Structural Dynamics

by Finite Elements

William Weaver, Jr.

Stanford University
Paul R. Johnston

Failure Analysis Associates

l'RI N ncr 11/\11 , INC., I nglc'wood Cliffs, New Jersey 07632

I 1111,111 1/ I ,111#111 1111,1/uNIIIN /11 /'11/1/1111//1111 /111111

WI ,v, 11 , 111 , Wll I lhM, (,1111~)

~11111 1111111 <1y,111111lu hy llnhc clc111cnl.

f1'1~1ilh 1 llnll civil cnijlncc, lng nod Cllijlnccrln
1111; illll lh II ~dl'oli}

1111h11h hlhll11K11111hlcs 1111d Index.

I %1111111,11 dy11111nlc,. 2. l'inite element method.
IL Title.
Ill ~" '"
IAMI WIJ 11/H/
624. 1'7 1
86-12257
l~IIN II I I H1 110H (1

I 11,11N11 1N, l'hlll ll .. (d111c)

Contents

I111l1111 l11l/prnducllon supervision

111111 l111t11l111 deNlgn: Kuthryn Pavelec
<'11vt11 d1Nl1111: l~dRnl Enterprises
M111111t 1w1111 l11g

huycr: Rhett Conklin

l'IUWl'H'IMIAl ,L CIVIL ENGINEERING AND ENGINEERING MECHANICS SERIES

W .I 1/111/, 111/ror

n, l'lll / hy Prentice-Hall, Inc.

Preface

A I >lvlNl1111 ol Sl111on & Schuster

I 1111lrw1111d <'llll'H, New Jersey 07632

1
1// 1/11/111 , , 1,1111w/, No part of this book may be
1,111111/11, ,.,1, lt1 1111v fmw or by a11y means,
11'///,11111111111/1,1/1111 /11 wrlring from the publisher.

IX

Introduction to Structural Dynamics

1.1 Structural Dynamics Concepts
1. 2 Dynamic Influences
4
I . 3 Discretization by Finite Elements
1.4 Computer Programs
9
References
9

l'il1111tl h1 1111, l l11il1d Stutes of America

Ill

lj

II

(I

5 4 3 2

2
t ~llN

1, '3- 853508-6

025

I 1ill'N Ill I II Al I. IN I I'll NATIONAL (UK) LIMITED, London

11/tl 'N II( I " "' ,. ()11 A USTRALIA J:>rv. LIM!TBD , Sydney
11111 NIii I II AI I ( 'ANAl)A INC'., Toro1110
11111 N 111 I I IAI I I IISl1ANOAMIHUCANA, S.A., M exico
1'111 N111 I' I IAI I. 011 INlllA P1UVATI! LIMl'l'l'.D , N1w {)('/!,/
1'111 N111 I I IAI I <II' JAPA N, IN(',, 'J'ok yn
1'11 1 N Iii I II AI I (Il l S0\ 11111.AS'I' AS IA 1111. I.ii>,, Sfl1111111m r
I 11111111 ,\ l'IIINIIC 'I' II AII llO IIIIASII , I ll >A, U/11t/1 .lr111r/10

w111 ,111A1, 111,,1,.,~ 1.1~11,11 ,, w,,1111111 11111,

N,,, 1tl/ll111,,,

Systems with One Degree of Freedom

2.1
2.2
2.3
2.4
2.5
2 6
27
2.8

Introduction
10
Free, Undamped Vibrations
12
Harmonic Forcing Functions
19
Effects of Damping
25
Periodic Forcing Functions
35
Arbitrary Forcing Functions
38
Step-by-Step Response Calculations
Response Spectra
51
Rl'fl'n.lllt'CN
59
Prohl1~n,11
59

10

45

Ill

Contents

Iv

Finite Elements and Vibrational Analysis

73

Contents

Normal-Mode Method
of Dynamic Analysis
4.1
4 .2
4.3
4.4
4 .5
4 .6
4 .7
4. 8
4 ,l)
4. I()

138

Introduction
138
Principal and Normal Coordinates
139
Normal-Mode Response to Initial Conditions
147
Normal-Mode Response to Applied Actions
152
Normal-Mode Response to Support Motions
157
Damping in MDOF Systems
164
Damped Response to Periodic Forcing Functions
168
Damped Response to Arbitrary Forcing Functions
172
Step-by-Step Response Calculations
175
Program NOMO for Normal-Mode Response
177
References
192
Problems
192

Direct Numerical Integration Methods

'i. I

Introduction

195

5.2 Extrapolation with Explicit Formulas

5.J
5.4

5.5
5. Cl
5.7

197

203
Iteration with Implicit Formulas
211
Direct Linear Extrapolation
2 17
Ncwmark's Generalized Acceleration Method
223
Numericul Stahility und Accurucy
225
J>rogrulll DYNA for Dynamk RlsponSl'
7'<,
Rl'lb l'lll'l'~
l I1
l '111llh111~

Framed Structures
6. 1
6.2
6.3
6.4
6.5
6.6
6.7
6.8
6.9

73
3. 1 Introduction
Stresses
and
Strains
75
3.2
Equations
of
Motion
for
Finite Elements
78
3.3
One-Dimensional
Elements
82
3 .4
95
3.5 Transformation and Assemblage of Elements
105
3.6 Vibrational Analysis
112
3.7 Symmetric and Antisymmetric Modes
118
3.8 Program VIB for Vibrational Analysis
References
124
Problems
125

195

Introduction
241
Plane Frames
244
Grids
249
Space Trusses
253
Space Frames
259
Programs for Framed Structures
Guyan Reduction
282
Constraints Against Axial Strains
Programs DYPFAC and DYSFAC
References
303
Problems
303

264
290
299

Two- and Three-Dimensional

Continua
310
7 .1
7. 2
7. 3
7.4
7. 5
7.6
7. 7
7 .8
7. 9
7. 10

241

Introduction
310
Stresses and Strains in Continua
310
N atura1 Coordinates
318
Numerical Integration
326
Isoparametric Quadrilaterals for Plane Stress and Plane
Strain
333
Program DYNAPS for Plane Stress and Plane Strain
340
Isoparametric Hexahedra for General Solids
345
Program DYNASO for General Solids
351
Isoparametric Elements for Axisymmetric Solids
357
Program DYAXSO for Axisymmetric Solids
365
References
369

Plates and Shells

8.1
8.2
8.3
8.4
8.5
8.6

8.7

370

Introduction
370
Element for Plates in Bending
371
379
Program DYNAPB for Plates in Bending
Element for General Shells
382
390
Program DYNASH for General Shells
Element for Axisymmetric Shells
394
406
Program DY AXSH for Axisymmetric Shells
410
Rcf'L-n:nccs

Contents

vii

Contents

111

')

Rigid Bodies within Flexible Structures

9. 1
9. 2
9. 3
9.4
9.5
9.6

to

Introduction
411
Rigid Bodies in Framed Structures
413
Program DYRBPF for Rigid Bodies in Plane Frames
Rigid Laminae in Multistory Buildings
425
Rigid Bodies in Finite-Element Networks
434
Program DYRBPB for Rigid Bodies in Plate-Bending
Continua
438
References
443

Substructure Methods

444

IO .1 Introduction
444
10.2 Guyan Reduction Methods
445
10.3 Modified Tridiagonal Method for Multistory
Buildings
457
10 .4 Programs DYMSPF and DYMSTB
463
10.5 Component-Mode Method
468
471
10.6 Component-Mode Method for Trusses
10.7 Programs COMOPT and COMOST
477
References
482

Notation

483

<~< 11wral References

Appendix A

Appendix 8

491

Systems of Units and Material

495
Properties
A. l Systems of Units
A. 2 Material Properties

495
497

Eigenvalues and
Eigenvectors

498

n.1

n.2

Inverse llcmtion
498
Tra11sfo1111ntion Methods
lhfo1l'lll'l'H
~ 17

505

411

424

Appendix C

Flowchart for Program

DYNAPT
519

Answers to Problems
Index

579

564

124

I lnltn I ln111n11t11111111 Vll1111tlw111I A11uly11l11

0

--~--

""

II " ,
I_ _ _
",--2!,JI_ _
(c)

c:tu1p. 3

125

1'111hlu11111

'I . Timoshenk o, S. P., Young, D. II. , nmI W(i:uvc1,. W., Jr ., Vibration Problems in
.
.
" I Wiley New York I n4.
4 th c,
f11g111('er111g,

'
'
E
2nd ed
,
.
W., Jr Matrix Algebra for ngmeers,
.,
IO. Gere,
J. M., and Wcaver,.
Brooks/Cole, Monterey' Calif., 1983.
.
N
Y k 1973
. Alge bra, 3rd ed . ' Macnullan, ew orD., V I.
Hohn F E. Elem entary Matrix
' P.' G., "Symmetry in Structural Meehamcs,
" ASCE ' J . Struct . 1v., o
ll. Glockner,
99, No . ST! , 1973, pp. 71- 89.

11

Figure 3.18 (cont.)

the truss is aluminum , we give parameters the following numerical values:

E = 6.9
L=5m

107 kPa

PROBLEMS

p = 2.62 Mg/ m 3

A = 6

10- 3 m 2

1.4-1.

. 1 I ment with a linearly distributed load (force per

Figure P3.4-1 shows an axia e e
- b + (b - b1)x/ L. Find the equivalent
i
unit length) given by the formula bx - .1
nodal loads Pb(t) = {pb1, Pb2} due to this mfluence. f"' { f , {, 1= [ i-{

-f.

where SI units are implied (again, see Appendix A).

To process this truss with Program VIBPT, we must analyze half the structure
twice. In the first analysis, restraints on the plane of symmetry allow only symmetrk
modes of vibration; and the second analysis uses restraint data for only antisymmetrk
modes. Figures 3.18(b) and (c) illustrate the mode shapes corresponding to the first and
second angular frequencies w 1 = 79.55 s - 1 and wz = 168.9 s - 1. We see that the first
mode is symmetric with respect to the plane of symmetry, while the second mode is
antisymmetric.

~ b,

c fo ~ 1,-,

xJ

:l J ,Y,u, f't-:."
0

],'

- I-------1 eb/f)=il I ["~"

P b2

Po,

b,

REFERENCES

Figure P3.4-1

1.4-2.

I. Weaver, W., Jr. , and Johnston, P. R., Finite Elements for Structural Analysis,
Prentice-Hall, Englewood Cliffs, N.J., 1984.

~'l

o
- I

( l

-T.. ),

l [h1+(kl,1)1 }

[(

L-)1) ~ b,+[k-l,').,

1, 1 +-(

b, - L1}-z::]

A parabolically d!stributedd ~oa:. (fo;~\:zer D~~~~::r:~e ~~sui:~e:~:~~~

bx = bz(x/L)2, as illustrate m ig.
.
loads Pb (t) = {pb1, p ,,2} resulting from this body force.

2. Zienkiewicz, 0. C. , The Finite Element Method, 4th ed., McGraw-Hill, Maidenhead, Berkshire, England, 1987.
3. Cook, R. D. , Concepts and Applications ofFinite Element Analysis, 2nd ed. , Wiley.
New York, 1981.
4. T imoshenko, S. P., and Goodier, J. N. , Theory of Elasticity, 3rd ed. , McGraw-Hill.
New York, 1970.
5. Weaver, W., Jr. , and Gere , J. M. , Matrix Analysis of Framed Structures, 2nd ed.,
Van Nostrand Re inhold, New York, 1980.
6. Oden, J. T. , Mechanics of Elastic Structures, McGraw-HiII, New York, 1967.
7. Archer, J. S. , "Consistent Matrix Formulations for Structural Analysis Using FiniteElement Techniques," AIAA J. , Vol. 3, No. 10, 1965, pp. 1910- 19 18 .
8 . Clough, R. W. , "Analysis of Structural Vibrations and Dynamic Response ," Rec.
Adv. M at. Methods Strucr. Anal. Des., ed. R. II. Onll ughe r, Y. Yamada, and J. 1~
Oden, University of Al abama Press, ll1111tsville, Alu ., I C
J7 I , pp . 25 45.

I ..I

.t

Figure P3.4-2
.
h
h
odes as shown in Fig. P3.4-3(a). In
Assume <hut an axial clement as t ree n
, 2 the uadratic displacement
l\l1ms of' th\l coordinate x measured from node '
q. f - (2x - L)x/ L z
' .
.
.
. , (h) (d) o f the figure are . 1 '
shupc
pc1rts
. e the 3 x 3 stiffness matrix
.
(I luncllons
i
4 I J) I,ini ind
(1 (2.1 I /,).,/ /,2 . 0 c, 11v
j~ l'w ;his l'll11m111( 1r't1w' nxinl ii gld lly FA is constant along the length.

.;

Finite Elements and Vibrationa l Analysis

130

1,5-7.

Figure P3.5-7 shows a two-element beam for which the parameters E, I, A, and
pare constant along the length. Assemble the stiffness matrix Ss and the consistent-mass matrix Ms (for translational inertias) in rearranged and partitioned
forms.

,I

0.6L

131

Problems

Chap. 3

CD
0----------0
2

0
0

/ ,k

0.8L

CD

\.

Figure P3.5-4

3.5-5.

.,

Figure P3.5-7

For the plane truss shown in Fig. P3.5-5, repeat Prob. 3.5-1. Assume that the
cross-sectional areas of members I and 2 are 0.8A, those for members 3 and 4
are 0.6A, and those for members S and 6 are equal to A.

I ~-It Repeat Prob. 3.5-7 for the two-element beam shown in Fig. P3.5-8.
y

CD

1~

,I

,/ ~,

~3

L~

Figure P3.5-8

I ~ 11, For the two-element beam shown in Fig. P3.S-9, repeat Prob. 3.5-7.
y

,---0.BL-- --.

i..l

Figure P3.5-5

3.5-6.

1
0

Figure P3.5-6

Repeat Prob. 3.5- 1 for the plane truss shown in Pig. P3.5-6. In this case let the
cross-sectional urcus for llH.lmhcrs I ll11011gh 4 tw NJUIII tn /\ , while those fo1
nw111lm1s 'i 111HI <, 11n l'q1111l to V? A,

G)

4ii

r,-3

.. - -L- -\- -L - - 1

-X

Finite Elements and Vibrational Analysis

134

Chap.3

3.6-7. Repeat Prob. 3.6-5 for the two-element continuous beam shown in Fig. P3.6-7.
Cross-sectional properties I and A are constant, and the free displacements at
joints 1 and 2 are both rotational.
y

Chap. 3

.t7-3.

135

Problems

For the symmetric continuous beam shown in Fig. P3.7-3, find the angular
frequencies and mode shapes for (a) symmetric and (b) antisymmetric distortions. Use only half the structure, assuming that each of the four elements has
the same values of I and A.
y

2___=2---~ _3- - x

, ____
(D=,_ _ _..

,/

i-1,/- L - ~ . - - - -- L ~
Figure P3.6-7

3.7-1. The fixed-end beam shown in Fig. P3.7-l consists of two prismatic flexural
elements having the same values of I and A. Using only half the beam, find the
angular frequencies for the (a) symmetric and (b) antisymmetric modes.
y

1.7-4. The symmetric plane truss shown in Fig. P3. 7-4 has cross-sectional areas for
members 1 and 2 equal to V2 A , whereas those for members 3, 4 , and 5 are
equal to A. Using only half the structure, calculate the angular frequencies and
the (a) symmetric and (b) antisymmetric mode shapes.

2
1~

/ 1

Figure P3.7-3

CD

ISym.

i-Sym.

3~

L= 2C

'l

Figure P3. 7-1

y

3.7-2. Figure P3.7-2 shows a prismatic continuous beam composed of four flexural
elements. Determine the angular frequencies and mode shapes for (a) symmetric
and (b) antisymmetric deformations, using only half the beam.
y

-~-...::.:~-6XX'.

0-2_0.::::::_4

01)/'.

l_,

~ L - 1 4l ~

Figure P3.7-4

For the symmetric plane truss shown in Fig. P3.7-5, determine the angular
frequencies 11nd the (n) symmetric and (b) antisymmetric mode shapes. Use only
hull' the structure, and assume that the cross-sectional areas of members 1 and
J llll' O.M . thut for mc111hcr 3 is 0.8A , und those for members 4 and 5 are equal
to A.

496

Systems of Units and Material Properties

Table A. I presents conversion factors for calculating quantities in SI units

from those in US units. The factors are given to four significant figures, which
usually exceeds the accuracy of the numbers to be converted. Note that stress is
defined in SI units as the pascal. That is,
1 Pa
TABLE A.I

1 N/m 2

Conversion of US Units to SI Units

Quantity

US Units

x Factor

Length
Force
Moment
Stress
Mass

inch (in.)
kilopound (kip or k)
kip-inch (k-in .)
kip/inch2 (k/in. 2 or ksi)
kip-sec2 /inch (k-s 2/in.)

