Mimw Rii>wW8afcmw^wj<*»--ftCT,.

« MaanHMHMMMMiMBH^

construction
engineering TECHNICAL REPORT M-132
June 1975

research Earthquake Effects on Structures

laboratory

MODAL ANALYSIS METHODS IN SEISMIC DESIGN

FOR BUILDINGS

by
William K. Stockdale

/D D C

■ ■■ ■■■
if

Approved for public release; distribution unlimited.

The contents of this report are not to be used for advertising, publication, or
promotional purposes. Citation of trade names does not constitute an
official indorsement or approval of the use of such commercial products.
The findings of this report are not to be construed as an official Department
of the Army position, unless so designated by other authorized documents.

DESTROY THIS REPORT WHEN IT IS NO LONGER NEEDED

DO NOT RETURN IT TO THE ORIGINA TOR
 ^J^^SS^SB^^J^iws^ .

UNCLASSIFIED
SECURITY CLASSIFICATION OF THIS PAGE (Whan Dim Bnlnrnd)

READ INSTRUCTIONS
REPORT DOCUMENTATION PAGE BEFORE COMPLETING FORM
2. GOVT ACCESSION NO 3. RECIPIENT'S CATALOG NUMBER

CERL-TR-M-132J
^
M. T.-r. ' fimirtflflT^ COVERED
jlWDALANALYSIS METHODS IN SElWHSESIGN FOR) FINAL /u^3
r
BUILDINGS# J ^-
\ li PMRF>RMIH« »WCr RBRIWT MUlinm

I
_ ftir^iiT-ifT<
Oil iiTitT) i 8. CONTRACT OR GRANT NUMBERf«)

//^William K./
Wi11iam K./stockdale)

9. PERFORMING ORGANIZATION NAME AND ADDRESS (0. PROGRAM ELEMENT, PROJECT, TASK
»pg. k ■*"""• ■■"IT NUMBERS
CONSTRUCTION ENGINEERING RESEARCH LABORATORY!
P.O. Box 4005
Champaign, IL 61820 |)y/rW671lT^ 02-004
tl. rnu-rant i IU« gmgg >jft|(» MM »OBUMMt

(jjßW^WSp^ ^ AGES

Xt. MONITORING AGENCY NAME ft AODRESSflf dlllumtt from Controllln« Olllc») *ITY CLASS, (of (/■<• raport;

Unclassified
15a. OECLASSIFICATION/DOWNGRADING
SCHEDULE

16. DISTRIBUTION STATEMENT (ot M» Rmpott)

Approved for public release; distribution unlimited.

17. DISTRIBUTION STATEMENT (ol lh» mbmtrmcl mtfnd In Block 20, It dlHannl from Report;

I*. SUPPLEMENTARY NOTES

It. KEY WORDS (ContJnu* on nrmttm ■Id« II n«c«««arr and Idmntlty by block numbmr)
modal analysis
seismic motjions
SEA0C code

20. ABSTRACT CAptlniM an r*nrM aMt II' inocooomrr

21
and Idanllly br block numbor)
TTiis report is the first
first step in preparing a change to the tri-services mttaual TM
5-809-10, Seismic Design far Buildings. Changes in this manual are necessai-y to pro-
vide guidance for the design of critical military facilities which must remain func-
tional after subjection to strong earthquakes. This report describes and discusses
modal analysis methods used in the dynamic analysis of structures in conjunction
with the earthquake response spectra and time history methods. Elastic and inelastic
conditions are discussed, as well as structural damping and assumptions and limita-
tions of the methods. Example calculations are included.-^
FOMH
DO I JAM 7J 1473 EDITION OF t MOV M IS OBSOLETE
£$11
UNCLASSIFIED
n.» W*lm*mtn
mum^^^.^

FOREWORD

This project was conducted for the Directorate of Military Construction, Office of
the Chief of Engineers (OCE), under RDT&E Army Program 6.27.19A, Project
4A"/6719AT05, "Initial Investigation in Military Construction Technology." Task 02,
"Engineering Design Criteria and Technology for Military Facilities," Work Unit
004, "Earthquake Effects on Structures." The OCE Technical Monitor was Mr.
W. A. Heitmann.

This study was conducted between August 1974 and February 1575 by the Struc-
tural Mechanics Branch, Materials Systems and Science Division (MS), Construction
Engineering Reseaich Laboratory (CERL). This report was prepared by COL W. K.
Stockdale, Department of Engineering, U.S. Military Academy, West Point, NY,
while on sabbatical at the University of Illinois, Urbana-Champaign. The work was
accomplished under the supervision of Dr. W. E. Fisher, Chief, Structural Mechanics
Branch.

COL M. D. Remus is Commander and Director of CERL, and Dr. L. R. Shaffer

is Deputy Director. Mr. J. J. Healy is Chief of MS.
 CONTENTS

DD FORM 1473 1
FOREWORD 3
LIST OF FIGURES AND TABLFS 5

1 INTRODUCTION 7
Purpose
Background

2 MODAL ANALYSIS METHODS 10

Description of Modal Analysis Method
Response Spectrum Method
Time History Method
Assumptions and Limitations
Input Motions
Linear Elastic Assumptions
Damping in the Structure
Basic Procedures
Idealization of Structural System and Calculation of Member
Properties and Maases
Assignment of Damping Values
Calculation of Mode Shapes and Frequencies
Number of Modes to be Considered
Calculation of the Response
Accuracy and Applicability of Method
to Various Structural Concepts

3 SUMMARY 20

4 DESIGN RECOMMENDATIONS 21

APPENDIX A: Review of Modal Analysis Theory 22

APPENDIX B: Significance and Use of Response Spectra 30
APPENDIX C: Extenaion of the Modal Analysis Method
to the inelastic Case 33

REFERENCES 35

DISTRIBUTION

3
t&tuntvmm***.

FIGURES

Number Page

1 Comparison of SEAOC Code and Modal Analysis Methods

of Seismic Design 8

2 Basic Design Spectrum Normalized to 1.0 g for 5 Percent Damping 9

3 Typical Vibrating Systems 10

4 Three Principal Mode Shapes of System Represented in Figure 3(b) 11

5 Displacement Response Spectrum for Single-Degree-of-Freedom

Systems Subjected to Earthquake Motion 12

6 Typical Response Spectrum 13

7 Basic Spectrum Normalized to 1.0 g Maximum Acceleration 14

8 Types of Structural Response 17

9 Possible Degrees of Freedom for a Three-Story Structure 18

Al Typical Single-Degree-of-Freedom System 22

A2 Forces Acting on the Mass for Single-Degree-of-Freedom System 22

A3 Plot of Displacement Versus Time for Undamped

Single-Degree-of-Freedom System 23

A4 Plot of Displacement Versus Time for Subcritically Damped

Single-Degree-of-Freedom System 23

A5 Plot of Displacement Versus Time for Supercritically Damped

Single-Degree-of-Freedom System 24

A6 Typical Multi-Degree-of-Freedom System 25

A7 Forces Acting on Masses for Three-Degree-of-Freedom System 25

A8 Example Three-Degree-of-Freedom System 26

A9 Principal Mode Shapes for Example System 27

81 Basic Design Spectra Normalized to 1.0 g 31

Cl Force-Displacement Relationship for Elastoplastic System 33

 m*w&*^~m*e^im**v**~&*mum«.nwKt».>

FIGURES (Cont.)

Number Page

C2 Design Spectra ,14

C3 Design Speclrum tor Elastoplastic System With ^ = 1.5 36

TABLES

1 Recommended Damping Values 16

2 Shear (in Kips) for 10-Story Hospital Subjected to Typical Earthquake 20

\
MODAL ANALYSIS METHODS IN SEISMIC
DESIGN FOR BUILDINGS

. code is Contained in the formula for the equivalent

base shear (V) given by
1 INTRODUCTION
V=ZKCW lEqU
Purpose
where
The purposes of this report are (1) to summarize
modal analysis techniques that can be used to assist Z = a coefficient dependent upon the relative in-
in the analysis of the dynamic response of buildings tensity of the ground motion at the site of the
subjected to seismic motions, and (2) to develop pre- structure
liminary recommendations for implementing modal K = a coefficient recognizing the effect of ductil-
analysis techniques to supplement the equivalent ity and energy absorption qualities of certain
static force method specified in the current version of types of construction which have historically
TM 5-809-10.» shown varying degrees of earthquake resis-
tance
Background C = a coefficient recognizing the effect of the
period of the structure on the response to the
ground motions
The seismic provisions of the Structural Engineers
W = the total weight of the structure.
Association of California (SEAOC) Code adopted in
TM 5-809-10 are designed to insure incorporation of
Furthermore, the SEAOC code distributes the base
acceptable minimum strengths into a structure.
shear vertkally along the height of the structure in a
These minima have been established by a consensus
linear manner that approximates the inertial loading
of engineers and building department officials who
imposed on the structure when it responds in its
have been guided by observations and calculations
fundamental mode of vibration. The largest force is
regarding the performance of various types of
applied at the top of the structure, with the force
structures.
decreasing to zero at the base (Figure 1(a)). This is
accomplished by use of the formula
Deficiencies in the code are revealed by unsatis-
factory performance of structures in past earth-
quakes, and improvements are brought about by (V-Ft)wxh,
Fx = (Eq2|
experience. For most aspects of engineering design,
experience is a rapid teacher and feedback is quick. Z wjhi
However, th-- relatively infrequent occurrence of i=l
strong and destructive earthquakes means that
experience is relatively slow in focusing attention on where Fx = force applied at any floor x (including
deficiencies in seismic design and construction. top)
When experience does occur, it may be too late to Fj = an extra force applied to the top of the
prevent significant loss of life and destruction of
structure = .004V (f^)»:=0 if 1J ^ <3
property. L's s
hn = height in feet to top of structure
The seismic design provisions specified in the Ds = dimension in feet of structure at base in
SEAOC code are based primarily on the first mode direction being analyzed
response of the structure; they substitute a set of hj = height in feet to any floor i (including
equivalent static lateral forces for the true dynamic top)
forces imposed on the structure by the seismic mo- WJ = weight in kips of any floor i (including
tion (Figure 1(a)). The basic concept of the SEAOC top).

