Appendix G: Elastic and

Inelastic Response Spectra

In the seismic design of structures, the maximum structural response subjected to
a design earthquake is of interest to practicing engineers. The maximum structural
response could be the maximum relative displacement with respect to the ground
motion displacement or the maximum absolute acceleration (i.e., inertia force) with
respect to the ground at its rest condition prior to the earthquake.
As shown in Figure G.1, for a single-degree-of-freedom (sdof) structure subjected
to earthquake excitation, the motion equation, based on force equilibrium (Clough
and Penzien, 1975), is

mx t (t) + cx(t) + kx(t) = 0 (G.1)

where
m, c, and k are mass, damping coefficient, and stiffness, respectively
ẋ and x are the relative velocity and relative displacement, respectively
superscript t represents the total displacement

The total displacement xt is

x t (t) = x(t) + xG (t) (G.2)

xG (t) is the ground motion displacement. Similar to Equation F.1, Equation G.1 can
also be expressed in terms of relative displacement, given as

mx(t) + cx(t) + kx(t) = −mxG (t) (G.3)

in which ẍG is the earthquake acceleration record expressed in terms of gravity, G.

Dividing Equation G.3 by the mass, m, leads to

x(t) + 2ρωx(t) + ω2 x(t) = −xG (t) (G.4)

where ρ is the damping ratio (or so-called damping factor) expressed as follows:

c c c
ρ= = = (G.5)
ccr 2 km 2mω

335

© 2012 by Taylor & Francis Group, LLC

 336 Appendix G

xt
m

k, c

xG x

FIg u r e g .1 sdof system subjected to ground motion.

where
ccr is the critical damping of the system
ω is the angular frequency expressed as follows:

k
ω= (rad/s) (G.6)
m

from which the structural natural frequency and the period can be calculated
by using

ω (G.7)
f = (cycle/s)
2π

and

2π
T= (s/cycle) (G.8)
ω

g .1 el ASt Ic r eSPo n Se SPec t r u m

A displacement response spectrum represents the maximum relative displacements
of sdof oscillators with different periods (or frequencies) of vibration correspond-
ing to a specified elastic damping ratio (typically 5%), subjected to ground motion,
ẍ G (t). The elastic displacement spectrum can be calculated using the step-by-step
numerical integration method described in Appendix F to solve either Equation G.3
or G.4. Since the abscissas of the spectrum represent the structural periods of vibra-
tion, in practice, it is convenient to use Equation G.4 to generate the elastic response
spectrum. The maximum displacement corresponding to each frequency, ωi, can be
obtained as follows:

Rd (ωi ,ρ) = max x(t, ωi ,ρ) ; i = the ith frequency or period (G.9)

© 2012 by Taylor & Francis Group, LLC

 Appendix G 337

To generate the acceleration spectrum, the equation of motion in terms of total

displacement should be used. By substituting Equation G.2 into Equation G.1, the
equation of motion becomes

mx t (t) + cx t (t) + kx t (t) = cxG (t) + kxG (t) (G.10)

From this, the absolute acceleration spectrum can be calculated from Equation G.10
by the step-by-step numerical integration method as follows:

Ra (ωi ,ρ) = max x t (t, ωi ,ρ) ; i = the ith frequency or period (G.11)

From Equations G.1 and G.5, the total acceleration can be expressed as follows:

x t (t) = −2ρωx(t) − ω2 x(t) (G.12)

In practice, the damping terms in Equation G.12 can be neglected considering

the damping force contribution to the equilibrium condition is small. Therefore,
Equation G.12 can be simplified to

x t (t) = −ω2 x(t) (G.13)

The approximate calculation of the absolute acceleration spectrum can be

formu- lated based on Equation G.13:

Ra (ωi ,ρ) = ω2 max x(t,ωi ,ρ) = ω2 Rd (G.14a)

Equation G.14a is called the pseudo-acceleration spectrum. Similarly, the pseudo-

velocity spectrum is given as follows:

Rv (ωi ,ρ) = ωRd (G.14b)

