while low annual probabilities of exceedence (Level 3 seismicity) are
 2.2.2 Design Elastic Spectra

generally dominated by larger earthquakes with significantly different

spectra shapes. The consequences of this difference are examined in
Example 2.1 below.
 Example 2.1
Design Spectra for a Site Adjacent to an Active Fault

We consider a rather idealized situation where the seismicity of a site is

governed by proximity to an active fault. The three design levels defined in
the previous section are to be considered.

A probabilistic seismic hazard analysis indicates that for a probability of

exceedence of 50% in 50 years (serviceability level design), seismic hazard
is dominated by an earthquake of magnitude 6.15 at a distance of 10 km.
For Level 2 seismicity (10% probability in 50 years), seismic hazard is
dominated by a M7.0 earthquake at 10 km, and for the life-safety
 Example 2.1
Design Spectra for a Site Adjacent to an Active Fault

considerations (level 3 seismicity; 2% probability in 50 years), the seismic

hazard is dominated by an M7.5 earthquake which is the maximum
considered feasible for the fault segment adjacent to the site, at a distance
of 10 km. The ratio of the magnitudes for the earthquakes with probability of
10% in 50 years and 50% in 50 years corresponds roughly to the average
applicable to the entire circum-pacific belt. Although very large variations
from this ratio could occur at different sites we consider the average to be
appropriate for this site, for the purpose of example.
 Example 2.1:
Design Spectra for a Site Adjacent to an Active Fault

With the information above, the design displacement spectra can be

generated for the three limit states, using Eqs.(2.3 and 2.5). The results for
corner period and peak response displacement are summarized in Table
2.1, and the spectra are plotted in Fig.2.10(a), based on firm ground
conditions.

Acceleration spectra can be generated from the displacement spectra

inverting Eqs.(2.2) and (2.7), provided an assumption is made about the
period TB at the end of the response acceleration plateau (see Fig.2.5(a)).
Table 2.1 Design Parameters for Elastic Spectra for Example2.1

Level 1 EQ Level 2 EQ Level 3 EQ

Prob. Of exceedence 50% in 50 years 10% in 50 years 2% in 50 years

Magnitude, MW 6.15 7.0 7.5

Corner Period, TC (sec) 2.13 4.25 5.5

Corner Disp. δmax 89 631 1995

(mm)
Period TB (Fig.2.5) 0.3 0.5 0.7
(sec)
Peak Ground Acc. (xg) 0.22 0.47 0.83

Peak Spectral Acc. (xg) 0.55 1.18 2.08

Fig. 2.10 Design Spectra for Example 2.1
 Example 2.1
Design Spectra for a Site Adjacent to an Active Fault

Values assumed for TB increase with earthquake magnitude, and are listed
in Table 2.1. A further necessary assumption is that the response
acceleration plateau commences at a period of TA = 0.15 sec., and that the
ratio of peak spectral acceleration to PGA is CA = 2.5. On the basis of these
assumptions the peak ground accelerations are found to be 0.22g, 0.47g,
and 0.83g respectively, with peak response accelerations 2.5 times these
values, as listed in Table 2.1. The resulting acceleration spectra are plotted
in Fig.2.10(b).
 Example 2.1
Design Spectra for a Site Adjacent to an Active Fault

Values for the PGA and peak response acceleration do not seem
unreasonable. Note that the ratio between PGA for the serviceability and
damage-control limit states (0.47) is higher than would apply to Eurocode
EC8 [X3] or the NZ loadings code [X1].
The very significant differences in the displacement spectra for the three
limit states are immediately apparent, and appear more significant than the
differences in the acceleration spectra.
 Example 2.1
Design Spectra for a Site Adjacent to an Active Fault

This is a consequence of the influence of variations in the corner period.

With a peak elastic response displacement of only 89mm (3.5in) most frame
buildings of four or more storeys would respond elastically, regardless of
elastic period, as discussed in more detail in Chapter 5.

