MODAL ANALYSIS OF LINEAR DYNAMIC SYSTEMS:

PHYSICAL INTERPRETATION

By Anil K. Chopra, l Member, ASCE

ABSTRACT: The modal analysis procedure for calculating the dynamic response of classically damped linear
structural systems is reformulated with the objective of developing the physical significance of the elements of
the solution and defining modal contribution factors. The resulting concepts are used to develop an approach
for predicting how the relative contributions of various modes to the response and the number of modes that
should be included in the solution depend on the vibration properties of the system, on the response quantity
of interest, and on (1) the spatial distribution and time variation of applied dynamic forces in the case of force-
excited systems, or (2) the response (or design) spectrum for an earthquake excitation. To illustrate this approach,
numerical results are presented for a five-story frame for a wide range of parameters and a base-isolated building.

INTRODUCTION RESPONSE HISTORY ANALYSIS: REFORMULATION

AND INTERPRETATION
The classical mode superposition method or classical modal
Equations of Motion
analysis method is widely recognized as a powerful method
for calculating the dynamic response of viscously damped A viscously damped system with classical damping and hav-
linear structural systems with classical damping. The method ing N degrees of freedom is considered. The response of the
is attractive because the response of a multi-degree-of-free- system described by the vector u of nodal displacements due
dom (MDF) system is expressed as the superposition of to external forces p(t) is governed by the N differential equa-
modal responses, each modal response is determined from the tions
dynamic analysis of a single-degree-of-freedom (SDF)
system, and these dynamic analyses need to be implemented mu + cti + ku = p(t) (1)
only for those modes with significant contribution to the re- where m, C, and k = mass, damping, and stiffness matrices of
sponse. the system. We will consider a common loading case in which
For analysis of earthquake-excited systems, two versions of the applied forces pj(t)-j = 1,2, ... ,N-have the same time
the procedure are in use: (1) the response history version, in variation p(t), and their spatial distribution is defined by s,
which the modal responses are computed as a function of time independent of time. Thus
and then superposed to obtain the response history of the sys-
tem; and (2) the response spectrum version, in which the peak p(t) = sp(t) (2a)
values of modal responses are determined from the response Eq. (1) also governs the vector u of nodal displacements
spectrum or design spectrum that characterizes the excitation, relative to the moving ground undergoing acceleration iig(t) if
and the peak response of the system is then estimated by ap-
propriately combining the modal peaks. (2b)
While classical modal analysis is described in many books where t = influence vector representing the displacements of
on structural dynamics and is widely applied in structural en- the masses resulting from static application of a unit-ground
gineering practice, it has usually not been formulated in a man- displacement, ug = 1. Thus, the spatial distribution of the ef-
ner that emphasizes the physical significance of the elements fective earthquake forces, (2b) is s = mt.
of the solution. The present paper is intended to be responsive
to this need. Modal Expansion of Displacements and Forces
Its objectives are: (1) to reformulate classical modal analysis
in order to provide a useful physical interpretation of the The natural frequencies W n and natural modes <Pn of vibra-
tion of the system satisfy the following matrix eigenvalue
method; and (2) to develop a methodology for predicting
problem:
how the relative contributions of various modes to the
response and the number of modes necessary in the solution (3)
depend on the vibration properties of the system, on the re-
sponse quantity of interest, and on either the spatial distribu- The displacement u of the system can be expressed as the
tion and time variation of the applied dynamic forces in the superposition of modal contributions un(t):
case of force-excited systems, or on the response (or design) N N
spectrum for an earthquake excitation. The physical interpre- u(t) = 2: un(t) = 2: <\>nqn(t) (4)
n-l n-l
tation of modal analysis presented in this paper is the central
concept utilized extensively in a recent book (Chopra 1995) where qn(t) = modal coordinates.
to analyze and understand the dynamic response of linear sys- A central idea of this formulation is to expand the spatial
tems. distribution s of applied forces or s = mt of the effective earth-
quake forces as
I Johnson Professor of Civil Engineering, 721 Davis Hall, Univ. of
California, Berkeley, CA 94720.
Note. Associate Editor: James M. Nau. Discussion open until October (5)
I, 1996. To extend the closing date one month, a written request must
be filed with the ASCE Manager of Journals. The manuscript for this
paper was submitted for review and possible publication on July 5, 1995. where
This paper is part of the ]ounuJl of Structural Engineering, Vol. 122,
No.5, May, 1996. ASCE, ISSN 0733-9445/96/0005-0517-0527/$4.00 (6a-c)
+ $.50 per page. Paper No. 11065.
JOURNAL OF STRUCTURAL ENGINEERING / MAY 1996/517

Downloaded 30 Nov 2011 to 129.187.254.46. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
 Eq. (6) for the coefficient f n can be derived by premultiplying (14)
both sides of (5) by <1>; and using the orthogonality property
of modes. The contribution of the nth mode to the vector S or Eq. (14), specialized for the displacement response, is equiv-
rna. is alent to (11) [Chopra (1995) page 470].