Typical elastic spectra for the 1940 El Centro earthquake N–S component with 5%
damping factor are shown in Figure G.2. The elastic displacement, velocity, and
acceleration spectra were generated by the linear acceleration numerical
integration method using Equation G.4, and they are in good agreement with
those shown in other references (Naeim, 1989).
The response spectra generated from a specified earthquake such as those in Figure
G.2 cannot be used for design, because the response of a structure due to this earth-
quake will be different from that due to another earthquake with similar magnitude,
and the local peaks and valleys are specific to the earthquake record and may not
rep- resent general peak responses. For this reason, in practical applications, the
response spectra from many earthquake records with common characteristics are
averaged to develop the design spectrum with a smooth curve or several straight
lines. Since the peak ground acceleration (PGA), velocity, and displacement for
various earthquake records differ, the computed response spectra from these
records cannot be averaged

© 2012 by Taylor & Francis Group, LLC

 338 Appendix G

1.0
Spectral displacement (ft)
3.0

Spectral velocity (ft/s)

0.8 2.5
0.6 2.0
1.5
0.4
1.0
0.2
0.5
0.0 0.0
0 1 2 3 4 5 0 1 2 3 4 5
Period (s) Period (s)

30
Spectral acceleration

25
20
(ft/s/s)

15
10
5
0
0 1 2 3 4 5
Period (s)

FIg u r e g .2 Elastic response spectra (5% damping).

on an absolute basis. Therefore, various procedures are used to normalize response

spectra before the averaging process is carried out. The most common normalization
procedure is to normalize each spectrum to the corresponding peak ground motion.
For example, a normalized design spectrum for a certain soil type in the AASHTO
bridge design specifications (1992–2008) is shown in Figure 1.4. This spectrum is
the average of many real earthquake spectra, developed by dividing their spectral
ordinates by the corresponding PGAs. For bridge design, the design acceleration
spectrum value, Ra in Equation G.14a, is equal to gCsm in which g is the gravitational
acceleration and Csm is the elastic seismic coefficient are shown in Equation 1.8.

g .2 In el ASt Ic r eSPo n Se SPec t r u m

Structures subjected to severe earthquake ground motion experience deformation
beyond the elastic range. The inelastic deformations depend on the hysteretic behav-
ior (i.e., load–deformation characteristics) of the structures.
Similar to the elastic response spectrum, the inelastic response spectrum can be
generated by the inelastic time history analysis described in Appendix F. An inelas-
tic displacement response spectrum represents the maximum relative displacements
of sdof oscillators with different periods (or frequencies) of vibration corresponding
to a specified ductility level, subjected to ground motion, ẍG (t). To illustrate how
to generate the inelastic response spectrum, a simple elastoplastic hysteresis model
shown in Figure G.3 is used here.
In the figure, xy and xm represent the yield displacement and the maximum dis-
placement of an sdof structure subjected to ground motion, ẍ G (t). The ductility is
expressed as follows:

© 2012 by Taylor & Francis Group, LLC

 Appendix G 339

Fe

Fy

Xy Xe Xm

FIg u r e g .3 Elastoplastic model for sdof system.

µ = xm (G.15)
xy

The step-by-step procedure below illustrates how to generate an inelastic spec-

trum with a target ductility level of μtarget and an elastic damping ratio ρ.
Step 1: Define a range of structural periods (T1 < T ≤ Tn), the incremental period ΔT,
and an assumed structural mass m.
Step 2: Select the ith period Ti, Ti = Ti−1 + ΔT; i = 1, n; T0 = 0 s.
Step 3: Perform an elastic time history analysis for Ti to find the corresponding elas-
tic strength demand Fe.
Step 4: Assume the structural yield strength, Fy, as a fraction of Fe (i.e., Fy = ratio × Fe).
The ratio is increased from 0.001 to 1. For each ratio increment, calculate the yield
displacement xy = Fy/k, in which k = ωi2 m and ωi = 2π/Ti, per Equations G.6 and G.8,
respectively.
Step 5: Using Fy, xy, and k perform a nonlinear time history analysis using the damp-
ing coefficient of c = 2ρ km in the following equation:

mx(t) + cx(t) + kx(t) = −mxG (t) (G.16)