2.540
4.448
1.130
6.895
1.751

X
X
X
X

10-2
10- 1
103
10 2

System

ST

us

Length

(GN/m2 or GPa), acceleration ii in meters per second squared (m/s2 ), and so on.
[Note that the force kilo newton corresponds to the mass megagram (Mg).] Also,
in US units we give force Pin kilopounds (kips o r k), length Lin inches (in.),
modulus E in kips per square inch (k/in. 2 or ksi), acceleration ii in inches per
second squared (in./s2), and so on.

= SI Units

meter (m)
kilonewton (kN)
kiloncwto n meter (kN m)
kilopascal (kPa)
megagram (Mg)

Consistent Systems of Units

Force

497

Material Properties

A.2 MATERIAL PROPERTIES

For any numerical problem in structural mechanics, we must use a consistent system of units. By this we mean that all structural and load parameters must
be expressed in the same units within each system. Some examples of consistent
units for force, length, and time appear in Table A.2. For instance, in SI(l) w1.
must express an applied force P in newtons (N), a length L in millimeters (mm),
the modulus of elasticity E in newtons per square millimeter (N/mm2 ), an
acceleration ii in millimeters per second squared (mm/s2 ), and so on.
TABLE A.2

Sec. A.2

Time

(I)
(2)
(3)

newton
kilonewton
meganewton

millimeter
meter
kilometer

second
second
second

(I)
(2)
(3)

pound
kilopound
megapound

foot
inch
yard

second
second
second

When programming structural dynamics for a digital computer, it is l'"

pecially important that the system of units for input data be consistent. Othl'I
wise, units would have to be converted within the logic of the program , thcrlhy
restricting its usage. For example, if in US units the length L were given in h-rt
and the modulus E were expressed in pounds per square inch, the program woulll
need to convert either L to inches or E to pounds per square foot.
For all of the numerical examples and problems in this book wc IISl' 1.11 hrr
S1(2) or US(2) in Table A. I . Thus, in SI units wc take force I' in kilo11l'Wl1111,
(kN), 11.!ngth f, in llll'lcrs (111), mocl11l11s /\' in gip1111l'Wl<llls 1w1 Nq 1111n 1mll I

To analyze solids and structures composed of various materials, we need to

know certain physical properties. For structural dynamics, the essential material
properties are modulus of elasticity E, Poisson's ratio 11, and mass density p.
Table A.3 gives these properties in both US and SI units for some commonly
used materials. Note that the shearing modulus G is not listed in the table
because it can be derived from E and 11.
TABLE A.3

Properties of Materials"

Modulus of Elasticity E
GPa

Poisson's
Ratio v

69
103
103
97
25
45
207
207
117
379

0.33
0.34
0.34
0.25
0.15
0.35
0.31
0.30
0.33
0.20

Material
k/in. 2
Aluminum
Brass
Bronze
Cast iron
Concrete
Magnesium
Nickel
Steel
Titanium
Tungsten

1.0
1.5
1.5
1.4
3.6
6.5
3.0
3.0
1.7
5.5

10
104
10
10
103

103

X
X

10
104
104

104

X
X
X

Mass Density p
k-s 2/ in. 4
2.45
8. 10
7.80
6.90
2.25
1.71
8 .25
7.35
4.20
1.80

X
X
X
X

X
X
X
X
X
X

10- 7
10-7
10- 7
10- 7
10- 7
10- 7
10- 7
IQ- 7
10- 7
10 6

Mg/m3
2.62
8 .66
8.34
7 .37
2.40
1.83
8 .82
7.85
4.49
19.2

"Numbers in this table are taken from J. M. Gere and S. P. Timoshenko, Me chanics of
Materials, 2nd ed . , Brooks/Cole, Monterey, Calif., 1984.

Preface

Structural Dynamics by Finite Elements represents a culmination of the two

topics identified in its title. Structural dynamics continues to grow as an essential
subject for structural engineers , and the best method for handling structural
dynamics problems is with finite elements. The solids and structures discussed
herein are subjected to time-varying influences that cause accelerations and
velocities as well as displacements, strains, and stresses. To analyze such a
problem, we discretize the structure (or solid) by dividing it into a network of
elements having finite sizes. Then equations of motion are written for the nodes
(or joints) of the discretized continuum, which include inertial and damping
actions. This finite number of differential equations may be solved on a digital
computer to find approximate time-varying nodal displacements and stresses
within the finite elements.
This publication is intended to be used as a textbook for a graduate-level
course on structural dynamics in civil, mechanical, or aeronautical and astronautical engineering. We have tried to present the material in a clear, forthright
manner for either a university student or a structural analyst in industry. As
background for studying this book, the user should have had the mathematics
and solid mechanics usually offered in schools of engineering at the undergraduate level. Included among the former courses are differential equations,
matrix algebra, and computer programming. Other desirable courses are vibration theory, matrix analysis of framed structures, and finite elements. Although it is not necessary to know the theories of elasticity, plates, and shells,
previous exposure to those topics will give the reader greater perspective on the
present work .
he

"

Preface

( 'hapter 1 introduces structural dynamics by compuring time-varying ~e11p1111Nl'S to dynamic loads against results for static loads. Next, we describe
11111111u>nly encountered dynamic influences, which are initial conditions, a~plild nctions, and support motions. Then discretization by finite elements 1s
dl l'l t'IISNcd for framed structures, two- and three-dimensional continua, plates,
111111 11hclls . Such analytical models are processed by digital computer programs,
111'1 11ll'lllioned in the last section of this introductory chapter.
In preparation for later work, we describe and analyze systems with only
mil' degree of freedom in Chapter 2. This material would be a useful review for
11 pl'l'Non who has had a previous course on vibration theory for one-degree
11yNll111s. It also serves as an introduction to the subject for anyone without such
11 h11ckground. Moreover, this book on structural dynamics becomes more selfl'1111t11i11cd by including material on one-degree systems.
We present derivations for finite elements and vibrational analysis in Chaph1 \, where all of the discretized structures have multiple degrees of freedom.
1~1111HY consistent stiffness, mass , and load matrices are developed for onedi11u111sional elements for later use in framed structures. From the homogeneous
101111 of the nodal equations of motion for an assembled structure, we can solve
1h1 l'lgcnvalue problem for vibrational frequencies and mode shapes. If planes
111 Hy111111etry exist, appropriate restraints at nodes on those planes allow us to find
11y11111mtric and antisymmetric modes, using only part of a structure. At the end
111 <'lluplor 3, we describe Program VIB for vibrational analysis of finite-element
mtworks, with specialization to Program VIBPT for plane trusses.
( 'huptcr 4 contains the normal-mode method for dynamic analysis of
dlN1 1l'li:1.cd structures. By this approach, we transform the equations of motion
111 11111111111 coordinates, where each flexible-body or rigid-body mode has unit
11111NH und responds as if it were a system with only one degree of freedom. We
d1 t1111sN normal-mode responses to initial conditions, applied actions, and sup11111 I 111otions for structures with or without damping. Step-by-step response
l 11il'11l111io11s ure explained and coded in Program NOMO for normal-mode
111111lyNis. This program is specialized to become NOMOPT for plane trusses,
wll1l'11 i11l'l11des VIBPT from Chapter 3.
111 C'lwpter 5 we cover direct numerical integration methods.for calculat~ng
dy1111111i1 r<.~sponscs. These approaches m~y all be. characterized a~ fim~edtlll' ll' lll'C approximations with respect to time. We discuss extrapolation with
,xplkil loa111ulas, iteration with implicit formulas, and direct linear extr~po!11111111 , with solution for incremental displacements. Newmark's generalized
11n'l'll'l'tllio11 method is extended by Hilber's approach and applied in Program
I >YNA lor dynamic responses of structures. Then this program is specialized to
t,,1111111 J)YNAPT for plane trusses, which includes VIBPT from Chapter 3 and
N< >M< WI ' fro111 Chapter 4. Thus, in several stages we construct a program that
wlll l11111dll 1101 only vihrntionul 111111lysis but ulso two types of dynamic response
l llll'll lllt iollN
<'011npl~ d!'V!'lop1d i11 p11vil111 Nl h11ptl1:, loa pl111ll' la llHHl1H nn: l'Xll11dld to

Preface

xi

all other types of framed structures in Chapter 6. The programs described are
DYNACB for continuous beams, DYNAPF for plane frames , DYNAGR for
grids, DYNAST for spaces trusses, and DYNASF for space frames. We also
discuss methods for reducing the number of degrees of freedom for beams,
grids, plane frames, and space frames. Guyan reduction may be used to eliminate joint rotations from the equations of motion for these four types of framed
structures. Also, constraints against axial strains can be imposed in plane and
space frames; and Programs DYPFAC and DYSFAC have been coded for this
technique.
Chapter 7 describes finite elements to be used for dynamic analyses of twoand three-dimensional continua. We employ isoparametric quadrilaterals for
solving two-dimensional problems in plane stress or plane strain and present
Program DYNAPS for that purpose. Isoparametric hexahedra are applied in
calculations for general solids, and we have coded Program DYNASO for
obtaining their dynamic responses. Axisymmetric solids require the use of ring
clements having cross sections that again are chosen to be quadrilaterals. Program DYAXSO for dynamic analyses of axisymmetric solids completes the set
of programs documented in this chapter.
In Chapter 8 the finite elements for analyzing plates in bending and general
11nd axisymmetric shells are based on those for general and axisymmetric solids
in Chapter 7. These specializations automatically include the effects of shearing
deformations and rotary inertias. The programs described are DYNAPB for
plates in bending, DYNASH for general shells, and DYAXSH for axisymmetric
Hhclls.
Chapter 9 explains the effects of including rigid bodies in the analytical
models for framed structures and other discretized continua. For framed structures the convenient approach is a member-oriented technique with rigid offsets
nl lhc ends of each member. On the other hand, for finite elements with more
lh1111 two nodes, a body-oriented method appears to be mandatory. Sample
paograms discussed in this chapter are DYRBPF for rigid bodies in plane frames
111,d DYRBPB for rigid bodies connected to plates in bending. We also describe
1IHid laminae in multistory buildings in preparation for the next chapter.
The topic of Chapter 10 consists of . substructure methods for dynamic
111111lysis. We divide the subject into Guyan reduction methods and the
111111ponent-mode technique. Within the first approach, emphasis is placed on the
111otli fled tridiagonal method, which is applied to multistory buildings. Programs
I lVMSPP and DYMSTB have been coded to handle multistory plane frames and
th I buildings with rectangular framing patterns. On the other hand, the theory
111 thl component-mode method is explained and implemented for the analysis
111 plnnl' und space trusses having only a few members. Programs COMOPT and
I '<>M< >ST represent the codes for such analyses.
At tlu: hnck of the book we give a list of notation, general references,
11ppl'11dil'l'H, 11ml 1111swl'l's to prnhlcms . Appendix A describes systems of units (SI
111111 l JS) 1111d physku l prop,~1tlt1, fol' v111l111111 mntcrinls . In Appendix B we divide

Preface

KIi

Nolution of eigenvalue problems into the topics of inVl'l'Sl' ill'rttlion (for few
11uidcs) and transformation methods (for many mod<.!:-.) . 'fhl' lutLer methods
include those of Jacobi, Givens, and Householder. Our finul approach consists
ol I louseholder transformations of the coefficient matrix to 11idiugonal form,
lollowed by iteration with the QR algorithm. Spectral shifting may be used to
l111prove the convergence of either inverse or QR iteration . Last, Appendix C
contains the detailed flowchart for Program DYNAPT.
As mentioned in Chapter 1, all of our computer programs are coded in
FORTRAN. These codes and data for examples have been assemb led on a
11111gnetic tape, a copy of which can be obtained from Paul R . Johnston for a
nominal fee. His business address is: Failure Analysis Associates, 2225 East
llayshore Road, Palo Alto, California, 94303.
We wish to thank graduate students and teaching assistants at Stanford who
hnve directly or indirectly contributed ideas for this book. Needless to say, our
wives Connie and Terry have shown much patience and consideration while we
wen: engrossed in its development. As before, Suzanne M. Dutcher did an
outstanding job of typing the manuscript, and working with her was a great
pleasure. Failure Analysis Associates of Palo Alto allowed us free computer
mage, for which we are most grateful. Also, a Ford Foundation grant from the
l'rnvost at Stanford provided funds to partially offset our expenses.

Introduction
to Structural Dynamics

1. 1 STRUCTURAL DYNAMICS CONCEPTS

WILLIAM WEAVER, JR.
PAUL

R.

JOHNSTON

i\(.'knowledgement

This book was written in collaboration with C. Lawrence Loh, Staff

E11gint!er, Engineering Information Systems, Inc., San Jose, California. Larry
rnmposed some of the computer programs, provided computer examples, and
dll'cked the Answers to Problems.

Il' a solid or a structure is loaded very gradually, it is said to be in a state of static

,quilibrium, for which static actions and reactions equilibrate each other. In such
11 situation, time has no significant influence; and static analysis may proceed
without considering this variable. On the other hand, if forces are rapidly
11pplied , the solid or structure is said to experience dynamic loads . In th is case
we can say that a state of dynamic equilibrium exists, for which time-varying
lll'lions and reactions equilibrate each other at every instant.
To examine some of the differences between static and dynamic analysis,
ht us consider the prismatic cantilever beam in Fig. 1. l(a). This beam is loaded
with a concentrated force P(t) in they direction at its free end (point 2) . If the
loud is applied slowly , the static displacement at point 2 given by elementary
ht11111 theory is

(1)

111 which m is lhc flexural rigidity of the cross section. The response curve
1,ilwlld I in Fig. I. I (c) shows that by gradual application, the load produces an
,1w111plotk value of (1J2).i, In add itio n, the displacement v (x) at any point along
IIH' lt11th of the beam is a function of r only.
Al tht olhl'I' l'Xlrc111t., suppose thut the force in Fig. I. I (a) were applied
l11~1111111111t11111,ly 111 tl11s l.'ll~l' thl' h l'HIII will not only displace but will also
1111 rlr111lt 111 l'Vl'IY p111111 111011 1h ll11pth , 'l'ht.1dort , the displacement D(x, t)

Preface

11111rn of' eigenvalue problems into the topics of invcl'Sl' lll'f"lllion (for few
ul1N) nnd transformation methods (for many modes) . Tlw laller methods
l111ll' those of Jacobi, Givens, and Householder. Our finul approach consists
11011,dlOlder transformations of the coefficient matrix lo 11 idiagonal form,
lh,wld by iteration with the QR algorithm. Spectral shifting may be used to
1prnv1 the convergence of either inverse or QR iteration . Lusl , Appendix C
11111111!-. the detailed flowchart for Program DYNAPT.
I\~ mentioned in Chapter 1, all of our computer programs are coded in
>l<'IRAN . These codes and data for examples have been assembled on a
111mlil' tupc, a copy of which can be obtained from Paul R. Johnston for a
111111wl Ice. His business address is: Failure Analysis Associates, 2225 East
11yi,ltrnc Road , Palo Alto, California, 94303.
We wish to thank graduate students and teaching assistants at Stanford who
uvr dirl'Ctly or indirectly contributed ideas for this book. Needless to say, our
1v1'1 C'onnie and Terry have shown much patience and consideration while we
111 111grnsscd in its development. As before, Suzanne M. Dutcher did an
11hl11mltng job of typing the manuscript, and working with her was a great
h11,111 l.'. Failure Analysis Associates of Palo Alto allowed us free computer
11111-11 , Im which we are most grateful. Also, a Ford Foundation grant from the
111,v11,1 111 Stanford provided funds to partially offset our expenses.

Introduction
to Structural Dynamics

1.1 STRUCTURAL DYNAMICS CONCEPTS

WILLIAM WEAVER, JR.
PAUL

R.

JOHNSTON

Atknowlcdgement

This hook was written in collaboration with C. Lawrence Loh, Staff

I t1Hlnl'l't , Engineering Information Systems, Inc., San Jose, California. Larry
111111pw,1d some of the computer programs, provided computer examples , and
1h11 k1d tlw Answers to Problems.

