129

DYNAMIC ANALYSIS OF STRUCTURES

Athol J. Carr1

ABSTRACT

This paper is the result of discussions of the study group on Dynamic Analysis established
by the New Zealand National Society for Earthquake Engineering and whose brief was to
provide some guidance to design engineers on the usage of the dynamic analysis methods
that are becoming more common in structural design with the stricter requirements of the
seismic loadings requirements of NZS 4203:1992. The paper first of all discusses the
concepts of dynamic analyses of single mass systems and then extends this to the modal
analysis methods for the analyses of multi-mass systems. The paper then covers a variety
of aspects that may be of concern in performing the dynamic analyses and in the modelling
of structures.

WHY CARRY OUT DYNAMIC ANALYSES? In the majority of analyses carried out by engineers the forces
are treated as static and even in situations where the forces are
Consider the single degree of freedom cantilever structure not actually constant with time they are considered to change
shown in Figure 1 where the mass M is supported on the top sufficiently slowly that dynamic effects are not significant. If
of the weightless column of length L and lateral stiffness K . the force P acting on the structure changes at such a rate that
If the column was prismatic with a flexural rigidity of EI, the inertial and damping forces have a significant effect on the
where E is Young's Modulus and I is the second moment of equation of equilibrium, then a dynamic analysis is required.
area of the section, then the stiffness K would be 3EI/L. If a
lateral force P is applied to the top of the column, then the In a dynamic analysis where the applied force P(t) is changing
lateral deflection x at the top would be obtained by solving the with time, the unknowns are the displacement, velocity and
equation acceleration of the mass but there is only a single equation of
equilibrium, albeit a second order ordinary differential
Kx=P equation,
Mx + Ci + Kx = P(t)
X where the dot denotes differentiation with respect to time.

In the case of an earthquake excitation, the external forces are,

in fact, constant but the ground on which the structure is
located is moving. If the foundation of the structure can be
considered to move as a rigid body when subjected to a ground
displacement xg then the following observations may be made.
The inertia force acting on the structure is proportional to the
total acceleration, i.e. the sum of the acceleration of the
L structure and the acceleration of the ground. The elastic force
in the structure is proportional to the deformation in the
structure, i.e. the displacements of the mass relative to the
foundation. It is generally assumed that the damping force is
also proportional to the relative velocity in the structure, i.e. the
velocity of the structure mass relative to the foundation. This
/ / / j.;!·-/-:.:. · / / ·////:; is expressed in the following equation."
M(i+ig) + Ci + Kx = P(t)
Figure J. Single Degree Of Freedom Model of a Structure.
This equation may be re-arranged to give
1 Department of Civil Engineering, University of Canterbury, Mi + Ci + Kx = P(t) - Mig
Christchurch (Fellow)

BULLETIN OF THE NEW ZEALAND NATIONAL SOCIETY FOR EARTHQUAKE ENGINEERING, Vol. 27, No.2, June 1994
130

In this case the applied force P(t) is constant and represents the t
static force PO and the second term on the right hand side of the
equation is often thought of as the equivalent earthquake force.
x(t) __!_Jx
W g
e-!.w(H) sinw (t-,) dt
D
0
This force is proportional to the mass of the structure and the
ground acceleration. If the maximum acceleration of the ground
is known the maximum equivalent lateral force can be This solution is usually obtained by a numerical integration
computed and the maximum displacement of the structure can such as the methods of Newmark [Clough, 1992] etc.
be found.
An alternative approach to the solution for the structure
If the structure is allowed to freely vibrate with no external subjected to the earthquake excitation is to use a response
applied forces and no damping in a simple harmonic motion spectrum method where, given the natural period of free-
then the displacement has a sinusoidal variation with time such vibration and fraction of critical damping of the structure, the
that maximum displacement and maximum acceleration of the
structure of the structure can be determined.
x(t) = Asinwt + Bcoswt
where A is the amplitude of the vibration and w is the natural Most engineers are familiar with the use of Response Spectrum
circular frequency of free-vibration whose units are radians per methods of carrying out the dynamic analyses of structures for
second. Substituting this solution into the equation of design. These Response Spectrum methods reduce the dynamic
undamped motion yields response calculations to that of observing the maximum
displacement or acceleration response of their structure to a
-w 2 Mx+Kx=O given earthquake for which the spectrum has been determined
so that or specified.

w
RESPONSE SPECTRA

and Response spectra are derived from the response of a single

degree of freedom system subjected to an earthquake ground
f = ~. T = _! acceleration excitation. The maximum absolute value of the
2rc ' f displacement of the mass is obtained from
where f is the natural frequency (Hertz) and T is the natural
period (seconds) of free vibration. = s = J_s = __!_s
d W v W2 a

In the above equations of free-vibration, the effects of damping where Sd is the Spectral Displacement, Sv is the Psuedo
have been ignored. In general, the observed equivalent amounts Spectral Velocity and S, is the Pseudo Spectral Acceleration.
of damping in building structures appears to be small where the The Pseudo Spectral Velocity is the maximum value of the
fraction of critical damping 11. is less than about 5%. This integral term above which is an integration of an acceleration
means that the difference between the undamped natural with respect to time and therefore has the units of velocity. It
frequency w and the damped natural frequency w 0 is is not the maximum velocity of the mass, the Spectral Velocity,
insignificant. If the structure has damping and if the damping but it is related to it. The Psuedo Spectral Acceleration usually
coefficient C is defined as differs from the Spectral Acceleration, which is the maximum
total acceleration of the structure, by only one or two percent.
C = V . wM
The period used in the calculation of the modal response is that
where 11. is the fraction of Critical Damping, then the response of the structure in undamped free-vibration. These spectra are
in a damped simple harmonic mo/ion is usually published for several levels of viscous damping, usually
0%, 2%, 5%, I 0% and 20% of critical viscous damping.
x(t) = e-J.w, {Asinw 0 t + Bcosw 0 t} Spectra in the various New Zealand design codes have been
where the damped natural circular frequency w0 is defined as based on 5% of critical viscous damping. With the low levels
of damping encountered in building structures the natural
WD = w/17 period is little affected by viscous damping.

If the structure is linearly elastic, the Principle of Superposition The maximum value of the base shear V can be obtained from
is valid and the solution for the response under a combination
of loads can be obtained by superimposing the response Vmax = KS d = w2 MS d = wMS v = MS a
obtained from each separate load. Therefore the static force
solution can be added to that for the earthquake excitation. This The maximum displacement of a structure with a given natural
paper will consider only the earthquake excitation response. period of free-vibration and a given fraction of critical damping
subjected to the El Centro ground accelerogram shov-m in
If the ground acceleration history is known, then the equation Figure 2, can be read off the Displacement Spectrum shown in
of motion may be re-,vritten as Figure 3.
Mx +ex+ Kx = -MxgCt) The maximum acceleration of the mass of the structure
Substituting for C and K and dividing through by M gives subjected to the El Centro ground accelerogram can be
obtained from the Acceleration Spectrum for the El Centro
X + 2 /1, w X + w2 X = -x g (t) accelerogram, shown in Figure 4.

The solution for this equation can be shown [Clough, 1992] to

be
 131

40
The Code specified response spectra are smoothed response
spectra based on an earthquake with a specified probability of
exceedance during the expected design life of the structure. If
a different level of risk is desired, or the expected life of the
.2 0
structure is much different from that of normal buildings, then
the spectral values may need to be adjusted accordingly.

.0 0

-.2 {
l"'V
II

l
I

r
11!

I ~ 1,
l ~
1J. t
~ 1~
I.Al .d
I\'~ I II~
Many building codes supply the design acceleration spectra
with the units being fractions of the acceleration of gravity so
that the engineer may use the weight of the structure rather
than the mass of the structure in computing the base shear.