(7) Total Response

which is independent of how the modes are normalized.
Eq. (5) may be viewed as an expansion of the applied force
distribution s or rna. in terms of inertia force distributions Sn
associated with natural modes (Crandall and McCalley 1961;
Veletsos, unpublished notes, 1977). The expansion of (5) has
the useful property that the force vector snp(t) or -SnUg(t)
produces response only in the nth mode but no response in
any other mode. This implies that the response in the nth mode
is due entirely to the partial force vector snp(t) or -SnUg(t).
Modal Equations
By using (4), (1) can be transformed to a system of un-
coupled equations in modal coordinates:
iin + 2t.wnt/. + w~q. = f.p(t) or -f.uit),
where tn is the damping ratio for the nth mode. The factor f n
that multiplies the force p(t) or ground acceleration ug(t) is
often called a modal participation factor, implying that it is a
measure of the degree to which the nth mode participates in
the response. This is not a useful definition, however, because
f n is not independent of how the mode is normalized, nor a
measure of the contribution of the mode to a response quantity.
Both these drawbacks are overcome by modal contribution
factors that will be defined later.
The solution for the modal coordinate qn (t) is
Combining the response contributions of all the modes gives
the total response:
N N
r(t) = 2: r.(t) = 2: r:A.(t) (15)
n-l "_1
Interpretation of Modal Analysis
In the first phase of the modal analysis procedure, the vi-
bration properties-natural frequencies and modes-of the
structure are computed and the force distribution vector S or
rna. is expanded into its modal components s. Fig. 1 shows
schematically the rest of the analysis procedure for earthquake
excitation to emphasize the underlying concepts; with appro-
n = 1,2, ... , N priate modifications it also applies to force-excited systems.
(8) The contribution of the nth mode to the dynamic response is
obtained by multiplying the results of two analyses: (1) static
analysis of the structure subjected to external forces, s., and
(2) dynamic analysis of the nth-mode SDF system excited by
ground acceleration iig(t). Thus, modal analysis requires static
analysis of the structure for N sets of external forces: s., n =
1,2, ... ,N, and dynamic analysis of N different SDF systems.
Combining the modal responses gives the dynamic response
of the structure.
MODAL CONTRIBUTION FACTORS

q.(t) = f.D.(t) (9) The contribution r. of the nth mode to response quantity r,
(14), can be expressed as
where D.(t) is governed by
(16)
D. + 2t.w.D. + w~Dn =p(t) or -ug(t) (10) st
where r = static value of r due to external forces s, and f.,
This is the equation of motion for the nth-mode SDF system the modal contribution factor for the nth mode, is defined as
(an SDF system with vibration properties-natural frequency (17)
W n and damping ratio tn-of the nth mode of the MDF sys-
tem) excited by ground acceleration ug(t). For force-excited These modal contribution factors have three important prop-
systems the nth-mode SDF system is defined as having unit erties. First, by definition they are dimensionless. Second, they
mass, and its response to the force p(t) is also governed by are independent of how the modes are normalized. This prop-
(10). erty becomes obvious by noting that r:
is the static effect of
Sn' which does not depend on the normalization, and the modal
Modal Responses properties do not enter into r st Third, the sum of the modal
contribution factors over all modes is unity, that is
The contribution of the nth mode to nodal displacements
u(t) is
(18)
(11)

To determine the forces in various structural elements- This important result can be proven by recognizing that S =
beams, columns, walls, etc.-from the displacements un(t), 2 s. [(5) and (7)], which implies that r sr = 2 r~'. Dividing by
st
we define the equivalent static forces associated with the nth- r gives the desired result.
mode response: f n = kun(t). Substituting (11) and using (3)
gives PEAK MODAL RESPONSES
f.(t) = s.A.(t) (12) Consider the displacement Dn(t) and pseudo-acceleration
An(t) of the nth-mode SDF system and define their peak values
where the pseudo-acceleration is defined as (denoted by subscript "0") as the maximum of their absolute
A.(t) = w~Dn(t) (13) values:

The nth-mode contribution r.(t) to any response quantity r(t) D no = maxID.(t)I; A.o = max\A.(t) I (19a,b)
is determined by static analysis of the structure subjected to
external forces fn(t). If r: denotes the modal static response, The corresponding value of rn(t), (16), is
the static value (indicated by superscript "st") of r due to
(20)
external forces s., then
518/ JOURNAL OF STRUCTURAL ENGINEERING / MAY 1996

Downloaded 30 Nov 2011 to 129.187.254.46. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
 Static Analysis of Dynamic Analysis of Modal Contribution to
Mode
Structure SDFSystem Dynamic Response

Forces
s,

I
(," fA""OJ" 1;,

~
_ii,<l)
,,(I) = ,;' A,(I)

::'l~ ::'l~ ::'l~

Forces
S2

7,'7,:z
Il2

7,'7,:z
" fA""~,1;2

~
_ii.(t)
'2(1) = ,~l Az(1)

::'l7,:z

. .
. .
. .
Forces
SN

IlN
"
fAJ"
OJN,1;N
'I'll) = ,~l AI'II)

7,'7,7,; ~~ ~7,7,; _ii.(I)

N
Total response r(1) = ::r(1)
FIG. 1. Conceptual Explanation of Modal Analyele

For force-excited systems we shall rewrite this equation in
terms of a dimensionless dynamic response factor. For the nth-
mode SDF system this factor is R dn = Dno/(Dn.s,)o, where
(Dn.s,)o is the peak value of Dn.s,(t), the static response, Ob-
tained by dropping the Dn and Dn terms in (10), Dn,,,(t) =
p(t)/w~ and its peak value is (Dn.s,)o = Po/w~. Therefore, (20)
becomes
(21)
The quantities r" and Fn depend on the spatial distribution s
of the applied forces but are independent of their time varia-
tion p(t); on the other hand, R dn depends on p(t), but is in-
dependent of s.
For earthquake-excited systems (20) can be rewritten as
(22)
where An == A no is the ordinate A(Tn, 'n) of the pseudo-accel-
eration response (or design) spectrum.
The algebraic sign of rno is the same as that of r',,' = r"Fn
because R dn and An are positive by definition. Although it has
an algebraic sign, rno will be referred to as the peak value of
the contribution of the nth mode to response r or, for brevity,
the peak modal response because it corresponds to the peak
value of Dn(t) or An(t).
EARTHQUAKE RESPONSE SPECTRUM ANALYSIS
The peak value r o of the total response r(t) is estimated by
combining the peak modal responses r no (n = 1, 2, ... , N)
according to the well-known modal combination rules: square-
root-of-sum-of-squares (SRSS) rule or complete quadratic
combination (CQc) rule, as appropriate. The algebraic sign of
rno is relevant in the CQC rule, but inconsequential in the
SRSS rule.
The response spectrum analysis (RSA) is a procedure for
dynamic analysis of a structure subjected to earthquake exci-
tation, but it reduces to a series of static analyses. For each
JOURNAL OF STRUCTURAL ENGINEERING / MAY 1996/519