From an economic point of view, the amount of

1
Seismic DeagH tor Buildings. TM 5-8(W-l() NAVFAC P-.V* investment that should be made in the seismic design
AFM 88-3, Chapter 13 (Departments of the Army. Navy, and Air of a structure is limited; that is. there is an optimum
Force. April IW). point beyond which the extra cost of the design effort
 wnKrew^WH^W'ir-'^V^vWffiaSMKMaB

-SS

7^0 y&F?
Lateral Forces Displacement
a) SEAOC Code
I
f. = Frequency
Xj = Mode Shape
Fj = Participation
Factor In
fr^W" r^Nw*
Total Vibration
Model Mode 2 Mode 3

b) Modal Analysis

Dis 2
X|F|U| X2F2U2 X3F3U3
p- VicxiF^)
.=1 ' ' '

c) Modal Analysis With Response Spectrum Method

"\

)=»Dlsp. = Z Fy.irrv(t)=
i:|
fprn*
/U^^. (WHILE SHOWN FOR THE TOP ONLY THIS
^irrf yW**9 J METHOD MAY BE APPLIED AT ALL LEVELS.)

d) Modal Analysis With Time History Method

Figure 1. Comparison of SEAOC code and modal analysis methods of seismic design.
exceeds the resulting increase in the structure's value notable have been the design standards adopted b>
oi the reduction in construction costs. The equiva- the Nuclear Regulatory Commission (NRC, formerly
lent static force method, with some adjustment in AEC). These standards have tended to specify the
the various coefficients comprising the method, seismic hazard in terms of a basic design spectrum,
approaches the optimum method for the majority of which is commonly normalized to a maximum hori-
conventional, non-critical, low-rise construction. zontal ground acceleration of 1.0 g, but is capable of
However, for critical military facilities which are re- being scaled to other acceleration levels to satisfy site
quired to support relief and rescue operations and conditions (Figure 2). Thus, these standards encour-
retain military defense integrity immediately after an age using response spectrum modal analysis tech-
earthquake, the use of equivalent static force niques for analyzing the dynamic response of a pro-
methods is not always adequate to insure surviv- posed or existing structure. The alternate approach
ability. For critical facilities, more reliable and re- is to use a family of actual past or artificial earth-
fined methods of design must be implemtnted to quake records scaled to specific parameters estab-
eliminate the uncertainty associated with the current lished during a site-dependent investigation to
code. perform a time history modal analysis to compute
the transient response of the structure.
Primarily as a result of the emphasis on safe
seismic design of nuclear reactors and the 1971 San Response spectrum and time history modal
Fernando earthquake, various improved procedures analysis methods are not new, but they have been
for the seismic design of facilities have recently been made practical with the availability of large capacity
proposed, and some have been implemented. Most computers and a number of general purpose com-

I
0.1 02 Q5 I 2 5 10 20 50 100
Frequency, cps

Figure 2. Basic design spectrum normalized to 1.0 g for 5 percent damping. From N. M.
Newmark and W. J. Hall, "Procedures and Criteria for Earthquake Resistant Design," Build-
ing Practices for Disaster Mitigation (Department of Commerce, February 1973).
puter programs2 developed for dynamic analysis of SQ2Z2ZZ2
structures (Figures Kb, c, and d)). Both methods a)
provide better understanding of the nature of the
dynamic response of a structure and a more practical
Planar Structure Idealized Model
method of analysis and design. Other time history Strucluie
methods may be used for analysis of more compli-
cated structures than those considered in this report.

J,,,,
ezzzzzzz
£, MODAL ANALYSIS METHODS
b)
rr/rf/tf ESm2
Description of Modal Analysis Method zazzzz i:: - :imi

'»rfrn»»/»'WfntF
All structures and their contents are complicated Three Story Planar Idealized Model
systems whose dynamic characteristics are not Structure Structure
amenable to exact evaluation. Similarly, earthquake
tremors are random motions whose direction and
magnitude are not accurately predictable. Thus an
exact analysis of how an existing structue will re-
spond to future earthquakes or if a designed new zzmza
structure can successfully withstand all future earth-
quakes is difficult if not impossible. To approach
c) raam
these analysis and design p* ^cesses rationally, zzW I 1 ni|
simplifying assumptions and engineering judgment R
l
are essential. The method currently used most often
for critical facilities is the modal analysis method. Thren Story Planar Idealized Model
Structure With Damping Structure

The terminology "modal analysis method" comes Figure 3. Typical vibrating systems.
from the concept of separating a vibration system
into its principal modes. All vibrating systems con- the system undergoes free vibrations. Since the
sist of a vibrating mass or masses and elements system has only one mass and one spring in the hori-
which tend to resist the motion or displacement of zontal direction, it has only one degree of freedom,
the mass or masses. The resisting elements are or one typical motior,; thus, it has one mode. Its
usually idealized as (1) cither elastic or inelastic motion can be represented by the equation
springs which tend to make the mass or masses re-
turn to a minimum displacement position, and (2) X(t) = Xniax sin wt [Eq3]
some energy-absorbing or frictional system which
tends to dampen the motion of the masses. Som; where X = the displacement of the mass relative
simple tyoical vibrating systems are shown in to the base at any time t
Figure 3. <max = the maximum displacement 'dative
to the base
Assuming the spring stiffness (k) of the system cu = the natural frequency of the system
represented by Figure 3(a) is linear, the motion of in radians per unit of time.
the mass will be sinusoidal with respect to time when
The system represented in Figure 3(b). however,
2
has three masses, each of which moves according to
AgbabianJacobsen Associates, fjcr J Guide for GENSAP
the forces acting on it at any instant of time. While
Code (U.S. Army Corps of Engineers, Huntsville Division, May
1972); and E. L. Wilson and H. H. Dove», Three Dimensional these masses are not independent, neither do they
Analysis of Building Systems—TABS. Report No. EERC 72-8 necessarily move as a unit. In fact, they would typi-
(University of California. December 1972). cally assume three different characteristic mode

10
 ■f£*?fW"!lVv«i«

tfffff?
TTTT, rmmrrm
a) Structure b) First Mode c) Second Mode d) Third Mode

Figure 4. Three principal mode shapes of system represented in Figure 3(b).

shapes (Figure 4). Each of these principal mode method. While the equation represents deflection
shapes would have a unique frequency of vibration only, similar relationships exist for velocities, accel-
(w) and each mass would have a particular displace- erations, shears, moments, and other linear func-
ment for each mode shape (Xnm)- The displacement tions of interest. Append'x A contains a more com-
represented by Figure 4(b) is typically referred to as plete discussion of modal analysis.
the first or fundamental mode, while (c) and (d)
represent the second and third principal modes, Response Spectrum Method
respectively. The exact values for Xnm for each mass
are functions of the masses themselves and the When a system is subjected to a forcing function
spring stiffnesses. applied to the system base, such as an earthquake
motion, its particular response is a function of the
if the system is vibrating freely in one of its prin- system characteristics and the properties of the forc-
cipal mode shapes, each mass will follow Eq 3 or ing function. A typical example of the displacement
response spectrum for single-degree-of-freedom sys-
Xnni(t) = (Xnm)max sin ü;nt (Eq 4] tems subjected to an earthquake motion is shown in
Figure 5. In this case, the natural frequency of the
If the system is disturbed by forces applied to the structure, f(f = w/ln) is plotted along the horizontal
masses themselves cr by base disturbance (such as axis, and the maximum displacement of the mass
an earthquake motion), each of these mode shapes (Xrnax) is plotted along the vertical axis.
will be excited to some extent. The amount that each
mode shape contributes to the total response is
known as its participation in the motion. The total The displacement, velocity, and acceleration are
free vibration of each mass relative to the base is related in the following way:
represented by
Displacement -- X(t) = Xmax sin tot [Eq 6]
Xm(t)= Z Fn(Xnrn)maxsin(V (EqS)
n-1 Velocity = X (t) = w (Xmax) cos cut [Eq 7]

where Fn represents the participation of mode n in Acceleration = V = X (t) = -aj2(Xmax) sin cot (Eq 8]
the total response of the system. The forced vibra-
tion response is discussed under "Time History where ' represents a derivative with respect to time.
Method." The maximum absolute values for the displacement,
velocity, and acceleration relative to the base are
The use of Eq 5 in the analysis of dynamic systems Xmax- wXmax> and w2Xmax- respectively. Const-
is generally referred to as the modal analysis quently, if the natural frequency and either the