If a solid or a structure is loaded very gradually, it is said.Jo be in a state of static

equilibrium, for which static actions and reactions equilibrate each other. In such
a situation, time has no significant influence; and static analysis may proceed
with?ut considering this variable. On the other hand, if forces are rapidly
apphed, the solid or structure is said to experience dynamic loads . In this case
we_ can say that ~ state of dynamic equilibrium exists, for which time-varying
actions and reactions equilibrate each other at every instant.
To examine some of the differences between static and dynamic analysis,
let us consider the prismatic cantilever beam in Fig. 1. l(a). This beam is loaded
with~ conc~ntrated force P(t) in they direction at its free end (point 2). If the
loud ts applied slowly, the static displacement at point 2 given by elementary
hcum theory is

(v2\1

PL 3
3/

(])

111 which EI is the flexural rigidity of the cross section. The response curve
lulwlcd I in Fig. 1. I(c) shows that by gradual application, the load produces an
11Ny111plotic value of (v2)<1. In add ition, the displacement v(x) at any point along
tlt1 l1ngth of the beam is a function of x only .
/\1 lhl' olhlr l'Xlrl'mc, suppose that the force in Fig. l. l (a) were applied
1111,l1111111111m1sly 111 1hi1, l'H/'ll' 1h1 lwa111 will not only displace but will also
1111 l'lrtilll' ,,1 rv11y p111111 al1111F tis l1nr11t Thl'tcforc, the displacement v(x, r)

111111111111 tlon to !itru1,111r1I Uvnmh 11

<.hnp. 1

,~
1

/ ~

P(t)

'2

(a)

(2a)

y, v

the inertial force,

P(t)

P(t)

m!2t

1~--~~~~~~---

,/

!1tr111 t111nl lly111011h1 <.11111 npt1

lwrn1ms II fu1K'lto1111f holh spm.'l' (.t) und time (t). Although it is possible to write
n pmliul di ffcrcntiul equul ion of motion for dynamic equilibrium of this simple
structure [ 11, * that approach will not be pursued in this book. Instead, we will
nlwuys discrctize the structure, as discussed later in Sec. 1.3.
A crude discretization of the cantilever beam problem is represented by the
unalytical model in Fig. l.l(b). Appearing at point 2 is a concentrated mass m,
representing some fraction of the distributed mass of the beam. Also shown at
point 2 is a hypothetical dashpot damper that generates a dissipative force in
proportion to velocity. The three types of forces in the figure opposing the
applied load are the elasticity force,

Y, V

v(x,t)

"" 1.1

c~

t
ilt

PE

(2b)

and the dissipative force,

X

PD = cv(t)

where c is a damping constant. From D' Alembert's principle we have

PM P0

(b)

(2c)

P(t) - Pe - PM - PD

=0

(3a)

which expresses dynamic equilibrium of the mass m. Substituting Eqs. (2) into
Eq. (3a) and rearranging yields
(3b)

0
(c)

li'IJ(lll't' 1.1 (u) Beam with distributed mass; (b) lumped-mass approximation; (c)
Nllllll 1111d dynumic rcspon~cs.

Solution of this ordinary differential equation of motion by the theory in Chapter

2 produces the responses in Fig. l. l(c) labeled 2 and 3. Curve 2 is the response
when the damping constant c is zero, and curve 3 represents a case of nonzero
damping.
For perspective, let us review what has been shown by this example. First,
we replaced a structure having an infinite number of degrees of freedom with an
analytical model having only one degree of freedom. That is, in the former case
the mass was distributed over an infinite number of points along the length of
the beam, whereas in Fig. l. l(b) the single concentrated mass m exists only at
the end. This simplification separates the variables of space and time both
physically and mathematically. Therefore, Eq. (3b) is characterized as an ordinary differential equation instead of a partial differential equation.
Second , we described the response at point 2 due to the applied force for
both gradual and instantaneous rise times. In Fig. l. l(c) we can see that the
maximum value of curve 2 is twice that of curve l. This figure dramatically
illuslruks lhe difference between responses due to static and dynamic loads of
~N11111h1'1~ 111 h111t' k1' I~

l11dl111tl'

11fc1cm11 111

lhl end of 1hc chapter.

l1111111h1t lh111 111 hl11111111nl llv11n111l1 11

< l111p I

I'

lht ~11111l' 11111g1111t11k . <'111w .> 11st ill11h ~ 111,11111 llw displm1d p11111111111 (r 1),1 wllh
11 p,1111d '/' L'quul lo lhc 11a111ral 11,,wrl t!/ 1i/1mtlo11 for lht 111111ly11c11I 111odd.
'l'hinl, we included the possih1ll1y of <.l ccuying motion L'Ullscd by the
pll'Nl'lll'C of u hypothetical damping mechanism . Curve 3 in Fig. I . I (c) shows

fill' cffct:l of a dissipative force is to suppress the response as time passes.

Allltough this influence is of interest to us, it is not nearly so important as the
1ll1l'I of lhe inertial force due to the presence of the mass.
This simple example also demonstrates the steps an engineer takes when
1o11lvi11g a slructural dynamics problem . They are:
111111

I . lktine the problem.

1

<'rnnposc an analytical model.

C'uk'u late the response.

ll y lh1 word re.wonse we mean time-varying displacements, stresses, internal

1H 1111111,,

1111d so on, that may be of interest. Such quantities may be expressed as

values, or response spectra, depending on the nature

111111 ltl i.. lol'ics, maximum

111 llt1 problem .

(el

1,2 DYNAMIC INFLUENCES

V111 irn1.~ 11nlurnl and man-made influences may cause dynamic responses in
Nl11wlt11l'S. The rnost common types of influences are initial conditions, applied
,1, t/011.1, 1111d s11pport motions . Figures 1.2(a) and (b) show what we mean by
111lli11 I 11111ditions that result in dynamic responses. In the first case, a crane

'
(f )

(g)

P(t)

(11)

lb)

le)
(11)

Figure 1.2

Dynamic innuences.

l1'l1111r11 1,2

(('(l11 / .)

Introduction to Structural Dynamics

Chap. 1

1111ddcnly drops its load and rebounds from an initial condition of static-load
tltsplucements. In the second instance , a truck backs into a loading platform at
so11,c initial velocity. Consequently, the frame and body of the truck respond
tlynumically.
In Figs. l .2(c) and (d) we illustrate two types of applied actions in the
lrnms of lateral loads that act on a building. Curve 1 in Fig. 1.2(d) depicts the
1111 cc of an air blast that rises very suddenly, decreases more slowly, and also
hl'l'Omcs negative. On the other hand, curve 2 represents the force of a wind gust
Ihat rises and falls gradually, but still fast enough to induce significant accelerntions. Some other types of applied actions appear in Figs. 1.2(f)-(h). The
machinery force acting at the third level of the frame in Fig. l .2(f) follows a
11i111plc harmonic function, as shown by curve 1 in Fig. l .2(g). On the other
hnnd , the moving load on the plate in Fig. l .2(h) may have any time variation.
Two kinds of support motions are indicated in Figs. l .2(c) and (e)-(g).
'lhc first type consists of rigid-body ground accelerations due to earthquake,
111,tninst which most modern buildings and other structures are designed. Figure
1,2.(c) shows a typical time history of such an aperiodic ground acceleration. A
N<'l'ond type of restraint motion is the independent displacement implied at the
111,t ht hund support in Fig. l .2(f). The time history specified for such an induced
d111placcment can be arbitrary. For example, it may vary according to the squarewnvc pattern labeled 2 in Fig. l.2(g).

1.3 DISCRETIZATION BY FINITE ELEMENTS

To Ntt up an analytical model for dynamic analysis, we will discretize solids and
NII lll'htrcs using the method of fmite elements. The applications include framed
11t111l'l11rcs [2J, two- and three-dimensional continua, plates, and shells [3]. The
1111111 tlernent method enables us to convert a dynamics problem with an infinite
1111111lll1 of degrees of freedom to one with a finite number in order to simplify
1h1 1mlution process. Ordinary differential equations of motion can be written for
11,lt'l'kd points (called nodes) on the analytical model. The primary objectives
ol dynamic analysis by finite elements are to calculate approximately the re11po11Nc11 ut such nodes or at other selected points.
Framed structures usually are automatically discretized by virtue of the
l'o1111111,n dcflnitions for members and joints. That is, a frame member may be
rnnsidcrcd as a one-dimensional finite element spanning the distance between
two Joints . On the other hand, a joint (or node) is defined as a point where
1111111lwrs join, a point of support, or a free end of a member. However, we need
11111 ll'Slrkt ourselves to these purticular definitions. Figure 1.3 shows a plane
l1111m l'on11lsling of 1hrcc members (indicutcd hy encircled numbers) and four
!111111 ~ II d,sirtd , wt could 1n1hdividt tu,h ancmhcr inlo four purls and view each
flql,llll'III IHI (I llt'W llll'llllll'I . Thon lilt' IIIClll' lt'hmcl IIIIUlylkul lllOdll would have
I) tlll' llllll'r~ und I I 1111111 ~ Sud, 111t'll11t'111t111 w1111ld p111vid1 IIIOll' po1111~ whl'lt'

Sec. 1.3

Discretization by Finite Elements

0..

2.,.__ _ _ 2 _ _ _ _3

---x

Figure 1.3 Plane frame with subdivided members.

masses ~ould be ~ocate~ for the purpose of writing more equations of motion.
Thus, ~1th these hberahzed definitions for members and joints we are able to use
~e finite-element method more effectively. Although framed structures are
discussed throughout the book, Chapter 6 is devoted exclusively to this topic.
Whereas framed structures fit rather neatly into the theory, finite elements
have much more powerful applications in two- and three-dimensional continua.
For exa~ple, ~he hfperbolic paraboloidal (or hypar) shell structure in Fig. 1.4
mar be d1scretized_mto qu~drilateral elements that are curved in space. We can
denve energy-consistent stiffnesses, masses, and equivalent nodal loads for such
clements and assemble them into an analytical model having a large but finite
number of degrees of_f~eedom. By so doing, we are able to simulate accurately
the response of the ongmal structure. The theory of finite elements appears first
z

1,1111111 1.4

I l y p111 ~hl'II tli~l 11'111111 hy

lilllll' CICllllllllN .

Introduction to Structural Dy namics

Chap. 1

in Chapter 3 for one-dimensional elements. Later, Chapters 7 and 8 contain

applications to two- and three-dimensional continua, plates, and shells.
Occasionally , we may encounter a structure in which one or more parts are
very stiff compared to other parts. In such a case, it may be convenient to inc_lu~e
rigid bodies within the analytical model. For example, the rectangular sohd m
Fig. l . 5 could be divided into three-dimensional finite elements, while the pl_ate
is modeled by two-dimensional elements . However, a more clever modeling
procedure would treat the block as a rigid body , which is connected to the
flex ible plate. The resulting model would have fewer degrees of freedom (?ue
to rigid-body constraints) and better numerical conditioning than the first choice .
This topic of rigid bodies within flexible continua is covered in Chapter 9.
z

Figure 1.5 Rigid body supported by discretized plate.

I( is also possible to analyze a structure by dividing it into substructures
1h11t 111t handled one at a time. Substructuring conserves core storage in a digital
t'1H11putcr nnd also allows several groups of analysts to work on diffe_ren~ parts
ol II structure simultaneously . Figure 1.6 shows half of a symmetnc aucraft
~t111ttmc di vided into five types of substructures that are joined at common
(i11tt1fun:) nodes. Five groups could analyze the substructures individually while
11 sixth group handles the assembly and solution process for the entire structure.

M111111 I .<,

!-:v 111111111 l, 111 1111111

.....

/
I

il1 vhh d 11 1111 - 1111~1111, 1111,,

Sec. 1.4

Computer Programs

The wing group may choose to model the engines as rigid bodies connected to
a flexible substructure. Chapter 10 contains descriptions of various substructure
methods that have proven to be useful for structural dynamics.

1.4 COMPUTER PROGRAMS

Structural dynamics by finite elements is a computer-based method, so we

include and explain digital computer programs that perform the calculations.
Our programming philosophy consists of starting in Chapter 3 with the relatively
simple task of vibrational analysis for plane trusses (Program VIBPT). In Chapter 4 we extend the vibrations program to include dynamic responses by the
normal-mode method (Program NOMOPT). Then we further extend the program in Chapter 5 to calculate dynamic responses by a direct numerical integration approach (Program DYNAPT). In subsequent chapters this program is
revised many times so that it applies to other types of framed structures and
continua that are discretized by finite elements.
All of our coding is written in FORTRAN, and we present FORTRANoriented flowcharts for the main programs in VIB , NOMO , and DYNA. We also
include a detailed flowchart for Program DYNAPT in Appendix C. Various
tables for preparation of data appear throughout the book in order to show how
a user must interpret problems to the computer. We show some computer output
in the form of line prints, but the amount of information for structural dynamics
Is so large that we decided to minimize printing. Instead, we put emphasis on
rnmputer plots to show time histories of forcing functions, displacements ,
stresses, and so on. A plot is much more informative than a print, so the more
desirable choice of output became obvious . Program notation is included in the
h~t of notation near the end of the book.

REFERENCES
T i111oshc nko, S. P. , Young, D. H. , and Weaver, W., Jr. , Vibration Problems in
l:'11gi11eeri11g, 4th ed ., Wiley , New York, 1974.
' Weaver, W ., Jr. , and Gere, J. M. , Matrix Analysis of Framed Structures, 2nd ed .,
Van Nostrand Reinhold, New York, 1980 .
Wr11w 1, W., Jr., and Johnston , P. R., Finite Elements for Structural Analysis,
l '1lntitc ll ull , Englewood Cliffs, N . J ., 1984.

Sec. 1.1

Introduction

11
D(t)

~
j.

D(t)

em
L

./

c_{

&

).

L
2

(a)

(b)

Systems
with One Degree
of Freedom

D(t)

/1-lom____
/

/
/

.,,

I
I

(c)

2.1 INTRODUCTION
/

1111 preliminary or approximate analyses, simple structures may often be ideal1/t<I us systems with a single degree of freedom (SDOF systems). Figure 2.1

Nhows a few cases where mass m or mass moment of inertia Ir is somehow

1111Nod11tcd with a single point or axis. In each case the translation of a point or
rnllltion about an axis varies with time. For the beams in Figs. 2. l(a) and (b),
tht 11111sscs may be either attached concentrations or some fractions of the
<IINtr ihutcd masses of the members. Each of the translational displacements is
ttulkutcd in the figures by a single-headed arrow.
On the other hand, the rectangular plane frame in Fig. 2. l(c) has a mass
111 t11kl'll to be the tributary mass at the horizontal framing level. Also, in Fig.
J l (d) we see a massive rigid body connected to framing members in space. In
lhl' lnttcr case, the mass moment of inertia Ir of the body is computed with
ll'spcct to the axis of rotation, for which the displacement is shown by a
tlouhlc hcuded arrow. Finally, Fig. 2. l(e) depicts a thin plate in bending with
1111 l'dgcs tlxcd and a single mass mat its center, associated with a translational
tll11plm:cmcnt.
Fl,1/bilitit.1 for the structures in Fig. 2.1 are obtained as displacements
(t11111sl111io11s or rotations) due to unit values of the corresponding actions (forces
111 1110111l' 11l11) . If thl' canlilcvcr bcum in Fig. 2. l(a) is prismatic, the flexibility 6
111

lht 1111 111<1 Ill

(d)

(e )

li'l1turc 2.1

Structures modeled as SDOF systems: (a) cantilever beam; (b) simple

lwum ; (c) plane frame; (d) space frame; (e) fixed plate.

wh11h. is th~ stati~ translation caused by a unit force applied in the positive
dl1n 111:11 of the dis~lace~~nt D ._ In this expression the symbol E represents
\ 111111p 1:1 modulus of elast1c1ty, I 1s the moment of inertia* of the cross section
111 llll' htn111 , and t, is its length.
S11//i11.1'.1'('.I'

nrc found as static holding actions, corresponding to the discaused by inducing unit values of those displacements. At

plt111 1111 111s 11how11 ,

I. '
11'/

' I It, 111111111111 ut hw11l11 111 1111 l lllHN Hrt'li1111 IH 11 1111N11omcr for the second moment of area
1,1 1111 1111~~ ~,, 111111 Wllh 11~1111 I lo 1hr 1111111111 uxiH

12

Systems w ith One Degree of Freedom

Chap. 2

Soc. 2.2

Free, Undamped Vibrations

13

lhl free end of the cantilever beam in Fig. 2.1 (a), the stiffness k is a static force
111 the direction of D that is required to produce a unit value of D. Thus, we have
k = 3EI =
L3

Nole that the stiffness k is the reciprocal of the flexibility 6.

Similarly, if we assume that the simple beam in Fig. 2.1 (b) is prismatic,
then its flexibility is
u(t),ii(t)
(a)

which is the translation at the center due to a unit force corresponding to D.

Furthermore, the stiffness of this beam has the value
k

= 48/
=!
L3
o

which is the holding force at the center required for a unit amount of D.
We could also find flexibilities and stiffnesses for the other structures in
Ftg. 2. 1. However, the analyses required would be more complicated than those
tor lh1.: statically determinate beams. Although other motions of the structures are
possi ble, we will restrict our attention in this chapter to systems idealized as
l111ving onl y one degree of freedom. Analytical models having coupled multiple
d111H'l'S of' freedom will be discussed in subsequent chapters.
Topics of interest for SDOF systems consist of free and forced harmonic
1110111111s with and without damping, response to arbitrary time-varying loads or
,.11ppo, 1 motions, and response spectra for dynamic loads. These subjects are
l'OVl'll' d in the ensuing sections of this chapter and will be used throughout the
hook .

--
ku

mii

u(t),ii(t)
(b)

Figure 2.2

(a) Mechanical analogue for SDOF system; (b) partial free -body dia-

grum .