However, these Response Spectrum methods are applicable

only to single degree of freedom structures. For a multi-degree
of freedom structure, which covers nearly all real structures, the
-.4 answer requires the simultaneous solution for the displacements
0123 • 567 8 9 10 11 12 13 14 15 16 17 18 19 20
of every degree of freedom of the structure as a time-history.
Time (Seconds) This time-history may then be enveloped to obtain the
maximum values of the displacements and member forces. The
Figure 2. El Centro May 1940, North-South displacement of the structure cannot now be described as a
Accelerogram. function of a single variable and the structure does not have a
single natural frequency of free-vibration and so the simple
0.60,------r------.-----.----,------, application of response spectra techniques is no longer
available to the designer. This means that for these structures
a different approach must be made to their dynamic analyses.
The following example of a building subjected to a horizontal
ground motion will illustrate some of the differences in
behaviour when compared with a single degree of freedom
system.
c
E a. 3 "l-----+-----+--==-+-+-'r--+------1

.
1:l
ii
STRUCTURES WITH MORE THAN A SINGLE MASS
:a 0.2't----t'n'-+-+---:l---,Ll-'-;._~...,.....t-=-"-"----:--',,,,....~=::----,
For the structure shown in Figure 5, the ground is assumed to
"E move initially to the right. The base of the structure has to
i
{JJ O l .----t-t-i.-+-s:-ne<--:,~----1-----,------j move with the ground but upper parts of the structure have not
yet followed the lower structure. It takes time for the shear
forces, caused by the deformation of the structure, to accelerate
the masses of the upper floors. The speed of propagation of
Natural period (Seconds) the wave motion up the structure is governed by the stiffness
of the stories and mass of the floors. By the time that the
upper part of the structure has started to follow its base, the
Figure 3. Displacement Response Spectrum.
ground may well have started to move back in the reverse
direction. The deformation pattern of the structure may take on
a complex form.
2.50

2 0n
l

A.=0% I
u ~
2%-~
;"-.:'.!~1

Q
I
/ 5% I_ ,1 \j'.
/ /\I'I
,-
'",_,,..... __. '!0?<
20%
,
I•-
', ,
--- M
0\
-~. ''c ""' ,,.---.__
··-.~~--~--=~~
-=-=--:_-:-_~~
o.o n ~--
0

Natural period (Seconds)

Figure 4. Pseudo Acceleration Response Spectrum.

Figure 5. Structure subjected to ground displacement.

132

The inertia forces acting on the various levels of the structure

are proportional to the masses and total accelerations of the (xi;= A{<l>l; sinwit
floors and are derived from Newton's Laws. These where A is an, as yet, undetermined constant and the equation
accelerations are the second derivatives of the displacements of undamped free-vibration becomes
with respect to time and exhibit a much greater variation with
respect to time than are shown by the displacements of the
structure.
The mode shapes, or vectors, have the very important property
The number of degrees of freedom in a structure is defined as in that they are orthogonal with respect to the mass and
the number of variables required to uniquely define the inertia stiffnesses matrices of the structure. This means that, provided
forces or the displacements. For any real structure with one assumes that the mode shapes are also orthogonal with
distributed mass, the number of degrees of freedom is infinite. respect to the damping matrix of the structure, one can consider
However, most computational methods reduce the number of the structure vibrating in any of the modes of free vibration as
degrees of freedom by restricting the displacements of the a single degree of freedom system with the natural frequency,
members, or elements, within the structure to a limited class of or period, of that mode. The normally coupled N degree of
functions. For example, for beam or column members in a freedom system behaves as N single degree of freedom
conventional frame analysis, the lateral displacements of the systems, each one associated with a natural mode of free
beam are usually assumed to have a cubic variation along their vibration. The shape of each mode of free vibration is unique
length so that their displacements can be uniquely defined by but the amplitude of the mode shape in undefined. The mode
the displacements of the joints, or nodes, at the member ends. shapes are usually normalized such that the largest term in the
In dynamic analyses, the number of degrees of freedom may be vector is 1.0 as is shown in the top of Figure 6. Alternatively
further reduced by lumping the mass of the structure at a the sum of the squares of the terms in the vector is 1.0 or, in
selected number of joints or levels. Those degrees of freedom the case of some of the computer programs in common
without mass are no longer independent variables but are engineering usage the vectors are normalized so that the
dependent on the displacements of those degrees of freedom generalized mass M. is 1.0 i.e.
that do have mass associated with them.
M\ = {<l>;IT [M] {<!>ii = 1.0
In most building structures with stiff in-plane floor diaphragms It must be noted in this last case that the 1.0 has the units of
the mass representation may be reduced to two horizontal mass.
inertias, the mass of the floor acting in two orthogonal
horizontal directions, and a rotational inertia of the floor about If the modes shapes are assumed to be orthogonal with respect
a vertical axis, generally through the centre of mass of the to the damping matrix as well as the mass and stiffness
floor. The latter is related to the polar moment of inertia of the matrices, then damping will have no effect on the mode shapes
floor mass about the vertical axis. but the natural frequency for each mode will be the damped
frequency as was discussed earlier for the single mass systems.
The equation of motion can now be written as
[M]{x} + [C]{x} + [K]{x} = {P} - [M]{r}xg w0i = W ~
iVJ. - .n.i

where [M], [C] and [K] are the mass, damping and stiffness where A; is the fraction of critical damping in the ith mode of
matrices, {x} is the vector of displacements. The vector {r} is free vibration.
the influence vector for the ground displacement where each
term is the displacement in the structure associated with a unit
foundation movement in the direction of the earthquake
excitation.

MODAL ANALYSES

The natural frequency of free vibration associated with a single

degree of freedom system vibrating in simple harmonic motion
has its equivalent in multi-degree of freedom systems. At each
natural frequency of free vibration the structure vibrates in
simple harmonic motion where the displaced shape, or mode
shape, of the structure is constant but the amplitude of the
displacement is varying in a sinusoidal manner with time. A
system with N degrees of freedom has N natural frequencies of
free vibration and N mode shapes of free vibration {q>} , one
associated with each natural frequency. This set of N mode
shapes forms a Basis set of displacement vectors in that any
displacement shape of the structure can be made up of a
combination of these linearly independent mode shapes.
Luckily, most engineers are generally only concerned with a
small number of these modes, i.e. those associated with the few
lowest natural frequencies of free vibration.
-1 0
When assuming simple harmonic motion the displacement m
mode shape can be written as
Figure 6. Mode shapes of free-vibration.
 133

If the structure has symmetry in both stiffness and mass with

respect to two orthogonal horizontal axes of the structure, then L •. L '.
{4>}-' Sd. {4>}-' _!__Sa.
the mode shapes will uncouple the motion in the two horizontal 'M*. l IM*.,w 2 I
1
axis directions and it may be possible to speak of, say, x and
y direction mode shapes as well as torsional modes. However,
if these symmetries are not present and eccentricity of mass Once the displacements have been obtained for the ith mode,
and/or stiffness is to be considered, then such a directional then the inter-storey drifts, the member forces and moments
uncoupling is unlikely. This means that any translation of the can all be obtained for the ith mode.
structure in an axis direction will involve translation in the
orthogonal direction and also rotation about the vertical axis. It can be shown that the maximum base shear for the ith mode
In design, even though the structure may appear to be can be written as
symmetrical, this is not guaranteed by the materials and in the
yield properties. This is one reason for code requiring (L '.)2
- - ' - Sa.
designers to make provision in the design for some torsion M'. I
1
about the vertical axis.
and if the equivalent inertia forces {f}; associated with each
The uncoupling of the degrees of freedom means that one can degree of freedom is required, then they are given by
use Response Spectra methods to obtain the response of each
mode in the structure and thus for each mode obtain the = [M]{4>}; V
displacements and accelerations or forces. The difficulty, {fh max L •. i max
I
however, is that Response Spectrum methods have lost parts of
the information relevant to the dynamic behaviour of the These forces may be applied to the structure and actions
structure. Unanswered questions are: when has the peak evaluated with an equivalent static analysis. This is unnecessary
response occurred, what sign should be associated with it and with most modem programs as the member forces in the
what any other mode with a different natural frequency was structure follow immediately from the computed displacements.
doing at the time the peak response occurred in the mode of
interest. It is thus impossible to combine these modal Mass Participation Factors.
responses to obtain the maximum response of the multi-degree
of freedom system. The problem is usually resolved by relying The multiplier of the ground acceleration, L'IM'· in the
on a statistical combination of the modal responses. The most equation of motion for the ith mode, is the participation factor
common method being the Square Root of the Sum of the for the ith mode of the structure. It is a function of the mode
Squares (SRSS) combination method which was used in the shape, the mass distribution of the structure, and the direction
earlier New Zealand loadings codes. of the earthquake excitation. If the vibration components of the
mode shape are orthogonal to the direction of the ground
These modal methods are only applicable to linearly elastic excitation, the participation factor for that mode is zero.
structures undergoing small deflection responses in that these Negative participation factors may also be observed. In general,
are the requirements for the Principle of Superposition to be the magnitudes of the participation factors diminish with
valid. If these conditions are not met, then the natural modes increasing mode number and at some point it may be
and frequencies of free vibration do not have a meaning considered that the remaining higher modes do not significantly
because the frequency and mode shapes depend on the affect the displacements.
amplitude of the displacement and the resulting free vibration
in each mode is no longer simple harmonic. Further, if the The participation factor shows how strongly a given mode
Principle of Superposition is no longer valid, then the contributes to the response of the structure when subjected to
combination of the modal responses to get the total structural a specified direction of ground acceleration. An important point
response is also invalid. to note concerning participation factors is that the numbers
obtained are dependant on the normalization method used in
computing the mode shapes of free-vibration. Thus the
MODAL ANALYSIS USING RESPONSE SPECTRA participation factors computed from one computer program
should not be compared with those from another program
The response of the structure in each mode can be written as unless they both use the same methods of normalizing the
mode shapes.. This normalization does not affect the solution
L *.
Y, + 2J.. 1.w 1 Y1. + w 2 Y. = ---' i (t) for the displacements or member actions as the modal
' M*. g amplitude Y; has to be multiplied by the mode shape and any
l
effect of the normalization is cancelled out. It is only the
where Yi is the amplitude of the ith mode where the modal amplitude Y; that is affected by the choice of
displacement of the structure in the ith mode may be expressed normalization method.
as
Number of Modes to be Used.