II
 500

JOS 0.1 0.2 05 10 20 50 10.0 20.0 50.0

Natural Frequency (Cycles per Second)

Figure 5. Displacement response spectrum for single-degree-of-freedom systems subjected to

earthquake motion. Constructed from N. M. Newmark and W. J. Hall, "Procedures and
Criteria for Earthquake Resistant Desi^r ' Building Practices for Disaster Mitigation
(Department of Commerce. February 1973).

maximum displacement, velocity, or acceleration of values v*Ml not occur at the same time. Thus, simply
a system for a particular forcing function are known, adding the Tiaximum values for the various modes to
the other two maxima are also known. A typical re- obtain the maximum for the system is not possible.
sponse spectrum used in earthquake analysis where Instead, they must be combined in such a way that
all three maxima are of interest is shown in Figure 6, the expected response of the total system results.
a replotting of the response shown in Figure 5. Three general methods are used: ,1) the funda-
mental mode alone, since it is usur.lly dominant in
For linear systems, it can be theoretically shown the response; (2) the absolute maxima of the several
that the response of two or more systems having the mode shapes, to achieve a conservative upper bound
same frequency (w) will be the same for the same of the response; or (3) a square root of the sum of the
forcing function. This fundamental concept of squares approach, which takes account of the prob-
modal analysis allows prediction of the maximum ability that the maxima of the modes do not all occur
behavior of large systems through studying the re- at the same time (Figure 1(c)). The problem is dis-
sponse of simple systems. The frequency is the link cussed more completely under "Basic Procedures."
between the two when they arc subjected to the same
motion. A further complication is the accuracy of the re-
sponse spectrum itself in predicting future motions.
A complication in the application of this method Each spectrum prepared to date has been based
to large systems occurs because large systems usually upon measurements taken at a particular point for a
have several principal mode shapes, each with its particular earthquake. One cannot now predict with
own w value. Since these mode shapes will generally reasonable assurance of being accurate what
not be in phase with each other, their maximum motions will occur at any particular geographical

12
WWwiUfcawi^iK.—

0.2 0.5 I 2 5 10
Natural Frequency (Cycles per Second)

Figure 6. Typical response spectrum. From N. M. Newmark and W. J. Hall, "Procedures and
Criteria for Earthquake Resistant Design," Building Practices for Disaster Mitigation
(Department of Commerce, February 1973).

point of interest. Additionally, experience indicates Use of the response spectrum method assumes
that even at the same geographical location the that an appropriate response spectrum for a particu-
ground is subjected to motions of various intensities lar site is available.
during various earthquakes; this is reinforced by the
recognition that earthquake sources vary in loca- Time History Method
tions. Thus no reliable way of predicting what
motions a particular system will experience in its life- Use of the response spectrum with the modal
time exists. analysis technique described above yields only the
maximum values of the functions such as displace-
However, the data accumulated so far have shown ment, acceleration, shear, or moment. In some
that there are typical patterns of response spectra cases, how a function varies with time during an
which result from earthquake motions. In fact, earthquake may also be of interest.
bounds on the maximum expected behavior can be
drawn as shown in Figure 7. Additionally, as more Again using the modal analysis method in con-
data are collected and microseismic techniques junction with an appropriate earthquake accelera-
perfected, it is expected that for any particular site, tion record, the response of several single-degree-of-
response spectra bounds will be related to the prob- freedom systems with frequencies similar to the fre-
ability of occurrence of those motions. This relation- quencies of the real system being analyzed are com-
ship will be similar to the way the runoff in a puted. Using a time history representation, these
particular stream can be related to the probability of computations are made in a series of time steps,
occurrence of a storm of particular intensity (a 100- starting with initial conditions and taking a small
year storm, a 10-year storm, etc.). time interval and computing the response at the end

13

1 «
 500

0.5 I 2 5 10 20
Natural Frequency ( Cycles per Second )

Figure 7. Basic spectrum normalized to 1.0 g maximum acceleratioi. From N. M. Newmark

and W. J. Hall, "Procedures and Criteria for Earthquake Resistant Design," Building Prac-
tices for Disaster Mitigation (Department of Commerce, February 1973;

ofthat At. The responses of the mode shapes of the or frequency of the system and the particular earth-
real system are then computed and combined using quake motion input.
appropriate participation factors to obtain the re-
sponse of the real system as a function of the time The response is then computed using another time
(Figure 1(d)). In equation form this is: interval. This operation continues until the maxi-
mum conditions are encountered or the end of the
n time period of interest is reached.
F
n xnm Mn^ |Eq9]
n=l The calculations are best done by computer since
they are long and repetitious.
where ^ is the response of the corresponding single-
degree-of-freedom system to the earthquake at the In most cases this method will give more exact
end ofthat At interval, and is a function of the period values for the maximum velocities, displacements.

14
■ISMtMMa»,

shears, etc. than the response spectrum method. While most effort to date has been spent on linear
However, the values calculated apply only to one elastic analysis, considerable effort is now underway
particular acceleration record. Thus while they may to extend the design procedures into the inelastic
approximate the motions for the prior earthquake, range.
they may have no meaning for the next. Conse-
quently, the analyst has to balance the increased cost The linear elastic assumptions lead to two major
of the computer time and the accuracy of the results problem areas:
achieved, and decide whether to use the response
spectrum method or the time history method. Addi- (1) Concrete, one of the most important struc-
tionally, while the time history method appears to tural materials, is elastic over only a small range of
yield more accurate results, there are several possi- strain. Distortion quickly results in a cracked
bilities for error, and results may not be as accurate section; this significantly changes its load-deflection
as they first appear. relationships. Additionally, its effective modulus of
elasticity changes drastically, and the concrete
Assumptions and Limitations crushes (loses all strength) if strained much at all.
These effects can be considered in analysis, but
When simplifying anything as complicated as the seriously complicate the modal analysis method
response of a structure to earthquakes, several described above.
assumptions and some limitations to the use of the
analysis methods are required. Some of these have (2) Inelastic action seriously affects the period of
already been discussed. vibration and may cause the structure to undergo an
entirely different kind of response than that pre-
Input Motions dicted by the elastic analysis methods described
above. In general, inelastic action extends the period
Some discussion of the input motions is in the pre- of the structure; that is, it takes longer to make one
vious section. When analyzing a particular building complete cycle of vibration. This might change the
for a particular site, knowing what motions can be response of the structure from one in which accelera-
expected when would be helpful. This is currently tion is critical to one in which deflection is critical.
impossible with any degree of accuracy. After
detailed study of the earthquake experience and the While these problems can be dealt with, they do
soil and rock conditions of an area, experts can pre- seriously complicate the linear elastic analysis.
dict with some degree of assurance a band of
motions which might be expected and how often they Damping in the Structure
might occur; however, an earthquake of a particular
intensity might occur tomorrow, or it might occur Every vibrating system, including structures sub-
100 years from now. Thus, when particular ground jected to earthquake motion, loses energy in some
motions are selected for investigation and use in the way. If it did not, it would continue vibrating forever
design process, a certain level of protection which once it is started. This loss of energy in structural
may or may not be exceeded in the structure's antici- systems is called damping. Damping can be of
pated life span is provided. several types, the most common of which are fric-
tional and viscous damping. Frictional or coulomb
Linear Elastic Assumptions damping is the kind that occurs when a chair is
pushed across a floor. The total energy used « a
The modal analysis method is based on the function of the force pressing the two surfaces
assumption that the structure remains elastic, or together, the coefficient of friction, and the distance
nearly so, during the entire earthquake. This means moved.
that there is either very little or no permanent
deformation in the structure. For small or moderate Viscous damping is a function of the velocity of
earthquakes, this is not a bad assumption for well- the mass and the characteristics of the system. Since
designed structures. However, for strong or very most of the energy absorbed in a system is internal to
strong motions, this may be a very poor assumption the structural materials themselves, this kind of
even for well-designed structures. damping best represents the energy loss of the

15
 „.-,* t&PtS. - "■' ■ *" O^fflfWiSRt.