Downloaded 30 Nov 2011 to 129.187.254.46. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
mode considered, static analysis of the structure subjected to m
forces S. provides the modal static response, r~, which is mul- Flexural rigidity
tiplied by the spectral ordinate A. to obtain the peak model
response r (22).
tlO ,
m ColumnsElc
Beams Elh
HOW MANY MODES TO INCLUDE
m
The response contributions of all the natural modes must be -I::
V)

included if the "exact" value of the structural response to dy- II

namic excitation is desired, but the first few modes can usually -I::
@i
m
provide sufficiently accurate results. The number of modes to V)

be included depends on two factors, modal contribution factor

i. and dynamic response factor R dn or spectral ordinate A., that m
enter into the modal response equations (21) and (22).
If only the first J modes are included, the error in the static
response is
,
e, = 1 - 2: i.
.-1
(23)
2h
For a fixed J the error e, depends on the spatial distribution S FIG. 2. Properties of Uniform Flve-Story Frames
of the applied forces. For any s the error e, will be zero when
all the modes are included (J = N) because of (18), and the
error will be unity when no modes are included (J = 0). Thus, Modal Expansion of Forces
modal analysis can be truncated when le,l, the absolute value
of e" becomes sufficiently small for the response quantity r Consider two different sets of applied forces: p(t) = sap(t)
of interest. and p(t) = SbP(t), where s~ = (0 0 0 0 1) and sr = (0 0 0 -1
The dynamic response factors R dn or spectral ordinates A. 2); note that the resultant force is unity in both cases (Fig. 3).
also influence the relative values of the modal responses and Substituting for m, <1>., and S = Sa in (6) and (7) gives the
hence the number of modes that should be included in the modal contributions S. shown in Fig. 3. The contributions of
analysis. Rdn and A. depend on the shape of the dynamic re- the higher modes, especially the second and third modes, to S
sponse factor plot or on the shape of the earthquake response are larger for Sb than for Sa, suggesting that these modes may
(or design) spectrum respectively, and on the vibration prop- contribute more to the response if the force distribution is Sb
erties T. and ~. of the nth mode. than if it is Sa' We will return to this observation later.

EXAMPLE 1: FORCE-EXCITED, FIVE-STORY SHEAR Modal Contribution Factors

BUILDING
System Considered
To illustrate the concepts developed in the preceding sec-
tions, consider the structure of Fig. 2: a five-story shear build-
ing (Elb = 00, i.e., flexurally rigid floor beams and slabs) with
lumped mass m at each floor, and same story stiffness k =
24Elclh 3 for all stories.
We first study how the modal contribution factors i. depend
on the spatial distribution S of the applied forces. Static anal-
ysis of the building subjected to forces S and S. gives r" and
r~, respectively, and (17) leads to i.; for each S = Sa and Sb
the S. vectors are displayed in Fig. 3. For the base shear and
roof displacement of the building, the modal contribution fac-
tors and their cumulative values considering the first J modes
are presented in Table 1. Consistent with (18), the sum of

0.356 0.301 0.106 0.029

0.327 0.195 0.077

0.250 0.055 0.101

= + 0.327
+ + 0.149
+ 0.093

0.101 0.250 0.272 0.179 0.055

Sa St ~ S3 S4 S5

2 0.385 0.508 0.564 0.135

0.354 0.403 0.746

0.679 0.475
= +0.553 + + 0.569
+ 0.646
0.110 0.739 0.685

Sb SI ~ S3 S4 Ss

FIG. 3. Modal Expansion of Excitation Vectors s. and Sb

520 I JOURNAL OF STRUCTURAL ENGINEERING I MAY 1996

Downloaded 30 Nov 2011 to 129.187.254.46. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
 TABLE 1. Modal and Cumulative Contribution Factors
FORCE DISTRIBUTION, s. FORCE DISTRIBUTION, Sb
Mode Roof Displacement Base Shear Roof Displacement Base Shear
nor J J J J
number
of modes, J asn 2:
n.1
aSn IIbn 2:
n.1
II bn
asn 2:
n.1
aSn IIbn 2:
n.1
IIbn
(1 ) (2) (3) (4) (5) (6) (7) (8) (9)
1 0.880 0.880 1.252 1.252 0.792 0.792 1.353 1.353
2 0.087 0.967 -0.362 0.890 0.123 0.915 -0.612 0.741
3 0.024 0.991 0.159 1.048 0.055 0.970 0.431 1.172
4 0.008 0.998 -0.063 0.985 0.024 0.994 -0.242 0.930
5 0.002 1.000 0.015 1.000 0.006 1.000 0.070 1.000