and The number of modes needed depend~ on the structure, the

direction of earthquake excitation and the degree of coupling
L\ = {4>}/[M]{r} between the translational and torsional modes. Sufficient
modes should be used to ensure that most of the structure's
To obtain the maximum displacement of the structure in the ith mass is contributing to the excitation in each direction of
mode, the procedure is very similar to that for the single degree excitation. One measure that can be used is to investigate the
of freedom system. Participation Factors for each mode. These will spow which
modes are participating in the particular direction of excitation,
134

which modes are orthogonal to the excitation and at which uniform moment resisting frame undergoing a shear-type sway
mode number the contributions from the higher modes are deformation the natural frequencies of free vibration increase
becoming insignificant. roughly in the ratios of I : 3 : 5 : 7 : 9 . etc. and whilst the
relative ratios between the frequencies of the higher modes tend
A more significant method uses the Effective Mass of each towards 1.0 the contributions of these higher modes is rapidly
mode (L')2JM' which is seen in the expression for the diminishing because their Participation Factors become smaller
maximwn base shear in each mode. The sum of the effective with increasing mode number.
masses for all modes must equal the total mass of the structure.
The effective mass associated with a mode usually diminishes In three-dimensional structures, different modes in different
rapidly as the mode number i increases. The nwnber of modes directions may have very similar frequencies. If one of these
used in the analysis should be large enough so that the modes modes is strongly excited by the earthquake at a given instant
used represent at least 90% of the total mass of the structure. then the other mode, with a similar frequency, is also likely to
be strongly excited at the same instant. In these cases Root
Sum Square combinations have been shown to give non-
COMBINATION OF MODAL MAXIMA conservative estimates of the maximwn likely response.

The use of response spectra techniques for multi-degree of With an increasing realisation that three-dimensional analyses
freedom structures is complicated by the difficulty of were important in the response of real structures to earthquake
combining the modal responses. The combinations are usually excitation and the increasing concern of torsional behaviour,
achieved by using statistical methods. Response spectra improvements in the modal combination methods were sought.
calculations have lost all information on sign or when the For very short period structures where all the natural
maximum displacement etc. occurred. Therefore, proper frequencies are low, and are therefore, in an absolute sense,
combinations of modes is not possible. In each mode close together, the few modes were often combined by taking
structural members are in equilibriwn and all actions in the maximum response as the sum of the absolutes of the
members have the appropriate of signs. However, what modal responses. In these structures, which are inherently
contribution or sign other modes should have at the same time massive, the excessively conservative nature of the
are unknown. combinations is unlikely to be a problem. In the 1972 Los
Angeles proposal [Los Angeles, 1972] it was suggested that the
Let Ri be the modal quantity (Base Shear, Nodal displacement, worst of the SRSS of a minimwn of three modes or the swn of
Nodal force. Member stress etc.). Values of R; have been the absolutes of any two modes be used for design. In a three
found for all modes (or for as many as are significant). Most dimensional structure it is likely that natural frequencies of
design codes require designers to use a sufficient number of modes in one translational direction will be similar to the
modes so that the error associated with the omitted modes is natural frequencies of modes in the orthogonal translational
only a small percentage of the total response. NZS 4203 direction or to the natural frequencies of the torsional modes.
requires at least 3 modes in each direction to be considered.
Other modal combination methods that have been proposed are,

Rmax = E: 1 Ri max would be true only if all maxima occurred

in general, all related to the statistical combination of natural
mode responses when the structure is subjected to a white noise
at the same instant of time and all had the same sign. excitation. It has been shown that these methods give
reasonable combinations when applied to earthquake excitation.
In general, Similarly, the method proposed by Humar was initially

Rmax ~ E: 1 R; max and m almost all cases the inequality

developed for the torsional response of structures subjected to
lateral ground excitation.
holds.
All the modal combination methods may be expressed in the
There are several accepted statistical combination methods. form

(a) Maximum Possible Response: Sum of Absolutes. This is Rmax = I{L LR; max PijRj max)
very conservative and is very seldom used except in some \l I J
codes for say two or three modes for very short period where Pij is the correlation coefficient which varies from
structures. combination method to combination method.

11
.L "max = 't""'N
L...,j,,.. l IR
l max
I CQC [Wilson et al, I 981] (Complete Quadratic Combinations) .

(b) Maximum Likely Response: Square Root of Sum of

Squares, SRSS or Root Sum Square. This is the most p .. =
commonly used method. It is NOT Root Mean Square. 'l (wf - wf)2
where Ai and Ai are the fractions of critical damping in the ith
andjth modes and roi and roi are the natural circular frequencies
The Root Sum Square method was initially used when two- of the modes. Further discussion of this method may be found
dimensional structural analyses were the norm. In a two- in the text by Clough and Penzien [Clough I 993].
dimensional structure no two lateral frequencies are close, and
so no strong correlation between modal responses is likely.
The Root Sum Square method implies no correlation between
the responses of the different modes, the maximum of each
mode is independent of the maximum in other modes. In a
 135

DSC [Rosenblueth and Elorduy, 1969] (Double Sum Root-Sum-Square implies p,, = 1.0 and Pu = 0.0 for i j and *
Combination). therefore it would be possible to have a computer program to
cover all of these combination methods, SRSS, CQC, DSC,
Humar and Gupta with the only difference being in the choice
of the correlation coefficients p,i•

Table 1 shows the correlation coefficients for the different

modal combination methods for a structure having 5% viscous
where damping in all modes. For the DSC methods, the product of
the duration of strong shaking S times the lower natural
circular frequency w equals 30, 60 and 120 respectively. For
the Humar method S is equivalent to infinity.

S is the Time Duration of the White Noise segment of the

earthquake record. For actual records, for example, the
duration of the strong motion segment characterized by
extremely irregular accelerations of roughly equal intensity
should be taken.
\

Humar [Humar, 1984]. Here Pu is as for DSC with S =

\
\
infinity, 1.e. \' = A, and if A, = \ = A (the usual assumption)
then

Frequency ratio: T=l.00, >.=5%

and if A is much less than 1.0 then
Figure 7. Modal Correlation Coefficient versus
Frequency Ratio.
Pii = [ ( 1 ( w. - w.)) 1
2
1 + ~ w:+w;
Figure 7 shows how the correlation coefficients change with
which was a mid-l 980s proposal for the revision of NZS 4203. the ratio of the frequency between any two modes and show
that once the frequencies of the two modes differ by more that
Gupta and Cordero [Gupta, 1990]. Proposed a modification about 20% the effects of the correlation are insignificant. The
to the DSC rule where p,j is given by only way to estimate the correctness of any modal combination
method would be to compute the response using the response
1 spectrum for a given earthquake and then carry out a elastic
time-history analysis to the same earthquake excitation with the
same excitation orientation and compare these responses with
the maximum responses from the modal combination method.
This would, in the view of the author, be about the only
practical use, in a general seismic design, of elastic time-history
where analyses.
cii = (0.16-0.51)(1.4-lwf-wf 1)2:0.0
In the use of these statistical combination methods. the modal
This method produces greater values of the correlation responses should be the final response required such as the
coefficient when the two frequencies are very close but reduces displacements, the inter-storey drifts, the inter-story shears, the
to the CQC or the Humar rules as the frequency difference base shear, the axial forces, the bending moments or member
increases. Gupta also suggests a modification to all of the stresses. No further combination of responses should be
combination formulae when the damping is different in the undertaken after the modal results are combined. This means
different modes. This effect tends to reduce the correlation that if the longitudinal stress is required in a column then one
coefficients but in most code models all modes have the same should not obtain the axial forces and bending moments on
fractions of critical viscous damping. The Gupta approach is each column, carry out the modal combinations of these actions
not shown in Table 1 as the coefficient c,j is not dependent on and then use the SRSS etc. values of the axial forces and
only the frequency ratio, but has a constant part independent of bending moments to obtain the longitudinal stresses in the
all frequencies. column. Rather, one should obtain the longitudinal stress in
the column for each mode using the modal axial forces and
In all of these modal combination methods, p i tends to 0.0 as
1 bending moments and then carry out the combinations on the
w, becomes very different from wi , i.e. modal longitudinal stresses.
0:::: p,J:::: 1.0
136

Table 1. Modal Correlation Coefficients.