system and is used most often in this type of analysis. (masses) as represented by the dashpots and masses
in Figure 3(c). Thus, the damping depends upon the
A difficult step in modeling a structure is deter- relative velocities between the masses or between the
mining the amount of damping to use to represent first mass and the foundation.
the energy loss. The amount of damping used is
usually specified in terms of critical damping, which To make modal analysis possible, the amount of
is the amount of damping that prevents oscillating damping between masses must be consistent wilh the
motion. The effects of subcritical and supercritical other structural characteristics, the masses, anc the
damping for a single-degree-of-freedom system are spring stiffnesses.
shown in Figures A4 and A5, respectively. Usually,
the amount of damping exhibited in real structures Appendix A contains a more complete discussion
is between 0 and 10 percent of critical damping, of damping.
depending upon the type of materials used and the
structural concept employed. Some currently recom- Basic Procedures
mended values are shown in Table 1.
The procedures discussed below provide for the
Table 1
analysis of typical buildings such as hospitals, ware-
Rccommcndt Damping Values* houses, offices, and billets. The analysis of more
Type and Condition PereenUg« of
complex structures such as dams, nuclear reactors,
Streu Level of Structure Critical Damping and nonlinear structures requires more detailed
evaluation than is provided by the following pro-
Working stress, no Vital piping 0.5 to 1.0 cedures.
more than about one-
Welded steel, pre-
half yield point
stressed concrete, well Analysis of structures subjected to earthquake
reinforced concrete motions requires several simplifying assumptions
(only slight cracking) 2
and considerable engineering judgment. In addition,
Reinforced concrete of course, it requires many calculations. While these
with considerable can be performed by hand, they are most often done
cracking 3 to 5
by a digital computer with programs written specifi-
Bolted and/or riveted cally for this type of analysis.
steel, wood structures
with nailed or bolted
joints 5 to 7 When using these prepared programs, however,
the analyst should be careful to understand the
At or just below yield Viial piping method of solution, the modeling techniques, and
Welded steel, pre- the assumptions employed in preparing the pro-
stressed concrete gram. Additionally, he should recognize that the
(without complete loss answers are accurate to only one or perhaps two
in prestress) significant figures.
Prcstressed concrete
7
«rlh no prestress left The procedures described below represent a
Reinforced concrete 7 to 10 general approach to analysis. With some structures
or some prepared programs, other procedures or
Bolted and/or riveted
steel, wood slructure>. steps may be required.
with bolted joints 10 to 15

Wood structures with

it is assumed at this point that a structure has
nailed joints 15 to 20 already been designed on a preliminary basis. A
structural concept must be formed with certain
•N. M. Newmark and W. J. Hall. "Procedures and Criteria for members and equipment determined or a fairly
Earthquake Resistant Design." Building Practices for Disaster accurate estimate of their mass and strength made.
Mitigation (Department of Commerce, February 1973),
Since design is an iterative process, the results of any
When systems with more than one degree of free- one dynamic analysis may require modifications in
dom are analyzed, the damping acts between floors the design of a structure under consideration. How-

16

-f"
'<"•»*»*»,-,„

d = Unit Deflection

77777 77777
b.
Shear Beam Cantilever Beam P-A Effect

Figure 8. Types of structural response.

ever, certain decisions must already have been made, sponse of the structure.
at least tentatively, before this analysis can be For modeling purposes, the masses must be
performed. lumped at discrete points. Usually the mass of walls,
partitions, and vertical structural elements are con-
Idealization of Structural System and Calculation solidated at the floor level. For these portions of the
of Member Properties and Masses structure, half the mass goes to the floor system and
half goes to the next higher floor (or roof) system.
As a first step in analyzing the response of a struc-
ture to earthquake motion, the engineer must model Determining the stiffness of the individual mem-
the structure. An overall evaluation of the structure bers and the structure as a whole is a more compli-
is required to define the vertical and lateral force cated problem. Some structures will respond like a
resistant system, and to determine where the stiff- shear beam (Figure 8(a)), while others will respond
nesses, masses, and energy-absorbing systems of the like a cantilever beam (Figure 8(b)). In some cases
structure are, and how they are linked together. both responses must be considered.
Assuming Figure 3(c) is a typical representation, the
engineer then determines the values of k, c, and m If a shear beam analysis is appropriate, the
for the stiffness, damping, and mass respectively, of analyst must calculate the force it takes to deform
the various parts of the structure. the members on the basis of a unit deflection (Figure
8(a)). Summing all these for one floor in one direc-
It is usually not too difficult to evaluate the mass tion gives the stiffness against motion (k) for that
of a typical structure. Al' dead load that is expected, mass in that direction. A similar analysis for any
all live load that can reasonably be expected to be other direction of interest and all other masses must
attached to the structure, and in some cases, such as also be performed.
a warehouse, some percentage of the design live load
must be included. Normally, the rest of the design If a cantilever beam analysis is appropriate, the
live load is considered to move around within a struc- analyst must calculate the resistance to motion of the
ture and does not contribute to the horizontal re- structure shown in Figure 8(b), including all rcsisl-

17
 ^wJWBWWWraMHMilWIMW^'W1 *«»to-:««..

ing elements. SimilaHy, this analysis must be re- and Vibration Problems in Engineering* by
peated for all floors and all directions of interest. Timoshenko. This becomes quite tedious if done by
When performing the analysis indicated above, hand for systems with several degrees of freedom.
the engineer must take into account any structural The Rayleigh method,5 an approximate method
details that i v<tld affect his analysis, such as connec- which will give satisfactory accuracy for the first few
tions that w'l not resist movement, the cracking of modes, is also available. If more than one or two
concrete sections, instability, and limits to the modes are required, however, the accuracy necessary
amount of deflection allowed before linear behavior in these hand calculations prohibits their use as a
ends. He must also consider the P-A effect; this is the design method. Consequently, this calculation is
effect of axial loads (P) which have been displaced usually accomplished with the computer and stand-
some distance (A) and then contribute to overturning ard programs.
moments, which in turn contribute to more deflec-
tion, etc. This is shown in Figure 8(c). In general, Number of Modes to be Considered •
the P-A effects ma; be ignored if the resulting
moments are less than 10 percent of the other design At this point it should be recalled that a simple-
moments. looking structure will have many frequencies and
mode shapes, since each mass and resisting element
Since it is long and repetitious, the analysis is has its own impact on the total system. Even if the
usually performed by a computer. However, the grossest simplifying assumptions are made, there
engineer must be able to input all of the values will be at least three motions for each mass—the x
necessary to perform his analysis, ach as the cross- direction, the y direction, and the z direction (Figure
sectional area, the modulus of elasticity, the moment 9). There will usually be torsion about these axes,
of inertia in both directions, and any details in the but for most structures, these torsional effects are
structure such as hinges. not emphasized. Thus, even for a three-story build-
ing there will be at least nine degrees of freedom,
Assignment of Damping Values and thus nine mode shapes and frequencies.
A present, less is known about the damping values
to use in real structures than about other parts of
these analysis techniques. Considerable research is
underway to develop better understanding of this
effect and to provide a more rational way of deter-
4^
mining correct values for various structures.

Pending the results of this research, it is appropri-

ate to perform several analyses of the structure with
various values for the damping factor. Currently 0,
0.5, 2, 5, and '.0 percent of critical damping are
appropriate. While this gives five separate results, it 7^
at least brackets the response the engineer should
expect from the real structure.

Calculation of Mode Shapes and Frequencies

After the structure has been modeled, the analyst Ms
should determine the mode shapes and frequencies
of the system. An exact method would involve the Figure 9. Possible degrees of freedom for a three-
solution of a series of differential equations, a pro- story structure.
cedure that is described in several references such as
Fundamentals of Vibration Analysis1 by Myklestad
"Stephen Timoshenko, Vibration Problems in Engineering
'N. O. Myklestad, Fundamentals of Vibration Analysis (Van Nostrand. 1955).
s
(McGraw-Hill. 19S6). Mvklestad.

18
■

Theoretically, tor a symmetrical structure these general predict the kind of motion expected. The
motions could be uncoupled and analyzed separ- methods are the response spectrum method and the
ately. Practically, however, because of nonuniform time history methods described unde "Description
materials, construction, loadings, and ground mo- ot Modal Analysis Vlethod."
tions these effects are interrelated and cannot be
considered separately without introducing some er- • Response Spectrum Method. In this method,
ror. Depending upon the type and design of the the analyst must know the response spectrum for
structure, this error may be relatively small and the systems with periods similar to the periods of the
structure may be uncoupled and analyzed separately structure being investigated. The expected responses
in the three principal directions. of these modes in the real structure can then be com-
puted using the appropriate participation factor,
"L"-shaped structures or those with the center of and combined by adding their absolute values or by
mass significantly displaced from the center of resist- taking the square root of the sum of their squares.
ance (5 percent of the major dimension in the plan As mentioned before, adding the absolute values
view of the structure) should be analyzed with mo- gives an upper bound for the expected response,
tions in all directions considered at the same time; w hile using the square root of the sum of the squares
the effects cannot be uncoupled. yields a more probable value for the response. In
cases where some of the modal frequencies are close
With a normal, fairly symmetrical structure, the together, the absolute sum may give a better result.
motions can be decoupled and effects considered in These calculations can be done by computer or by
only one direction at a time. Even so, as many mode hand.
shapes and frequencies as there are floors in the
structure must be considered. Fortunately, not all of • Time History Method. This method requires
the frequencies and mode shapes need to be evalu- that the time history responses of single-degree-of-
ated to have an adequate evaluation of a structure freedom systems having the same periods as the real
because: structure be calculated for the entire duration of the
input motion. The responses of these systems, multi-
(1) The participation factors for the high frequen- plied by the participation factors for the mode
cies are usually much smaller than those for the first shapes and the mode shape of the real structure
three or four modes. itself, are then added together for each time interval
to yield the time history response of the real struc-
(2) With higher frequencies the masses must ture. Since a very large number of calculations are
move faster. Since the damping force is directly pro- required in this method, it can only realistically be
portional to the velocity, these higher mode re- done with a computer.
sponses are damped out faster.
Accuracy and Applicability of Method
(3) As shown in Figure 6, as the frequency goes up to Various Structural Concepts
(or the period goes down) the responses of the higher
modes to the typical earthquake in terms of velocity It is generally assumed that the response spectrum
and displacement are insignificant, and the accelera- method of analysis described above is applicable to
tion responses approach a constant value. most structures, as long as they remain linearly
elastic. However, in using this method with certain
Thus analysis can usually be confined to the first structural systems and configurations, the results
three or four modes, or the ones that have frequen- may appear to be in error until they are reconciled
cies <20cps. with the behavior of the structure and the modeling
assumptions. For example, if the structural system
Calculation of the Response employs both shear walls and a moment-resisting
space frame, interactions between the two systems
The analyst can now predict the response of a will occur; that is, in the lower stories, the shear wall
structure, assuming the structure remains linear or will support the space frame and in the upper stories
nearly linear in its response, and assuming he can in the space frame will support the shear wall. As a