At each incremental time step, check the structural demand force F(t). If F(t) < Fy,
the elastic condition controls; if F(t) ≥ Fy, the post-yield condition controls. Adjust
the unbalanced force so that F(t) = Fy. Since the elastoplastic model is used for this
example, use k = 0 and ωi = k /m = 0 for the next incremental time step. It is noted
that, for a structure with the hysteresis model other than the elastoplastic model,
the tangent post-yield stiffness, k, should be used in accordance with the hysteresis
model, and F(t) adjusted accordingly. From the time history analysis, find the maxi-
mum displacement, xm = max|x(t)|.

© 2012 by Taylor & Francis Group, LLC

 340 Appendix G

1.0 3.0
Spectral displacement (ft)

Spectral velocity (ft/s)

0.8 2.5

0.6 2.0
1.5
0.4
1.0
0.2
0.5
0.0 0.0
0 1 2 3 4 5 0 1 2 3 4 5
Period (s) Period (s)

35
Spectral acceleration

30
25
20
(ft/s/s)

15
10
5
0
0 1 2 3 4 5
Period (s)

FIg u r e g .4 Inelastic response spectra (μ = 4, ρ= 0.05).

Step 6: Calculate ductility μ = |xm /xy |. Compare μ with μtarget. If |μ − μtarget| is more
than a specified tolerance, increase values of ratio and go to step 4. If |μ − μtarget|
is less than a specified tolerance, xm = max|x(t)| corresponding to μtarget is obtained.
Similarly, the absolute maximum acceleration is ẍ m = max|ẍ t(t)|, where ẍ t(t) is calcu-
lated from Equation G.1. Per Equation G.13, for simplicity, the pseudoacceleration
of ẍ m = (Ti/2π)2|xy | can be used instead of using absolute maximum acceleration and
pseudovelocity as given as ẋm = Ti/2π|xy|.
Step 7: If Ti = Tn, the inelastic displacement spectrum is completed. If Ti < Tn, go to
step 2 and select the next period Ti, and repeat steps 3–6.
Inelastic spectra for the 1940 El Centro earthquake N–S component generated
using ductility μ = 4 are shown in Figure G.4. These spectra were generated by the
Wilson-θ method using Equation G.3. The mass and the initial damping ratio are 20
kip-s2/in. and 0.05, respectively.

g .3 Fo r ce-r ed u c t Io n R-FAc t o r
SPec t r u m
As described in Section 1.2.3, the force-reduction factor R is the ratio of the elastic
strength demand to the inelastic strength demand of a structure subjected to the
ground motion, ẍ G (t). Therefore, the force-reduction R-factor spectrum represents
the ratio of the elastic strength demand to the inelastic (or yield) strength demand,
corresponding to a specified ductility demand, within a range of periods of vibration.
By performing the elastic and inelastic response spectrum analyses described in the
precious sections, both elastic and inelastic strength demands can be obtained (i.e., Fe
and Fy). A typical Fe(μ= 1) and Fy(μ = 4) spectrum for the 1940 El Centro earthquake

© 2012 by Taylor & Francis Group, LLC

 Appendix G 341

N–S component is shown in Figure G.5, and the corresponding force-reduction

R-factor spectrum (i.e., Fe/Fy) is shown in Figure G.6.
Figure G.6 clearly shows that the R-factor is a function of structural type and
period. The R-factor is less than the ductility demand μ = 4 in the short period
range (T < 0.5 s), while between 0.5 < T < 5, the R-factor varies significantly. This
observation is only based on one earthquake acceleration record. However, simi-
lar to the design response spectrum, a design force-reduction R-factor spectrum
should be generated based on a large number of ground acceleration time his-
tory records, soil conditions at site, initial damping, and the hysteretic behavior
of structures.