Period CQC DSC DSC DSC Humar

Ratio Seo;= 30 Seo;= 60 Seo; = 120

1.000 1.000 1.000 1.000 1.000 1.000

0.990 0.995 0.998 0.996 0.994 0.990

0.980 0.970 0.992 0.985 0.977 0.961

0.970 0.928 0.983 0.967 0.950 0.915

0.960 0.872 0.970 0.943 0.914 0.857

0.950 0.807 0.953 0.912 0.870 0.792

0.940 0.739 0.932 0.876 0.821 0.723

0.920 0.603 0.882 0.795 0.715 0.590

0.900 0.485 0.882 0.706 0.610 0.474

0.850 0.279 0.653 0.498 0.394 0.276

0.800 0.168 0.491 0.340 0.254 0.168

0.750 0.108 0.360 0.233 0.169 0.109

0.700 0.072 0.261 0.162 0.116 0.074

0.600 0,035 0.138 0.083 0.059 0,038

0.500 0.019 0.074 0.045 0.033 0.022

0.400 0.010 0.041 0.025 0.019 0.013

0.300 0.005 0.022 0.015 0.011 0.009

0.200 0.002 0.012 0.008 0.007 0.006

0.100 0.001 0.006 0.005 0.004 0.004

0.010 0.000 0.003 0.003 0.003 0.003

Whatever response R, is being computed, be it Base Shear, is not a great choice in the available software. A level of skill
overturning Moment, Displacement, Bending Moment, Shear is also required in interpreting the results. A major limitation
force, Axial Force, member stress etc., the quantity must be is that individual earthquake accelerograms have to be used and
found for EACH mode BEFORE the combination method is the choice of appropriate accelerograms is difficult to relate to
used. the design spectra provided in the design codes. These aspects
will be covered in a subsequent paper.

DESIGN CONSIDERATIONS In general, designers prefer to use elastic analyses for seismic
design and apply appropriate modification factors to adjust the
All of the preceding discussion has been related to linearly solutions to allow for the inelastic behaviour of the structures.
elastic structures undergoing small displacements. In most In seismic design, allowances for these inelastic effects are
seismic codes, design to resist earthquake excitation implies a made by modifying the elastic response spectra to get an
degree of ductile behaviour in the structures so that they exhibit inelastic design response spectra.
a tough rather than brittle behaviour under a major earthquake.
Such a ductile structure, at the design level of response, is no Inelastic Design Spectra.
longer linearly elastic and thus the Principle of Superposition
is no longer valid and modal analysis methods are not The inelastic design spectra used by most design codes are
applicable. The inelastic methods of dynamic analysis are generated from the elastic spectra using the Equal Displacement
usually based on deterministic time-history analyses of the and Equal Energy concepts developed over the past three
structure subjected to known ground motions. These methods decades. An inelastic structure subjected to a design level
are costly in data preparation in that a large amount of data is earthquake is assumed to sustain a structural displacement
required and this is further complicated by the fact that alt the ductility of µ where µ is defined as the ratio of the inelastic
data is not available until the design is complete. The analyses displacement L'iu of the structure to its yield displacement ti,Y
are also expensive, the computer models of the structural
members require a level of skill in their use, and as yet, there
 137

The concept of the equal displacement principle appears to date displacement µ times that of the elastic system. For systems
from a paper by Newmark [Newmark, 1960] at the 2nd World with a natural period in an intermediate range the concept of
Conference on Earthquake Engineering, Tokyo, 1960. It was equal energies or equal velocities has become accepted.
observed in a series of inelastic analyses that the inelastic
structures had similar magnitudes of maximum displacement to In the case of equal energies the areas under the elastic and
those of identical structures which were constrained to remain inelastic force displacement plots are equal. This means that the
linearly elastic. Assuming that this concept has some validity, ratio of the inelastic force to the elastic force R is given by
in general, then an elastic structure has approximately the same
displacement L',.e as the A,, of a yielding structure when R = Inelastic Force =

subjected to the same earthquake excitation. If a single degree ElasticForce

of freedom inelastic system sustains a displacement ductility of and the ratio of the inelastic structure displacement to the
µ when subjected to the earthquake excitation, then from elastic structure displacement is given by
Figure 8, the inertia force, or base shear, generated in an elasto-
plastic single degree of freedom system is 1/µ of that in the .6.u µ
matching elastic system.

For systems with very short natural periods, and in particular These three principles have been adopted in the modifications
as the natural periods tends to zero, both inelastic and elastic of elastic response spectra for application to ductile structures.
systems appear to undergo equal accelerations and hence the The application is most readily apparent in the tri-partite
forces generated in the systems are similar. This implies that spectra used in ATC-3 and most other U.S. codes. Figure 10
the inelastic system with a ductility of µ undergoes a is one of the design spectra from NZS 4203 showing how the

Force Force
Elastic Structure Fe Elastic Structure

"'-----..;...__________--4-.,isplacement

_6.e Llu
Figure 8. Equal Displacement Concept. Figure 9. Equal Energy Concept.

1.0
t-
I \
a, I \
;;-e. \
0.8
C
/µ=1
Q)

0
I 1.25
Q)
0
0.6 I
0 I
0
E
(/)

Q)
0.4
U)

0.2

0 1.0 2.0 3.0 4.0

Period (seconds)

Figure 10. Basic seismic hazard acceleration coefficient for intermediate soil sites.
138

elastic design spectra are modified to allow for ductile 10% to 20% when member rigid end-blocks were included.
behaviour. This will be most marked where coupling beams frame into
wide members such as structural walls and columns connect to
spandrel beams. The computational cost of modelling rigid
MODELLING OF THE STRUCTURE end-blocks is negligible and programs should incorporate such
features. In reinforced concrete frames subjected to lateral
Effective Modulus for Reinforced Concrete Structures. forces, very large shear stresses and hence shear strains are
developed in the beam-column joints. In such cases it is best
Experimental work on reinforced and prestressed concrete to remove the presence of end-blocks.
bridges on the West Coast [Moss, 1979] has indicated that the
effective modulus of elasticity of the concrete was somewhat Modelling can be considerably enhanced if the analysis
greater than recommended design values. In order to match the program also considers lateral offsets as well as rigid end-
experimental frequencies of free vibration and mode shapes blocks. In frames with different girder depths in different
over a range of modes, computer studies had to use values spans, the locations of the neutral axes of the girders relative
increased by approximately 20% for the reinforced concrete to the floor level will not be constant. With a lateral off-set the
and approximately 40% for the prestressed concrete when nodes, or frame joints, may be considered to be at floor level
compared with the design values. The model of the bridge and not at the girder mid-depth as is assumed in most current
assumed that the full cross-section was intact and all analyses. In reinforced concrete structures subjected to large
components acted as a unit even though there was some lateral forces and in steel frames with bolted connections, joint,
evidence that there was discontinuity between the precast and or panel, flexibility may also need to be considered.
in-situ concrete. If this discontinuity was also allowed for then
it is implied that the effective moduli may have been even Mass Representation.
greater. The stiffness of concrete does increase with age but
design values are often based on the 28 day strength. For most building structures a lumped mass model, usually
with the whole storey mass lumped at the floor level, is
Cross-Section Properties. generally all that is required. In a two-dimensional analysis the
floor mass provides a single degree of freedom at each floor
The values for the section properties used during the analysis level. For a three-dimensional analysis, this mass acting in two
should represent the likely properties exhibited by the structure orthogonal horizontal directions together with a rotational
during the earthquake, whilst remaining elastic as the inelastic inertia about a vertical axis, usually through the centre of mass
effects are taken into account in modifying the spectral values on the floor, is adequate provided that the floor diaphragm can
used in the lateral load analysis. This means that to gauge the be considered rigid in its own plane. If the displacements of
inelastic response of concrete structures to rigorous shaking the each floor are not those of the centre of mass, a coupling of the
properties assumed should not be those of pristine un-cracked inertia terms will occur. Some of the more recent computer
members but should represent the effects of cracking, averaged programs assume displacement degrees of freedom to be those
out in some way along the member. This means that in of an arbitrary point on the floor plan, though most current
reinforced concrete structures, different properties should be programs seem to require that the displacement degrees of
used in the analyses for a serviceability limit state than those freedom are those of the centre of mass which results in a
which are used in analyses for the ultimate limit state. One diagonal mass matrix. The rotational inertia of the floor is that
should also allow for shear deformations if these are felt to be of the floor rotating as a rigid body about the vertical axis. The
significant, particularly if the beams are short and deep such as expressions for the inertia terms are found in most texts on
in coupling beams of coupled walls, wall members and in dynamics and engineering mechanics [Clough, 1992]. If the
spandrel beams. code eccentricity of mass is used then the rotational inertia of
the floor should be adjusted using the parallel axis theorem.
The analysis should also take into account representative parts NZS 4203 allows the designer to treat the rotational inertia as
of the floor slab in estimating girder flexural stiffness. It must a constant when the centre of mass is shifted.
be stressed that in real structures the effective second moment
of area will vary along the member's length depending on the Foundation Compliance.
amount of cracking and the sign of the curvature at points
along the member. However, this will vary as the direction of Most structures are located on a foundation medium that is not
loading changes. Most current analysis techniques assume that rigid. Not withstanding this fact, at the present time most
an averaging of the second moments of area will be analyses of structures subjected to earthquake excitation are
satisfactory. As will be seen later, the significance of accuracy analyzed assuming that they are located on an infinitely rigid
in predicting stiffnes~ diminishes when inelastic response is to foundation material. An allowance for foundation compliance,
be evaluated. which is covered in many texts on soil-structure interaction or
foundation design, and which may be modelled by the
The commentary to the latest rev1s1on of the New Zealand judicious use of springs and dashpots and/or masses, can make
Concrete Code, NZS 3101, provides some guidance to a significant change in the natural frequencies of free-vibration.
engineers in the selection of effective section properties and This will affect the lateral forces determined from the response
Table 2 presents some of these recommendations. spectra. The effect is likely to most marked in the fundamental
modes which make the major contribution to the inertial forces
End Blocks. and also to the lateral displacements of the structure. There are
research investigations under-way to study the base-isolation
Assumptions in respect of rigid, or semi-rigid, end-blocks of effects that may be available to the structure from its flexible
members can have a significant effect on the stiffness of the foundations [Ahn and Gould, 1990).
frame, its natural frequencies of free vibration and on the
response of the structure to dynamic excitation. Some analyses
have indicated a decrease in the natural periods of the order of
 139