\9
 K^mr^mrmmmmgmmmKmniiiKamm^mj^^^.

result of this interaction, opposing forces will be This phenomenon has been noted most often when
present in a given story, leading to ir.uch larger there are few degrees of freedom. As the number of
values for shear in this transition zone than occur degrees of freedom increases, the time history re-
immediately above or below it. A similar observation sponse approaches the values from the spectral
has been noted where structural parts of widely analysis. While neither method can give exactly
divergent stiffnesses have been joined fogether, such accurate results, the discrepancy shown in the table
as building setbacks. Thus, while the response spec- indicates that the spectral analysis (square root of
trum method yields accurate results, these results the sum of the squares) method should be used with
sometimes lead to questions about the appropriate- extra care when there are only a few degrees of
ness of the results; time history analysis is then re- freedom involved in the response.
quired for clarification.
While the above discussion indicates that time
Another limitation of the response spectrum history response may be necessary in some cases, it
method is that only maximum or expected maximum should be recognized that this evaluation is expen-
values of a function result from the analysis. Some sive due to the considerable cost of computer time
equipment or mechanical or structural components necessary for the calculations and the time needed to
attached to the structure may need to be evaluated evaluate and assimilate the results.
independently for their response. This can only be
accomplished if the time history response of the When performing any of these calculations, the
structure itself is known. In effect, the structural analyst must not become enamored by the numbers
motion becomes the earthquake and the component that result. While the computer can produce num-
becomes the structure in an analogy to the structural bers to six or seven figures, only about one or
response calculations. perhaps two digits at most are significant. A great
many assumptions are involved in the modeling of
A further complication in the response spectrum the structure and the use of past earthquakes to pre-
method is that while the sum of tie absolute values dict future ground motions. Even the time history
of responses predicts an upper bound for a function, modal response calculations are no more accurate
the square root of the sum of the squares often pre- than the assumptions used in setting up the scenario
dicts a value of the function which is 'ower than the and modeling the structure. While continued re-
value calculated from a time history evaluation search is expected to improve this situation, con-
(Table 2). siderable engineering judgment is currently required
in using the results of any of these evaluations.
Table 2

Shew (In Idpe) for lO-Stoiy HoeplUl

Subjected to Typical Earthquake*
O SUMMARY
Urne Spectral 1968
Hbtory Analysis SEAOCCode
From the discussion above, it can be seen that the
Su*y Analysis
«VVV1 Design Value
modal analysis method can be applied to most struc-
10 7,517 6,511 725 tures provided realistic, simplifying assumptions are
9 11.675 10,085 1,194 made and engineering judgment is used in evaluat-
8 17,246 14,080 1,546 ing the results. In general, with the uncertainties in-
7 20.547 16,470 1,780
volved in understanding the energy dissipation in the
6 23.090 18.380 1,899
5 24,894 20.600 2,919
structure and in predicting future motions, it
4 25.890 22,760 3,779 appears unreasonable to try to be very accurate with
3 30.8% 29,020 6,429 most analyses. Rather, the response spectrum
2 37,559 32,780 7,814 method provides sufficient accuracy for the general
1 39,000 33,290 8,554
case using approximately the first four modes for
*Agbabian Associates, Existing Capacity and Strvngthrning evaluation. If, however, the structure is more com-
Concepts for Letterman and Hays Hospital {Tusk M), draft report plicated in its concept—it has reentrant corners,
(Construction Engineering Research Laboratory. April 1974). large setbacks, or has a combination of structural

20
■MHH

systems with widely different stiffnesses—a more

detailed time history analysis may be justified. Addi- H DESIGN RECOMMENDATIONS
tionally, a time history response should be accomp-
lished if a detailed analysis of the response of For most evaluations, it is recommended that the
supported equipment or subsystems is required. response spectrum method be used to investigate
structures subjected to earthquake motions. For
In either case, access to a computer with relatively structures having reentrants, large setbacks, combi-
large capacity is essential, although much more nations of structural support systems, or compli-
computer time is required for the time history re- cated equipment systems, the time history method of
sponse than for the response spectrum approach. analysis should be considered.

21

-, , , 1. r ■ ■;
 ...-T^.vr'Kr^z*.**«^-"'?*;33»*S^^^^^ -

APPENDIX A

REVIEW OF MODAL ANALYSIS THEORY Using Newton's Second Law (F = ma) and sum-
ming forces gives
This Appendix reviews the theory associated with
modal analysis methods and amplifies the discussion -kX-cX=mX.or [EqAll
in the main body of the report.
mX+cX + kX=0 |EqA2|
Single-Degree-of-Freedom System
Dividing by m and letting w2 = k/m and t, = c/2cüm
In the analysis of structural dynamic systems— gives
structural systems subjected to vibratory motion—
the basic building block is the single-degree of- X+24wX + a)2X =0 [EqA3]
>eedom system (Figure Al). This system includes a
base or support system, a mass which moves or A general solution for this, appropriate to struc-
vibrates, a stiffness or restoring system which returns tural dynamic problems, takes the form
the mass to its undetlected position, and a damping
or energy-absorbing system. X = jffrrt [A cos (vAj . £J) ^ + B sin (V 1 - £') ^t)

|Eq A4]

I
',
k
-vwwv
m
If there is no damping in the system (4 = 0), then Eq
A4 reduces to
D-
c X = A cos wt + B sin cut |Eq AS]

or
Y.Y, Y •x.x.x
X = C sin (wt + a) [Eq A6|

Figure Al. Typical single-degree-of-freedom sy ^em. This is the equation of free vibration in which C is
the amplitude of motion, co is the circular natural
Figure Al also shows a coordinate system in which frequency (radians/second), and a is a phase angle
X, X. and X represent the absolute deflection, dependent upon the initial conditions of motion. The
velocity, and acceleration of the mass respectively vibration represented by Eq A6 is shown in Figure
and Y, Y, and Y similar values for the base. A3. The period of the system is the time it takes to
make one complete cycle, (T = 2n/üj). The frequency
Assume for the moment that the base is stationary is the number of complete cycles per second (f =
(i.e., Y = Y = Y = 0) and that some force has dis- a>/2n=l/T).
placed the mass from its at-rest position. Also
assume that the damping is directly proportional to If ^ is not equal to zero in Eq A4, then damping
the relative velocity of the mass with respect to its exists in the system, and the motion can be repre-
base. Then if X and X are positive, the forces shown senttd by Figure A4. In this case, the motion is grad-
in Figure A2 act on the mass. ually damped out. The damped circular frequency
(ay) is now cpv 1 - £2, and thtsdamped period Tj is
In/w^ 1 - £2. Experience with real structures indi-
cates that they have values of i, equal to or less than
kX 0.2 (20 percent damping). Thus there is minimum
m effect on the frequency (wj ^ 97.98 percent of w)
CX and the period (Tj ^ 1.02T). For all practical pur-
poses, the co and T for the undamped systems can be
Figure A2. Forces acting on the mass for single- used to represent the frequency and period of the
degree-of-freedom system. damped systems.

22
TWTWfftyTW^^'W^niiii MiijiixiHH j

Io
0)
o Time
Q.
w

Figure A3. Plot of displacement versus time for undamped single-degree-of-freedom system.

Plot of A Cos(-/]77* )ü»t + B SinCyfTp )iüt

Plot of e

Q. Time
M
Q
Supercritical Damped Vibration ( r<l)

Figure A4. Plot of displacement versus time for subcritically damped single-degree-of-freedom
system.

If the values of C and w are known in Eq A6, then An understanding of these relationships is essen-
the maximum values of X, X, and 3C are also known: tial when dealing with modal analysis methods.

^max -C |Eq A7] Returning to Eq A3, it should be remembered that

w2 = k/m or the circular natural frequency <D =
v k/m. If k is increased or m decreased, w and thus
x
max " ^ [Eq A8| the maximum values of X and X in Eq A7 are in-
creased. Conversely, if k is decreased or m increased,
w and thus the maximum values of X and X are
x
max " w2c [Eq A91 decreased.