11111tial ~estoring force mu also arises due to the presence of the mass. From the
po1l111l free-body diagram in Fig. 2.2(b), we see that
- ku - mu

2.2 FREE, UNDAMPED VIBRATIONS

Figure 2. 2(a) shows a mechanical analogue for any of the SDOF systems
dtslnbcd in the preceding section. The mass m and the spring constant k are

d1ll'1111incd from the properties of a given elastic structure , and the symbol u (t)
dlnotes the single displacement coordinate. The figure also indicates the accel11at ion ii (t), which is the second derivative of u (t) with respect to time
(II
d 2 u/ dt 2). If the displacement is rotational, then m, k, and u may be
Hplt1l'l'd by /,. k,., and 60 (a smnll angle of rotation).
We assume that the sys1c111 in Fig. 2.2(a) is initially at rest inside or outside
11 ~rnvi1111 io1111l field I 11. Tht11 ltl 11s disturb it in some manner from its position
ol .,,111k l'qtdlihr i11111 1>111 lo th, d1Hpl11n111cnt II relative to the static position ,
1t ll'S1t11111 ~ t1111t 1q1111l 111 fo ill' vrl11p1, 111 tlw s p1111 g, 11s shown in Fig. 2.2(h) . An

=O

(1)

\\ hid, is. an applicatio_n of D' Alembert's principle for dynamic equilibrium.

l~111 111111grng Eq . (1) slightly, we have the following equation of motion:

mu+ ku _=

(2)

I lr h1 Nt'l'l~nd-order differential equation may be simplified further by introducing

1111 IIO(H(I Oll

(J) 2

= -

(3)

= 0

(4)

whit h prncluces

u + w 2u

',111 Ii 1111 1q1111tio11 will be satisfied if we take the general solution

11

<'1 cos

Ml

I C, sin "'t

(5)

Systems with One Degree of Freedom

14

Chap. 2

Sec. 2.2

Free, Undamped Vibrations

where C 1 and C2 are arbitrary constants. Thus, the response consists of the sum
of two harmonic functions that repeat themselves with time. This oscillatory
motion is called free vibration, for which the angular frequency w (rad/sec) is
the constant

15

(6)

w=~
as given by Eq. (3). The natural frequency f (cycles/sec) is

f=..!!!.._=_1 {k
21T

(7)

21T "{;;;

and the natural periodJ (sec) becomes

T

= = 2: = 21r ~

(8)

(a)

The last expression is the time for which the vibration repeats itself.
To determine the constants C 1 and C 2 in Eq. (5), we must consider the
initial conditions of a given problem. Assume that at the time t = 0 the mass m
has an initial displacement u0 from its position of equilibrium and that its initial
vl'locity is u0 . Substituting t = 0 into Eq. (5), we obtain

C1

= Uo

If we also take the derivative ofEq. (5) with respect to time and substitute t

= 0,

we find that
(b)

A cos (wt - a)

Thus, Eq. (5) becomes

Uo.

u = u0 cos wt + - sm wt

(9)

1lcncu, the cosine term depends only on the initial displacement, whereas the
sine term depends only upon the initial velocity. E2ch of these harmonic terms
c:un be represented graphically, as shown in Figs. 2.3(a) and (b), by plotting
displacements against time. The total displacement u of the vibrating mass mat
uny time t is equal to the sum of the ordinates of the two curves, yielding the
third harmonic curve shown in Fig. 2.3(c).
An allernative way to represent vibrations is with rotating vectors. Imagine
two orthogonal axes u and u/ w, as shown in Fig. 2.4, which define the phase
pl,1111. Let a vector OP of mngnitudc u0 rotatu with constant angular velocity w
11101111<1 thl' flxtd point O JI at tillll' t
O till' wctrn OP coincides with the 11
nxi~. tht 1111gk it 11111kts with tht t11111w nxis nt 1111y 1111111 tinw tis l'qunl to (I)(. The
p11111lti111111I tl11'i v1rl111 1111thr111a,w1 i'i l''lll1tl l111111 l 1111 ,,,, 1111d up1tst11ts till' flr~t

te

_l_
0

(c)

l'l1111n 2 ..\

RC'sponhl' funl'tionN: (n) cosine; (b) sine; (c) combined.

Systems with One Degree of Freedom

10

Chap. 2

17

Free, Undamped Vibrations

Sec. 2.2

te

(13)

= -(J)

This lag time is indicated at the upper left in Fig. 2.3(c).

R

Example 2.1

/
/

&"/

/
/'

The_ rectangular plane frame in Fig. 2.5(a) has a very stiff girder of mass m and rather
flexible columns, each having length Land constant bending rigidity El. Neglecting the
muss of the columns and their axial strains, find the values of w, f, and T for this
structure, treating it as a SDOF system.

~ sin wt
w

---

COS

(wt - a)

p/

'o
Uo
w

"\
\

\
w

I
i , - - - - - - - m_ _ _ _ _--,1 -

u(t)

\
0

Figure 2.4 Phase-plane .representation.

in Eq. (9). If we also take another vector OQ of magnitude u0 / w and

1w1 plndicular to the vector OP, we see that its projection on the u axis gives the
Hl'l'1111d h.:nn of Eq. (9) . The total displacement u of the vibrating mass is found
hy 11dding the projections OJl the u axis of the two perpendicular vectors OP and
00, rnluting with angular velocity w.
'l'hc sumc result will be obtained if we consider the rotating vector OR (the
H11111 ol wctors OP and OQ) and take the projection of this vector on the u axis.
llrn111 l'ig . 2.4 the magnitude A of this resultant vector is

/ -

111111

-x

'

(a)

(10)

11ml till' angle that it makes with the u axis is wt - a, where

a= tan- 1

uo

(11)

WUo

2r

'l'ltw,, we cun express Eq . (9) in the equivalent form

11

/\

cos (wt

a)

(12)

wlll'll' (hl' {llllf)littul< <fvi/m1tio11 /\ is giwn hy liq . (10), und thc plwse an1:le a
IM 11lhmd by l\q . ( 11 ). Nott th11t till' 1t1mlt1111I vl'ltor ()It in Fig. 2.4 lags the
10111p11111111 v11'llll OP hy tlw pllllMl' 1111)'11 cv Pm thl N1111s011 th1 tolul vihrntion
111 Fi)' J \(1) l11pl'I 1111 111111prn1111I 111 tlu 11111tlllll 111 l'l p , \(11) hy th1 111111

11111111 11\ 1,x11111ph~ (11) 111111Nl111lu1111l ;

tlti 1111,1111111111

ill)

,y111111111

1ft

with 01111 l>nu1nn 111 I 1nnil11111

19

I ln11n1111l1 I 1111 1t1u I 111111111111

Cl111p, 'I

HARMONIC FORCINO FUNCTIONS

lly 1runsl111ing the girder u unit tllHluncc (u
Ntlll11w1H l'Onstunt us the following holding force:
k

I) 111 1h1

= 2 (12EI) = 241
LJ

L3

dl111 liou , we tlnd lhc

'1111 111 lhl 111011.1 in~port.unt types of forcing functions that we must study is the
l111pli h111111on1c f unc~1on P sin Ot (or P cos Ot), where n. is the angular

(a)

111 q111 my ol the funcllon. The first of these functions appears in Fig. 2.6(a) ,
11ppl lid lo Ih~ undarn.ped ~DOF system of the preceding section. From the partial
Ii, 1 hody d1ngrum JO Fig. 2.6(b), we can write the differential equation of
11111111111 Ior this case, as follows:

' lfa111 tht 11n~ul11r frequency w from Eq. (6) is

w= 'J;
ik=IL 'J-;i
f6Ei
1111!1

(b)

mu + ku = P sin Ot

tlll' 1111turnl frequency f from Eq. (7) becomes

I II i.l111plif'y thi s equation, let us divide both sides by m to obtain

f=.!:!...=-1 {6Ei
21r

1rL

'J -;;;f,

u+

(c)

w2u

= Pm sin Ot

/\IH11 , the notuml period T from Eq. (8) is seen to be

T=7= 1rL~

Pm= -

(d)

I h 11 lhl symbol p,,. represents a force per unit of mass.

1r.x11m11h.12.2
S11pp0Hc thnt the rigid disk in Fig. 2.5(b) has a mass moment of inertia lry with respect
lo 1h1 y nxis. The disk is supported at its center by a massless flexible rod of length L
h11vl111t II drculur cross section with radius equal to r. If the disk can rotate about the y
11xlN without trunslation, find w, f, and T for this problem as a SDOF system.
Rolntlon of the disk a unit amount (80 = 1) about they axis yields the stiffness
l'IHllltlllll IIS the holding moment

(e)

k = GI

'

111 whk h <, is the shearing modulus and

1rr4
J=2

(a)

(f)

h1 1h1i p11lt1r moment of inertia of the cross section of the rod. Replacing m and k with l ry
111111

k, in Eqs. (6), (7), and (8), we find that

(g)

111111
2

I-t&G

w
r
J -=-=21T 21T -21,y/,

(h)

---

ku

mu

- P s i n flt

u(t),iJ(t)

(ll)

1111d
'/ '

271

(1)

{?t,. /,

,.J \/

,,c;

(i)

1<111111,, 1.t1 (11) ll111111onlt' fo1d 111-1 f'unlllon npplicd to SDOF system; (b) partial
111 1 liudy dlllHIIIIII,

(2)

(3)

y1111111111 with 01111 1>11u11u1 111111111110111

20

Chnp. 'J

Thu total solution for Llq . (l) rnnsists of lhe su111 111 llll' ~1mrul solulion
IOI' the homogeneous equation, plus the particular solul1011 lhnt :mtislies the
whole equation . The form of such a particular solution is

= C3 sin

11111..

Harmonic lorol1111 I 111wll111111

2.3

/3
3

(4)

flt

Suhstituling this expression and its second derivative with respect to time into
l\ll , (2), WC have
2

21

I
-

I
I
I
I
I
I
I
I
I

'l'hncforc,

C3

Pm
W

(5)

n2
,H

I
I

'l'lwn the total solution becomes

= C 1 cos

wt

+ C2 sin

wt

+ C3

sin flt

(6)

Till' first two terms on the right-hand side of Eq. (6) are called the free part of
till' response (with angular frequency w), and the third term is referred to as the
/on'1d part (with angular frequency fl).
Let us rewrite the forced part of the solution in Eq. (6) as

= w2

'l'hc11, by using Eq. (3) and w 2 =

p,,,
-

fl2 sm :ut

(7)

k/m, we may cast this expression into the form

u = [l-(~/w) 2 ]fsinflt

(8)

'l'hc second factor on the right-hand side of this equation is the displacement of
llll' mass in Fig. 2.6(a) if the forcing function were applied statically. The
11hsol11tt.: value of the term in brackets is called the magnification factor /3. Thus,
Wl' h11ve

f3

_ _ _ _ _ _ J_ _ _

= \ 1 - (~/w) 2

(9)

whkh reprt.:scnts the ratio of the dynamic response to the static response. The
p111 I of lhe motion in which this factor appears is called steady-state forced

1'ihl'lltio11.
Figure 2.7 shows values of the magnification factor f3 plotted against the
t1c.qucncy ratio fl/ w, which is the ratio of the impressed angular frequency fl
ol the disturbing force to the natural angular frequency w of the SDOF system.
I 11)1' s,nall values of the frequency ratio , the magnilicut ion factor is approximately
1111i1y ; und tht.: response is about the same ns t'or the slatic application of the
h111111011ic.: load . When this mt io 11pproadws 11111ty , however, the magnification

I
o'---~~---1.~~~--L~~~...L_~~____J~~~_J_~-

fUw

Figure 2.7 Magnification factor.

factor a~d th~ amplitude of forced vibration rapidly increase and theoretically
become mfimte for the case of
= w. This state is known as the condition of
res~na_nce. The infinite response at resonance implies that there is no damping
to dis~ipate energy and suppress the response. However, in practical structures
there 1s always some damping, as will be discussed in Sec. 2.4.
Wh~n th_e frequency of _the d!sturbing force increases beyond the frequency
of free vibration, the magnification factor again becomes finite. Its absolute
v~lue ~iminishes with the increase of the ratio fl/wand approaches zero when
this ratio becomes very large. Thus, when a harmonically varying force of high
frequency acts on a SDOF system, the response is very small; and in such cases
the mass may be considered as remaining stationary.
Considering the sign of the expression I/[ 1 - (fl/ w )2], we see that for
~ < ~ the sig~ i~ positive. Physically, this means that the displacement of the
v1bratmg mass JS m the same positive sense as the disturbing for.ce. On the other
hand, when _fl > w _the expression is negative, meaning that the displacement
of the mass is opposite to that of the force . In the first case the vibration is said
to be in phase with the excitation, while in the latter case the response is said
to be out of phase.
.
It is also possible to cause forced vibrations with harmonic ground motions
instead of applied forcing functions. Two types of ground motions of interest in
~-tructural .dy~amics. are. ground displacement ug (t) and ground acceleration
~g(t), as mdicated m Fig. 2.6(a). A harmonic version of the first of these
influences may be expressed as

ux d sin flt

(10)

y11"1111 with

01111 DM\11 ""

111 I 1m11l11111

C:hnp. ')

wht'll' till' symbol d represents thl' 11111plltuuc or grou11d di.~plun1m11t. For this
,1111t Wl' write the fo llowing differential equation of motio11.
mu

k (u - Ug)

=0

( I I)

111 whkh ,,
u is the displacement of the mass m relative to the ground.
S11h111it11ti11g Eq.\JO) into Eq. (11) and rearranging the latter gives
mu

ku

kd sin flt

(12)

111

~. I

I 11111110111!

11*

u+ w

(17)

lt1 th1'Nt' 1tlutionships the symbol ulf< represents the relative displacement of the
11111 N with rnspect to the ground, and u* is the relative acceleration. Substituting
,,
// 111HI Ii from Eqs. ( 17) into Eq. (11) and rearranging, we obtain
mu*

mu*

= pg sin Ot

(13)

u*
(14)

ku*

(15)

= -ma sin Ot

\\ 111

+ w 1 u*

(19)

Pi sin Ot

(20)

11

Pi

Th INv11 h1c of p11 is an equivalent force per unit of mass, due to ground displace111t111 .
( 'olllparing Eq. (12) to Eq. (1), we see that kd has replaced P. Therefore,
1t IH l'Vident that occurrence of the harmonic ground displacement d sin flt is
,q11i v11 1L1Jl to applying a force kd sin flt directly to the mass m. All previous
r111wl11sions regarding the harmonic force may now be applied to the case of
111111111111 k grnund displacement. That is , if we replace Pin Eq. (8) with kd and
t 'lllll'l' I k, the result is

(18)

111111 di vision of Eq . (19) by m produces

wht'll'
kd
Pc= m

+ ku* = -mug

II \VI' 11 IH0 use Eq . (16) in Eq. (18), the result is

'l'lw11 division of both sides by m yields

2

23

I 011 111111 I 11111 lilll in

= -a

(21)

l ltl11 v11 lm! of pf is an equivalent force per unit of mass, due to ground acceler,11 h 111

l(q11111ion (19) is of the same mathematical fonn as Eq. (1), except that u*
1111~ 111pluccd u and - ma has replaced P. Thus, the case of harmonic ground
11 11 It rnl ion a sin flt is equivalent to applying a force -ma sin Ot directly to the
1111111N 111 . Using - ma instead of Pin Eq. (8), we have
u*

1
]ma."'
= - [ 1 - (O/w)2
ksmut

(22)

,, hh Ii ii, lhc displacement of the mass relative to the ground. This relative motion
flnding the force in the structure,* which is represented by the spring
111 th1 SDOF system. Note that the minus sign before the brackets in Eq. (22)
11111111 tlw n.ilati ve response of m out of phase with the ground acceleration.
I IINr t II I for

'l'ht tl1111 d si n Ht in this expression is the motion of the mass when the ground
dl1,pl111..c11wnt occurs very slowly (or statically). The premultiplier in brackets is
!Ill' 1,111111 f'uctor as that discussed before [see Eq. (9)]. Thus, we need only
1111111id11 th\: displacement of the mass due to the static displacement of the
gi 111111d in order to calculate the steady-state forced response of the SDOF
Nyt1t1111.
111 1111111y cases it is more convenient to deal with ground accelerations than
I'' 111u1d dl splucemcnts because a measuring device called an accelerometer [2]
hns ht1111 IIH\:d 10 obtain information about the ground motion. For example,
1111 tliq1111k1 round motions are usually measured and reported in terms of three
111 lh11prn11tl coinponents of ground acceleration, which are north-south, eastw1111, 1111d Vl'l'lical. Therefore, we shall reexamine the ground motion problem by
Np11 ilyi11g II hurmonic ground acceleration , as follows :

a~-

(I

sin

n,

( 16)

wh1111 1lw symbol <I denotes the un,plitu(k ot gm111ul t1l'l'tl1rn1ion . In order to use
lhlNtxp11ssio11 in Hq . ( 11 ), Wt' 11111 st d111111l' 1111111q1111tlo11 to lhc rt'lmi v, ,oor1/11111/1

I,111u11lt 2.3
'11p1 111N1 th111

the rectangular plane frame in Example 2.1 is subjected to a harmonic force

I' 1111 \lt, 11s indicated in Fig . 2.8. Calculate the steady-state forced vibration of the mass

II O
~w/6.
l(1111lllng that the stiffness constant for this frame is k = 241/L 3 , we substitute
1111 v111i11s o l' !land k into Eq. (8) to obtain
.