Table 2. Effective Section Properties.

Type of member Ultimate Limit Serviceability Limit State

State
µ=1.25 µ=3 µ=6

I.Beams
(i) Rectangular 0.40 lg lg 0.70 lg 0.40 lg
beams
(ii) T, L beams 0.35 lg lg 0.60 lg 0.35 18
2.Columns
(i) P ) 0.5 f~g 0.80 lg lg 0.90 lg 0.80 lg
(ii) P = 0.2 f0 A8 0.60 lg lg 0.80 18 0.60 lg
(iii) P = -0.05 f~g 0.40 lg lg 0.70 lg 0.40 lg
3.Walls
(i) P = 0.2 f0 Ag 0.45 lg ; 0.80 A8 lg; Ag 0.70 lg ; 0.90 Ag 0.45 lg ; 0.80 Ag
(ii) P = 0.0 0.25 lg ; 0.50 Ag lg ; Ag 0.50 lg ; 0.75 Ag 0.25 1g ; 0.50· Ag
(iii) P = -0.1 f~g 0.15 18 ; 0.30 A8 lg; A8 0.40 lg ; 0.65 Ag 0.15 lg; 0.30 Ag
4.Coupling beams
..
(i) Diagonally 0.401. 1. 0.101. 0.401.
reinforced
1.7 + 2.7(-i':)' 1.7 + 1.3(~)2 1.7 + 2.7(~) 2 1.7 + 2.7(~)2
L L L

(ii) Conventionally 0.401. 1, 0.701, 0.4018

reinforced
1 + 8(~)2 I+ 5(~)' 1 + 8(~)2 1 + 8(~)'
L L L L

•• The effects of shear deformations and strain penetration into walls along beam bars have been included.

Soil-Structure Interaction. the half-space of the foundation. Simple fixed boundary

conditions applied to a finite element mesh act as reflective
This section considers the foundation medium that, in general, boundaries and special techniques, such as the "silent"
lies below the structural foundation that may be considered as boundary, usually using dashpot members, may be needed to
part of the structure proper. In most cases it is the soil medium represent the real semi-infinite nature of the real domain.
between the structure and bed-rock. Much work has recently been done on such problems in both
the frequency and the time domains [Wolf, 1985, Zhao, 1990,
Soil-structure interaction may be considered in one of several Wolf, 1988].
ways, the approach and the required accuracy depending on the
relative lateral stiffnesses of the structure and its supporting In some cases the problem may be modelled by a multi-step
soils. solution [Newmark, 1971]. The foundation is modelled by
finite element or simpler methods depending on whether or not
The simplest approach is to ignore the phenomenon altogether. it can be represented adequately as a one, two or three
This is reasonable if the foundation material is stiff relative to dimensional problem, to obtain the response of the free surface
the structure. This can be thought as being when the free-field without the structure being present. This modifies the input
vibration periods of the soil are short when compared with the motion for the effects of the underlying soil. The modified
lateral vibration natural periods of the structure. input can now be applied to the analysis of the structure which
should account for any flexibility, or foundation compliance,
If the foundation is complex and if the structure is of similar that the foundation may provide for the structure. This may be
dimensions to the foundation, then the most appropriate method done by a finite element or boundary element model [Zhao,
is to model the structure and the foundation together using 1990] or by impedance coefficients such as those suggested in
finite element methods to model the continuum of the [Newmark, 1971, Table 8.1] or [Idriss, 1980, Tables 1 and 2].
foundation. [Newmark, 1971, Idriss, 1980]. The foundation These allow for the flexible soil effects that may lead to a
model in these cases should somehow represent the effects of significant lengthening of the natural periods of the structure
radiation damping, the loss of energy which radiates out into and which has been shown to have a major effect on the
response of structures on soft foundations [Zhao, 1990].
140

Research by Zhao [! 990] has shown that if the fundamental shift of the centres of rigidity from those locations computed
period of the structure is less than the fundamental period of if one assumed that the base of the structure was level.
the site, ignoring the soil-structure interaction may sometimes
be dangerous while if the fundamental period of the structure Torsion.
is longer than the fundamental period of the site, the soil-
structure effects would reduce the response of the structure Torsion is considered in most analyses to be caused by the
even without the effects of radiation damping. If the eccentricity of the mass with respect to the centres of rigidity
fundamental periods of the structure and the site are similar, in each storey.. This may be due to eccentricity of the mass,
the displacement response of the structure relative to the free as the actual distribution of which at any particular time in the
field responses are generally very large. If foundation structure's life is unknown, or due to eccentricity in the
compliance results in significant rocking of the structure, the stiffness or strength properties of the members. Even though
result may be the inter-structure pounding observed in some the members may be meant to be symmetric in their stiffnesses,
earthquakes. It has been found that in most cases, the period the variability in yield properties of ductile structures or of the
shift is the more important factor affecting the structural material stiffnesses may mean that some eccentricity is likely
response than is the energy dissipation by plastic deformations under an earthquake excitation. Because of such uncertainties,
in the structure and radiation of energy into the flexible soil. most codes require the designer to consider some assumed
Radiation damping is only significant when the natural torsion as being present in the structure. This is usually
frequency of the soil-structure system is greater than the natural achieved by re-positioning the mass.
frequency of the site itself.
Most computer programs do not consider the possibility of a
It has been suggested that soil-structure interaction is not torsional input ground motion. From a consideration of
necessary for small to medium size buildings, or buildings that continuum mechanics, the concept of a torsional input ground
are built on very rigid soils (soils with a shear wave velocity motion is difficult to define but may be considered as the result
greater than 700 mis). Soil-structure interaction is necessary of a travelling wave moving across the width of a structure
for tall and/or massive buildings (e.g. smoke stacks, nuclear with a reasonably wide foundation. In this case the effect of
power plants, offshore platforms), or buildings that are built on the torsion is to decrease the mean translational effects of the
soft soils (soils with a shear wave velocity less than 200 mis). earthquake as not all the width of the structure is experiencing
SEASOC [1977] suggest that if the shear wave velocity divided the same maximum ground translation at the same instant of
by the product of the structure's fixed base fundamental natural time but the eccentrically applied translational accelerations
period and the height to the centroid of the inertia forces in the will increase the response of the torsional modes. This
fixed base first mode is less than 20 then soil-structure torsional response may well increase the translational shear
interaction studies should be carried out. Further suggestions deformations in the perimeter members of the structure. It
on the modification to natural periods, inertia forces and must be noted that in frames where torsional actions are being
displacements may be found in Chapter 6 of ATC-3 [ATC, considered, the torsional and translational modes will not be
1978]. distinct unless the building has two axes of symmetry in plan
and the centres of mass coincide with the shear centres in each
The NZS 4203 allows for the effect of the modification of the storey. Even in the case of such a structure, codes require that
excitation by the soil foundation by providing a series of design the centres of mass be moved to an eccentric position to allow
spectra for hard, intermediate and soft soils. If these spectra for accidental torsional effects.
are being used, the designer must still consider what degree of
foundation compliance should be taken into account in Orientation of Seismic Motion.
modelling of the structure.
The designer should consider two orthogonal horizontal
One of the greatest difficulties in allowing for the full soil- excitations. If thought necessary because of the presence of
structure interaction model including the structure and, say, a large beam spans, prestressed concrete girders or long
finite element model of the foundation medium, is that the cantilevers, a vertical excitation should also be considered. The
required input motion is that of the bed-rock at the bottom of most significant response, or the major torsional response in the
the foundation medium. Most earthquake accelerograms have structure may not be exhibited along the convenient horizontal
been recorded at the ground surface which is on top of a soil x or y axes of the structure, particularly if the structure is
medium which will have modified in input bed-rock motion. asymmetric. The orthogonal horizontal axes should chosen to
The question now is, what is the bed-rock motion? The be the major and minor principal axes of the lateral stiffnesses
difficulty in trying to derive the bed-rock motion is that under of the structure, in plan.
strong lateral shaking the soil medium is likely to behave in a
markedly non-linear manner. This difficulty in determining an Concurrent Excitation.
appropriate bed-rock motion is most likely to occur in
deterministic non-linear dynamic analyses. Where designers Most codes at the present time do not appear to consider
are using response spectral methods the effects are usually concurrency of orthogonal earthquake excitation. Most
accounted for by developing different spectra for hard, consider a uni-directional excitation in any direction and
intermediate and soft soil media as is shown in Figure 4.6. l of require analyses be carried out separately in two orthogonal
NZS 4203. directions. A few codes, such as the Norwegian Offshore
Requirements, require concurrent excitation, using SRSS
Sloping Foundations. combinations of the modal responses from the two orthogonal
inputs. Other codes require the concurrent input from one
Sloping ground may have a major impact on the torsional direction together with 30% to 40% of the response from the
response of the structure to lateral ground excitation acting in orthogonal input. Arguments for the applicability of this
a direction along the slope. The differing flexibilities of the approach may be found in the text by Clough and Penzien
structure on the up-hill and down-hill sides will cause a large [Clough, 1993, Chapt. 26].
 141