modal contribution factors over all modes is unity, although Rdn in (21). In Fig. 4 R d for harmonic force of period T is
the convergence mayor may not be monotonic. For the struc- plotted against TnfT for SDF systems with natural period Tn
ture and force distributions considered, the convergence is and two damping ratios: , = 5 and 70%; R d for a half-cycle
monotonic for roof displacement but not for base shear. sine pulse force of duration td is plotted against Tnftd in Fig. 5
The data of Table 1 permit two useful observations pertain- for undamped SDF systems.
ing to relative values of the modal responses: How Rdn , for a given excitation p(t), varies with n depends
on where the natural periods Tn fallon the period scale. In the
1. For a particular spatial distribution of forces, the modal case of impulsive excitation, Fig. 5 shows that Rdn varies over
contribution factors for higher modes are larger for base a narrow range for a wide range of Tn and could have similar
shear than for roof displacement, suggesting that the values for several modes. Thus, several modes would gener-
higher modes contribute more to base shear (and other ally have to be included in modal analysis with their relative
element forces) than to roof displacement (and other response contributions, (21), determined primarily by the rel-
floor displacements). ative values of fn' the modal contribution factors. The same
2. For a particular response quantity, the modal contribution conclusion also applies to highly damped systems subjected to
factors for higher modes are larger for force distribution harmonic force because as seen in Fig. 4, several modes could
Sb than for Sa, suggesting that the higher modes contrib- have similar values of Rdn However, for lightly damped sys-
ute more to a response in the Sb case. Recall that the tems subjected to harmonic excitation, Fig. 4 indicates that Rdn
modal expansions of Sa and Sb had suggested the same is especially large for modes with natural period Tn close to
conclusion.
1 0 r - - - - - - - -........- - - - - - - - - - ,
How many modes should be included in modal analysis?
We first examine how the number of modes required to keep
the error in static response below some selected value is influ-
enced by the spatial distribution S of the applied forces. If the ..-!<
objective is to keep lell < 0.05 (5%) for the base shear, the
data of Table 1 indicate that three modes suffice for the force
distribution Sa, whereas all five modes need to be included in
the case of Sb' For the same accuracy in the roof displacement,
- +---==:..-....
.!
C>
::s
II 1 ::--+---'r-------1
~
two modes suffice for the force distribution Sa, but three modes
are needed in the case of Sb' More modes need to be included
for the force distribution Sb than for Sa because, as mentioned
earlier, the modal contribution factors for higher modes are
larger for Sb than for Sa'
We next examine how the number of modes required is
influenced by the response quantity of interest. If the objective 0.1 '-,-_-'----'---''--'-....L..J...LJ...l-_-'-_..Iol._J.....JL..L..L.l...LJ
0.1 1 10
is to keep lell < 0.05 (5%), three modes need to be included Tn _ Undamped natural period
to determine the base shear for force distribution Sa, whereas T- Forcing period
two modes would suffice for roof displacement. To achieve FIG. 4. Dynamic Respons. Factors for Harmonic Force; t =
the same accuracy for force distribution Sb, all five modes are 5% and 70%
needed for base shear, whereas three modes would suffice for
roof displacement. More modes need to be included for base 2r---------,-----------,
shear than for roof displacement because the modal contribu-
tion factors for higher modes are larger for base shear than for
roof displacement. ..-!<
It is not necessary to repeat the preceding analysis for all 1,
response quantities. Instead, some of the key response quan- ~ 1 ~::::::::::::::....----.:~--__+----~--_l
II
tities, especially those that are likely to be sensitive to higher
modes, should be identified for deciding the number of modes ~
to be included in modal analysis.
o,:---'---'---'-.....................~---'----'---'--'- .........~
Dynamic Response Factor 0.1 1
'[g = Undamped natural period
10

We now study how the modal response contributions de- td Force duration
pend on the time variation of the excitation. The dyamic re- FIG. 5. Dynamic Response Factor for Half-Cycle Sine Pulse
sponse to p(t) is characterized by the dynamic response factor Force; t=O

JOURNAL OF STRUCTURAL ENGINEERING / MAY 1996/521

Downloaded 30 Nov 2011 to 129.187.254.46. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
the forcing period T. These modes would contribute most to
the response and are perhaps the only modes that need to be
included in modal analysis unless the modal contribution fac-
tors f n for these modes are much smaller than for some other
modes.
It is not necessary to compute all the natural periods of a
system having a large number of degrees of freedom in as-
certaining which of the R dn values are significant. Only the
first few natural periods need to be calculated and located on
the plot showing the dynamic response factor. Then the ap- (a) (b) (c)
proximate locations of the higher natural periods become read- FIG. 6. Deflected Shapes: (a) p = OJ (b) P = 118; and (c) p = 00

ily apparent, thus providing sufficient information to estimate

the range of R dn values and to make a preliminary decision on
the modes that may contribute significant response. Precise
5..-----------------...,
values of Rdn can then be calculated for these modes to be
included in modal analysis.
The reduction in computational effort achieved by consid-
ering only the first few modes may not be significant in dy- OIl

namic analysis of systems with a small number of dynamic ..(

degrees of freedom, such as the five-story shear frame consid- c
0
.~
ered here. However, substantial reduction in computation can ...
be achieved for practical complex structures that may require 0
'0
u
hundreds or thousands of degrees of freedom for their ideali- ~
zation. 6
"0
::l 0.1
0
EXAMPLE 2: EARTHQUAKE-EXCITED, FIVE-STORY '"
p..
BUILDINGS
Systems Analyzed
To illustrate the concepts developed in the preceding sec-
tions, consider the structure of Fig. 2, a single-bay, five-story
frame with constant story height = h and bay width = 2h. All
the beams have the same flexural rigidity, Elb , and the column 0.1 1 10
rigidity, EIe. does not vary with height. The building is ide- Natural vibration period Tn- sec
alized as a lumped-mass system with the same mass m at all
FIG. 7. Elastic Pseudo-Acceleration Design Spectrum for
the floor levels. The damping ratio for all five natural vibration
= =
Ground Motions with UI/O 1 g, ulIO 48lnJsec, and ul/O 36 In.; t =
modes is assumed to be 5%.
Only two additional parameters are needed to define the
= =
5% (1 in. 25.4 mm)