111

1
PL3
u = 1 _ (S/ 6) 2 24! sin Ot
=

3PL3
- - sin Ot
22/

(a)

N111, th11t lh1 vn lul.! of thl.! mngnifi cution factor is f3 = I36/ LL I, in which the positive sign
lhl' vi hrntlon is in ph11sl.l wi th the forcing function.

1111pll1 N 111111

1 w, wlll 111111111 lo thlN 1111111<,1 o l 1111,,11lutl1111 tl11i 111l111ivc displ uccrncnt of the mass w ith
11 ~1111 I 111 1hr p1111111d 111 llx11111ph 1 1.1

:u

Vh1t111 with

0110

Chup. 'J

IJ11111nn 111 I 11111!111111

I.

'" ~ ,1
I

11111111 111 I >1011pil111

26

I) pt rnlitl'l'S

p,,,!l
w2 - 112

11(1)

I
.-------m--------, -

I hll N,

P sin !lt

= uo _ p,,,11/w
w2

(25)

112

111111 1111 lolnl solution in Eq. (23) becomes

. wt +
Pm l'\ (sm
. I'\
11 sm
. wt)
sm
~ t - + -iio
w
w 2 - 3.2
w
11 111,, 11111ial conditions are u0 = u0 = 0, Eq. (26) reduces to
11

uo cos wt

u = w2 P,,,
_ 112 (('\
sm Ht

1
1 .wt)
w sm

(26)

(27)

I q1111l 11111 (27) represents the response of the SDOF system to the forcing

Figure 2.8 Frame with harmonic forcing functions.

11t, and we see that it consists of two parts . The first part is
1111 !iltndy state response discussed previously, and the second part is called
1111 11,111si111t re.sponse. This name refers to the fact that the term dies out in the
1111 ~t lll'l' or damping, as do all free vibrations. The sum of the two terms is not
1 h111111011 ic motion even though it is composed of two harmonic functions,
1111 1111N
1 lhll components have different frequencies, 11 and w.
When the forcing function is P cos 11t instead of P sin 11t, the term cos 11t
11 11l1111s sin !lt in Eq. (23). In this case the initial conditions result in the
l 11ll11w 111g constants of integration:
I 11111 11011 I' sin

IC1mm1>lt1 2.4
11 lh1 Nt1111c frumc as that in Example 2.3 has a harmonic ground acceleration ii8

a sin

0, (Nl'l' Fig. 2. 8), find the steady-state forced vibration for O = 3w.
In this case the relative response, as given by Eq. (22) is:

l
mal3 .
u* = - I - (3)2 24/ sm Ot
maL3
= --sm Ot
192/

(b)

lk1, w1 Nl'l" tlrnt the value of the magnification factor is /3 = l-1/8 1, so that the
vllu11tlm1 ls 11pp11rcntly out of phase with the ground acceleration. However, the two
11111111~ Nig11s l:llllCCI lo yield a positive sign in Eq. (b), meaning that in this situation the
vlh1ntio11 ts ul'lunlly in phase with the ground acceleration.

C I = Uo -

(28)

l'\2
~L

,11liict1t111ion of' these values into the total solution gives

II

111 tht explanation preceding the examples, only the third term on the
il1,1 hl h1111d side of' Eq. (6) was used for studying forced vibrations. Now let us
111111111m the free vibrations associated with the first two terms, which will be
dilhll' ttl frolll those in Sec. 2.2. Substituting Eq. (5) into Eq. (6), we have

p,,,
2

. wt + - Pm
uo cos wt + -Uo sm
2
W

II 1111 111illn l conditions are taken to be

u0 =

AZ

u0 =

(
= w 2 p,,,
AZ cos ut .I'\

(cos

I'\

ut -

cos wt)

(29)

0, this equation becomes

cos wt)

(30)

II

C I cos

WI

.
+ C 2 s in

wt

w2

p,,,

. I'\
112 Sll1 Hf

(23)

111 1111 11 111slancll the transient part of the response has the same amplitude as the
"lllly Hllltt purl , regardless of the ratio 11/w.

I >illln111inting this equation with respect to time yields

ti

C 1w sin

CJJI

I C1 cJJ cos"'' I

,,llln
,

n l

(1/

cos

{\

(24)

al

flu1111 il1l' i11iti11l rn11dilio1111

11 11 111 litm I
0 , l \q ( .> \) giws C,
iN illl' hllllll' llh lill1>1l' I loWl'Vl' I , H1tl11,li1t1ti1111111 fi
ti11 Ill tiilll' I

which
0 into Eq .
110,

F CTS OF DAMPING

111 pH v11111s tllstussions or free und fon:etl vibrations for SDOF systems, we did
11111 1 c111N1dt1 lht l'ffttts ol disslpuliVl' i11lh1tnct!s, such as friction or air rel 1111111 <'011stq111111tl y, wt l1rn11d 111111 llw n111pllt11dt of' frllt! vibration remains

Gy IMl!lft

with

01111 l>MIJtnM 111

I 111Md11111

Chup, 'J

11111Nl11111 wllh lime; but experiem:c shows that the 11111plit11d1 d1111i11l shes with
ll1w 1111d that the vibrations arc graduall y damped out. Si111ll111'1 y, for undamped
l111 c1d vihrutions the theory indicates that the amplitude cun grow without limit
111 11so1111 11cc. 1lowcver, we know that because of damping there is always some
1111111 11111plitu<.lc of steady-state response, even at resonance .
'l'o bring our analytical discussion of vibrations into better agreement with
111d 11y, do111f)i11g f orces must be included. These forces may arise from several
11111s1N, such as fri ction between dry or lubricated sliding surfaces , air or fluid
11NtNl1111c1, clectric impedance , internal friction due to imperfect elasticity of
1111111,1luls, und so on. Among all of these sources of energy dissipation, the case
wh1H lhL' dumping force is proportional to velocity, called viscous damping, is
llw i;it11pkst to deal with mathematically. For this reason resisting forces of a
rn111pl k uted nature are usually replaced, for purposes of analysis, by equivalent
1/,1,011s rlo111f)i11g I 1]. This equivalent damping is found by equating the dissip111 lo11 of" energy per cycle of vibration to that for viscous damping . For example,
.1tr 11,11m ,! da111pi11g due to internal friction can be treated by this approach.
Wt shall now consider the case of a SDOF system that includes viscous
tl11111pi11p i11 lhl) form of a dashpot damper , as shown by Fig. 2 .9(a) . Assuming
lh111 11 viscous fluid in the dashpot resists motion in proportion to velocity, we

HII ,

:,!.11

I 11111:11 ol l>nt11pl1111

27

w, 111 the hon1oeneous equ ution for free vibration as

ma

cit

=0

ku

(1)

1111 rnc l"ficient c in this equation denotes the damping constant, which has
dl11ll't11,ions of force per unit velocity . Note that in the partial free-body diagram
IN1t Fi. 2.9(b)] the damping force acts in the direction opposite to the velocity.
I >lviding Eq. (I ) by m produces
ii

+ 2nu +

w 2u = 0

(2)

111 which we define the new constant

2n = -

(3)

1111 rnnvenience in the ensuing solution.

To solve Eq. (2), we assume a harmonic function in the following form :

(4)

u = Ce 51

111 which C and s are constants that satisfy Eq. (2). Substituting Eq. (4) into Eq.
I

1) ,

we find that
s2

2ns

=0

1111111 which

-n \!n 2

w2

(5)

II lhe value of n is less than that of w, then the quantity

wa= w2 IM positi ve,

n2

and we obtain for s two complex roots, as follows:

~u,U,iJ

(a)

wh1re i = \/=1. Substituting these roots into Eq. (4) , we find two solutions of
h 1 (2). The sum or difference of these two solutions multiplied by any constant
will ulso be a solution. Thus,

cu -

1\ddin these formulas , we obtain the general solution of Eq. (2) for damped free
1/11mtio11s as
~

11, 11, 1/

(6)
(Iii

l'hl 1111101 1 "' i11 lhis t'<111111io11 d11uus1H wi lh li1IIL', so the vibrations originally
~w1111111td wtll lw HI 11d1111ll y d11111111d 11111

Systems with One Degree of Freedom

28

Chap. 2

Sec. 2.4

Sine and cosine functions appearing in the parentheses of Eq. (6) are of the
same general form that we found before for free vibrations without damping [see
Eq. (2.2-5)]. However, now they have the angular frequency
wd

= V w2 -

n2

= w ~1 -

~)2

29

Effects of Damping

(7)

which is called the angular frequency of damped free vibrations. The corresponding period of damped free vibrations is
(8)

If the damping term n is small, then wd = wand~ = T.

The constants C 1 and C 2 in Eq. (6) must be determined from known initial
conditions. Assume that at time t = 0 we have initial displacement Uo and initial
velocity u0 . Substituting these quantities into Eq. (6) and its first derivative with
respect to time, we find that
C2

_ uo
-

nuo

(9)

wd

-A

.-- ---

---

--- --- --

Putting these constants into Eq. (6) yields

(10)

The first harmonic term in this equation depends only on Uo, but the second
dl.lpends on both u0 and uo.
Equation (10) can also be written in the equivalent form
u

= Ae-"

Vcf + c~ =

(15)

In this case the solution is not periodic, so it does not represent a vibratory
2
Uo

(uo

+ nuo)
2
WJ

(12)

und
ar1

In the foregoing discussion of Eq. (2), we assumed that n < w. For the
opposite case of_n > w, both of the roots in Eq. (5) become real and negative.
Substituting them into Eq. (4), we obtain two solutions of Eq. (2), and the
general solution becomes

(11)

cos (wdt - ad)

111 this expression the maximum value is

A=

Figure 2.10 Damped_free vibration.

= tan - I -C2 = tan _,uo +

nuo

motion. The viscous resistance is so large that when the mass is displaced from
its equilibrium position, it merely creeps back to that position.Jn such a case the
system is said to be overdamped, anq the motion is called aperiodic. The
constants C 1 and C2 in Eq. (15) may be evaluated by substituting the initial
conditions at time t = 0 into the equation and its first derivative, yielding

(13)

Wr1Uo

C 1

We may regard Eq. ( 11) as a pseudoharmonic motion having an exponentially

dei.:rcasing amplitude Ae 'U, an angular frequency Wr1, and a phase angle a,1.
Figure 2. 10 shows this damped free vibration with u lag time of

n,,

,,,,,

from which
(16)

Thus, Eq. ( 15) becomes

( 14)
II

111di1111td Ill 1111' 11ppt' I liII

1iu
,1' 1

,\', llo

"''

t\I

(I 7)

Systems with One Degree of Freedom

10

Chap. 2

'l'ht general appearance of a graph of Eq. (17) depends on the parameters n, uo ,

Sec. 2.4

from which the solution is

1111d 1.i0 .

Between the underdamped and overdamped cases lies the special case of
11
w, which is the level of damping where the motion first loses its vibratory
l hn rncler. Using Eq . (3) for this condition , we have
Ccr

(18)

= 2nm = 2wm = 2'Vkm

ln which the symbol cc, denotes critical damping . For the critically damped
SDOF system, Eqs. (5) and (7) show that s1 = s2 = -wand wd = 0. Neither
l\q . (6) nor Eq. (17) constitutes the solution, which in this particular case of
11 pealed roots takes the form

u = e-w1(C1

C2t)

(19)

Suhst ituling the initial conditions into Eq. ( 19) and its first derivative, we find
lhnt
nuo

(20)

(uo + nuo)t]

(21)

C2 = uo

(w2 - fi2)Pm
Q - (w2 - fi2)2 + 4n2fi2

= e-w [u 0 +
1

(25)

By substituting these constants into Eq. (24), we can obtain the particular
solution of Eq. (23).
The total solution of Eq. (23) consists of the sum of the particular solution
and the general solution derived previously as Eq. (6). Thus , considering only
subcritical damping, we have
u = e- "1(C 1 cos wdt

+ C 2 sin

wdt)

Q cos fit

+ R sin

u = A cos (fit - 0)

= y Q 2 + R2 =

\!(w 2

Pm

iV) 2

(27)

(2nfi) 2

Pmf w 2

htt'OlllCS

cu

= P cos

ku

+ 2nu +

Y[l - (fi/w)2]2 + (2nfi/w 2)2

(22)

fit

I >ivlding this equation by m produces

ii

w u

= Pm cos

(23)

fit

() =

wlw11 () und R are constants. To determine these constants, we substitute Eq .

()I) 1111d its derivatives into Eq. (23) and obtain

( U'() I 211fiR + w 2 Q - Pm) cos fit

(-fi 2R - 2nfiQ

w 2R) sin fit

=0

'l'hiH tquation can be solved for all values oft only if the expressions in the
p11rn1Hhcscs vttnish. Thus , for calculating Q and R we have two linear algebraic
,q1111llo11s, as follows :
()

'2.11 HU I ,,,i (J

oi Ii

1
1,1 Ii

l 11 O(! I

fi 111
II

R
2nfi
= tan- 1 ---,.----,,
Q
w2 - fi2

tan- 1 -

- tan

(24)

+ R sin fit

u = Q cos fit

(28)

und
2

llll' which all of the notation has been defined previously. A particular solution
111 Hq . (23) can be taken in the form

(26)

where

which ugain depends on n, u0 , and Uo.

Turning now to damped forced vibrations , we take the applied forcing
h111ttion in Fig. 2.9(a) to be P cos fit. Then the equation of motion for the mass
mii

fit

Because of the factor e-nt in the first part, the free vibrations gradually subside,
leaving only the steady-state forced vibrations represented by the last two terms.
The harmonic force maintains these latter vibrations indefinitely, so they are of
great practical importance.
The steady-state response in Eq. (24) may be written in the equivalent
phase-angle form as

'l'hlll the general solution becomes

u

31

Effects of Damping

2nfi/w 2
1 - (fi/ w )2

(29)

Thus, we see that steady-state forced vibration with viscous damping is a simple
hunnonic motion having constant amplitude A , phase angle (), and period
lj 2n/!l.
Using the values of w2 = k/m and Pm = P /m and introducing the symbol
'Y for the damping ratio ,

n
w

(30)

y =-=w1 11111y

Ccr

s11hst itull l~q. (28) into Eq. (27) to obtain

" t1 I',. l'Os m,

o)

(3 1)

Systems with One Degree of Freedom

J2

In which the magnification factor

f3 for damped forced

1
(0/ w)2]2

/3 = Y[l -

Chap. 2

Sec. 2.4

33

Effects of Damping
4.0 ,--,----,--,--..--.-rT--ir--.-----,---,

vibrations is

y= O

(32)

+ (2y0/ w)2

f3

Also , Eq. (29) for the phase angle becomes

_

e-

-I

tan

2y0/w
1 - (O/w)2

(33)

Thus, lhe amplitude and the phase angle both depend on the damping ratio y as
wd l as the frequency ratio 0/ w.
Figure 2 .11 (a) shows the magnification factor f3 plotted against the ratio
U/ <u for various levels of damping. As for the undamped case, the value of {3
is approximately unity for small values of 0/ w; and {3 approaches zero for large
vu lucs of 0/ w. However, as the value of O approaches that of w (that is, 0/ w
11pproaches unity), the magnification factor grows rapidly. Furthermore , the
vulue of f3 at or near resonance is very sensitive to the amount of damping. Also,
note thal the maximum values of {3 occur at frequency ratios less than unity.
Sltling the derivative of {3 with respect to 0/ w equal to zero, we find that the
111uximum occurs when

1.0

1.0
(a)

1.5

2.0

(34)

- = V1 - 2y 2

Fm small damping ratios ( y :5 0.20) the maximum value of {3 occurs very near
to tl sonunce, and taking the value of {3 at resonance as the maximum provides
1,111 lldcnl accuracy for engineering analysis. Then, from Eq. (31) we have

y=O

Amax

= /3,esk = 2y k =

2n w2 m

= cw

(35)

'l'hus , wh ile damping has only a minor effect when the system is remote from
1N10111111ee, il has a dramatic effect at or near resonance. In structural dynamics
thl' inl1uencc of damping is crucial for this case and represents its most important
1q1pl k 11tion . In metal structures the damping ratio y is usually in the range 0.01
to 0 ,0); whereas its range for reinforced concrete structures is about 0.05 to
0, I0 . ' l'hc value of y would always be less than 0.20 for practical structures.
While the phase angle is of less consequence, we also show in Fig. 2.1 l(b)
plots of () versus the frequency ratio 0/ w for different values of the damping
llltlo. The phys ical meaning of the phase angle in damped forced vibrations is
thnt the respo nse of the system lags the forci ng function by the angle 0. For the
c11Sl' of zero dHmping, the forced vibrations are exactly in phase (0 = 0) with the
d1st11rhi11g force I'm nil vn lues of H/ <u I , On the other hand , they arc a half
l'yrk out of ph11Sl' (0
1r) for ull v11 l1tl'S of 0 / <i>
I . Also, f'or zero damping
thr pl111Hl' llll)' ll iH l11dlt(1111ii11111l' 11! t l'Nlllllllll'l', whl'tl' U
,,,. Wlwn d111nping is
11011 tl' 111 , Wl' t1t1ll' 111 ,1111111111111 h11111w 111 /J11Mtlw 11111110 / ,,1 v11il1H llut , tl'p11rdkss

Fll(urc 2.11 (a) Magni fication factors;

th) phnsc unglcs.

