EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS. VOL.

23, 507-521 (1994)

CONSISTENT INELASTIC DESIGN SPECTRA: STRENGTH AND

DISPLACEMENT

TOMAZ VIDIC*, PETER FAJFAR’ AND MATEJ FISCHINGER~

Department of Civil Engineering, University of Ljubljana. famova 2, Ljubljana. Slovenia

SUMMARY
A procedure for the determination of inelastic design spectra (for strength, displacement, hysteretic and input energy) for
systems with a prescribed ductility factor has been developed. All the spectra are consistent (interrelated and based on the
same assumptions). This is the first of two companion papers which deals with the ‘classical’ structural parameters:
strength and [Link] input data are the characteristics of the expected ground motion in terms of a smooth
elastic pseudo-accelerationspectrum. Simple, approximate expressionsfor the strength reduction factor R are proposed.
The value of R depends on the natural period of the system, the prescribed ductility factor, the hysteretic behaviour,
damping and ground motion. Fairly accurate approximations to the inelastic spectra for strength and displacement can
be derived from the elastic spectrum using the proposed values for R.

INTRODUCTION
The design spectrum approach is the most commonly used method of analysis in earthquake engineering.
Although theoretically limited to the linear elastic behaviour of structures and single-degree-of-freedom
(SDOF) inelastic systems, it can, in many cases, also be applied approximately to the analysis of the inelastic
behaviour of buildings. Elastic design spectra are often given in aseismic codes or, in the case of important
structures, are determined by a seismic hazard analysis at the site. Inelastic spectra depend not only on the
characteristics of the expected ground motion at the site, but also on the non-linear characteristics of the
structural system, which severely complicates the problem.
An inelastic design spectrum is usually synonymous with the strength spectrum which basically represents
a reduced elastic pseudo-acceleration spectrum. The degree of reduction is based mainly on experience and
judgement, and depends primarily on the provided or tolerated ductility and on the overstrength. For
a rational design procedure, however, data on other parameters, in addition to required strength, are needed.
Maximum relative displacement is the structural response parameter most widely used for evaluating the
inelastic performance of structures, at both the serviceability and the failure limit state. Recently, the
importance of cumulative damage has been recognized. Input energy spectra are related to the cumulative
damage potential of ground motions, and dissipated hysteretic energy is the parameter which can be
correlated to cumulative damage. Consequently, according to the opinion of the authors, strength, displace-
ment, input and hysteretic energy spectra represent the basic spectra which define seismic demand in
earthquake-resistant design. This paper is restricted to strength and displacement spectra. Energy spectra are
discussed in a companion paper.’ All these spectra are consistent, i.e. interrelated and based on the same
assumptions.
Inelastic strength spectra have been one of the research preoccupations of several researchers for the last 20
years or so. Two main different approaches have been investigated: (a) reduction of the elastic spectra
through the use of reduction factors, and (b) direct derivation of inelastic spectra through statistical studies

*Research Engineer.
‘Professor.
+AssociateProfessor.

CCC 0098-8847/94/040507-15 Received 5 November 1992

0 1994 by John Wiley & Sons, Ltd. Revised 30 April 1993
508 T. VIDIC, P. FAJFAR AND M. FISCHINGER

of the spectra obtained by the non-linear dynamic analysis of structures subjected to available ground
motions. The second approach yields more accurate results, provided that the site-specific ground motion
can be predicted with reasonable accuracy. According to Bertero et a[.' this approach can be considered as
the method of the future. For the time being, however, the approach (a) has attracted more attention. Among
the methods based on approach (a), the most popular is perhaps the Newmark and Hall procedure,3 which
was originally proposed in the early 1970s. Several proposals, which were mainly intended to provide
improved reduction factors, followed. (The reduction factor is the ratio of elastic strength demand to inelastic
strength demand for a specified ductility ratio.) References 4-1 S represent a selected list of publications.
A critical review of the early proposals can be found in Reference 16. Different variants of the approach (a)
have been adopted in several national codes.
Much less attention has been paid to displacement spectra. Usually it is assumed that, in the medium- and
long-period region, the inelastic displacement spectrum is approximately the same as the elastic displacement
spectrum. This assumption, like many other simplifications in earthquake-resistant design, should be applied
to the mean values corresponding to a group of records. Large deviations can be observed in the case of some
particular records, as is shown in Reference 2. Realistic approximations are more difficult to find in the
short-period region, where the displacement spectrum is highly influenced by the characteristics of the
non-linear behaviour of the structure. Theoretically, the inelastic displacement spectrum can be obtained
from the inelastic strength spectrum [using equations (6) and (3) or (S)]. In the short-period region, however,
the inaccuracies in the strength spectrum are increased, which usually results in an unrealistic displacement
spectrum. Very few authors have explicitly proposed general-purpose displacement spectra or general
procedures for the determination of inelastic displacements (References 6 and 17-20).
In the development of the procedure presented in this paper two novel features were used:

Several interrelated spectra (reduction factors, strength, displacement, spectra related to energy) were
investigated simultaneously. Approximate formulae for the reduction factor R (and for two other
non-dimensional parameters related to energy, described in the companion paper') were supposed to
yield results of reasonable accuracy not only for the R-factors but for all the spectra under considera-
tion.
Different groups of records representing ground motion of basically differing types were used as the
lnput motion.

result of the study, in this paper simple bilinear expressions are proposed for the R-factors. The
proposed bilinear R-spectra take into account the frequency content of the ground motion. Periods at which
the R-factors become constant are a function of the characteristic period of the ground motion and of
ductility. The maximum R depends not only on ductility but also on the hysteretic behaviour and damping
and is not necessarily equal to the ductility factor. Previously proposed bilinear R-spectra (References $ 6 . 1 1,
14, 15) took into account only some of these features, or none of them.
Approximate inelastic spectra for strength and displacement can be derived from elastic pseudo-acceler-
ation spectrum using the proposed R-spectra. The procedure was validated by a comparison of approximate
spectra with the 'exact' spectra obtained by non-linear dynamic time-history analysis.
Recent ~ t u d i e s ~ 'have
- ~ ~ demonstrated that the application of R-factors in the case of extremely
narrow-band ground motions, like those recorded on very soft soil deposits, needs special consideration. The
proposed procedure is not intended to be directly applicable to such ground motions.

BASIC RELATIONS
The structural system will be simulated by an equivalent SDOF model with a bilinear force-deformation
envelope. The following relations apply:
F, = rnw2D, (11
 CONSISTENT INELASTIC DESIGN SPECTRA 509

where F, is the strength of the system (yield strength), m its mass, w the natural frequency, D, the yield
displacement, D the maximum displacement and p the corresponding ductility factor.
A structure with a given strength F,, which relies on energy dissipation through inelastic deformation,
should have a limit deformation capacity which exceeds seismic demand in the case of severe earthquakes. It
is well known that seismic demand in terms of ductility [which is related to deformation according to
equation (2)] and seismic demand in terms of the strength F, are interrelated. The problem can therefore be
stated in a different way: assuming that a certain deformation capacity is provided (or, in the case of
a serviceability limit state, a certain deformation is tolerated), the strength F , of the system should at least be
equal to the required strength. This approach is actually used in design procedures and can be written in the
form

where A, is the value in the elastic pseudo-acceleration spectrum and R is the strength reduction factor,
which is equal to the elastic strength demand mA, divided by the inelastic strength demand F,. R depends
mainly, but not exclusively, on the assumed ductility and on the period of the system. Expressions similar to
equation (3) can be found in various seismic codes. However, an important difference should be noted
between equation (3) and the expressions in the codes. In equation (3), F , represents the actual strength,
whereas the seismic forces in the codes correspond to code strength which is, as a rule, lower than the actual
strength. This difference is due to different reasons, one of them being overstrength, which is an inherent
property of properly designed, detailed, constructed and maintained highly redundant structures.’. 24
In the paper the non-dimensional strength q will be used, which is defined as
q = - FY
(4)
mag
where ag is the peak ground acceleration. Using equation (3), equation (4) can be rewritten in the form

The maximum relative displacement of the system, D,can be determined from equations (1)-(3):

Equations (1)-(6)represent the relations between the main ground motion, the structural parameters and the
response parameters (with the exception of energies). In the procedure presented in this paper, the elastic
pseudo-acceleration spectrum A, and the ductility factor p are assumed to be prescribed. Starting from these
values, and using the given formulae for the reduction factor R, the spectra for the required strength F , and
for the maximum relative displacement D can easily be determined.

ELASTIC SPECTRA
In this paper simple formulae for the determination of approximate R spectra are proposed. These formulae
can be applied for the calculation of F, and D in combination with any smooth elastic spectrum (with the
exception of extremely narrow-band spectra). In order to validate the proposed formulae, a special type of
elastic spectrum has been used. The ‘exact’elastic spectra corresponding to different records were simplified
in accordance with the Newmark-Hall representation of ~ p e c t r a Constant
.~ values of the pseudo-acceler-
ation A,, the pseudo-velocity V, and the displacement D, were assumed in the short-, medium- and
long-period range of the spectrum, respectively:
A , = c,,a,, 04T1< T < TI (7)
V, = c,u~, TI < T < TZ (8)
D, = Cddg, TZ < T (9)
5 10 T. VIDIC, P. FAJFAR AND M. FISCHINGER

Table I. Amplification factors for elastic

spectra with 5 per cent damping (the
standard group represents the combined
group of U.S.A.--California and Mon-
tenegro records)