system completely: the fundamental natural vibration period TABLE 2. Modal Contribution Factors for Vb and V.
T 1 and the beam-to-column stiffness ratio p, where p = h/4/e
The latter parameter indicates how much the system may be Base Shear Vb Top-Story Shear V.
expected to behave as a frame. For p = 0 the beams impose Mode p=o p = 1/8 p=oo p=o p = 1/8 p=oo
no restraint on joint rotations, and the frame behaves as a (1 ) (2) (3) (4) (5) (6) (7)
flexural beam [Fig. 6(a)]. For p = 00 the beams restrain com- 1 0.679 0.796 0.879 1.38 1.30 1.25
pletely the joint rotations, and the structure behaves as a shear 2 0.206 0.117 0.087 -0.528 -0.441 -0.362
beam with double-curvature bending of the columns in each 3 0.070 0.051 0.024 0.204 0.211 0.159
story [Fig. 6(c)). An intermediate value of p represents a frame 4 0.033 0.026 0.007 -0.080 -0.089 -0.063
in which beams and columns undergo bending deformation 5 0.012 0.009 0.002 0.020 0.023 0.015
with joint rotation [Fig. 6(b)]. As an example for the frame of
Fig. 2, P = 1/8 represents I b = le12, which implies a frame with TABLE 3. Modal Contribution Factors for Mb and Us
columns stiffer than the beams, typical of earthquake-resistant
construction. The parameter p controls the vibration properties Base Overturning
of the frame [Roehl (1971) and Chopra (1995) pages 642- Moment Mb Top-Floor Displacement u.
644]. Mode p=o p = 1/8 p=oo p=o p = 1/8 p= 00

(1 ) (2) (3) (4) (5) (6) (7)

Design Spectrum 1 0.898 0.985 1.030 1.009 1.027 1.030
2 0.Q78 -0.003 -0,035 -0.009 -0.030 -0,035
The earthquake excitation is characterized by the design 3 0.016 0.014 0.006 0.0005 0.003 0.006
spectrum of Fig. 7, multiplied by 0.5, so that it applies to 4 0.006 0.003 -0.001 -0.00005 -0.0005 -0.001
ground motions with a peak ground acceleration UgO = 0.5 g, 5 0.002 -0.001 0.0003 0.000005 0.00007 0.0003
peak ground velocity ugo = 24 in.lsec, and peak ground dis-
placement ugo = 18 in. In the design spectrum shown for 5%
damping, the acceleration-sensitive, velocity-sensitive, and quantity. For four response quantities-base shear Vb, top-
displacement-sensitive regions are identified. story shear Vs , base overturning moment M b , and top-floor
displacement us-the modal contribution factors, are presented
Modal Contribution Factors in Tables 2 and 3 for p = 0, 1/8, and 00; these results are
independent of T!. Consistent with (18), for each response
We first study how the modal contribution factors depend quantity and each p the sum of modal contribution factors over
on the beam-to-column stiffness ratio p and on the response ail modes is unity, although the convergence mayor may not
522/ JOURNAL OF STRUCTURAL ENGINEERING / MAY 1996

Downloaded 30 Nov 2011 to 129.187.254.46. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
be monotonic. For the class of structures considered the con- frames (p = 00) with fundamental natural periods T I = 0.5 and
vergence is monotonic for base shear, but not for the other 3.0 s, respectively, are identified. For the building with T I = 3
three response quantities. The data of Tables 2 and 3 permit s, the An values for the higher modes are larger than A 1 for
three useful observations that have a bearing on relative values the fundamental mode, whereas for the building with T 1 = 0.5
of the modal responses: s, the An (n 2: 2) values are either equal to or smaller than
AI' Thus, the higher-mode response, expressed as a percentage
1. For a fixed value of p and each of the four response of the total response, should be larger for the building with T 1
quantities, the modal contribution factor for the first '1 = 3 s than for the T1 = 0.5 s building. In general, for the
mode is larger than the factors 'n
for the higher modes,
suggesting that the fundamental mode should have the
spectrum selected, as T 1 increases within the velocity- and dis-
placement-sensitive regions of the spectrum, the higher-mode
largest contribution to each of these responses.
2. For a fixed value of p, the absolute values of T. for the 1.5....-----------r-.-r----r--..,---------,
second and higher modes are larger for V~ than for Vb'
and the values for Vb in turn are larger than those for M b
and u~. This observation suggests that the second- and
higher-mode response contributions should be more sig-
nificant for base shear Vb than for the base overturning
moment M b or top-floor displacement u~. Among the 0.5
story shears the higher-mode responses should be more
significant for the fifth-story shear than for the base OIl

shear. ~ a
c:
3. As p decreases, the absolute values of the higher-mode 0
'.0 (b) Tl = 3 sec
contribution factors Tn for V~, Vb' and M b increase (but e p=oo
for minor exceptions), especially in the second mode ~(,)
(Tables 2 and 3). This observation suggests that the (,)
<U
higher-mode contributions to any of these forces should 0
"0
::3
become a larger fraction of the total response as p de- Q.)
0.5
creases and should be largest for a flexural beam with p '"
p..

= O.
a
Influence of T, on Higher-Mode Response
In this section we use the preceding concepts and data to
predict how the modal response contributions depend on the
fundamental natural period T 1 of the structure. For this purpose
we examine the three factors that enter into (22) for the peak
0.51----/
modal response: (I) The static value r" of r is a common factor
in all modal responses and therefore does not influence the
Ts
relative values of the modal responses; (2) as mentioned ear- aL- ~.........J.......J..L-...J...... .......L..~o.L_~__L~.........=

lier, for a fixed p the modal contribution factors '. are inde- 0.01 0.1 10
pendent of T 1 ; and (3) the pseudo-acceleration spectrum or-
dinate An is the only factor in (22) that depends on T1 and Natural vibration period Tn. sec
period ratios T 1IT.; for a fixed p, T1ITn do not depend on T 1 FIG. 8. Natural Perloda and Spectral Ordinatea for Three
This is illustrated for the selected design spectrum in parts (a) Caaea: (a) 7; 0.5 a, p= (b) 7; 3 a, p ="";
and (c) T, 3 a, = =""; =
and (b) of Fig. 8, wherein the natural periods T. of two shear p=O