It is suggested [Clough, 1993] that if two concurrent inputs are account of inelastic effects. If an inelastic analysis is being
being considered, the strongest component should be applied performed then the inelastic effects should be taken into
along the critical axis of the structure and the second account in the hysteretic modelling and not by equivalent
component should have about 85% of the strength of the viscous damping effects.
strongest component of the excitation. SEASOC [1977] gives
the formula of 100% in one direction with 40% in the It must be noted that certain types of structure, such as bare
orthogonal directions. steel frames, prestressed bridges etc. may exhibit very low
levels of viscous damping. As ductile behaviour is being
The effects of concurrent loadings will be most significance in accounted for in other ways, the damping must not be adjusted
the design of columns and joints and will have a smaller effect to account for energy dissipation by plastic work.
on the design of beams unless the ensuing torsion will affect
the flexural strength of the beams. Mode Shapes for Inelastic Structures.

Damping. One of the most significant shortcomings of the modal methods

of analysis is that the mode shapes of linear structures are not
It must be noted that the use of viscous damping is a typical of the deflected shapes of inelastic structures. In elastic
mathematical convenience. The damping behaviour of real structures, the higher modes have a rapidly diminishing
structures is not well understood and the damping mechanism contribution to the displaced shape as the mode number
is not known. Damping forces in structures probably follow a increases. For inelastic structures, the higher mode effects
hysteretic pattern, i.e. they are functions of the displacements become more significant. The first, or fundamental, natural
of the members and not velocities as is implied by viscous period of free vibration lengthens and as a result, the
damping. There is also damping associated with the radiation fundamental mode may have a smaller contribution to the
of energy away from the structure into the foundation soils. To deflection and the deflected shape of the structure. In an
model damping correctly requires a non-linear material analysis elastic structure with reasonable uniform masses and stiffness
with more elaborate member models than are currently distributions up the height of the structure, the maximum
available. There is also an enormous amount of research displacement profile is nearly linear with height. If the structure
required to obtain the correct hysteretic models. All structures shows a ductile response, then the greatest inelastic
show a damped response to dynamic excitation. The amounts deformations occur in the lower levels on the structure
of damping are generally small and can thus be modelled resulting in greater inter-storey drifts in these storeys. The
reasonably well by a viscous damping model. inelastic response in the lower storeys limits the shear forces
being transferred to the upper storeys and as a result, the
Damping in the structure is often assumed to be 5% of critical maximum displacement profile becomes more parabolic in
viscous damping in all modes. This is implicit in the response shape in that the inter-storey drifts near the top of the structure
spectra provided in the NZS 4203 which provides only the 5% remain small while those in the lower storeys increase and may
damped curve. If damping values other than the assumed 5% approach twice the average inter-storey drift. The result is that
is more appropriate for the structure then the spectral values the modal participation factors, and in particular that of the
need to be adjusted for the expected level of viscous damping. fundamental mode, increase as the level of structural
The proposed Los Angeles code of 1972 gave tables of displacement ductility increases.
multiplying factors for Spectral Displacement, Pseudo Spectral
Velocity and Spectral Acceleration for different damping levels. For this reason, there is some justifiable concern over the
Some of the modal combination correlation factors are also precise application of modal analysis techniques to ductile
functions of the amount of damping in each mode. structures. The implied structural behaviour does not fit the
linear elastic and small deflection assumptions implicit in the
Table 3 shows the reduction factors to be applied to the modal analysis method.
tripartite undamped response spectrum to get the damped
response spectra. [Los Angeles, 1972] Lack of Equilibrium following modal combinations.

Table 4 [Berrill, 1980] gives multiplicative factors for spectra In a modal analysis, the forces in the frame associated with
for 2% and 10% damping when the supplied spectra is for 5% each mode shape are in equilibrium. However, when the forces
of critical viscous damping. This would enable a designer to or stresses etc. are combined by one of the modal combination
generate spectra for more or less fractions of critical damping methods to find the most likely maximum forces etc., the
than that provided by NZS 4203 if it is felt that the implied 5% apparent equilibriwn is lost. The combinations, such as SRSS,
is not appropriate for the structure under consideration. lose all sense of sign and hence the concept of which forces go
with other forces to make up equilibrium. As far as the
The values for the fractions of critical damping in structures analysis is concerned, this causes no difficulty. The results give
suggested in some design tables have been increased to account the most likely response in each member of the structure,
for the reduction in the response due to inelastic behaviour in which is what they are meant to do. This does, however, lead
the structure. If the design response spectra have already been to some concern in capacity design strategies where an implicit
modified to account for this inelastic behaviour, then the use of equilibrium is used to provide protection for columns. In a
these higher damping values would doubly account for the capacity design, the girder moments . from the modal
effects of the inelastic behaviour on the response. combinations are factored up and the members designed. Once
the over-strengths of the girders are known then the total over-
Table 5 [SEASOC, 1977] gives typical damping values for strength moment must be distributed to the columns above and
various types of structure for varying levels of displacement. below the joint in some accepted manner. Traditionally,
It should be noted that these fractions of critical damping seem following a equivalent static analysis the joint moments are in
to be applicable for carrying out elastic analyses and engineers equilibrium and the girder over-strength moments are
are using the damping to reduce the response in a way to take distributed to the columns using the same proportions as
142

Table 3. Spectrum Multipliers

Damping% Acceleration Velocity Displacement

0 1.00 1.00 1.00

0.5 0.91 0.90 0.88

1 0.81 0.80 0.80

2 0.67 0.70 0.72

5 0.41 0.48 0.56

7 0.30 0.38 0.48

10 0.23 0.33 0.44

15 0.20 0.30 0.41

20 0.19 0.28 0.40

Table 4. Multipliers for Differing % Damping

Percentage Critical Damping Multiplier

2 1.4

5 1.0

10 0.8

Table 5. Typical Values of Damping.