Group c, cti cd

Standard 2.5 2.0 1.8

[J.S.A. 2.5 1.8 1-6
Montenegro 2.5 2.2 2.0
Friuli 2.5 2.0 1.8
Banja Luka 2.5 1.6 2.2
Chile 2.5 26 2.0

where ug,ug and d , are the peak ground motion parameters (acceleration, velocity and displacement) and
are the corresponding amplification factors. For very short periods ( T < 0-03 s) the spectral
car c,,, c d
acceleration is equal to ug. A linear increase of A , was assumed in the period range from T = 0 0 3 s to
T = 04T1.
The periods T I and T2 can be obtained directly from the spectrum, and this feature makes the New-
markpHall types of elastic design spectra especially convenient for the proposed procedure. The following
relations apply:

Cddg
T, = 271- (1 1,
I "g

In the reported study the amplification factors c,, c , and cd were determined by a trial-and-error procedure
for all the investigated groups of records (Table I). The condition of approximately equal areas under the
exact and simplified spectrum was applied. The resulting amplification factors were rounded off. New-
mark--Hall-type spectra were computed for each record in each group separately by applying the amplifica-
tion factors corresponding to that group.
In Figure 1, the mean spectrum determined from the approximate spectra is compared with the mean
spectrum computed from the exact spectra for each group. It should be noted that, in the subsequent
computations discussed in this paper, approximate elastic design spectra of the Newmark-Hall type for each
record. rather than mean spectra, were used.

PARAMETRIC STUDIES
Over the last ten years or so extensive parametric studies of SDOF systems have been carried out at the
University of Ljubljana. In these studies, input ground motion, as well as the initial stiffness (expressed by the
natural period or frequency), strength, ductility, hysteretic behaviour and damping of SDOF structural
systems, was varied. Seismic demand, expressed in terms of the maximum relative displacement D, nor-
malized strength q, reduction factor R, and various energy parameters and ductility factors, was studied. The
results, obtained at different stages of the study, have been presented in several publications (e.g. References
25- 27). In these publications, as well as in the Ph.D. thesis of the first author,28a more detailed description of
the parametric studies can be found. Here only the most important data and some of the most recent results,
which have been used for the validation of the proposed procedure, arc given.
The standard records from California and the records from Montenegro, 1979 (20 records altogether), were
chosen as being representative for 'standard' ground motion (i.e. severe ground motion at moderate
epicentral distances, with a duration of strong ground motion in the range of 10 30 s; predominant periods
 CONSISTENT INELASTIC DESIGN SPECTRA 511

-
____ STANDARD.
STANDARD,
'EXACT"
APPR. ____ U.S.A. +
U.S. A . ,
'EXACT'
APPR.
- STANDARD, DESIGN

[Link] 0.1 1 10 0.01 0.1 I 10

PERIOD (sl PERIOD [sl

____ MONTENEGRO. " E X A C T '

MONTENEGRO, APPR.
-
____ CHILE.
CHILE, 'EXACT"
APPR.
~ FR IU L I , ' EXACT' - BANJA LUKA, 'EXACT'
____ FRlULl, APPR. ____ BANJA LUKA. APPR.
\ f,*

0.01 0.1 1 10 0.01 0.1 1 10

PERIOD [sl PERIOD Is1

Figure 1. Mean 'exact' and approximate elastic spectra for 5 per cent damping (all ground motions are normalized to ug = 50 cm/s)

are mostly in the range from 0.3 to 0.8 s). The study of the influence of different structural parameters was
performed using these records. The influence of input motion has been studied using five different groups of
records representing ground motions of basically differing types. In addition to the records from California
and Montenegro, ground accelerations obtained in Friuli, 1976, Banja Luka, 1981, and Chile, 1985, have
been used. (Some additional results, corresponding to a Mexico City, 1985, group, which includes ground
motions recorded on significantly different soil conditions, can be found in References 28 and 29.) The main
characteristics of the Friuli and Banja Luka records are the short duration of the strong ground motion and
the short predominant periods. The duration of the Chile records is long, but their predominant periods are
512 T. VIDIC, P. FAJFAR AND [Link]