100 ...,-----------...-------------,----------~loo
(a) p = 0 (b) P = 1/8 (c)p=oo

-.,
c
80 80

~
0.
oS
'"c 60 60
.,8-
'"
.,....
'8 40 40
e
.,.:.
..c
OIl
:E
20 20

0
0.1 10 0.1 10 0.1 10
Fundamental natural period T b sec

FIG. 9. Higher Mode Reaponseln Vb' V., Mb, and u. for Uniform Five-Story Framea for Three Valuea of p

JOURNAL OF STRUCTURAL ENGINEERING / MAY 1996/523

Downloaded 30 Nov 2011 to 129.187.254.46. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
response is expected to become an increasing percentage of of the spectrum, the ratio AnfA I increases (or more precisely,
the total response. does not decrease) with decreasing p, and this trend should
This prediction is confirmed by the results of dynamic anal- lead to increased higher-mode response.
yses. The peak values of the response of a frame with specified This prediction is confirmed by the results of dynamic anal-
T I and p is determined by considering (1) all five modes, and yses in Fig. 9, which demonstrate that for each response quan-
(2) only the first mode. The difference between the two results tity the higher-mode response is least significant for systems
for the peak value is the higher mode response, i.e, the com- behaving like shear beams (p = 00), becomes increasingly sig-
bined response due to all modes higher than the first mode. nificant as p decreases, and is largest for systems deforming
Such analyses were repeated for three values of p - 0, 1/8, like flexural beams (p = 0).
and oo-and many values of T I The higher mode response, The influence of T I and p on the higher mode response has
expressed as a percentage of the total response, is presented been studied previously (Cruz and Chopra 1986), but it was
as a function of T I in Fig. 9 for the four response quantities. not as elegant without the concept of modal contribution fac-
The higher-mode response is negligible for TI in the acceler- tors introduced in this paper.
ation-sensitive region of the spectrum and increases with in-
creasing T I in the velocity- and displacement-sensitive regions. How Many Modes to Include
Such results are useful in evaluating the lateral force provi-
sions in building codes [Chopra (1995) Chapter 21]. We first examine the error e2 in static response if two modes
are included. Table 4 shows this error computed from (23) and
Influence of p on Higher-Mode Response numerical values for the modal contribution factors in Tables
2 and 3. The error e2 is below 0.15 or 15% for the four re-
In this section we predict how the modal response contri- sponse quantities when the first two modes are included. For
butions depend on the beam-to-column stiffness ratio p. For a fixed p the error varies with the response quantity. It is
this purpose we examine the three factors that enter into (22) smaller in the base overturning moment M b relative to the base
for the peak modal response: (1) The static value ,Sf of , is a shear Vb, and in Vb compared to the top-story shear V~. The
common factor in all modal responses and therefore does not error is much smaller for the top-floor displacement u~, and it
influence the relative values of the modal response; (2) as p is less than 3% if the first mode alone is considered. For a
decreases, the absolute values of the higher-mode contribution particular response quantity the error e2 varies with p, being
factors Pn for the base shear and top-story shear increase, es- smallest for p = 00 (i.e., shear beams) and largest for p = 0
pecially in the second mode (Table 2); and (3) the pseudo- (i.e., flexural beams). The top-floor displacement displays
acceleration spectrum ordinates depend on T I and on TlfTn ; trends opposite to the forces in the sense that e2 increases with
the latter becomes larger as p decreases [Roehl (1971) and increasing p, but e2 is so small that higher modes are of little
Chopra (1995) pages 642-644] and the Tn values are spread consequence. These data suggest that the first one or two
out over a wider period range of the design spectrum. This is modes may provide a good approximation to the total re-
illustrated in parts (b) and (c) of Fig. 8. Both frames have the sponse, with the accuracy depending on the response quantity
same T 1 = 3 s, but they differ in p-one is a shear beam (p and on p.
= 00), and the other a flexural beam (p = 0). As a result, the We next examine how the spectral ordinates An influence
ratio A 2 for the second mode- generally the most significant
of the higher modes-to Al for the first mode is larger for TABLE 4. .. =1 - ~~.1 f n
buildings with p = 0 than for the p = 00 case. Thus, putting
these two reasons together, both the modal contribution factor Response p=o p = 1f8 p=oo
(1 ) (2) (3) (4)
Pn and the spectral ordinate An for the second mode are larger
for the p = 0 frame; therefore, the higher-mode response is V, 0.144 0.144 0.110
expected to be more significant in this case than for the frame Vb 0,115 0.086 0.033
Mb 0,024 0.18 0.005
with p = 00, In general, for the design spectrum selected and
u, 0.0004 0,003 0.005
for T I within the velocity- and displacement-sensitive regions

10 10
(a) p = 0 (b) P = 1/8 (c)p=oa

No of modes
5

.~-
......
;:,.""