550
500
450
400
µ= 1
Yield strength (kip)

350
300
250
200
150
µ= 4
100
50
0
0 1 2 3 4 5
Period (s)

FIg u r e g .5 Elastic strength demand (μ = 1) and inelastic strength demand (μ= 4) spectra
for the 1940 El Centro earthquake N–S component (initial damping ρ= 0.05).

8
7
6
5
factor

4
R-

3
2
1
0
0 1 2 3 4 5
Period (s)

FIg u r e g .6 Force-reduction R-factor spectrum for the 1940 El Centro earthquake N–S
component (initial damping ρ = 0.05).

© 2012 by Taylor & Francis Group, LLC

 342 Appendix G

g .4 el ASt Ic d ISPl Acemen t SPec t r u m w It H

eQu IVAl en t VISco u Sd AmPIn g Fo r
d d Bd
As described in Chapter 2, the calculation of the inelastic displacement demand is
essential to the performance-based seismic design of highway bridges. In the direct
displacement-based design (DDBD) procedures, the inelastic displacement demand
and the corresponding equivalent viscous damping, ξeq, are calculated, so that the
effective period, Teff, of the substitute sdof system can be obtained from the elastic
design displacement spectrum. A typical elastic design displacement spectrum is
shown in Figure 2.2.
The concept of equivalent viscous damping was first proposed by Jacobsen (1930).
He developed the equivalent viscous damping coefficient, ceq, of a linear system,
which can be used to estimate the maximum nonlinear displacement of a nonlinear
system with high power of velocity of motion, subjected to harmonic motion (i.e.,
replace the equation of motion of a nonlinear system, mẍ + cn(˙)x n + kx = Q sin ωt,
with the linear system, mẍ + ceqẋ + kx = Q sin ωt). By equating the work dissipated
by the nonlinear system to the work dissipated by the linear system, the equivalent
viscous damping coefficient, ceq, can be obtained.
Jacobsen’s concept was adopted by many researchers to obtain the equivalent vis-
cous damping, ξeq, of an elastic substitute system to estimate the peak displacement
response of an inelastic hysteretic system. For example, equating the energy dis-
sipated in one cycle by an sdof bilinear hysteretic system (see Figure G.7) under the
steady-state harmonic motion between the positive and negative maximum displace-
ments to the viscous damping energy dissipated by the associated elastic substitute
system undergoing the same displacements, the equivalent viscous damping, ξeq, can
be expressed as (ATC-40, 1996)
2(µ −1)(1 − r)
ξ eq = (G.17)
πµ(1 + rµ − r)

Elastic substitute system

Fy rKi

Keff
Ki
– Xm
Xy +Xm
r –r+1
Keff = ki

Teff = Ti
r –r+1

FIg u r e g .7 Idealized equivalent viscous damping (bilinear sdof system).

© 2012 by Taylor & Francis Group, LLC

 Appendix G 343

in which r is the postyield stiffness ratio as shown in Figure G.7. It can be seen that
the equivalent viscous damping, ξeq, is strongly dependent on the ductility demand,
μ = xm /xy, of the inelastic system. Since the actual earthquake motion is not a steady-
state harmonic motion, the magnitudes of many small hysteresis loops due to earth-
quake are considerately lower than the maximum displacement, xm. Many modified
ξeq − μ models have been developed (Gulkan and Sozen, 1974; Iwan and Gates, 1979;
ATC-40, 1996; Kwan and Billington, 2003; Dwairi et al., 2007; Priestley et al.,
2007), based on a large number of ground acceleration time history records. From
the ξeq − μ relationship, an inelastic displacement spectrum such as the one in Figure
4. can be replaced by an elastic displacement spectrum with equivalent ξeq.
To demonstrate that the elastic displacement spectrum corresponding to ξeq can
be used to estimate the maximum displacement of an sdof inelastic system, the fol-
lowing three ξeq − μ models are used to generate elastic displacement spectra repre-
senting the maximum inelastic displacement of an elastoplastic hysteretic system
using the 1940 El Centro earthquake N–S component. These spectra will then be
compared with the inelastic displacement spectrum in Figure G.4.