2
(b)

D.
w

of the amount of damping, the phase angle is always equal to 1r/2 at resonance.
'1'11111 is, ut resonance the response lags the force by a quarter cycle.

ll,x111111,h 2.5
I It till' 11vt11111t11l11r pl11nl' l'ru1m 111 Fig. 2. 8 hnvc u hnrmonic force P cos D.t in place of
1111 loll'<' 11N1d 111 l~x11111pl1 l. \, AINo, ln1111tim lhal lhi: ground has an acceleration

Systems with One Degree of Freedom

34

Chap. 2

1/M
a cos flt instead of the sine function in Example 2.4. Calculate separately the
11h111ly -state forced responses of the mass m due to these innuenccs, assuming that
y
0.02 and fl = 0.9w.
For this example the magnification factor in Eq. (32) becomes

5 17
f3 = \/[1 - (0.9)2] + [(2)(0.02)(0.9)]2 =
2

Sec. 2.5

where
(h)

and
_

(a)

llsing this value and k = 24E// l 3 in Eq. (31), we calculate the response to the force as

35

Periodic Forcing Functions

ad=

tan

Qn

+ RO

(i)

Wd

In the last expression the symbol ad represents the phase angle for the damped system.

PL
u = (5.17) 24/ cos (flt - 8)
2.5 PERIODIC FORCING FUNCTIONS

Pl3
= 0.2151 cos (flt - 8)

(b)

<'lmNldcring the ground acceleration, we need only replace Pin Eq. (31) with - ma to
ol,111111 the relative response as
mal
u* = -0.215 El cos (flt - 8)

(c)

1,:x111111>IC 2.6
1 IH 111111sil!11t response of a SDOF system with subcritical damping may be found by
N11hst1t11t ing initial conditions into Eq. (26). Determine the free-vibrational response of
NIil I, 11 system due to the forcing function P cos flt.
Using u = u 0 and u = iio (at time t = 0) in Eq. (26) and its first derivative with
11Np1r t to time, we evaluate the constants of integration as

C1 =

C2

Uo -

_
-

uo+ n(uo -

Q) - R il

(d)

Wd

S11hNl1t11t111g these constants into Eq. (26) gives

u= e-"'(uocos wdt + :dnuosin wdt)

10
'

+ Q[cos

flt -

+ R(sin

flt - e_,,,

e- '"(cos

ll jlll'N~ III!'

!Iii \

1111n ~i111t

(' "' ( Q cos

in

phUNl'

P (t)

= ao + a 1 cos flt + a 2 cos 2flt + .

+ bi sin flt + b2 sin 2flt + . ..
= ao +

L (ai cos iflt +

b; sin iflt)

(e)

~ sin wd9

t,i11

11111!11 101111 ,

1,0

= 0, the transient portion u1r of the

Qn I RH .

,,,.,

Wl'

sin w,,t

(f)

hnw

(.
l/1,

111

,,,.,

l II ~ (

t1l,,f

ti ,/ )

(1)

The period of the applied force is Tr = 27r/fl, and the symbols a0 , a; , and b;
represent constants to be determined. To calculate these constants , we may use
lhc procedure described next.
p

wdt + ;dsin wd1 J

II th1 1111t1ul conditions arc taken to be u0 =

1111111111111g tl'sponsc is

11,,

In previous discussions of forced vibrations , we assumed simple harmonic

functions proportional to sin fit or cos fit. It is also possible to encounter
general periodic functions that are more complicated. For example, Fig. 2.12
shows a repetitive triangular function with period Tr= 27r/ fi. In this section we
learn how to calculate the response of a SDOF system to such a function.
We can represent a periodic dynamic load of any kind by decomposing it
into a trigonometric (or Fourier) series, as follows:

(g)
lo'IMtt11

.2. 1.2 l'ri l11dl1 ltm 11111 f 111w1in11 .

i,y11111111N wllh U1111 l>11111"'1 11f

lfl

I II nd11111

Chllft, ')

Assuming thal a 1 in Eq. (I) is dcsir<.!d , post11111ll1ply hnth sidl'S ol' th<.!
1q1111t 1011 hy t.:os i Ht dt and integral<.! from t O to t 1;. 'l'hl'II W<.! S<.!C thul

L''/' a

cos iOt dt

J(1'/

=0

Ju

(1i

Jo a; cos2 iOt dt

wl111t / und j arc integer numbers I , 2, 3,

1111d thnt
a;

= -2 l1j P(t)
I;

, oo.

jI;

By using these results , we

cos iflt dt

(3a)

~ (1i P (t) sin iOt dt

7; Jo

(3b)

.!.

r,

(3c)

(1i P(t) dt

7; Jo

is simply the area under the sawtooth diagram for one cycle. Since this area is
zero, the constant ao in Eq. (3c) is also equal to zero. Next, Eq. (3a) involves
the multiplication of the forcing function by cos iflt dt and integration from
t = 0 tot = 7; = 277/0. From the antisymmetry of P(t) and the symmetry of
cos iOt with respect tot = 'TT/fl, we conclude again that the integral in Eq. (3a)
is zero, so that a; = 0. Finally, considering Eq. (3b), it is apparent that P (t) from
t = 0 to t = 'TT/0 is symmetric about t = 'TT/20 , while from t = 'TT/0 to
t = 27T/O it is symmetric about t = 37T/20. However, when i is an even
integer, corresponding parts of sin iOt are antisymmetric with respect to
t = 7T/2fl and t = 37T/20. Thus, for all even values of i, we conclude that
b; = 0.
When i is an odd integer, both P(t) and sin iOt are antisymmetric with
respect to t = 'TT/0, and Eq. (3b) gives
7T

'l'l111s, l!y using Bqs. (3), we can calculate the Fourier coefficients a;, b;, and ao
1111 1111y known periodic function.
.
.
Assuming that the forcing function has been decomposed mto ~ Fo~ner
H1iitN, W<.! t.:1111 now write the equation of motion for damped forced v1brat1ons
I

cu + ku

= a0 + a 1 cos Ot +

+ .
+ b 1 sin flt + b2 sin 2flt + . . .
a2 cos 2flt

'l'ht 101111 solution of this equation consists of the sum of free and forced
vihrnfio11s assot.:iated with initial conditions and the terms on the right-hand side.
ll1c11111w of damping, the free vibrations will disappear with time, leaving only
thr lorttd vibrations. Each harmonic term in the Fourier series contributes a
1,lt111ly sfnt<.! forced vibration of the type described in the preceding section.
'l'lm1t'1111t , we conclude that forced vibrations with large amplitudes can occur
wh111tvcr th<.! period of any term in the series is the same as (or close to) the
11111111111 pcfiod of the SDOP system. This always happens when the period 7; of
tlw lrnl'ing function is equal to , or an even mu ltiple of, the natural period -r, of
lht dn111pl'd syst<.!111.
I .tl 11s now apply the 1111: thod of this sttlio11 to lht picclwise-linear forcing
llrm 11011 ill11slrntl'd i11 Fil\ , 2. 12. To <kto111pos1 th1 l1111ctio11 into Fourier compom11t s, wt haw tlw 10111111111s in Eqs , ( \) St111 ti11 u wllh l\q . ( k), we st.:c thul

40

7T

l1r/20 P(t) sin iOt dt

0

2
P(t) = -P1 0t

(5)

(6)

7T

where P, is the maximum value of the forcing function. Substituting this expression into Eq. (5) yields
b;

(4)

l21r/OP(t) sm. iOt dt =

In the time interval O ::s t ::s 'TT/20, the function in Fig. 2.12 has the formula

IIN
111fi

37

fi
b; = -

1111111lly , 111ultiplying Eq. (1) by dt and integrating yields

a0 =

the i 111tgrnl

111 11 si111ilar manner, multiplication of Eq. (1) by sin iOt dt and integrating
pt rnhlt'l'S

b; =

111111111111 I 1111:l1111 I 11111Jtl1111

P (t) dt

a1 cos jOt cos iHt di =

(2)

('' h1 sin j!lt cos iOt dt

tine , J.l,

2
BP, 0 l1r/20 . .
8P1 li1r/2 .
=-2t sm 1 Ot dt = Tz
v sm
7T

v dv

17To

in which the new variable is v = iOt. Integrating this expression by parts and
substituting the limits produces

i'TT = BP, (- )Ci-IJ/2

b'. = 2BP1z sm
2
2 2
I
l 7T
l 7T

(7)

In this formula the odd values of i are 1, 3, 5, 7, and so on. Thus, the Fourier
series that approximates the periodic function in Fig. 2.12 is
8P1 ( sin

1 sm
. 30t + 1 sm
. 50t - ...)
P (t) = -:;;i
Ot - 32
52

(8)

und we need only superimpose the sine curves with odd numbers of periods in
1he intervul t
O to t
27T/ H .
If we omit do111ping und use II frcq11<.! ncy ratio 0/w = 0.9 , the magnifirntio11 talto, for tlw 111 st lt:1111 in Eq . (H) is

Syste m s with One Degree of Freedom

.Ill

Sec. 2.6

Chap. 2

Arbitrary Forci ng Functions

39

/3, =

1
1 - (O/w)2

= 5.26

whereas that for the second term becomes

/33

1
=

1 - (30/ w)2 = -0.159

1<'1111hermore , a multiplier of 1/3 2 appears in the second term ofEq. (8). Therefm1., we conclude that using only the first term for the response
U

8 -P1 /3 Slll
. :Ht
= -7T2
k
t

(9)

l'llltscs an error of less than 0.4% in the solution.

Suppose now that the applied force in Fig. 2.12 is replaced by a periodic
Ht'm111d acceleration of the same form. If the peak acceleration a replaces the
11111xi111um force Pi, we need only use -ma instead of Pi in the response calculnlions. For example, the result in Eq. (9) would be changed to

. at
0
u = - - 8 -ma/3 , sm
7T2

Figure 2.13 Arbitrary forcing function.

(10)

(3)

whk h is only the first term of the series.

2,8 ARBITRARY FORCING FUNCTIONS

Wl' now consider dynamic loads that have no periodic character and may v~y

which is drawn from Eq. (2.4-10). Because each incremental impulse between
t' = 0 and t' = t has such an effect, we obtain the total displacement due to the
applied force as the integral

111 1111y nianner with time. Such arbitrary forcing functions must be handled m
11 11plci11I way , as described in this section.
,
.
.
Fi gure 2. 13 shows a general forcing funct~on P (t ) that 1s express~d ~n
It 1111s of a new ti me variable t'. The value of t' 1s less than that of t, which 1s
thl time when the response is to be calculated. If the function is applied to a
tl111111wd SDOF system, the differential equation of motion becomes

mu + cu + ku =

P(t')

(1)

/\1 11uy instant o f time t' , we may calculate an fncreme~tal impulse P dt' ,
ll'j)ll'Scntcd by the hatched strip in Fig. 2.13. That impulse 1m~arts to the mass
11, 1111 instantaneous increase in velocity (or incremental velocity) equal to

d,:i

p dt'
=- = p,,, dt,

= -e-nt
Wd

(4)

(2)

l'llllrlt1d1 111111 lhl /11r111111t1(///

IIIIY 11111'1 111111 f wtll lw
WI'

em' Pm sin wit - t') dt'

This mathematical form is known as Duhamel' s integral.

Equation (4) gives the complete displacement produced by the forcing
~u~ction P(t'), acting during the time interval from Oto t. For periodic forces
1t includes both steady-state and transient terms and is especially useful in
studying the response of a SDOF system to any kind of short-term (or transient)
load. If the function P (t ') cannot be expressed analytically, we can always
evaluate the integral in Eq. (4) by some numerical method. To include initial
l'Onditions for the damped system, we need only add their effects to Eq. (4), as
lollows:

'l'his l'X pr1.ssion is vu lid rcgnrdlcss of what other forces (such us the spring f?rcc)
11111y hl nl'ling on tlw mnsi., nnd n:g11rdlcss of i( i,; displ11c.tmcnt t~n~I . velocity . at
thl 1l11w 1 1 'l'Hat\111-1 1lw i11tn111l11t11l Vl'lol'lty us ii 11 Wl'll' 1111 1111t111I velocity
(111 till' lillll' ( 1 ),

l'

,//,1'/lftll 't'lllt'III

of till' lllllSH

Ill

t
I htN l'q11111io11 11pll'li1' 11I N till'

lol11l 1wlulio11

<-11,,

l' e"',
o

p,,, sin w"(t - t') dt'

or l!q . {I) .

(5)

Systems with One Degree of Freedom

40

Chap. 2

If damping is neglected, we haven == 0 and wd == w, and Eq. (4) reduces

Sec. 2.6

Arbitrary Forcing Functions

41

Pm == P1/m is constant and acts for an indefinite time. For the case of zero
damping, Eq. (6) gives

to

l' . ( ')

u == -1

W O

Pm sm w t - t

(6)

dt ,

.
u = Uo cos wt+ -u0 sm
wt+ -1
W

l'

W O

l11duding initial conditions without damping also gives a simpler form of Eq.
('I), which becomes

u == Pm
-

l' .

, ,

Pm sm w(t - t ) dt

Integration of this expression yields

u

= p'"(l
2

(7)

sin w(t - t') dt'

COS

wt)=

Pi(l k

COS

wt)

(8)

From this result we see that a step force P1 produces free vibrations of amplitude

'l'l11s briefer equation for the total response can be used whenever damping is not

Pi / k superimposed on a static displacement of the same magnitude, as depicted

Nigni !icant.
To demonstrate use of the Duhamel integral, assume that a constant force
/ 1 Isec Fig. 2.14(a)] is suddenly applied to the mass of the SDOF system in Fig.
1
J, <)(u) . This condition of dynamic loading is called a step function, where

in Fig. 2. 14(b). Thus, the maximum displacement due to a suddenly applied

force is twice as large as that caused by the same force acting staticaJly.
Next, let us consider a step load that acts only for a period of time t 1 This
type of forcing function is called a rectangular impulse and is illustrated in Fig.
2. lS(a). During the time when the force is nonzero, the response of an undamped SDOF system is the same as that given by Eq. (8). On the other hand,
the response after time t, may be found by evaluating the Duhamel integral for
two ranges: 0 to t, and t 1 to t. Only the integration over the first range will
produce nonzero results, because the forcing function is zero in the second
range. Altogether, the solution for this case is summarized as follows:

u =
u
0

kPi (1

(8)

- cos wt)

= Pi
k[cos

w(t - t 1)

cos wt ]

(9)

t'

The same results may be obtained by considering the rectangular impulse

in Fig. 2.15(a) to consist of the sum of the two step functions, as indicated in
Fig. 2.lS(b). The first step function (of magnitude P1) begins at time t == 0,
while the second step function (of magnitude -P1) begins at time t = t 1
A third method for determining the result given as Eq. (9) involves finding
the displacement and velocity of the SDOF system at time t,, using Eq. (8) and
its first derivative with respect to time. Thus,

(a)

(10)
II these two quantities are treated as initial conditions at time t 1, the ensuing
I1t'l' vihrali onal response may be calculated from
II

M111111 .1,1,1

1111 "1t, 11111111111111. 1111 ,,

fil"'""

11,, cos w(I -

11)

u,,

+ - sin
w

w(t - t 1)

(11)

S111I1-111111ti111111l 1x pnMsio11:, ( 10) 11110 Eq. ( 11 ), followed by trigonometric manip11111111111, y1rld, thr ~11111r 11,1111 11, lhnt 111 hi , ( 11)

v11111111 with 01111 l>11u1nn "' I 1n11!111111

<.111111 'J

/l

,111 ,

in

A1ltl1tury I 111111111I1111111111111

43

d1~pl11tt111t11t 11>1< ol tlw 11111ss relative to grnund . Thus, f'or a step acceleration a,
ol llw ground, l:q. (8) bccomcs

u"' - - ~(
I - cos wt) = - ma , (1 - cos wt)
2

(12)

111 addition, the second part of the response to a rectangular function for ground
1wteleration is

ma,
u* = - k[cos w(t - t 1)
0

t'

t,

cos wt]

(13)

which is a modification of Eq. (9).