similar to those of the standard records. It should be noted that the groups are rather small, consisting of 10
records (California, Montenegro), 8 records (Chile) or 6 records (Friuli, Banja Luka). When computing
displacement spectra, all the records were normalized to ug = 50 cm/s. In the case of the non-dimensional
quantities (R, q ) no normalization is needed.
The parametric study has shown that, as far as the influence of the hysteretic models simulating
predominantly flexural behaviour on the response parameters is concerned, the bounds have been generally
’ both cases 10 per cent hardening of
obtained with the bilinear model and stiffness degrading Q - m ~ d e l . ~(In
the slope after yielding was assumed. This value is considered to be appropriate for an equivalent SDOF
system with a bilinear force-deformation envelope, representing the overall behaviour of a MDOF structure.
The unloading stiffness degradation coefficient in the Q-model amounted to 0.5.) Consequently, these two
models were used for development of design spectra. It should be noted, however, that a hysteresis with
considerably lower energy dissipation capacity (e.g. Q-hysteresis with smaller post-yielding and unloading
stiffness) may yield results outside of the bounds set by the two chosen hystereses. Similarly, different results
can be expected in the case of a strength degrading system.
The concept of viscous damping, which is used to include the effects of energy dissipation other than
hysteretic energy, is, in the case of non-linear analysis, not as clearly defined as in elastic analysis. This
problem has so far only seldom been recognized (e.g. References 31 and 32). In our study two mathematical
models of damping were investigated, which place the bounds on the dynamic response. The first, so-called
‘mass-proportional’ damping assumes a time-independent damping coefficient based on elastic properties. In
this case, the effective (instantaneous) damping ratio (the actual damping coefficient divided by the damping
coefficient at critical damping) increases with decreasing values of the instantaneous (tangent) stiffness. The
second model is ‘instantaneous stiffness-proportional’ damping, where a time-dependent damping coefficient
based on tangent stiffness is assumed. In this case the effective damping ratio decreases with increasing values
of the tangent stiffness. The most realistic type of damping might be ‘frequency-proportional’ damping, which
assumes a time-independent effective damping ratio. It yields results which are between the bounds set by the
mass- and instantaneous stiffness-proportional damping. It was not considered in the present study. More
details on the influence of damping model are given in Reference 33. In the majority of cases. 5 per cent
damping was assumed. Some structural systems with 2 per cent damping were also analyzed.
Most of the results in this paper are presented by means of mean spectra for the period range 0.1- 2.5 s. The
spectra were computed by assuming a constant target ductility factor throughout the whole period range.

T
(a)
____
___.___
R-FACTUR,
STRENGTH
0 ISPL..
0.
~ = 4 . 0
fl=4.0
u=4.0 ;
l
(bl
____ STRENGTH 0. )l=1.0
z l
DISPL., J l = l .fl n

./-. .
’-----
._ - .
-
*
< Q-

‘.
1 ) O L O
4 6
> n
;

z =
L.

‘1 c, I
, I I I I , I I , I , I t -‘i T ‘ I -
0 0.5 1 1.5 2 2.5 n 0.5 1 r
PERIOD is1 I i k

-
Figure 2 Coefficients of vanation (Q-model, 5 per cent mass-proportional damping). (a) for R-factor, strength q and displacement
D (standard records); (b) for R-factor ( p 4)
 CONSISTENT INELASTIC DESIGN SPECTRA 513

Some examples of the coefficients of variation are shown in Figure 2. The average coefficient of variation
for the R-factors is about 0.3, which is a value similar to typical values observed in earthquake-resistant
design. For strength and displacement, the scatter of the inelastic spectra is similar to the scatter correspond-
ing to elastic spectra ( p = 1). It can be clearly seen that scaling by the peak ground acceleration, used in the
case of the non-dimensional strength tj [equation ( 5 ) ] , results in a large scatter of results in the medium- and
long-period region of a spectrum.

APPROXIMATE RELATIONS FOR THE REDUCTION FACTOR R

In the proposed procedure the inelastic spectra for strength and displacement are derived from the elastic
pseudo-acceleration spectrum using the reduction factor R [equations (3) or (9,and ( 6 ) ] . Based on the
results of the parametric studies, approximate formulae for the dimensionless factor R are proposed in this
section.
The results of our parametric studies have generally confirmed the main conclusions drawn by several
other researchers. The reduction factor R is, in the medium- and long-period region, only slightly dependent
on the period T, and is roughly equal to the prescribed ductility p. In the short-period region, however, the
R-factor depends strongly on both T and p. The moderate influence of hysteretic behaviour and damping can
be observed in the whole period region. Peaks in the R-spectrum correspond to peaks in the elastic
pseudo-acceleration spectrum (for this reason smooth R-spectra are applicable in combination with smooth
elastic spectra). The transition period from the period -dependent part to the more or less period-independent
part of the R-spectrum is near the period T1 (which approximately represents the predominant period of the
ground motion). A more precise value of the transition period depends on the prescribed ductility factor. It
should be noted that T I is an important characteristic of the ground motion, and varies greatly for the
different groups of records investigated in the study. The influence of other characteristics of the ground
motion on the R-factor has not been considered in this study.
Having in mind the large uncertainties inherent in parameters involved in earthquake-resistant design, and
the relatively small samples used in the present study, no attempt has been made to perform rigorous
statistics in the derivation of the simplified [Link] main idea was to produce formulae which are, on
the one hand, as simple as possible and, on the other, produce appropriate results for all the derived basic
parameters, including energies which are discussed in the companion paper.' Consequently, trial-and-error
procedures, combined with some simple statistical methods, have been used. The group of standard records
was used when determining the approximate expressions. All constants in the approximate relations were
intentionally rounded off. The relations obtained have been checked using the other groups of records.
Based on the above-mentioned observations, a bilinear curve is proposed for the R-spectrum. The
spectrum is divided into two period regions. In the first region, which roughly corresponds to the short-
period region, R increases linearly with increasing period from R = 1 to a value which is equal to the value of
the ductility factor (R = p) or near to it. In the remaining part of the spectrum (roughly in the medium- and
long-period range) the R-factor maintains a constant value:
T
R = c , ( ~ -l)El-+ 1, T S To
TO
R = C ~ ( P l- P + 1, T a To
To = CZP'T TI
7' is defined in equation (10). The constants c l , c 2 , c R and cT depend on the ..ysteretic behaviour and
damping. In Table 11, values for combinations of two hysteretic models and two damping models are
proposed. Five per cent damping applies in all cases. Some examples of R-spectra are shown in Figure 3.
The authors are aware of the fact that the reduction factor computed according to equation (13) with
constants c1 and cRdifferent from 1.0 does not fulfil the physical boundary condition R = p for T = [Link]
discrepancy, however, has little practical importance, because it only affects structures with very long periods.
Strength [equation (3)] and displacement [equation (6)] as well as energies' are derived from the elastic
514 T. VIDlC, P. FAJFAR AND M. FISCHINGER