0,1 0,1

0.1 10 0,1 10 0,1 10

Fundamental natural period T I sec

FIG. 10. Normalized Base Shear In Uniform Flve-5tory Frames for Three Values of p

524/ JOURNAL OF STRUCTURAL ENGINEERING / MAY 1996

Downloaded 30 Nov 2011 to 129.187.254.46. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
the number of modes that should be included in the analysis.
For a fixed p and T) in the velocity- or displacement-sensitive (24a)
regions of the spectrum, the ratio An/A) is larger for frames
with longer fundamental period T) [Fig. 8(a) and (b)]. Thus
for the same desired accuracy, more modes should be included and
in the analysis of buildings with longer T) than the number of
modes necessary for shorter T) buildings. For a fixed T) in the t - Cb (24b)
b - 2(M + mb)wb
velocity-sensitive or displacement-sensitive regions of the
spectrum, the ratio An/A) is larger for frames with smaller p in which M = 5m is the mass of the building excluding the
[Fig. 8(b) and (c)]. Thus, for the same desired accuracy, more base slab. The five-story building on a base isolation system
modes should be included in the analysis of buildings with is a six-degree-of-freedom system with nonclassical damping
smaller p compared to the number of modes necessary for because damping in the isolation system is typically much
buildings with larger p. more than in the building. In this example the base slab mass
These expectations regarding how T) and p influence the mb = m and the stiffness and damping of the isolation system
number of modes that should be included in earthquake re- are such that Tb = 2.0 s and ~b = 10%.
sponse analysis are confirmed by the results of Fig. 10, where, The natural periods and modal damping ratios of both sys-
for each p value, five response curves for base shear are iden- tems are presented in Table 5.
tified by indicating the number of modes included in the re-
sponse spectrum analysis. It is clear that the first two modes Analysis Procedure
provide a reasonably accurate value for the base shear in
frames with T 1 in the velocity-sensitive region of the spectrum, With ground motion characterized by the design spectrum
and one mode is sufficient in the acceleration-sensitive region. of Fig. 7, scaled to agO = 0.5 g, the RSA procedure summarized
This conclusion is also valid for shears in all the stories and earlier will be used to analyze two systems: (1) the building
overturning moments at all floors. The first mode alone pro- on a fixed base; and (2) the same structure supported on an
vides accurate results for Us over the entire range of Th and isolation system. In applying the RSA procedure to the isolated
for all p values, as indicated in Fig. 9.
In light of the preceding observations, it is instructive to No
examine the 90% rule for participating mass specified in the
Uniform Building Code. Because the effective modal mass is
V:
equal to the modal static response n for base shear [Chopra
(1995) pages 476-484], the prior rule implies that enough- m mi
say i-modes should be included so that eJ for base shear is }

less than 10%. However, as noted earlier, eJ varies with the

response quantity, and therefore this error may exceed 10%
for other response quantities such as shears in upper stories
ml ml
and bending moments and shears in some structural elements. 1

mb Isolation
EXAMPLE 3: EARTHQUAKE-EXCITED, BASE- 7J.~ 7,:~
~!'fi. ~" system
ISOLATED BUILDING
(a) (b)
System Considered and Parameters FIG. 11. N-Story Building: (a) Fixed-Base; (b) Isolated

The building to be isolated [Fig. II(a)] is a five-story shear-

frame (i.e., beam-to-column stiffness ratio p = 00) with mass TABLE 5. Natural Periods and Modal Damping Ratios
and stiffness properties uniform over its height: lumped mass
mJ = m = 100 kips/g at each floor, and stiffnesses k for each Fixed-Base Building Isolated Building
story; k is chosen so that the fundamental natural vibration Mode Tnl (s) tnl (%) Mode Tn (s) tn (%)
period TIf = 0.4 s. The damping matrix is proportional to the (1) (2) (3) (4) (5) (6)
stiffness matrix with 2% damping in the fundamental mode. - - - 1 2.030 9.58
As shown in Fig. l1(b), this five-story building is mounted 1 0.400 2.00 2 0.217 5.64
on a base slab of mass mb' supported in tum on a base-iso- 2 0.137 5.84 3 0.114 7.87
lation system with lateral stiffness kb and linear viscous damp- 3 0.087 9.20 4 0.080 10.3
4 0.068 11.8 5 0.066 12.3
ing Cb' Two parameters, Tb and ~b' are introduced to charac- 5 0.059 13.5 6 0.059 13.6
terize the isolation system:

TABLE 6. Calculation of Base Shear In Fixed-Base and Isolated Buildings

Fixed-Base Building Isolated Building
Mode An/g V~~/m Vb/W Mode An/g V:'/m VblW
(1 ) (2) (3) (4) (5) (6) (7) (8)
- - - - 1 0.359 5.028 0.361
1 1.830 4.398 1.609 2 1.291 -0.021 -0.005
2 1.272 0.436 0.111 3 1.058 -0.005 -0.001
3 0.859 0.121 0.021 4 0.792 -0.002 -0.000
4 0.700 0.038 0.005 5 0.682 -0.0005 -0.000
5 0.638 0.008 0.001 6 0.635 -0.0001 -0.000
SRSS - - 1.613 SRSS - - 0.361