(a)

1r.1mmple 2. 7
ll11d the undamped response of a SDOF system to the linearly increasing force, known
mmp function, given in Fig. 2. l 6(a). The rate of increase of the force P (t ') per unit
111 lime is P,/t1.
As indicated in the figure , the forcing function for this example is

IIN 11

/ P=P,
P,

(a)
~11

lhc force per unit of mass becomes

(b)

Applyi ng Eq. (6) to this case, we have

l'

u = -P,- t' sin w(t - t') dt '

mt1W o
l111l11,1111ting this expression by parts yields

/
P, - - - - - - - - -

P= - P,

(b)

Figure 2.15

(a) Rectangular impulse; (b) step-load simulation.

Depending on the duration time t 1 for the rectangular impulse, the response
SDOF system may take different forms. However, the maximum response
l'II II ,wvcr be more than the value 2P1/ k. Instructions for determining loci of
rt1ipo11st muxima due to various types of forcing functions appear in Sec. 2.8.
I I the dumped SDOF system in Fig. 2.9(a) is subjtctcd to arbitrary wound
,u,,l,mtio11.1, we need only rcplucc lhc f11111;tio11 p111 (t ') in liqs. (4) through (7)
wilh tht l11nltion fl~ (t ')
ii~ (11 ). In imch t'lllWH thtst tquutions yield the

111

11

. wt~
u = -P1 ( t - -l sm
kt,
w

-'-------"'-------

(c)

11111N, we sec that the response to a ramp function consists of the sum of a linearly
l111 w11Ning static displacement P1 t/ kt1 and a negative sinusoidal free vibration of ampli111d1 l'1 /kt1 w, as shown in Fig. 2.16(b).
The velocity at any time tis equal to the first derivative of Eq. (c) with respect to
llllll That is,
Li

= .!'.!_(
1kt,

cos wt)

= .!'kt,.!_ (1

- cos 2T7Tt)

(d)

I 111111 !his 1xprcssion we conclude that the velocity is zero at times t = 0, T, 2T, 3T, and
11 1111 'l'lll'wf'on.', the slope of the displacement curve in Fig. 2. 16(b) is zero at t = 0,
1
I 1, 11; 1111<1 .~o on . Furthermore, the velocity is always positive and has a maximum
v11l111 111 1/ '1/At1 11t tinws r T/2, 3'/'/2, 57'/2 , und so on rscc Fig. 2.16(b)].

II w1 linw 11 11111:p round 11,t'lk'rution i11sllud of an applied force, it may be

Systems with One Degree of Freedom

44

Sec. 2.7

Chap. 2

45

Step-by-Step Response Calculations

Similarly, the velocity in Eq. (d) is replaced by

p

u*

(g)

= - mai (1 - cos wt)

kt1

The term - ma 1 appears in these results instead of A.

Example 2.8
Suppose that the ramp function in Example 2. 7 is terminated at time 11, as shown in Fig.
2.17. Determine the response of the undamped SDOF system in the time ranges
0 :S t :S t1 and 11 :S t.
p

p = P,t'

(a)

t,
P,

:IP,T

kt,

P,t
kt,~

Figure 2.17 Triangular impulse.

The function appearing in Fig. 2.17 is called a triangular impulse. For the first
lime interval (O :S t :S t 1), the response is the same as for the ramp function, as given
by Eq. (c). In the second time interval, we must integrate over two ranges, as for the
rectangular impulse. Doing this and summarizing the results produces

/ '/I 2P,
~ kt,
1

I', I

(c)

U = -A~t - -1 Sill
. Wt

kt,

kt,

2T

3T

(b)

.!. sin wtl

(t, :S t)

(h)

If the triangular impulse is a ground acceleration with maximum value equal to

Il'Sponscs in Eqs. (c) and (h) have u* in place of u and -ma, in place of P1.

a,, the

u =
0

.fl [t,

kt,

cos w(t - t,)

+ .!. sin w(t - t,) w

Figure 2.16 (a) Ramp function; (b) response.

'J.,7 STEP-BY-STEP RESPONSE CALCULATIONS

l'II J>l\'NNl'd LIS

(c)
wh111 11 IH thl' f11'01lnd lll'l'l' lo l'llli11111111lnw 1, . Thon 1l1l' n.isponHC in llq . (c) is converted
1

tn lh1 11l111l w 11111111111

1('

I ~Ill 1111)

,11

(I}

111 11111t1y practical problems the forcing functions are not analytical expressions
!tut urc represented by a series of points on a diagram or a list of numbers in a
l11lill, 111 such cases it may be feasible to replace the data with certain formulas
hy l'ttt Vl' titting 111ethods und then lo use those formulas in the Duhamel integral.
I l11wrw1 11 11101l' l,ll'lll' l'III m~thod fol' tvnlunting the response consists of using
111 11111 1
u111pll 111ll' 1p11l11lim1 1'1111ctl1111 in II rnpt1titiw 1wdcs of calculations. The
1

Systems with One Degree of Freedom

46

Chap. 2

Sec. 2.7

47

Step-by-Step Response Calculations

2

latter approach is discussed in this section for piecewise-linear interpolation

.fimctions.
Figure 2.18 shows a forcing function of general shape that is approximated
by a series of straight lines. For a particular interpolating line in the time interval
1 -c;; t :s tj+ 1, the response of a damped SDOF system may be written as the sum
1
of three parts, as follows:
U

Using the definition t'

= U1 +

U2

U3

tiI'j [ w2 t' - 2n + e- nr (2n cos wdt' - w~ wd

- n sm
.
= !itjkw
wdt ' ) ] (Zc)
2

This formula is available as the solution of Prob. 2.6-3 .

By differentiating Eqs. (1) and (2) with respect to time, we can also find
velocity expressions in three parts. Thus,
(3)

(1)

U3

where

t - tj, we have for the first part

(4a)
(2a)

'l'his equation contains the free-vibrational motion of the system due to the
displacement uj and the velocity uj at time t = tj (the beginning of the interval).
The formula for this portion of the response is drawn from Eq. (2.4-10).

and
(4b)
Also ,

ih = ~[1

u~k

[1 -

e-ni'

(cos wdt' +~sin WJt')]

~

(4c)

At the end of the time interval !itj, the displacement expressions in Eqs. (2)
become

1'11

( u 1) i+ 1

_ 11 !!.,.(

"

\u.j cos WJutj +

(u 2)i+1 = l1
k [1 - e-11!!.1-1 ( cos

I 'I

Uj + nuj .
" \
WJ
sm wdutj)

"
n sm
. wdut
" 1\)
wdutj
+ WJ

(Sa)

(Sb)

(Sc)

111 addition, the velocity expressions in Eqs. (4) are rewritten as

Figure 2.18

= e- "l!. 1 [-

(u2)1, 1

= ljw

(,i 1)1 1 1

= ~~i [1

(ujwd + n ui :dnui) sin wd!iti + u1 cos wd!iti]

(6a)

Piecewise-linear interpolation of forcing function.

The other two parts of the response in Eq. (1) are associated with the
11t111ight line rorcing function in Fig. 2.18. The one caused by the rectangular
l!llpulse of magnitude ~ is
"i

(u1 )1+ 1

ljk I

, , 111'

( -os <tJ,1t' I

II

'""

sin

,,,,,1 ')

kwt1

e - "tJ.'1

(6b)

sin wd!it-1

- e ntJ.r1(cos w,,titj +

;d sin wd!itj)

(6c)

(2b)

wl11l'11 c1m1111 1111111 1111 s11l11ti111111t 1'111h ) ,, ) < >11 tht nth1~1 h1111d , till' tlltcl of
th, II 11111pul111 l111p11l111 111 11111v111t11d1 Ali
'i 1 1 , ; 1!11111111 11

h111111i1111s (5) and (6) <.. rnistitute rec111T<'IIC(' formulas that may be used to calcu11111' tlw d11111pvd ns p1111 Nl' tit the cud of sttp j and to provide initial conditions at

th1 hr i,ti1111i111-1 ol st,p ./ I I ,

48

Systems with One Degree of Freedom

Chap. 2

Sec. 2.7

Step-by-Step Response Calculations

If damping is neglected, Eqs. (5) for displacements simplify as follows:

(u1)i+t = ui cos w!l.ti
p
t(l
-

(u2)j+I =

COS

+ !!i. sin w!l.ti

(7a)

w!:J..tj)

(7b)

(u3\+1 = A!l.Pki (w!l.ti - sin w!l.tj)

49

P,

P, sin flt

(7c)

utj W

und Eqs. (6) for velocities become

(u1}j+1 = - ui w sin w!l.ti

+ ui cos w!l.ti

(8a)
(8b)

- P,

(8c)

(a)

Equations (7) and (8) are simple enough for hand calculations to obtain approxi111utc results.
Of course, we need not take the shaded impulse in Fig. 2.18 as the sum
ol' n rectangle and a triangle. Alternatively , it could be divided into two triangles,
ns indicated by the dashed diagonal line in the figure. Then it would be possible
to express the second and third parts of the response in terms of ~ and ~ +1.
Furthermore, if the time step !l.ti is constant, the coefficients of uj, uj, I'j, and~
(or 0, ,) all become constants for both the displacement and the velocity expressions. I lcnce, these coefficients need be computed only once and then used
tl'pclitively throughout the numerical solution [3].

- 1

Ex111111>lc 2. 9

- 2

Fi1tt1t l' 2. t 9(a) shows a forcing function P = P, sin flt that is applied to an undamped
Sl>Oli system. The function is discretized by piecewise-linear interpolation into 20 equal
tl111l' steps of duration !::.t = 'Fj-/20. Using the method of this section, calculate approxi11111tely the response of the system. Assume that the initial conditions are uo = Lio = 0 ,

u
2

T,

T,

- 3
4

thl' vulucs of P1 and k arc both unity, and the frequency ratio is 0 /w = 0.9.
From Prob. 2.6- 1, the exact solution to this problem (with zero damping) is

r.
= kP, (.sm a/
-

111 which the nrngnificution fuctor

/J

I1

A1111pp111x1111111r ~11!11111111 IN 111111111

){3

(b)

0.
. wt
w sm

(a)
Figure 2.19

f3 hus the value

(~l/,,,)J

I -1- !-0-,1-,)l. ~.) <, I

hy 11pply111p ht~ ( /) 111111 (H) 11'1 III NIVl'ly

(h)
Ill I()

HJll'l'ifll'd

(a) Sinusoidal forcing function; (b) approximate responses.

tl11w stlps. Results of such calculations (by hand or computer) are summarized in Table
1 I Also riven in the table arc exact displacements obtained from Eq. (a) . As expected,
till' 11pprnx1111Ull' displut'l' t11entH urc slightly less than their exact counterparts, because
11111111 111h1 pul11t1011 ol' llw siiw t111vc is imptrf'ect. Decreasi ng the step size would, of
111111~1, lrncl to 1xm1 v11lmN (l'Xl'l'J)I Im 1rn11HI ofl l'crms) .

Systems with One Degree of Freedom

110
TABLE 2.1

P;

Uj

0.309
0.588
0.809
0. 951
1.000
0.95 1
0.809
0.588
0.309
0

I
'},

1
~

(1

I
K
1)

10

0.006
0.048
0.154
0.338
0.593
0.896
1.203
1.461
1.613
1.607

Exact
0.006
0.049
0.156
0.341
0.598
0.903
1.213
1.473
1.626
l.620

Chap. 2

Solution for Example 2.9

j

11
12
13
14
15
16
17
18
19
20

Uj

Exact

1.407
1.000
0.404
- 0.338
- 1.151
-1.945
- 2.616
-3.068
- 3.220
- 3.020

1.418
1.001
0.407
-0.341
-1.161
-1.961
-2.634
-3.094
-3.246
-3.045

P;
-0.309
-0.588
-0.809
-0.951
-1.000
-0.951
- 0.809
-0.588
-0.309
0

51

Response Spectra

Sec. 2.8

0.8

0.6

0.4

0.2

The approximate undamped response in Table 2.1 is plotted as the solid curve in
H u, 2. I 9(b) . Also shown in this figure is a dashed curve that results from recursively
11ppl y111g Eqs . (5) and (6) with the damping ratio y = n/ w = 0.05.

t,.

t,o

1,:,1m1111>hl 2. lO

(b)

l 1lp1111 }.. 20(a) gives a series of plotted points simulating a blast load that impinges on
1111 1111d11 mpcd SDOF structure, such as the building frame in Fig. 2.l(c). Note that the
hlnNI lon:c rises quickly to the maximum value P1 and then diminishes more slowly (and
1v1~11 htcomcs negative for awhile). In this case we have 16 equal time steps, each of
wltl\'11 hns the value /:J,,t = T /30. Apply the method of this section to find the approximate
11Npo11st of the structure. Let the values of both Pi and k be unity, and assume that the
l11it11il rnnditi ons are uo = uo = 0.
A I lcr using Eqs. (7) and (8) recursively to obtain the time history of the response ,
wu 1h1111 list the calculated displacements in Table 2.2. For this problem there are no exact

Figure 2.20 (cont.)

TABLE 2.2 Solution for Example 2.10
j

1
2
3
4
5
6
7
8

/l

P;

Uj

1.000
0.850
0.720
0.590
0.475
0.360
0.250
0.155

0.007
0.050
0.127
0.230
0.350
0.474
0.594
0.699

9
10
11
12
13
14
15
16

P;
0.070
0
-0.050
- 0.080
-0.100
-0.080
-0.050
0

Uj

0.780
0. 830
0.844
0.819
0.755
0.654
0.521
0.363

111~11l ts with which to compare these numbers. Nevertheless, we have great confidence
111 thl'ir validity. Figure 2.20(b) shows a plot of the time history of undamped response
11111 Nolid curve) , which appears to be quite reasonable. Also appearing in the figure is
11 1l11Nh1d curve that gives the effect of using Eqs. (5) and (6) with 'Y = 0.05.

,II R SPONSE SPECTRA

()

dy11u111k louds discussed in this chapter cause vibrational responses in

I 111tli iiyilll' lltS, nnd the maximum values of these responses may be less than,
, 111111 1111 , 111 g n nl l' I' 1h1111 the corresponding static responses. For a SDOF system,
1111

11 11 11111111 11 1 pt" ind

1111

11'1111111

10

(II) i l1N1 l 1 il11 d iilllNI 11111d , t i,) ll)llt11,l11t1tl1

II

~1'1111"1I

11,

,1

(rn Ill' l[lll' lll'Y ) is llw ch11rnc1cristic that dclermines its response
ptv, 11 1111, 1111 h111ll11111 111 11ddit1l111. !Ill' slrnpl' tllld durntinn of' the forcing

Systems with One Degree of Freedom

62

Chap. 2

function itself play important roles in the response. Plots (or loci) of maximum
icsponse values against selected parameters of the structure or the forcing funclion are called response spectra. Such diagrams are useful in design because they
provide the means for predicting the ratio of the maximum dynamic stress in a
slructure to the corresponding static stress. The time at which the maximum
1l.lsponse occurs is also of interest, and plots of this variable usually will accompnny the response spectra discussed in this article.
Let us reconsider the rectangular impulse shown in Fig. 2.2l(a), which
wns applied to an undamped SDOF system in Sec. 2.6. If the duration ti of the
i 111pulse exceeds the value T / 2 (half the natural period), the maximum response
is always equal to 2. On the other hand, if the time ti varies from zero to T/2,
lhc maximum response goes from zero to 2. Figure 2.2l(b) shows the early
singes of a series of such responses for ti = O. IT, 0.2T, 0.3T, 0.4T, and 0.5T.
Note that the maximum response always occurs in the first excursion, which is
lhc only part for each curve exhibited in the figure . In all such cases , the
muximum displacement occurs after the impulsive action terminates, because
lhc velocity at time ti is positive.
Thus, to find the maximum value of the response and the time when it
Ot'l'lll'S, we must examine Eq . (2.6-9), for which ti s t. That equation may be
w1 illcn in dimensionless form as

= cos w(t - t 1)

cos wt

(1)

Sec. 2.8

63

Roaponoo Spoctro

P, l - - - - - - - - ,

0
(a)

Ust

whl.lrl.l the static displacement due to P1 is

Ust

o.Jr I

P,

(2)

I
I

l>ilfcrcntiating Eq. (1) with respect to time, we obtain

~=

--.~o.2r

lly setting the term in brackets equal to zero, we find an expression involving
lht time t,,,, at which the maximum response occurs. That is ,

= sin w(t,n

sin wtrn

ti)

I
I

'l'hl following relationship satisfies this equation:

wt,,,

7r

Wf1

2+ 2

(3)

I ltnl'l' . t,,, is II lillL'Ur fun<: tion of t 1 Substituting lhc formula for wt,,, from Eq. (3)
11110 Eq . (I) yillds
l S Iii

111/ 1

\ )( l

I
I
I
I
I

w[sin wt - sin w(t - t 1)]

Ust

II

o.s7

.L,,.-"r---.10.47

(4)

7
4

2
(b)

1,1Murl.' 2.21

(11) Rl'Clllnl!uli11 impulse; (b) response curves.

ll4

fly1111111 wlllt Ottn l>nu11tn 11I I 1nnclo111

1;111,p. 'J

ll11npo11111 ! ,p111 I111

55

tl~ing Hq11 . ( 1) 111HI (I), w, c11111011111111111:1.c the rcspo11111 Np11 l111111 lot the rcct
1111~111111 i111pulsc us follows :
Um
-us,
= 2 sm. -7Tf1
T

t,,,

2T

(5a)

I
~

-=-+-1

(5b)

HIid

1
II

(5c)
t/11

- =T 2

I >l11gl'll111s of' these dimensionless expressions appear in Fig. 2.22 , where u,,,/us,
1111d t,,,/'J' nre plotted against t 1/T. From Eq. (5a) we see that if the time t 1 is less
llu111 '/ '/6, the dynamic response is less than that caused by applying the load Pi
Nl11lk11lly. On the other hand, if the time t 1 is between T/6 and T/2, the value
ol 11,, / 11H1 is between 1 and 2.
Al this point we note that plots of magnification factors for forced vil>rnllous constitute response spectra, as defined in this section. Figure 2.1 l(a)
1011t11ins u series of curves for /3 = um/ U 51 plotted against the frequency ratio
U/ ,,,. We recall that these curves represent only the steady-state part of the
11sponsc nnd that a different curve is obtained for each level of damping. If the
l11111skn1 rurls of the forced vibrations were to be included, the response spectra
111 Fi1', 2. 11 (ii) would be somewhat higher, but this effect is of little significance.
u,,,

4
(b)

Figure 2.22 (cont.)