Stress Level Type of Structure % Critical Damping

Low, much less than 0.25 Steel,Concrete, 0.5-1.0

yield Wood, No cracking or joint
slip

Working Stress, less Welded Steel 2.0

than 0.5 yield Prestressed Cone.
Well Rein. Cone. 3.0-5.0
Cracked R.C.
Bolted/Rivetted Steel 5.0 7.0
Nailed/Bolted Timber 5.0-7.0

At or just below yield Welded Steel 5.0

Concrete (PC or RC) 7.0-10.0
Bolted/Rivetted
Steel,Bolted Wood 10.0-15.0
Nailed Wood 15.0-20.0

Beyond yield. Welded Steel 7.0-10.0

Permanent strain > Concrete 10.0-15.0
yield strain Bolted/rivetted steel and 20.0
wood

Rocking of Entire Structure Rock v, > 1800 mis 2.0-5.0

Firm v,?. 600 mis 5.0-7.0
Soft v, < 600 mis 7.0-10.0
 - - - - - - - - - ~ - ...,..._·""· . . . . .. --~.
~

143

obtained from the static analysis. Following a modal analysis considered as rigid in their own planes, particularly if the
where the joint no longer shows a moment equilibrium, the laterally stiff members are well separated from each other. The
girder over-strength moments could be distributed to the different deformation characteristics of wall and frame
columns above and below the joint in the same ratio as the structures when subjected to lateral excitation may result in
column moments from the modal analysis. The column very large shear forces in the floor diaphragm between the
moments from the modal analysis are not required, just the walls and frames.
ratio between them.
Parts and Portions.
Base Isolation.
Generally, response spectra should be produced for the section
Until recently, base isolation has only been used in design of the structure where the part or portion is to be located.
following lengthy deterministic time-history analyses because This, in theory, may be obtained by locating a single degree of
there were no simple reliable design methods available that freedom oscillator in that part of the structure and subjecting
would predict the distribution of forces through the structure the structure to the design ground motion. This analysis is then
when it was subjected to a lateral earthquake excitation. repeated as the natural frequency of the oscillator is varied.
Recent work [Andriono and Carr, 1991], has led to the This is similar to getting the response spectra at the ground
proposal of a relatively simple design method for base-isolated surface when the overlying soil medium is subjected to a bed-
buildings. Further comprehensive discussion on base isolation rock motion. Should the part have a natural period similar to
techniques and design methods may be found in the recent that of the structure, then a considerably enhanced response
book by Skinner et al [1993]. will be observed. It is for this reason that building codes apply
multipliers to the design spectra when the design of parts and
Rocking Structures. portions are being considered.

The effects of base rocking is very similar to those effects • .oo_ _ _ _~ - - - - - - - . . , - - - - - - . . , - - - - - - ,

caused by flexible foundation compliance except that in the
case of rocking structures, a considerable increase of potential
energy is made as the structure uplifts as it rocks to one side. Mode 3 Mode2
This is potential energy is released as the structure returns to
the upright position and there may be considerable forces
developed, with an additional high frequency response, as the
structure impacts on its foundation prior to rocking in the other
direction.

Modelling Irregular Structures.

Podium structures should not cause any great difficulty in

structural modelling. There will, however, be a discontinuity in
the mass distribution and in the stiffnesses. In some structures,
there may be the consideration given as to whether the floor
diaphragm should be considered rigid in the neighbourhood of
the tower where it meets the podium. A major difficulty will
Natural period (Seconds)
be experienced in attempting to reach a large fraction of the
total mass participating in the modes considered as most of the
.mass will be near the base of the structure and will not strongly Figure 11. Acceleration Response Spectra at Floor 12.
participate in the lower modes of free vibration. The difficulty
is that the large mass of the podium undergoes only small
displacements in the lower modes of free-vibration of the Figure 11 shows the acceleration response spectra obtained
structure as these lower modes reflect the greater flexibility of from the total acceleration history for the top floor of a twelve
the upper part of the structure. To achieve a 90% participation storey frame subjected to the El Centro excitation shown in
of the total mass may require a large number of modes to be Figure 2. This spectra could be used for the design of a part to
considered. It may be more appropriate to relax the mass be located at the top of the frame and should be compared with
participation rule in these cases. the equivalent design spectra for the frame itself if it is to be
subjected to the same motion and which is shown in Figure 4.
Structures which have unusual floor plans where the floor The results of the displacements and forces in a part based on
cannot be regarded as rigid in its own plane will cause this spectra are almost identical with those of an analysis where
difficulty in programs like ETABS and may require modelling the part was included in the structural model of the frame.
in more general finite element or space frame programs to
allow for the flexibility of the floor slab in its own plane.
EXTENDED MODAL ANALYSIS FEATURES.
Sloping ground will cause the greatest difficulty in modelling,
especially in programs such as ETABS. Ritz Vectors.

Hybrid structures should cause no real difficulty provided the Ritz vectors have been used by engineers for a long time in
programs used have the ability to model shear deformations, dynamic and stability problems and they are the basis of the
allow for rigid end-blocks, or have a suitable library of Rayleigh-Ritz method of reducing the number of degrees of
appropriate finite elements available. Again, consideration may freedom in a dynamic analysis. Traditional Ritz methods
be required as to whether the floor diaphragms may be attempted to use a set of approximate mode shapes· as Ritz
144

vectors and these methods, though satisfactory for a general { xJ/[MJ { x}i
dynamic solution, may use many vectors that are orthogonal to c. =
J
the response and thus have no contribution to the solution. In {xJ/[MJ{x)i
a more recent approach [Wilson et al, 1982] the vectors are The vector {x}; may now be normalized as for the earlier
recursively computed from the shape of the applied force vectors. Suppose that the M Ritz vectors have been generated
vector. This means that all computed vectors contribute to the as above, then the vectors may be arranged in the N by M
response and have a better chance of forming a set of mode matrix [X] where every column of the matrix is a Ritz vector.
shapes that make a significant contribution to the response of
the structure. The disadvantage is that a new set of Ritz The N degree of freedom equation of dynamic equilibrium
vectors has to be computed for each direction of excitation.
The eigenvalue problem that has to be solved is of the order of [M]{xl + [C]fxl + [K](xJ = {p(t)J
the number of Ritz vectors used. These mode shapes have to
may be modified with the Ritz approximation
be computed as the Ritz vectors do not have the property of
orthogonality whereas the mode shapes of free-vibration do {x} = [X]{y}
possess the property of orthogonality with respect to the mass
and stiffness matrices. Without the use of the property of where {y} is a vector of generalized coordinates.
orthogonality, response spectra and modal methods cannot be
used. The response spectra are computed for single-dcgree-of- The transformed equation may be written as
freedom systems and each natural mode of free vibration
uncouples the multi-degree-of-freedom structural system into [M']{y} + [C'] {y} + [K']ly} = {p'(t)}
an equivalent one-degree-of-freedom system. where
If the normal modes are used for dynamic analysis, then some [M '] = [X]r[M] [X]
of the mode shapes may be orthogonal to the excitation and
those modes will have no significant contribution to the [C '] = [X]r[C][X]
response. This sometimes makes it very difficult to determine
[K '] = [X]T[K][X]
how many modes should be included in the analysis. Even
examining the modal participation factors is not always a good I p '(t) I = [ X]T I p(t)}
way of determining the significance of a mode as the forces are
a function of the mode shape times the frequency squared and Note that the M by M matrices [M"], [C'] and [K"] are fully
as the frequencies increase with mode number, the higher populated matrices as the Ritz vectors [X] are not orthogonal
modes are more significant for forces than they are for with respect to the mass and stiffness matrices.
displacements.
If the mode shapes {z} and natural frequencies ro are found
One suggestion made by E.L. Wilson et al. starts with the from the equation of free vibration, we obtain a new set of Ritz
assumption that the excitation can be represented by the vectors [X 0 ] which are orthogonal to the mass and stiffness
product of a spatial vector and a time function so that matrices, and they are assumed to be orthogonal to the
damping matrix. At the same time, they are related to the form
l p(t)J = l fl g(t) of the loading in that none of them will be orthogonal to the
applied load
The first vector {x} 1 in the series of Ritz vectors is then
obtained by solving the equation [X°] = [X] [Z]

[K]{x), = {fl If the transformed equation of equilibrium is now re-computed

with the new Ritz vectors then the matrices [M"], [C'] and [K']
Vector {x} 1 which essentially represents the static response of are now diagonal matrices and the response to each generalized
the structure to the applied forces is then normalised. Wilson degree of freedom may be found by, say, Response Spectra
normalized the vector to be ortho-normal to the mass matrix methods and then transformed back to the original coordinates
but in the example here it may be normalized in any way that {x}.
the user wishes.
Static Correction to Modal Analyses.
Subsequent vectors in the series are members of the following
Krylov series The modal response of a structure in mode i is given by the
equation

.. 2' . z... pi'(t)

y. + ,-..w.y. + w. J = - -
These vectors are obtained from the recurrence relationship 1 I 11 I M.,j.
I