JOURNAL OF STRUCTURAL ENGINEERING 1 MAY 1996/525

Downloaded 30 Nov 2011 to 129.187.254.46. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
structure we are ignoring the coupling of modal equations due Effects of Base IsolatIon
to nonclassical damping. The peak response due to the nth
mode of vibration is given by (22), which is specialized for In this section we use the preceding concepts to predict
the influence of base isolation on the base shear in the build-
the base shear Vb in the building and the base displacement or
isolator deformation Ub ing, and to identify precisely the underlying reason for this
effect. The peak modal response is the product of two parts:
(25a,b) the modal static response V~n and the pseudo-acceleration
An. Each part is examined for the first mode of the base-
where D n = Anlw; is the deformation spectrum ordinate. The isolated building and of the fixed-base building; this is the
peak modal responses are combined by the SRSS rule. mode that provides most of the response in each case. Ob-
serve that V~I =5.028m for the isolated building is somewhat
larger than V~I = 4.398m for the fixed-base building. How-
Modal Static Responses ever, the pseudo-acceleration Al = 0.359 g for the isolated
=
building is only one-fifth of Al 1.830 g for the fixed-base
The modal static responses V~n and u~n for Vb and Ub, re- building; as a result, the first-mode base shear coefficient of
spectively, are presented in Tables 6 and 7. It is clear that 36.1 % for the isolated building is much smaller than the
V~n has significant values in the first two modes of the fixed- 160.9% for the fixed-base building. The isolation system re-
base structure, and these modes are expected to contribute sig- duces the base shear primarily because the natural period of
nificantly to the response. However, for the isolated building, the first mode, providing most of the response, is much
V~n is small in all the higher modes and the response in these longer than the fundamental period of the fixed-base struc-
modes should be negligible, i.e., the first mode with the dom- ture, leading to a smaller spectral ordinate, as seen by com-
inant value V~I should provide most of the response. paring Figs. 12 and 13. This is typical of design spectra on
firm ground and fixed-base structures with fundamental nat-
ural period in the flat portion of the acceleration-senstive re-
Peak Modal Responses gion of the spectrum.
This prediction is confirmed by the results of dynamic
The preceding predictions from the modal static responses analysis, considering all modes, in Table 6. These results
are confirmed by the results of dynamic analysis in Tables 6
and 7. The peak value of the earthquake response due to each TABLE 7. Calculation of Isolator Deformation
natural mode of both systems is determined from (25), where
the spectral ordinates Anlg are shown in Figs. 12 and 13. The Mode Dn w~u:' u""(in.)
dynamic response of the isolated building due to all higher (1 ) (2) (3) (4)
modes is negligible. The first mode alone produces essentially 1 14.470 0.971 14.045
the entire response: isolator deformation of 14.045 in. and base 2 0.597 0.022 0.013
shear equal to 36.1 % of W, the 500-kip weight of the building 3 0.133 0.005 0.001
4 0.050 0.002 0.000
excluding the base slab. The first two natural modes provide 5 0.029 0.001 0.000
significant response of the fixed based building; however, 6 0.022 0.0001 0.000
the second mode contributes little to the combined SRSS SRSS - - 14.045
value.

2.0
~=2% 1.830g

00
\0
o
o
0.0
0.02 0.05 0.1 0.2 0.5 2 3
Natural vibration period Tn. sec

FIG. 12. Design Spectrum and Spectral Ordinates for FIXed-Base Flve-5tory Building

526/ JOURNAL OF STRUCTURAL ENGINEERING / MAY 1996

Downloaded 30 Nov 2011 to 129.187.254.46. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
 2.0
1.5
Oll
..:
0::
0
'll 1.0
ju
1;j
6
"0
""
'"
Po.
0.5
0.0
0.02 0.05
FIG. 13. Design Spectrum and Spectral Ordinates for Isolated Flve-8tory Building
demonstrate that base isolation reduces the base shear coef-
ficient from 161.3% to 36.1 %.
CONCLUSIONS
With the reformulation of classic.al modal analysis and phys-
ical insight contributed in this paper, the following conclusions
result:
1. The contribution of the nth natural vibration mode to the
dynamic response of a multi-degree-of-freedom structure to
applied forces p(t) = sp(t) or ground acceleration iig(t) can be
interpreted as the product of the results of two analyses: (1)
static analysis of the structure subjected to external forces Sm
the contribution of the nth mode to spatial distribution s of
p(t) or ml of the effective earthquake forces; and (2) dynamic
analysis of the nth-mode SDF system excited by the force
p(t) or ground acceleration iig(t).
2. The modal contribution factors f n, which depend on s,
are dimensionless and independent of how the modes are nor-
malized; their sum over all modes is unity.
3. The relative contributions of various modes to the dy-
namic response and the number of modes that should be in-
cluded in modal analysis to achieve the desired accuracy can
be determined from numerical values of the modal contribu-
tion factors and of the dynamic response factor for p(t) or the
earthquake response (or design) spectrum for iig(t).
4. Examples have been presented to illustrate that the phys-
ical interpretation of modal analysis and the concept of modal
contribution factors, developed in this paper, are useful in pre-
dicting how the relative contributions of the various modes to
Downloaded 30 Nov 2011 to 129.187.254.46. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
~=2%
5%
1.291g
10%
0.1 0.2 0.5 2 3
Natural vibration period Tn' sec
the response and the number of modes necessary in the solu-
tion depend on the vibration properties of the system, on the
response quantity of interest, and on: (1) the spatial distribu-
tion s and time variation p(t) of the applied dynamic forces;
or (2) the response (or design) spectrum for an earthquake
excitation.
5. Another example has been presented to demonstrate that
the physical interpretation of modal analysis is useful in pre-
dicting that the base shear in a building is reduced due to base
isolation and in identifying precisely the underlying reasons
for this reduction.
ACKNOWLEDGMENTS
This paper is based on Chapters 12, 13, 18, and 20 of a recent book
(Chopra 1995). The author is grateful to Dr. Rakesh K. Goel for imple-
menting the calculations and preparing the figures, and to Dr. Gregory L.
Fenves for valuable discussions.
APPENDIX. REFERENCES
Chopra, A. K. (1995). Dynamics of structures: theory and applications
to earthquake engineering, Prentice Hall, Englewood Cliffs, N.J.
Crandall, S. H., and McCalley, R. B. Jr. (1961). "Numerical methods of
analysis." Shock and vibration handbook, Chapter 28, C. M. Harris
and C. E. Crede, eds., McGraw-Hili, New York, N.Y.
Cruz, E. E, and Chopra, A. K. (1986), "Elastic earthquake response of
building frames." J. Struct. Engrg., ASCE, 112,443-459.
Roehl, J. L. (1971). "Dynamic response of ground-excited building
frames," PhD thesis, Rice Univ., Houston, Tex.
Uniform building code. (1994). Int. Conf. of Build. Officials, Whittier,
Calif.
Veletsos, A.S. (1977), Structural Dynamics, class notes, Univ. of Calif.,
Berkeley.
JOURNAL OF STRUCTURAL ENGINEERING / MAY 1996/527