1. Model No. 1 (Dwairi et al., 2007):

 
 eq   elas   hyst  5  C   1 % (G.18)
  

in which C = 85 + 60(1 − Teff) if Teff < 1 s, and C = 85 if Teff ≥ 1 s. The equiva-
lent viscous damping is the sum of elastic damping, ξelas, and hysteretic
damping, ξhyst · ξelas = 5% is typically used here for concrete structure. Teff is
the effective period of the substitute elastic sdof system.
2. Model No. 2 (Priestley et al., 2007):

 
 eq   elas   hyst  5  67   1 % (G.19)
  

3. Model No. 3 (ATC-40, 1996):

   1  
 eq   elas   hyst  5    200   % (G.20)
    


κ = 1.13 − 0.51(μ− 1/μ), if ξhyst > 16.25; κ = 1.0, if ξhyst ≤16.25.

Substituting μ = 4 into Equations G.18, ξeq = 25.3% for Teff ≥1 s, and

38.2% ≤ξeq ≤26.7% for 0.1 s ≤Teff ≤0.9 s. Similarly, substituting μ = 4 into
(G.19) and (G.20) leads to ξeq = 21% and 40.7%, respectively. The elastic
displacement spectra corresponding to the above calculated ξeq s, generated
from the elastic time history analysis (see Appendix F for elastic time his-
tory analysis), are shown in Figure G.8.

© 2012 by Taylor & Francis Group, LLC

 344 Appendix G

1.0

Spectral displacement (ft) 0.8

0.6
Model 2
Model 1
0.4
Model 3
0.2

0.0
0 2 4 6 8 10
Effective period, Teff (s)

FIg u r e g .8 Elastic displacement spectra with equivalent viscous dampings.

Since these spectra represent the maximum responses of a substitute elastic sdof
system, the period on the abscissa of Figure G.8 is the effective period, Teff, not the
initial elastic period, Ti. In order to compare Figure G.8 with the inelastic displace-
ment spectrum in Figure G.4, which has the elastic period, Ti, on the abscissa, the Teff
in Figure G.8 needs to be shifted to Ti. From Figure G.7 for the elastoplastic system
(i.e., r = 0)

Teff Teff
Ti = = (G.21)
µ=4 2

Figure G.9 shows the comparison between the inelastic displacement spectrum and
the elastic displacement spectra with equivalent viscous dampings calculated from
the ξeq − μ models after shifting the period per Equation G.21. It can be seen that

1.0

0.8
Spectral acceleration (ft/s/s)

µ= 4
0.6
Model 2
Model 1
0.4
Model 3
0.2

0.0
0 1 2 3 4 5
Period (s)

FIg u r e g .9 Elastic and inelastic displacement spectra.

© 2012 by Taylor & Francis Group, LLC

 Appendix G 345

Models 1 and 2 provide a good estimate of inelastic response when Ti is less than
1.5 s. For Ti > 1.5 s, the estimates of all of the models are less than the inelastic dis-
placement values from the inelastic displacement spectrum. As shown in the figure,
Model No. 3 significantly underestimates the inelastic response due to overestimat-
ing the equivalent viscous damping. Note that models 1 and 2 were developed based
on results from extensive time history analysis, using many ground motion records.
Based on just one ground motion record, comparing the elastic displacement spec-
tra with ξeq s with an inelastic displacement spectrum is not adequate. However,
the main purpose of the above comparison is to demonstrate that the elastic dis-
placement spectrum with appropriate ξeq can be used to estimate maximum inelastic
displacement.
Instead of using elastic displacement design spectra with associated equivalent
viscous damping, it is possible to develop inelastic displacement design spectral
curves in terms of ductility demands for the DDBD. However, the disadvantage of
using inelastic displacement design spectra is that the equivalent ductility demand
of the substitute sdof system is not easy to obtain, due to the difficulty of estimating
the equivalent yield displacement of the substitute sdof system, especially for bridges
with non-regular geometry or nonuniform distribution of weight and stiffness.

© 2012 by Taylor & Francis Group, LLC