Table 11. Values of the constants in equations (12)-(14) for 5 per cent
damping
Model

Hysteresis Damping ('1 Cn cz c,

Q Mass 1 .o 1.0 0.65 0.30

Q Instantaneous 075 1.0 065 030
stiffness
Bilinear Mass 1.35 0.95 0.75 0.20
Bilinear Instantaneous 1.10 0-95 0.75 0.20
stiffness

Tl PER I00 0 1

Figure 3 Approximate spectra for R-factor (5 per cent damplng):(a) the influence of ductility (Q-model, mawproporttonal damping),
(b) the influence of hysteretic model and damping model ( p = 4)

spectrum, which, in the case of very long periods, is highly uncertain (as a consequence of the uncertainty of
the maximum ground displacement which governs the elastic spectrum in the long-period region and which
is highly influenced by the processing applied to the records).
As far as damping is concerned, the R-factor depends not only on the mathematical modelling of the
damping but also on the level of the assumed damping coefficient. The quantitative influence of this
parameter is still under investigation and has not yet been included in the formulae. Some data can be
obtained from Figure 4(b), where a comparison of R-factors for 2 and 5 per cent damping is presented. It can
be seen that the R-factors increase with decreasing damping. In the case of the Q-model, the R-factor
corresponding to 2 per cent damping is about 10 per cent higher for the mass-proportional model and about
20 per cent higher for instantaneous-stiffness-proportional damping. As a conservative approximation, the
R-factor for 5 per cent damping can be used instead of an R-factor corresponding to a lower damping
percentage.
The influence of damping on the strength and displacement spectra [Figures 4(c) and 4(d)] is a combina-
tion of damping effects on the elastic spectrum [Figure 4(a)] and on the R-factor [Figure 4(b)]. In the case of
the Q-model and mass-proportional damping, the approximate ratio between the values corresponding to
2 and 5 per cent damping is 1.15, for both strength and displacement. A lower ratio (1.05) was obtained in the
case of instantaneous-stiffness-proportional damping.
More information on the influence of damping can be found in Reference 33.
 CONSISTENT INELASTIC DESIGN SPECTRA 515

j
V
u o

Lo
3
W

a 0

-
____ 2% [Link].

0 I I J 1 , 1 1 \ I / I ( I I J J l l l j l l t l J

0 0.5 1 1.5 2 2.5

i! ____ 5% MASS
5% I NST. ST I FF.
-2 % MASS
_ _ _ _ 2% [Link].

w
Ly
I-
ffl
0
0

d/ ____
__ 2 % M A S S
5 % MASS
5% I N S T . ST I F F .

___ 2% I N S T . S T I F F .
I

1
o--lm
0 0.5 1 1.5 2 2.5
PERIOD Is1

Figure 4. Influence of damping on mean elastic and inelastic spectra ( p = 4, Q-model): (a) elastic spectra for the group of standard
records (normalized to ug = 50 cm/s): (b) spectra for reduction factor R; (c) spectra for strength 4;(d) spectra for displacement D

VALIDATION OF THE PROCEDURE

The proposed procedure (including approximate expressions for the R-factor) has been validated by
comparing the approximate spectra with the corresponding mean spectra computed in the parametric
studies using non-linear time-history analysis (exact spectra). Some results are shown in Figures 5-7.
A system with a bilinear hysteretic model, with a ductility factor equal to 4, and with 5 per cent
mass-proportional damping, subjected to the standard group of ground motions, has been chosen as the
basic system. The influence of variations in ductility, the hysteretic model and the mathematical model of
damping, as well as the input motion, is shown in the figures (additional results are shown in References 28
and 29). A linear scale has been used in all the figures, with the exception of displacements, where
a logarithmic scale was applied in order to enlarge the short-period region. Exact and approximate spectra
are presented at the left- and right-hand side of the figures, respectively. For each record approximate spectra
were determined according to the procedure proposed in this paper. After that, the mean spectra for all
records in the group under investigation were calculated. It should be noted that the periods TI and T2 vary
516 T. VIDIC, P. FAJFAR AND M. FISCHINGER