23
Discussion of Damping of Single-Degree-of- ary between cyclic and aperiodic behavior of the
Freedom Systems system. If £ < 1, then cyclic response occurs (Figure
A4). If £ > 1. then aperiodic motion results
All vibrating systems lose energy in some way; this (Figure A5).
is called damping of the system. Energy can bo lost
from a structural system in several different w.^ys. A> mentioned above, realistic values of 4 for struc-
For example, if two surfaces are rubbed together, tural systems are equal to or less than 0.2. Values
coulomb, or dry friction, damping occurs. If the less than 0.1 are usually applicable.
damping force is proportional to the relativ? velocity
of the mass, viscous damping occurs. In some Multi-Degree-of-Freedom Systems
systems, damping is proportional to the square of
the relative velocity; other concepts for damping are Almost all systems of interest are more compli-
also used. cated than the single-degree-of-freedom system de-
scribed above. Consequently, this section describes a
In real structures, the total damping is a combina- method of analyzing more complex systems. Figure
tion of different types of damping, but damping A6 shows a rather simple multi-degree-of-freedom
always resists the relative motion of the mass. From system which is representative of most structures.
the standpoint of theoretical analysis of these Some structures, such as nuclear reactors, are much
systems, the easiest type of damping to deal with is more complicated in that mass two might be con-
the one in which the resisting force is a linear func- nected with the base as well as with masses one and
tion of the velocity. Experience has shown that this three. Other complications are also quite possible,
assumption results in reasonably accurate results, but for the general building. Figure A6 is acceptably
and it is the kind of damping that is used in this representative.
report.
In Figure A6, each mass has only one component
The damping factor £ used in Eq A3 and subse- of motion; therefore, this system has three degrees of
quent work equals c/lwm, and is a dimensionless freedom, three frequencies, and three mode shapes.
number giving a multiple (or fraction) of critical In a real structure, each mass usually has the ability
damping. Critical damping is defined as the value of to move in three orthogonal directions and twist
damping for the system which represents the bound- about these three orthogonal axes. If this were true

Supercritical Damped Vibration( r> D

c
«
E

Q. Time
M
o

Figure AS. Plot of displacement versus time for supercritically damped single-degree-of-
freedom system.

24
 -k^X,)- m 'KplX^Xi)

•^w
—wwwv—r
kgCXg-X,)' m
C!
U-Y.Y.Y IUx^.X, T "- T
Ux2.X2.X2 U-XyXj,

■vw m

Figure A6. Typical multi-degrce-of-freedom system.

for Figure A6, the system would have 18 degrees of

freedom with 18 frequencies of vibration and 18 Figure A7. Forces acting on masses for three-degree-
separate mode shapes. of-freedom systemC, = C, = Cj = 0 and Xj > Xj > X^

Generally, this complication does not present a X^DjsinM + a) [Eq A18]

significant problem, since the responses of a struc-
ture in the various mode shapes can be decoupled Substituting these values and their derivatives into
and treated separately. Eq A13, A14, and A15 gives

To find the natural frequencies and mode shapes, sin (art + a) {[ - m,^ + (k, + k,)] D, - k, D,} - 0
the same type of analysis is performed as was done
for single-degree-of-freedom systems. As a starting [Eq A19]
point, assume that the base is stationary and that
Xj > X, > X, with forces shown in Figure A7. Using sin (art + aH - k, Dj + (-m, w2 + (k, + k,)]
Newton's Second Law and summing forces gives
Dj-k,D,}-0 [EqA20]
-k,X, + MX, - X,) « m, X, [Eq AlO]
sinM + a){-kJD, + [-m,u>2-»-kJD,}»0
-MXj-XjKMXj-Xp-m^, [EqAll]
[EqA21]
-kjCXj-XjJ-mjX, [EqA121
For these equations to be true for all values of t and
Rewriting these gives to have nontrivial solutions (Dk # 0, and Dj # 0), the
following determinant must be equal to zero:
mjX, + (k, + k,) Xj - k, X, -0 [Eq A131
((k. + k,) -m.o)1! -k, 0
mjX,-kjXj + Ckj + k^Xj-MXp-O fEqAM)
-k, [(kj + k^-m^1] -k,
mjXj-kjXj + kjXj-O [EqAlS]
0 -k, [kj-m,<ü2l
Since these equations are similar to Eq A2, solutions
are expected of the form [Eq A22]

X, = D, sin (tut + a) (Eq A16] Performing this evaluation yields three values of a>2
in terms of k/m, which when evaluated wi'l give the
X| - D| sin (art + a) lEqAl?) three natural frequencies of vibration for tl>e system.

25

„..
 3k
-wwwv- 2m \—vwww 2m —VWAMr— m
i
I
Figure A8. Exan>ple three-degree-of-freedom system.

As an example, take a three-story structure and First Second Third

assume that the mass of the second and third floor is Mode Mode Mode
twice the mass of the roof. Using a typical code, the
relative stiffnesses would be as shown in Figure A8. D, 1 1 1

Substituting these values into Eq A22 gives 5

D2 2 1
7
(S'/J k - 2müj2) -2,/3k 0
2
-2I/3k O'/jk -2ma>2) -k =0 Da 3 -2
7

0 -k (k - tW) These mode shapes are shown in Figure A 9.

Solving for CD2 gives CDJ = =— Referring to Eq A16, A17, and A18, it can be seen
that these mode shapes represent relative values only
and that the displacements for any mode n at any
. 3k time t are

X, = Dj sin (a)nt + a) ^ 1 |Djsin (wnt + a)|

,_7k
fEq A23|

Xj = D, sin {(unl + a) = pj1 (D, sin (ajnt + a)|

or a), = = 0.5774./-
(Eq A24|

1^ Xj = Dj sin (ajn( + a) = ^r1 |D, sin (a)nt + a)J

Wj = = 1.2247 /—
m (Eq A25j

0^ = = 1.8708 As a multi-degree-of-freedom system with independ-

ent or uncoupled modes is subjected to a base dis-
turbance, each mode responds in its own way. Since
To find the mode shape, values for cu, m, and k are O,. Dj, and Oj above are arbitrary values, the total
substituted in Eq A19, A20, and A21. While exact response of a system can be represented by Eq 5
values are not available, values for 02/0] and D,/D, (from main body of report) which is reproduced
are. Assuming that D, is equal to some unit displace- here:
ment, a representative mode shape can be obtained. 1 F
For the example, the following values result:
Xni(i) n «tW sin »^
n=l

26
*$&ie&:&*mffW' -w«. „ _

2/7

-5/7

Ar
Second Mode Third Mode

Figure A'?. Principal mode shapes for example system.

where Dmn is the maximum displacement of mass m 6m

in mode n caused by the particular disturbance, and 2m(l) + 2m( V7)+m(2/7) 7 21
F3 =
Fn is the participation of mode n in the total re- 2m(l)2 + 2m( V7)2 + m(2/7)2 152m 76
sponse. The phase angle a has been omitted since it 49
does not affect the magnitude of the response.
If other arbitrary values for D, are judiciously
It has been shown* that for earthquake motions, chosen, values for all Fn equal to I could result. This
Eq A26 allows evaluation of Fn: would simplify analysis methods.

J While the discussion above is based on displace-

Z mjDj ment alone, similar relationships exist for any linear
i=l function, such as shears, moments, rotations, accel-
Fn = [Eq A26|
j erations, etc.
I mjlDi)2
i=l Damping of Multi-Degree-of-Freedom
Systems
where j represents the total number of degrees of
freedom of the system. For the example, the values As mentioned earlier, damping is a complicated
for the participation factor Fn are as follows: phenomenon. To understand it and its effects on
multi-degree-of-freedom systems, defining what a
mode is when damping is involved is necessary.
2m(l) + 2m(2)+m(3) 9m 36
Fx-
2m(l)2 + 2m(2)2 + m(3)2 19m 76 In the previous section, mode shapes were de-
scribed; it was found that with multi-degree-of-
2m(l)+2m(l) + m( 2) _2m 19 freedom systems, the masses vibrate in such a way
F^
2m(l)2 + 2m(l)J + m( 2)2 8m 76 that the ratios of their displacements are always the
same. Assume that

X^DjsinttDt a)
'G. W. Housner. "Earthquake Resistant Design Based on
Dynamic Properties of Earthquakes," Proceedings. Vol 53
(American Concrete Institute. 1956-57). pp 85-98. X2 = D2sin(iut a)

27
 X3 = D3sin (ait a) a system is established, it is fixed lor all other modes
lor that system, since the a;n values are dependent
then only upon the masses and stiffnesses for that system.