The spectra developed by the approach outlined in this example are rather
different from spectra resulting from simple scaling of a standard spectral
shape.
 2.2.3 Influence of Damping and Ductility on Spectral
Displacement Response

Displacement-based seismic design using a secant stiffness representation

of structural response requires a modification to the elastic displacement
response spectrum to account for ductile response. The influence of ductility
can be represented either by equivalent viscous damping or directly by
inelastic displacement spectra for different ductility levels. Figure 2.11 shows
both options for a firm-soil site, with PGA = 0.4g, and corner period of TC = 4
sec. The use of spectra modified by different levels of damping requires
relationships between ductility and damping to be developed for different
 2.2.3 Influence of Damping and Ductility on Spectral Displacement
Response

structural hysteretic characteristics as discussed in detail in Section 3.4.3,

but enables a single design spectrum to be used for all structures (e.g.
Fig.2.11(a)). The use of spectra modified by different levels of ductility (e.g.
Fig.2.11(b)) is perhaps more direct, but requires the ductility modifiers to be
determined for each hysteretic rule considered.

If a relationship between initial-period elastic displacement and inelastic

displacement such as the equal-displacement approximation is assumed,
the inelastic spectra of Fig.2.11(b) can be directly computed.
Fig. 2.11 Representation of Inelastic Response by Displacement Response
Spectra
 2.2.3 Influence of Damping and Ductility on Spectral Displacement
Response

The procedure is considered in detail in Section 3.4.3(e). Using this

procedure, and assuming validity of the equal displacement approximation
the data in Fig.2.11(b) have been calculated for an elasto-plastic response.
Different relationships apply for different post-yield stiffnesses, but not to
different hysteretic energy absorption within the loop, provided that the equal
displacement approximation is assumed to be valid. However, recent
research presented in Section 4.9.2(g) indicates that the equal-displacement
approximation is in fact non-conservarive.
 2.2.3 Influence of Damping and Ductility on Spectral Displacement
Response

The consequence is that inelastic spectra based on displacement ductility

(e.g. Fig.2.11 (b)) must be calibrated for each different hysteretic rule.

Because of these reasons, the equivalent viscous damping approach,

presented in Fig.2.11 (a) is preferred, since codified specification is simpler.
However, there appears to be still some uncertainty amongst seismologists
as to the appropriate form of the damping modifier Rξ to be applied to the
elastic displacement spectrum for different levels of damping ξ.
 2.2.3 Influence of Damping and Ductility on Spectral Displacement
Response

A commonly used expression was presented in the 1998 edition of

Eurocode EC8 [X3], and is shown below in Eq.(2.8):

 0.07 
R    (2.8)
 0.02   

Where α = 0.5. In the 2003 revision to EC8, Eq.(2.8) was replaced by

0.5
 0.10 
R    (2.9)
 0.05   
 2.2.3 Influence of Damping and Ductility on Spectral Displacement
Response

In both Eq.(2.8) and (29), ξ is expressed as a damping ratio related to the

critical damping. An alternative expression proposed by Newmark and
Hall[N2] in 1987 is defined by Eq.(2.10):

R  (1.31  0.19 ln(100 )) (2.10)

The three expressions are compared in Fig.2.12 for different levels of

damping.
Fig. 2.12 Damping Modifiers to Elastic Spectral Displacement
 2.2.3 Influence of Damping and Ductility on Spectral Displacement
Response

It will be seen that the expression given by Eq.(2.10) appears very

conservative compared with the other two expressions. Our analyses using
both spectral compatible earthquake records and actual accelerograms
supports the 1998 EC8 expression for accelerograms without near-field
forward directivity velocity pulse characteristics. A comparison is shown in
Fig.2.13 for the average response of a suite of seven spectrum-compatible
records used for design in the Port of Los Angeles[E1], California. The design
elastic displacement spectrum was determined from a site-specific
Fig. 2.13 Comparison of Spectral Analysis Results and Eq.(2.8) for a Suite
of 14 Spectrum-Compatible Accelerograms
 2.2.3 Influence of Damping and Ductility on Spectral
Displacement Response
probabilistic seismic hazard analysis, and seven two-component
accelerograms from earthquake records appropriate to the site, distance from
fault and expected magnitude were manipulated to match the spectrum for a
clamping ratio of ξ = 0.05. The fourteen individual modified accelerograms
were then subjected to spectral analysis to determine the effect of different
levels of damping. The averages for the records are shown in Fig.2.11 by the
solid lines. The dashed lines represent the spectra resulting from modifying the
design spectrum for the damping ratio of ξ = 0.05 in accordance with Eq.(2.8),
 2.2.3 Influence of Damping and Ductility on Spectral
Displacement Response