.-

.-

--v-' ' 1 ' 1 " ~ ' 1 ' 1 I " ' 7 7

iJ .5 1 1.5 2 2.5

?.-

,,, ----

____
~
B I L I N E A R . MASS
B I L I NEAR, I NST.
0. M A S S
ST I FF.

- -1 ~ 0. i N S T . S T I F r

1
I
____
,-,

-...._*
CHILE
U.5.A.
MONTENEGR@
i
_ _ FRiULl
-_ _ - BANJA LUK4

,4 V T T , I I , , , I , 1 I ' , , 1 T ' , , \
-1
I T ? - n 7 - - T

n G 5 1 1.5 2 2.5 0 0 5 1 1.5

PERIOD Is1 PFFI

Figure 5 Comparison of 'exact' and approximate mean inelastic spectra for reduction factor R (5 per cent damping) (a) the influence of
ductility (bilinear model, mass-proportional damping), (b) the influence of hysteretic model and damping model (p = 4). (c) the
influence of ground motion ( p = 4, bilinear model, mass-proportional damping)
 CONSISTENT INELASTIC DESIGN SPECTRA 517

r-

c
--
I/
/\
\\
-
____ BILINEAR,
B I L I NEAR.
MASS
I NST. S i 1 F F .
-
____ BILINEAR.
BILINEAR.
MASS
[Link].

____
, - , CHJLE -
____ CHILE
U.S.A.
_______
__
U.S.A.
MONTENEGRO
FRlULl
____ BAN34 LUKA

0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5

PERIOD [sl PERIOD Is1

Figure 6. Comparison of 'exact' and approximate mean inelastic spectra for strength (5 per Cent damping):(a) the influence of ductility
(bilinear model, mass-proportional damping); (b) the influence of hysteretic model and damping model ( p = 4); (c) the influence of
ground motion ( p = 4, bilinear model, mass-proportional damping)
518 T. VIDIC, P. FAJFAR AND M. FISCHINGER

L
u

1 .o
2.0
3.0
4.0
5.0
5.0
8.c
' C .

- /j ____
.-, -
B I I NEAR. MASS
B I L I N E A R . INST. ST FF
-~0 , M A S S
1 _ _ _ _ 0. [Link].
I

I-
- LHILt.

I ' I u a

"EP 110 < I c

Figure 7 Comparison of'exdct' and approximate mean inelastic spectra for displacement D ( 5 per Lent ddmplng) (a) the influence o r
ductility (bilinear model, mass-proportional damping), (b) the influence of hysteretic model and damping model ( p = 4) (cr the
influence of ground motion ( p = 4, bilinear model, mass-proportional damping)
 CONSISTENT INELASTIC DESIGN SPECTRA 519

for different records [equations (10) and (ll)]. Therefore, the approximate mean spectra shown on the
right-hand side in Figures 5-7 are smooth in spite of the fact that a bilinear R-spectrum was used at the
record level. Very good agreement between the exact and approximate spectra can be observed. The
correlation would have been even better if the constants in the approximate relations had not been
intentionally rounded off. The largest discrepancy can be observed in some parts of the spectra for R-factors
where the proposed values are smaller than the exact values. These discrepancies, however, have little
practical importance as is clearly demonstrated in Figures 6 and 7, where q and D-spectra are shown. Due to
the smoothing effect of inelastic behaviour (the peaks in the exact R-spectra correspond to peaks in the elastic

0 0.5 1 1.5 2 2.5

PERIOD Is1

Figure 8. Approximate spectra for strength v, based on a mean elastic design spectrum of the Newmark-Hall type (the group of
standard records normalized to ug = 50 cm/s, 5 per cent damping): (a) the influence of ductility (bilinear model, mass-proportional
damping); (b) the influence of hysteretic model and damping model ( p = 4)

- -
E -
u - ,--

i-I
1.0
2.0
3.0
4.0
5. 0
6.0
8. 0
10.

-
.0 1 L 1 1 1 I 0 ' I 1 , I , , , > I
0.1 1 10 0.1 1 I0
PERIOD Is1 PERIOO Is1

Figure 9. Approximate spectra for displacement D based on a mean elastic design spectrum of the Newmark-Hall type (the group of
standard records normalized to u4 = 50 cm/s, 5 per cent damping): (a) the influence of ductility (bilinear model, mass-proportional
damping); (b) the influence of hysteretic model and damping model ( p = 4)
520 T. VIDIC, P. FAJFAR AND M. FISCHINGER