X3 = Djsin(cüt a) = Dj It can also be shown from this work that the value
X, D, sin(ajt a) D, ofthe damping coefficient (e) is established from the
relationship
which is a constant for any particular frequency.
Similarly, 2Mn
X D
v2 =n2 = a constant,
Xj Uj where i refers to the relative location in the system
and n refers to the frequency and mode numbers.
Extrapolating this to damped systems, for mode Having established or selected the damping fraction
shapes to exist with damping, the ratios of the dis- for one frequency, and knowing all the k's and n's
placements of the masses should always be the same. for the system, all of the damping coefficients are
If 4 — portion of critical damping for one frequency. determined and the damping fraction for each of the
a; = that frequency, and wj = the damped frequency other modes is fixed.
icvfi = u)\/ 1 - i,1), then the following relationships
would exist The above discussion is based upon theory. In
practice, however, most computer programs allow
X, = D,e ^sinlwdt a) the separate establishment of damping factors for
each mode. This is considered appropriate for real
X2 = D2e ^'sin (a;dt - a) structures for the following reasons:

X3 = D3e ^wtsin((i;dt a) (1) Structures generally have six degrees of

freedom for each mass; deflection in the x, y. and /
Then directions; and rotation about each of these axes.
The stiffnesses and damping coefficients in these
X j _ Dje ^sin (wdt a) _ Dj =
- = a constant various directions need not be related to each other.
X
i D,e ^sinl^dt a) Di Therefore, no relationship between the damping
and appropriate to one frequency of a structure and that
of another is necessary.
X2 _ D2 _
v - iS = a constan
f 'or any particular system. (2) Viscous damping is only an idealization of the
actual damping that exists. Some forms of actual
damping depend upon deflection only and are not
Substituting these values into the equations of
functions of velocity. Therefore, judgment should be
motion for the masses and solving, it can be shown
used in applying the theory discussed above: in-
that for modal response to be appropriate with inter-
creasing the portion of critical damping for the
floor damping.
higher frequencies may not be necessary,

= a constant for a particular system (J) Generally, increasing the damping results in
k, kj k3
less displacement and less response. Therefore,
using smaller amounts of damping for the higher
with a particular portion of critical damping. It can
modes is conservative and does not result in large
also be shown that
errors in analvsis.
1
= — =y = a constant, In summary, modal analysis can be used with
CDj OJj (1)3
damped multi-dcgree-ol-lreedom systems. Results
Once the portion of critical damping for one mode of will be sufficiently accurate even though percentages

26
of critical damping for each mode are arbitrarily Damping in these systems will usually not exceed 10
established (using experience and engineering judg- percent of critical damping.
ment) independent of the frequency relationships.

29
 APPENDIX B

SIGNIFICANCE AND USE Since the time of the maximum response is not re-
OF RESPONSE SPECTRA corded in this procedure, when the maximum
response occurs is unknown.
The response spectrum, a standard tool used in
Until relatively recently, there were few recorded
the analysis of vibrating systems, is usually a graph
earthquake motions because there were lew acccl-
or plot of the expected responses of systems to a
erometers emplaeed to measure them; the El C entro.
certain input motion. Figure 6 shows a typical earth-
CA, earthquake of 1940 was the most severe earth-
quake response spectrum. In some cases, the fre-
quake recorded and was used as the basis lor much
quency (f) is plotted along the horizontal axis, while
analytical work. Recently, however, many more
in others the period of the system (T) is the hori-
earthquakes have been measured. Maximum re-
zontal coordinate, f = 1/T. Thus some spectra
corded accelerations have gone from about 0.32 g for
appear to be the mirror image of others, while in fact
the El Centro earthquake to values greater than 0.5
they are the same except they have a different coordi-
g. It is expected that even larger values will be meas-
nate system.
ured as more instruments are placed closer to
epicenters of active earthquakes.
It can be theoretically shown that the response of
an undamped system to a particular motion is a
Earthquakes consist of a series of random ground
function of the motion itself and the natural fre-
motions. Usually the north-south, east-west, and
quency of the system. While this seems to neglect the
vertical components of accelerations are measured.
stiffness and mass of the system, it does not, since
Currently, no accurate predictive method has been
the frequency is directly proportional io the square
developed which allows description of the particular
root of the stiffness and inversely proportional to the
motion that a site can be expected to experience.
square root of the mass,
Thus, it is better to use a consolidated response spec-
A
(i) =,/—
V m
trum which incorporates the consolidated spectra for
several earthquakes with the primary variable, as far
.-1 as the earthquake motion is concerned, being the
Thus if the stiffness and the mass of a system are maximum acceleration. Such a prediction is shown
doubled, the frequency remains the same and the in Figure Bl. Also shown in that figure are the maxi-
same response can be expected for the new system as mum responses expected when various amounts of
for the old. critical damping are applied to the system.

In general, response spectra are prepared by Three generalizations can be made about the re-
calculating the response of single-degree-of-freedom sponse spectrum shown in Figure Bl. When the fre-
systems with various amounts of damping to meas- quency is low (less than 0.2 cps), the displacement
ured earthquake motions. Mathematical integration response is fairly constant. This corresponds to a
methods are available which apply the measured system having a relatively small stiffness with respect
motions to the base of a system, with integration over to its mass. When the frequency is large (greater
short time intervals, and calculation of the response than 2 cps), the acceleration is relatively constant.
of the mass. They proceed in a step-by-step process This corresponds to a system having a large stiffness
until the total earthquake record has been com- compared to its mass. In the midrange, the pseudo-
pleted. The largest value of the function of interest is velocity (or velocity of the mass relative to its base)
recorded and becomes the response of that system to remains fairly constant with a change in the fre-
that motion. Changing the parameters of the system quency. This is consistent with the response expected
to change the frequency, the process is repeated and from such systems.
another response recorded. This process is repeated
until all frequencies of interest have been covered The following is a demonstration of the use of the
and the results plotted. This becomes the response response spectrum method in conjunction with the
spectrum for that motion. Since no two earthquakes example structure analyzed in Appendix A.
are alike, this total process must be repeated for all
earthquakes of interest. Assume that the value of relative stiffness (k) used

30

t .t,.
 »»««S»«!^^^™^.,

500
■

200

100

2 5
Frequency, cps

Figure Bl. Basic design spectra normalized to 1.0 g. From N. M. Newmark and W. J. Hall,
"Procedures and Criteria for Earthquake Resistant Desigi," Building Practices for Disaster
Mitigation (Department of Commerce, February 1973).

in that example is 500 kips/in., and the value of the

relative mass (m) is 2 kip-seconds2/in. This mass is 0)3=1.8708 1.8708 V 250 =29.5799 rps
m
equivalent to a weight of about 773 kips. Then the
frequencies will be = 4.71 cps

Wl = 0.5774/^ = .5774/^= 9.1295 ^i^i Assuming zero damping with these frequencies
m second
and using Figure Bl, displacements of 22.0, 6.8. and
= 1.45 cps 2.9 in. for modes 1, 2, and 3 respectively result.

Using these values in conjunction with the modal

w2 = 1.2247\/^= 1.2247 Vr250 = 19.3642 rps
m displacements and participation factors calculated
in Appendix A gives the following values for the
= 3.08 cps modal responses:

31
 Fint Second Third First Second Third
Mode Mode Mode Mode Mode Mode

Displacement, mass one 10.4 in. l.7in. 0.8 in. Acceleration, mass one 2.27 g I,h3jj I.^Og
Displacement, mass two 20.8 in. I.7in. -0.6 in. Acceleration, mass two 4.55 g Ih3>i - l.28js
Displacement, mass three 3l..1in. -J.4in. 0.2 in. Acceleration, mass three h,82g - 3.25g I1 ?l >!

These displacements are relative to the base of the Combining these in the same way as above gives:
structure.
First Maximum Expected
To predict the maximum displacement of this Mode Absolute Value Value
structure, either the first mode, the maximum abso-
lute value, or the maximum expected value (the Mass one 2.27 g 5.70 g 3.32 g
Mass two 4.55 g 7.4ft g 5.00 g
square root of the sum of the squares) can be used.
Mass three h.82g 10,58g 7.57 g
The values obtained using the three figures are:

Flnt Maximum Expected The responses shown above appear to be quite

Mode Absolute Value Value large, partially because the design spectrum is
normalized to a maximum ground acceleration of
Mass one 10.4 h. 12.9in. 10.6 in.
Mass two 20.8 in. 23.1 in. 20.9 in.
I.Og. If the maximum ground acceleration expected
Mass three 31.3 in. 34.9in. 31.5 in. at a particular site is 0.3 g, then each of the values
shown above would be multiplied by the factor 0.3.
If the relative deflections between the masses, the thus reducing the responses to realistic values. Addi-
shears, or some other linear function of the structure tionally, small amounts of damping would signifi-
are desired, it would be necessary to start with the cantly reduce the results, as can be seen from
correct functional value (relative deflection) but use Figure BI.
the same participation factor and same value from
Figure BI, combining them in the same way. For The dominance of the first mode in the responses
example, the acceleration responses would be 4.6 g. shown above should be noted.
6.5 g. and 6.5 g for modes 1, 2, and 3 respectively.
In summary, the response spectrum method pro-
Using these values with the modal displacements vides a fast, reasonably accurate tool to help in the
and participation factors from Appendix A gives the analysis of structures subjected to earthquake
following values for the modal accelerations: motions.