but with α = 0.5. It will be seen that the agreement with the curves from
spectral analysis is very good, on average. Similar agreement has been
obtained from analyses of real records[K8].
It would also be desirable to have an equivalent expression for sites where
forward directivity velocity pulse characteristics might be expected. It has
been suggested[P3] based on limited data, that a modification to the 1998
EC8 expression given by
0.25
 0.07 
R    (2.11)
 0.02   
 2.2.3 Influence of Damping and Ductility on Spectral Displacement
Response

might be appropriate. lt will be seen that Eq.(2.11) is the same as Eq.(2.8)

but with α = 0.25 instead of 0.50. The effect of this modification is to
increase the value of Rξ compared with the value applying for “normal”
accelerograms. The reduction factor resulting from Eq.(2.11) is included in
Fig.2.12 for comparison with the other expressions, and Fig.2.14 compares
the dimensionless displacement modifiers for the 1998 EC8 expression
(Eq.2.8) and the expression suggested for near-field forward directivity
conditions (Eq.(2.11)). The data in Fig.2.14 are based on the shape of the
Fig. 2.14 EC8 (1998) Damping Reduction factor, and Tentative Factor for
Forward Directivity Effects
 2.2.3 Influence of Damping and Ductility on Spectral
Displacement Response

displacement spectrum for firm ground plotted in Fig. 2.5(a). Some qualified
support for Eq.(2.11) is available in work by Bommer and Mendis[B8] who
provide additional discussion of this topic. Their work indicates that the
scaling factors may be period- dependent, which is not currently considered
in design.

It will be shown in Section 3.4.6 that use of displacement spectra for near-
field forward-directivity effects results in a requirement for higher base-shear
strength when compared to requirements for “normal” conditions. This
 2.2.3 Influence of Damping and Ductility on Spectral
Displacement Response

requirement has been recognized for a number of years, in particular since

the 1994 Northridge earthquake, and is incorporated empirically in recent
force based codes [e.g. X4, X8]. This is an example of conditions where it
has been recognized that existing displacement-equivalence rules are
inadequate in force-based design. With displacement-based design, the
influence of near-field effects are directly incorporated in the design,
provided the reduced influence of damping (and ductility) in modifying
displacement response is recognized by graphs such as Fig.2.14(b).
 2.2.3 Influence of Damping and Ductility on Spectral
Displacement Response

As discussed earlier in this chapter, displacement spectra for design are in a

developmental stage, with rather rapid progress being made by
seismologists. The same statement can be applied to developments in
definition of acceleration spectra. lt is probable that displacement spectra,
and modifiers for damping, period and ductility will change from the tentative
values suggested in this Chapter. However, this should not be taken to
reduce the utility of the work presented in subsequent chapters. The
approaches developed are independent of the displacement spectra, and
 2.2.3 Influence of Damping and Ductility on Spectral
Displacement Response

several different shapes will be used in different examples, to illustrate the

flexibility of the direct displacement-based design method.
It should also be noted that the material presented in this chapter is at least
as valid as acceleration spectra currently used for force-based design (see
p53). Taken together with the extensive research described in Chapters 3
and 4 on relationships between ductility and damping, the procedures
developed are significantly less susceptible to errors than resulting from
current force-based design, and much better adapted to achieving specified
limit states.
 2.3 CHOICE OF ACCELEROGRAMS FOR TIME-HISTORY
ANALYSIS

The most reliable method at present for determining, or verifying the

response of a designed structure to the design level of intensity is by use of
nonlinear time-history analysis. The selection and characteristics of
accelerograms to be used for this requires careful consideration. The reader
is referred to Section 4.9.2(h) where this is discussed in some detail.