spectra) no distinctive peaks appear in the exact displacement and strength spectra. Very similar results can
be obtained with the proposed procedure using smooth elastic design spectra and smooth R-spectra.
In order to test the applicability of the proposed procedure for a design situation, approximate spectra
were computed using a typical elastic design spectrum of the Newmark-Hall type (see the section Elastic
spectra). Trapezoidal line fitting was carried out on the mean elastic spectrum for the group of standard
records (Figure I). Note that such a procedure yields, due to the variation in the knee periods T I and T 2 ,an
elastic design spectrum which differs from the spectrum computed as the mean of the spectra by fitting
trapezoidal lines to each spectrum of the ensemble.34 The chosen elastic spectrum is defined by the peak
ground motion parameters (a, = 525 cm/s2, tj, = 50 cm/s, d , = 15 cm) and by the amplification factors
(cu = 2.4, cL = 1.9, cd = 1.7). The computed spectra for strength and displacement are shown in Figures 8 and
9. Fair agreement with the mean spectra shown in Figures 6 and 7 can be observed. The main difference is in
the peak values for strength which, in the mean spectra, are smaller due to the smoothing effect of the
averaging process. The step decrease in strength in the range from T = 0 to T = 0.03 s (which will typically
not be taken into account in a practical design procedure) is due to the constant value of the elastic design
spectrum in this range. Such an assumption is probably not realistic, but it is widely used due to the lack of
reliable data in the very short period range. A similar phenomenon can be observed in the design spectra
according to the new Chilean [Link]

CONCLUSIONS
Based on an analysis of the results of extensive parametric studies of the inelastic response of SDOF systems,
the non-dimensional reduction factor R has been chosen as being the most convenient for approximate
representations by simple formulae. Knowing the characteristics of the input motion given by the elastic
pseudo-acceleration spectrum, and the R-spectrum, the strength and displacement spectra can easily be
determined. This procedure can be extended to the determination of hysteretic and input energy spectra, as is
shown in the companion paper.’
Due to the surprising accuracy of the approximate spectra, only partly demonstrated in this paper, the
authors believe that the proposed procedure is soundly based. It may be, however, that further studies, using
larger statistical samples, will suggest some small changes in the proposed R-spectrum. The authors also
believe that, for use in practical design procedures, some additional simplifications of the R-spectrum are
possible without jeopardizing the accuracy which is reasonable for seismic design procedures.

ACKNOWLEDGEMENTS
’lhc results presented in this paper are based on work supported by the Ministry for Science and Technology
of the Republic of Slovenia and by the U.S. NSF, in cooperation with the University of California, Berkeley,
and Stanford University. This support is gratefully acknowledged. The authors are indebted to Professor H.
Krawinkler from Stanford University for stimulating discussions and helpful suggestions throughout the
work. The authors also welcomed the suggestions of the reviewers.

REFERENCES
1 . P. Fajfar and T. Vidic, ‘Consistent inelastic design spectra: hysteretic and input energy’. Earthquake eng. .srrur’[. [Link] 23, 52.7 ‘37
( 1994).
2. V. 1’. Bertero, J. C. Anderson, H. Krawinkler and E. Miranda, ‘Design guidelines for ductility and drift limits’, R r p o r / %-(t
L:CB/EERC-YI/IS, Univ. of California, Berkeley, 1991.
3 . N. M. Newmark and W. J. Hall, Earthquake Spectra and Design, EERI, Berkeley, CA, 1982.
4. R. Riddell and N. M. Newmark, ‘Statistical analysis of the response of nonlinear systems subjected to earthquakes’. S t r u ~ ~ r u m l
Research Series N o . 468, Civil Engineering Studies, University of Illinois, Urbana-Champaign, IL, 1979.
5. E. Rosenblueth, ‘Seismic design requirements in a Mexican 1976 Code’, Earthquake eng. struct. dyn. 7, 49-61 (1979).
6. .I. B. Berrill, M. J. N. Priestley and H. E. Chapman, ‘Seismic design of bridges, Section 2, Design earthquake loading and ductility
demand’, Bull. New Zealand natl. SOC. earthquake eng. 13, 232--241 (1980).
7 S.-S. P. Lai and J. M. Biggs, ‘Inelastic response spectra for aseismic building design’, J . struct. diu. ASCE 106, 1295 ~1310(1980).
8. G . J. Al-Sulaimani and J. M. Roesset, ‘Design spectra for degrading systems’, J . struct. eng. ASCE 111, 2611-2623 (1985).
9. F. E. Elghadamsi and B. Mohraz, ‘Inelastic earthquake spectra’, Earthquake eng. sfruct. dvn. 15. 91-104 (1987).
 CONSISTENT INELASTIC DESIGN SPECTRA 521