[K] {x\ = [M] {x}i-i where A; is the fraction of critical damping in the ith mode, w;
is the natural circular frequency (radians/second) of the ith
Each vector {xi} is then orthogonalized with respect to the
mode which has a mode shape vector {~;} and where Pt
and
vectors obtained earlier. The pure vector {x}, is obtained from Mt are given by
the following relationship
A'(t) = {<l>liT {p(t)}
i-1
I x}i = {x}i - 1:Cil x}i M/ = l<til? [MJ l<l>li
j=l
where
 145

and the displacement {x} of the structure 1s obtained by displacements or the same rotations about horizontal axes. If
combining all N the modal contributions the structure is modelled as a single three-dimensional frame or
if there are several spatially isolated frames inter-connected
N
only by the floor diaphragms there is no difficulty.
{x} = L l<Pl;Y;
i=aJ
2. The program, at least in its earlier versions, cannot
allow for in-plane flexibility of the floor diaphragms. This will
If only M of the N modes are used then the M modes must cause some modelling problems where long narrow floors are
include all the frequencies contained in {p(t)} that are encountered, or floors having an L or U shape in plan. Similar
significant [Humar,1990). If the frequency W; of mode i is difficulties may be encountered in precast floors with a low in-
much greater than the frequencies found in {p(t)} then the plane shear stiffness. These difficulties will be more
response of that mode can be shown to be basically static. In significant in hybrid structures where the different deformation
the above equation of equilibrium for y, the last term on the characteristics of wall and frame systems can result in large
left hand side will dominate the equation and the first two shear forces being transmitted in the floor diaphragms.
terms will become relatively insignificant. The equation above
is used to solve the first M modes and the remaining values of 3. The program only allows the modelling of vertical
y are found from columns. Some more recent structures have had raking
columns.
Y; =
w/M;' 4. Only horizontal floor systems are readily modelled
The total response can now be obtained but this is not very which causes difficulty with parking structures and with
helpful in that all N mode shapes and frequencies have to be sloping foundations. All columns are assumed to be connected
found to compute the solution. at each floor level causing difficulties where some columns
may extend over more than one storey before again being

{x} = t {<Pl;Y; + i=tl {<Pl;w;~;'

connected at a level.

5. Vertical inertias and rotational inertias about horizontal

axes are not considered. For squat buildings this is not a
The solution {x},..ti, can be defined as problem but the lack of vertical inertial effects will affect the
natural frequencies on tall narrow buildings and hence have an
effect on the lateral response of the structure. Vertical
earthquake excitations are not able to be considered.

The solution {x} can then be rewritten as 6. Time-history analyses can be carried out for linear
elastic structures but these are likely to be useful only in
M M •
confirming the appropriateness of the modal combination
{x} = L l<Pl;Y; + {x}static - L l<P};+ methods used in the response spectra analyses unless one
i=l i-1 W; Mi
wishes to compute acceleration histories at upper floors in
or
order to produce floor spectra.
M M P'.
{x} :E l<Pl;Y; _+ rKr 1 1p1 - :E l<Pl;- 2 '-.
i=I t=I W; M i ACKNOWLEDGEMENTS

where the last two terms are the static correction to the The author wishes to acknowledge the help given by other
response obtained with the first M modes. members of the study group, Dr R.D. Sharpe, Dr B. Davidson
and Mr G. Sidwell, for their deliberations on the points covered
in the paper and to Prof. T. Paulay for his suggestions for
ETABS MODELLING OF STRUCTURES improvement in its presentation.

The ET ABS program is widely used by the engineering

profession in design and in teaching, for the analysis of three REFERENCES
dimensional buildings subject to earthquake excitation and has
become almost the de-facto standard analysis software in New Ahn K. and Gould P. 1990. Aseismic Base Isolation Using
Zealand design offices. There are, however, a few points that Natural Elements in the Interactive Soil-pile-structure
may need consideration by the users of the program when System, Struct. Div Report No.86, Department of Civil
modelling structures. Engineering, Washington Univ. in St. Louis.

The major difficulties with program's structural model are: Andriono T. and Carr A.J. 1991. Reduction and Distribution of
Lateral Seismic Inertia Forces on Base Isolated Multi-
1. The lack of compatibility between inter-connecting Storey Structures. Bull. NZ Nat. Soc. Earthquake Eng
frames. Only the two horizontal displacement degrees of 24(3):225-237.
freedom and the rotation about the vertical axis are forced to
be compatible between all frames at any floor level. A Andriono T. and Carr A.J. 199 I. A Simplified Earthquake
problem arises if, as was implied in the original TABS model, Resistant Design Method for Base Isolated Multi-Storey
the structure is modelled as a series of inter-connecting plane Buildings. Bull. NZ Nat. Soc. Earthquake Eng 24(3):238-
frames which are then required to have the same lateral 250.
displacements at each floor. The result is that the inter-
connecting joints in each frame do not have compatible vertical
146

Applied Technology Council. 1978. ATCJ-06, Tentative NZS 4203: 1992. 1992. Code of Practice for General Structural
Provisions for the Developmenl of Seismic Regulations for Design and Design Loadings for Buildings. Volumes 1,
Buildings. The Code and Volume 2, The Commentary. Standards
New Zealand, Wellington.
Berrill,J.,Priestley,M.J.N. and Chapman,H.E. 1980. Design
Earthquake Loading and Ductility Demand. Bull. NZ Nat. Rosenblueth, E. and Elorduy, J. 1969. Responses of Linear
Soc. Earthquake Eng. 13(3):232-241. Systems to Certain Transient Disturbances, Proc. 4th World
Conf on Earthquake Engineering. Santiago, Chile.
City of Los Angeles. 1972. Proposed Code Amendments, 1972.
City of Los Angeles, Building and Safety Commission. SEAOSC. 1977. Seismic Analysis by Computer. Report by
Electronic Computation Committee, Structural Engineers
Clough, R.W. and Penzien, J. 1993. Dynamics of Structures. Association of Southern California.
2nd Edition, McGraw-Hill, New York. 738p.
Skinner, R.I. 1966. Earthquake-Generated Forces and
Gupta, A.K. 1990. Response Spectrum Method In Seismic Movements in Tall Buildings. Bulletin 166. NZ DSIR,
Analysis and Design of Structures. Blackwell Scientific, Wellington.
Cambridge, Mass., l 70p
Skinner, R.l., Robinson, W.H. and McVerry, G.H. 1993. An
Humar, J.L. 1984. Design for Seismic Torsional Forces. Proc. Introduction to Seismic Isolation. Wiley, Chichester, 354p.
Can. J Civ. Eng. 11:150-163.
Wilson, E.L., Der Kiureghian, A., and Bayo, E.P. 1981. A
Humar, J.L. 1990. Dynamics of Structures. Prentice Hall, Replacement for the SRSS Method in Seismic Analysis.
Englewood Cliffs, New Jersey, ISBN 0-13-222068-7. (See Earthquake Engineering and Structural Dynamics
Chapter 13) 9:187-194.

Idriss et al. 1980. Analysis for Soil-Structure Interaction Effects Wilson E.L., Ming-wu Yuan and Dickens J.M. 1982. Dynamic
for Nuclear Power Plants. Report by Ad-Hoc Group on Analysis by Direct Superposition of Ritz Vectors. J Eq.
Soil Structure Interaction, Structural Division, ASCE. Eng. and Struct. Dyn. 10:813-821.

Lysmer, J. 1978. Analytical Procedures in Soil Dynamics, Wolf, J.P. 1985. Dynamic Soil Struc/ure Interaction. Prentice
Report No. UCB/EERC-78/29, College of Engineering, Hall Inc., Englewood Cliffs, New Jersey.
University of California, Berkeley.
Wolf, J.P. 1988. Soil-Structure Interaction Analysis in the Time
Moss, P.J. and Carr, A.J. 1979. Vibration Tests on the Buller Domain, Prentice-Hall, Englewood Cliffs, New Jersey.
River Bridge at Westport. Research Report 79-16,
Department of Civil Engineering, University of Canterbury, Zhao, X., Seismic Soil-Structure Interaction, Ph.D Thesis,
33p. Department of Civil Engineering, University of Canterbury,
1990.
Newmark, N.M. and Veletsos, A.S. 1960. Effect of Behaviour
of Simple systems to Earthquake Motion. Proc. 2nd World Zhao, X, Carr, A.J. and Moss, P.J., Soil-Structure Interaction
Conf on Earthquake Engineering. Tokyo, 11-18 July 1960 Using Boundary Elements in the Time-Domain, Proc. 11th
895-912. Australasian Conf. on the Mech. of Struct. and Mat., 1989

Nev.'IIlark and Rosenblueth. 1971. Fundamentals ofEarthquake

Engineering. Prentice Hall.