32
 •:*B-a!Sv<i»»m..„__

APPENDIX C

EXTENSION OF THE MODAL ANALYSIS

METHOD TO THE INELASTIC CASE

Most well designed and constructed buildings are «/>

able to withstand minor earthquakes with little or no (75
damage. However, a structure built in seismic areas
O
will generally be subjected to large or severe earth-
quakes sometime in its life span. While structures 8
can be designed to resist these earthquakes, it is not
economically feasible or realistic to design all build-
ings to withstand elastically the greatest possible <«y(Yield Deflection)
■

earthquake. Thus, in general, the analyst must con-

sider how his structure will respond inelastically to Deflection (or Strain)
earthquakes he might reasonably expect it to experi-
ence. Figure Cl. Force-displacement relationship for
elastoplastic system.
Discussion of Inelastic Action
in Vibrating Systems tionaily, while the energy is still represented by the
area under the force deflection line, only part of this
While most structural materials behave elastically work (represented by A in Figure Cl), contributes to
for small displacements, they sooner or later under- subsequent vibration. The rest of the work has been
go inelastic action. The inelastic force-displacement lost as far as vibrations are concerned. Conceptually,
relationship that is usually used in earthquake this is somewhat similar to the energy lost by
analysis is shown in Figure Cl; it is referred to as damping.
elastoplastic action. If the force or deflection is re-
moved prior to the occurrence of yielding, the As far as displacements are concerned, larger dis-
material will return along its loading line to the placements might be expected since it takes a greater
origin. If the force continues long enough, or if the deflection under the inelastic condition to store the
displacement exceeds the yield value, some perma- same amount of work as it does under the elastic
nent deformation will occur, and the structure will condition. This does not appear to be the case in
be permanently deformed unless it is subsequently actual calculations for earthquakes, however. The
deformed the same amount in the opposite direction. energy lost in the inelastic action reduces the total
energy available to the system for subsequent vibra-
In a typical single-degree-of-freedom system (Fig- tions and thus reduces the deflections which are sub-
ure Al) that is responding elastically, the response sequently observed.
continues to act along the elastic action line shown in
Figure Cl. This is the basis for the analysis con- Comparison of Elastic and inelastic
tained in Appendix A. The force on the mass, or the Response Spectra
measure of the acceleration, is directly proportional
to the deflection. The period of the structure and the In Figure Bl the elastic response spectrum for a
frequency are inversely related, and the energy of the typical strong motion earthquake is reproduced.
system ('/J mV2) is related to the triangular area Also shown are the bounds of the ground displace-
under the elastic action line. ment, velocity, and acceleration. As noted in Appen-
dix B, some generalizations can be made about this
With an inelastic system, these fundamental rela- spectrum. When the frequency of the structure is
tionships do not hold true for the entire vibration. small—the mass is large with respect to the stiffness
For instance, the force on the mass can never exceed —the maximum deflection of the system is equal to
the force which occurs when yielding occurs. Thus the maximum ground deflection. This occurs be-
when yielding occurs, the acceleration is reduced cause the mass remains relatively stationary as the
from what it would be for the same system with the ground moves; thus, the relative deflection ap-
same deflection if the system remained elastic. Addi- proaches the value of the deflection of the ground.

33
 DISPLACEMENT

ti/ ACCELERATION-^ X.

Figure C2. Design spectra. From N. M. N:wmark and W. J. Hall. "Procedures and Criteria
for Earthquake Resistant Design," Building Practices for Disaster Mitigation (Department of
Commerce. February 1973).

At the other extreme, when the frequency is large For the reasons discussed above, the displace-
—the mass is very small relative to the stiffness—the ment, velocity, and acceleration spectral values are
acceleration of the mass equals the acceleration of no longer related by Xmax = wXmax and Xmax =
the ground. This occurs because the spring is so stiff w2 Xniax. but certain of the relationships still apply.
that there is little relative deflection between the In the low frequency region the maximum displace-
mass and its base, and the accelerations applied at ment is still equal to the maximum ground deflec-
the base are directly felt by the mass. tion, and in the high frequency region the maximum
acceleration is equal to the maximum ground accel-
Between these two extremes is the region of ampli- eration. In between, however, there is a transition
fication where for any one frequency, the displace- region that needs explanation. Before discussing this
ment, velocity, and acceleration of the ground are region, the relationship of displacement and acceler-
amplified by the dynamic system. .iti'.vn at the extremes will be discussed.

Figure C2 shows a typical design spectrum cur- In the low frequency region where the displace-
rently recommended for use when inelastic action is ment is preserved, the acceleration of a structure is
anticipated. In this figure, the line D-V-A-A0 repre- reduced. Since the force tor an clastoplastic struc-
sents the elastic spectrum shown in Figure Bl. ture does not increase when vicldinp occurs, the

34
■HMM*

acceleration reaches a maximum value then also case, however, the designer should check whether
(Figure CD. Thus the acceleration is reduced by the the structure did in fact achieve the established ^
factor 1 /\i where \i is the ratio of X maximum to X factors. If it did not, the structure or w factors should
elastic. At the other end of the spectrum, it can be be modified and calculations redone.
shown that the accelerations are preserved; that is.
the maximum accelerations of the system equal the In this way, an inelastic analysis of a structure can
maximum ground accelerations, while the deflec- be performed.
tions are greater than the elastic deflections. Be-
tween these extremes, the energy in the system must Example
be preserved; the lines in Figure Cl differ by a value
\l 2\i'\ which is derived from conservation of en- The following example illustrates the use of the
ergy methods. design spectra in the inelastic case. Assume the same
example structure used in Appendices A and B and
In summary, the line D'-V'-A'-AQ represents the the ground motion used in Appendix B. Also assume
plot of maximum accelerations with inelastic action. (hat a ductility factor of 1.5 is allowed, resulting in
The line D-V-A"-A0" represents the plot of maxi- the design spectra shown in Figure C3.
mum displacements with inelastic action, and the
line D-V-A-A0 represents the interrelated values of The following results are achieved:
displacement and acceleration when elastic action
applies. Dbplaccniciit Accclcntlon

First mode 22.0 in. 3.2 g

Inelastic Modal Analysis Second mode 6.7 in. 4.5 g
Third mode 2.8 in. 4.5g
The application of modal analysis to inelastic
structures is similar to that for elastic structures, Combining these as before gives the following dis-
except that for elastoplastic spectra the results can placements:
be used only as an approximation of the expected
First Maximum Expected
response.
Mode Abtolulc Value Value

One of the first values the designer must establish Mass one 10.4 in. 12.9in. 10.6 in.
is the value of the ductility (JJ) that will be allowed. Mass two 20.8 in. 23.11.1. 20.9 in.
For structures that must continue functioning after Mass three 31.3 in. 34.9in. 31.5in.
an earthquake, a value of 1.1 to 1.2 is appropriate.
For values much larger than this, significant struc- and accelerations:
tural damage may result and the structure will not be
Flnl Maximum Expected
functionally effective. While values up to S may be Mode Abiolute Value Value
allowed before collapse occurs, the damage when \x >
1.2 to 1.3 will generally be too great to allow func- Mass one 1.52 g 3.9 g 2.3g
tional use after the ground motion. Mass two 3.03 g 5.1 g 3.4 g
Mass three 4.55 g 7-2 g 5.1 g

Having established p. the designer then prepares

the response spectrum by using the p factors and the
elastic response spectrum. The elastic periods of the The displacements expected are comparable to
modes of interest are then calculated and the maxi- those calculated in Appendix B. but the accelera-
mum value of the expected response from the spec- tions are considerably reduced because the allowed
trum determined. These values should then be used ductility reduces the maximum force the structure
as they would be for an elastic evaluation. In this experiences.

35

. .v ■... . ■ ■■
..■>:■" -. » *^
 1000

II 02 0.5 I 5 10 20 50 100 200 500 1000

FREQUENCY, cps

Figure C3. Design spectrum for elastoplastic system with ^ = 1.5.

REFERENCES

Agbabian Associates, Existing Capacity and Newmark, N. M. and W. J. Hall, "Procedures and
Strengthening Concepts for Letterman and Criteria for Earthquake Resistant Design."
Hays Hospitals {Task 8), draft report (Con- Building Practices for Disaster Mitigation (De-
struction Engineering Research Laboratory, partment of Commerce. February 1973).
April 1974).
Seismic Design for Buildings. TM 5-809-10/NAVF-
Agbabian-Jacobsen Associates, User's Guide for AC P-355 AFM 88-3. Chapter 13 (Departments
GEN SAP Code (U.S. Army Corps of Engineers, of the Army, the Navy, and the Air Force. April
Huntsville Division, May 1972). 1973).

Housner, G. W., "Earthquake Resistant Design Timoshenko. Stephen. Vibration Problems in Engi-
Based on Dynamic Properties of Earthquakes," neering (Van Nostrand. 1955).
Proceedings, Vol 53 (American Concrete Insti-
tute, 1956-57). pp 85-98. Wilson. E. L. and H. H. Dovey. Three Dimensional
Analysis of Building Systems—TABS. Report
Myklestad, N. O., Fundamentals of Vibration No. EERC 72-N (University of California. De-
Analysis (McGraw-Hill, 1956). cember 1972).

36