10. M.-H. Peng, F. E. Elghadamsi and B. Mohraz, ‘A simplified procedure for constructing probabilistic response spectra’, Earthquake
spectra 5, 393-408 (1989).
1 1 . R. Riddell, P. Hidalgo and E. Cruz, ‘Response modification factors for earthquake resistant design of short period buildings’,
Earthquake spectra 5, 571-590 (1989).
12. P. A. Hidalgo and A. Arias, ‘New Chilean Code for earthquake-resistant design of buildings, Proc. 4th [Link]! earthquake eng.,
Palm Springs, EERI, 2, 927-936 (1990).
13. H. Krawinkler and A. Nassar, ‘Damage potential of earthquake ground motions’, Proc. 4th [Link]: earthquake eng., Palm Springs,
EERI, 2,945-954 (1990).
14. M. J. N. Priestley and G. M. Calvi, ‘Towards a capacity-design assessment procedure for reinforced concrete frames’, Earthquake
spectra 7 , 413437 (1991).
15. W. K. Tso and N. Naumoski, ‘Period-dependent seismic force reduction factors for short-period structures’, Can. j . ciu. eng. 18,
568-574 (1991).
16. S . A. Mahin and V. V. Bertero, ‘An evaluation of inelastic seismic design spectra’, J . struct. diu. ASCE 107, 1777-1795 (1981).
17. W. D. Iwan, ‘Estimating inelastic response spectra from elastic spectra’, Earthquake eng. struct. dyn. 8, 375-388 (1980).
18. K.S. Shimazaki and M. A. Sozen, ‘Seismic drift of reinforced concrete structures’, Special Research Paper;Civil Eng. Dept., Univ. of
Illinois, Urbana, 1985.
19. H. Hwang and J. W. Jaw, ‘Statistical evaluation of deflection amplification factors for reinforced concrete structures’, Proc. 4th U.S.
cant earthquake eng., Palm Springs, EERI, 2, 937 -944 (1990).
20. X. Qi and J. P. Moehle, ‘Displacement design approach for reinforced concrete structures subjected to earthquakes’, Report No.
UCB/EERC-91/02,Univ. of California, Berkeley, 1991.
21. H . Krawinkler and M. Rahnama, ‘Effects of soft soils on design spectra’, Proc. 10th world cot$ earthquake eng., Madrid, Balkema,
Rotterdam, 10, 5841-5846 (1992).
22. E. Miranda, ‘Nonlinear response spectra for earthquake resistant design’, Proc. 10th world conf: earthquake eng., Madrid, Balkema,
Rotterdam, 10, 583555840 (1992).
23. F. Tarquis and J. M. Roesset, ‘Inelastic response spectra for narrow band earthquake’, Proc. 10th world conf: earthquake eng.,
Madrid, Balkema, Rotterdam, 2, 64-50 (1992).
24. C.-M. Uang, ‘Establishing R (or R,) and Cd factors for building seismic provisions’, J. struct. eng. ASCE 117, 19 -28 (1991).
25. P. Fajfar;T. Vidic and M. Fischinger, ‘Seismic demand in medium- and long-period structures’, Earthquake eng. struct. dyn. 18,
1133-1144 (1989).
26. [Link], T. Vidic and M. Fischinger, ‘A measure of earthquake motion capacity to damage medium-period structures’, Soil dyn.
earthquake eng. 9,236242 (1990).
27. P. Fajfar, T. Vidic and M. Fischinger, ‘On the energy input into structures, Pacific conf: earthquake eng., Auckland, Vol. 1, 1991,
pp. 81-92.
28. T. Vidic, Inelastic seismic response of SDOF systems, Ph.D. dissertation, University of Ljubljana, 1993 (in Slovenian).
29. T. Vidic, P. Fajfar and M. Fischinger, ‘A procedure for determining consistent inelastic design spectra’, in H. Krawinkler and
P. Fajfar (eds.),Nonlinear Seismic Analysis and Design of Reinforced Concrete Buildings, Supplementary proceedings of a workshop
held in Bled, J. A. Blume EEC Report No. 103, Stanford Univ. and Univ. of Ljubljana, 1992.
30. M. Saiidi and M. A. Sozen, ‘Simple nonlinear seismic analysis of RC structures’, J. struct. div. ASCE 107, 937-952 (1981).
31. S. Otani, ‘Hysteresis models of reinforced concrete for earthquake response analysis’, [Link] eng. Tokyo Univ. (B), XXXVI,(2),
125-159 (1981).
32. P. E g e r and S. Dussault, ‘Seismic energy dissipation in MDOF structures’, J . struct. eng. A X E 118, 1251-1269 (1992).
33. P. Fajfar, T. Vidic and M. Fischinger, ‘Influence of damping model on the seismic response of nonlinear SDOF systems’, in T. Moan
et al. (eds.), Structural Dynamics - EURODYN ’93, Balkema, Rotterdam, 1, 77-84 (1993).
34. A. K. Jain and S. Pal, ‘Probabilistic amplification factors for response spectra’, J . struct. eng. ASCE 117, 24642776 (1991).
35. P. Hidalgo, ‘New earthquake-resistant design regulations in Chile after the 1985 earthquake’, Proc. IDNDR int. symp. on earthquake
disaster reduction technol., Tsukuba, Japan, 1992, pp. 227-246.