I 1111h,1111ore, while damping is of great importance in the problem of forced

, 1l1111flo11s, it is often omitted as a consideration in response spectra due to

f111pt1INiVt' excitations. Small values of damping have littl~ ef~ec~ on such re111111111 111nxima, which usually occur before much energy 1s d1ss1pated. How ,., 1, 11 group of damped response spectra can always be co~structe? for any
1t111 l11u function, with a different curve for each level of dampmg. This may be
,11 111111plishcd for simple cases by deriving the appropriate analytical functions ,
11111 Im rn1nplicated situations a numerical approach must be used.
1,11111ph 2. 11

u.,

')

t,
o'---- -~,----~,----3:------:---- T

(5d)

---

- - - .;;-------------

I IJ'llll' ) 23(u) shows a ramp-step Junction that increases linearly from zero to Pi in the
11 11 11 11 1111<1 is constant thereafter. From Prob. 2.6-7, the response of an undamped SDOF
vt 111 In lhis excitation is

= !:2_ (.!. _ sin wt)

k

II -

(O ::;; t ::;; t,)

wt1

t,

Pi
[ I + ---'-----sin w(t - !1) - sin wt]
k

(a)

(b)

wt1

11 11 11hji11lvc of this example is to determine the response spectrum and the correspondt11i 111111 l1111c1ion.
.
.
0

---,----"',----3
._____ ....____ t,T

ll y i11spcction of' Eqs. (a) and (b), we see tha~ th~ max_1mum response will always

11l1t1 111111: 11 Therefore, only the latter equation 1s of mterest here, and 1t may be
, 111, MN1d 111 llw dinwnsionless form
111,111

II
(11)

1/
1

1'1111111 2.22

(11) RtNpttONl' Nl)l'L'l111111 fm 11111111~111111 l111p11IN1: (h) tlim ol' 11111xl 11111m

l<ut, I sl u 1,,(r

/1) -

11>(1

ll' NPIIIIM' .

1tllh 1, 11111111111{ l\q .

(1)

with 11sp11f lo 1111w, w, oh111i11

sin wt]

(c)

56

Systems with One Degree of Freedom

Chap. 2

Soc. 2.7

Response Spectra

57

- = -[cos w(t - ti) - cos wt]

Ust
ti

Setting the term in brackets equal to zero gives an expression containing the time tm.
Thus,
COS Wtm = COS w (tm - t1)
This equation is satisfied by
wt,,, =
(a)

wt1

1r

+2

(d)

As for the rectangular impulse, t,,. is linearly related to ti. Substitution of wtm from Eq.
(d ) into Eq. (c) produces

Um = l + ~ sin wti = 1 - - V 2(1 - cos wti )

u,.
wt,
2
wt,

(e)

'l'hese expressions represent both maxima and minima, which depend on the value of wt1.
Summarizing for maxima, we have

u,,,

Ust =

lm

T
' - - ~ ~~--'-~~~~'--~~~-'-~~ ~~.l.._~~ o
2
3
4

t,

(b)

I sm-T I
.

1Tl1

1rt.

(f)

= .!_ + ~
2

(g)

2T

Figure 2.23(b) and (c) contain dimensionless plots of the response spectrum from
liq. (f) and the time of maximum response from Eq. (g) . We see that the response
Hpcctrum has its highest value of u,,,/u,, = 2 when t, = 0, for which the input becomes
11 step function. Whenever t1 ~ Tthe value of umdoes not exceed u, , very much, and with
11 large rise time the loading is essentially static.
11:xumple 2.12

,,,

A dumped SDOF system with harmonic ground displacement u8 = d cos flt appears in
1111-1. 2.24(a). Let us derive an expression for the response spectrum of the steady-state
frn cc in the spring due to that influence. For this purpose, we need only find the
Nl111dy-state relative displacement u* = u - ug and multiply it by the spring stiffness k.
By taking the second derivative of the ground displacement with respect to time,
Wl' obtain the absolute ground acceleration as

u8

1
'1

l
1
'J:

(c:)

Mw1111 1.z.1
ll' ~pttllHI'

(h)

I h11cfore, the equation of motion in the relative coordinate u* becomes

mu*

= -fl2 d cos flt

( 111 H11111p Hh'p 111111111111,

t,
----'----- T

"

(hl 11~p1111 H1 Hiu, 11,1111 , (, > 111111 " ' 11 11i,d 11111111

cu*

ku* = -mug = m02 d cos flt

(i)

'11111pudng lhis relationship with Eq . (2.4- 22), we see that u* has replaced u and mfl2 d
h11Htl'pluccd P. llence, the steady-state response [see Eq. (2.4-31)) becomes

11

en
k cos ~1
/3 111!l2d

0)

(j)

111 wh11'11lh1 11111~111hl'lllior1 Im Im /I 111111 lhl ph11s1 11111-1k 0 111c given hy Eqs. (2 .4- 32) and
( I I I\)

68

Systems w ith One Degree of Freedom

Chap. 2

Chap. 2

59

Problems

Dividing this equation by F,, = kd gives

F
- = {3* cos (Ot - 6)

(n)

F,,

<)f course,

the sine function used in place of the cosine function would produce similar
l'SUlts.
Figure 2.24(b) shows dimensionless plots of the response spectrum {3* = F,,./ F,,,
11g11inst 0 / w for various levels of damping. When the frequency ratio 0 / w is zero, the
Npring force is zero; and when the ratio is large, the spring force approaches the static
vuluc, where {3* = l. On the other hand, at resonance (0 = w ), the magnification factor
W becomes the same as {3.
I

~ uu

= dcosnt

u
(a)

The forcing functions treated in this section lead to explicit expressions for
11111 /
4.0

1-ttncral, it is difficult to identify the time range within which the maximum
ll'Nponse occurs . In addition, the equation containing tmis usually transcendental
1111tl cannot be solved explicitly. Under these circumstances , values of unJus, and
t,,,j'r must be found by exhaustive calculations, using a series of values for the
tt111c ratio t 1/T. For each ratio the expression for u/u,, in terms of t/ T may be
plotlcd, and a value of u,./ us, as well as tm/T may be obtained therefrom.

1..-- 'I

r,

=0

.-- 0.10

3.0
..- 0.15

13*

"'~

2.0

-~
\

Jl'-'

1.0

0.25

REFERENCES

~0.375

~~

0.50 _.-,;::
.__....

1.00

us, and tm/T. but it should be mentioned that these are exceptional cases. In

1.0

2.0

3.0

4.0

Ti moshenko, S. P., Young, D. H. , and Weaver, W. , Jr., Vibration Problems in

l:'11Rineering, 4th ed., Wiley, New York, 1974.
Wilson, J. S. , "Performance Characteristics and Selection of Accelerometers," J.
Sound Vib., Vol. 12, 1978, pp . 24-29 .
C'rnig, R. R., Structural Dynamics, Wiley, New York, 1981.

,!l
w

PROBLEMS

(b)

l'l~ure 2.24 (a) Dam~d SDOF system with harmonic ground displacement; (b)
1cspo11sc spectra fo r spring force.

To improve the form of Eq. U), we use w2 = k/ m and define the new
11111g111fkntion factor, {3 *, as follows :
{3* -

{3(~)2 = Vl

(0/w) 2
(0/w)r lr I (2'Y0 / wP

,i I. For the prismatic cantilever beam in Fig. 2.l(a), find approximate expressions
for the angular frequency w, the natural frequency f, and the natural period T
:t,~.2. D1:lcrmine approximate expressions for the values of w,f, and T for the simply
~upportcd prismatic beam in Fig. 2. l(b).
:t,l I , Assume that the overhanging beam shown in Fig. P2.2-3 has constant flexural
11gidity W , and find approximate values of w, f , and T.

(k)

D(t)

lh111 11w11k Eq . (j) in the ndutcd form

II*'

' ll11

~h11dy

N( illl'

fl+ ,I l'ON({ lt

lrnn Ill 1111 ~1111111 11111y

llllW

fl)

{{')

/ fi A,/ , IIH ( 111

//)

~ ----1
...~ - - ~ ~,
- - Im

---- ----+ - - -l - j

Ill" tllkl'II IIN

(111)

11'1111111 l'l.J. .I

60

Systems with One Degree of Freedom

Chap. 2

Chap. 2

61

Problems

2.2-4. Figure P2.2-4 sho~s a body of mass m hanging from a massless cable having
length L, c~oss-sect1onal area A, and effective modulus of elasticity E, . What are
the approximate values of w, f, and T for this SDOF system?

Figure P2.2-6

2.3-1. Suppose that the SDOF system in Fig. 2.6(a) is subjected to the harmonic
ground displacement u8 = d cos flt . What will be the steady-state forced vibration of the mass m due to this influence?
2.3-2. If the SDOF system in Fig. 2.6(a) experiences the harmonic ground acceleration
ii8 = a cos flt, determine the steady-state forced response of the mass m relative to the ground.
2.3-3. A harmonic force P sin flt is applied to the mass m at the center of the fixed
beam shown in Fig. P2.3-3. Find the steady-state forced vibration of the mass,
assuming that the flexural rigidity El of the beam is constant along its length and
that fl= w/2.

Psin Ot

Figure P2.2-4

center tt 1s connected to a flexible, massless rod that has a circular cross section
w!th radius rand len_gth L, as indicated in Fig. P2.2-5. Find approximate values
ol w, J, and T for this problem, assuming that the bar can only rotate about the
uxis of the rod.

tm

l,25. A rigi~ b_ar AB of mass density p has cross-sectional area A and length L. At its
Figure P2.3-3

L
2

.1.

L
2

.1

2.3-4. Figure P2.3-4 shows a rigid bar AB attached at its center to a flexible rod of
length L, having a circular cross section with radius r. Determine the steadystate rotational response of the bar to the harmonic moment M cos flt, where
fl= 2w.
A

-2r

M cos Ot

2r

')

')

-~

Flflurc P2.2-5

l,l<,. For lhl'. 11,wd (ll''.1111 Nhow11 111 111H , P.2.} I! , dt1h11111l1w 1lpproxl1111,tl' cxpn:ssions
lot 1,1, / , 1111d / , lw lhl Np11q111H1' llI 1111' ll11x 11111I ilHldlty /o'I hr 1011Nt1111t ulonH the
l1111Hlh ol llu- lu-11111

62

Systems with One Degree of Freedom

Chap. 2

2.3-5. Let the prismatic cantilever beam shown in Fig. P2.3-5 be subjected to a ground
translation Da = d sin Ot, as indicated in the figure. Calculate the steady-state
response of the mass m for O = 2w/3 .

m- ------------~ f0

Chap. 2

Problems

63

2.3-10. If the left-hand support of the overhanging beam shown in Fig . P2.3-10 has the
harmonic displacement d cos Ot, what is the response of the mass m? Assume
that the beam has constant flexural rigidity and that O = 6w.

td

cos Ot

(t),6, (t)

-L----.l

I t + - - -

f - - - - - L - - --

Figure P2.3-S

2.3-6. Repeat Prob. 2.3-5 with ground acceleration Da = a cos Ot and .0 = 4w.
2.3-7. The rigid disk shown in Fig. P2.3-7 is attached at its center to a flexible rod of
length L with a circular cross section of radius r. Find the steady-state response
of the disk caused by a rotational ground displacement Dg = 88g sin .Ot if
.0 = 3w/4.

#;,
L-----+i

Figure P2.3-10

2.4-1. Considering a damped SDOF system subjected to the forcing function P sin .Ot,
derive the expression for the steady-state response in phase-angle form.
2.4-2. Derive the steady-state response (in phase-angle form) of a damped SDOF
system that experiences the ground acceleration ii8 = a sin Ot.
2.4-3. For a SDOF system with subcritical damping, determine the transient response
due to the forcing function P sin Ot. Give the solution in a form similar to Eq.
(2.4-f).

2.4-4. Repeat Prob. 2.3-3, assuming that the damping ratio is -y = 0.01 and that the
frequency ratio is .0/ w = 0.8.
2.4-5. Repeat Prob. 2.3-4, but assume that the damping ratio is -y = 0.03 and that the
frequency ratio is .0/ w = 1.1.
2.4-6. Repeat Prob. 2.3-6, but let the damping ratio be -y = 0.02 and take the frequency ratio as 0 / w = 0.95.
2.4-7. Calculate the magnification factor {3 at resonance for values of the damping ratio
-y equal to O.Gl, 0.02, ... , 0.20.
2.5-1. Expand the square-wave function shown in Fig. P2.5-1 into a Fourier (or
trigonometric) series.
p

Figure P2.3-7

Da = 88g cos Ot

2.J-8. Repeat Prob. 2.3-7 with rotational ground acceleration

n = 5w, assuming that the mass moment of inertia of the disk is !,.

and

P,

2.J-9. Let the right-hand support of the simple beam shown in Fig. P2.3-9 oscillate in
accordance with the harmonic displacement d sin Ot. Determine the steadyslate response of the mass m, assuming that the beam is prismatic and
7w/8.

,r

:n

2,r

d sin Ht

W,~,- --, ----9}1

m

I~-

- ~~

1<1111111 I' l, I 11

p1

I
l'l111m 1'2.~- 1

3,r

4,r

Systems with One Degree of Freedom

64

Chap. 2

2.5-2. In place of the periodic forcing function shown in Fig. P2.5-2, determine a
Fourier series.
p

,~

'IT

2'lT

3'lT

4'!T

n
- P,

I--

Figure P2.5-2

Chap. 2

65

Problems

2.5-5. Derive the general expression for steady-state forced vibrations of a damped
SDOF system due to the forcing function given as the Fourier series in Eq.
(2 .5-1) .
2.5-6. Using a frequency ratio of D,/ w = 0.9, find the undamped response of a SDOF
system to the first term in the series from Prob. 2.5-1.
2.5-7. Determine the undamped response of a SDOF system to the first term in the
series from Prob. 2.5-2, assuming a frequency ratio of 0/w = 0.95.
2.5-8. For a frequency ratio of D,/w = 1.05 , calculate the undamped response of a
SDOF system to the first term in the series from Prob. 2.5-3.
2.6-1. Rederive the expression for undamped forced vibrations of a SDOF system
subjected to the harmonic function P = Pi sin fit'.
2.6-2. Find the damped response of a SDOF system to the step function P = A.
2 .6-3. Derive the expression for damped response of a SDOF system to the ramp
function P = Pit' / ti.
2.6-4. Determine expressions for the undamped response of a SDOF system to the
forcing function shown in Fig. P2.6-4.

l.5-3. Decompose the piecewise-linear periodic function shown in Fig. P2.5-3 into a

Pourier series.
p

- P, -

-'------

Figure P2.6-4
,(1!i.

Figure P2.5-3

2.!i-4. 11or the sawtooth function shown in Fig. P2.5-4, substitute a Fourier series.

Assuming that the forcing function shown in Fig. P2.6-5 acts on a SDOF
system, find expressions for the undamped response.
p

P,

()

/11
II

1111
II

IJ11
II

1111
II
'111411111 I' L. t, ~

1111

1..,,.,,.

Cht1p. 'I

I )llllVC l!Xl)l llSShHIH 1111 till' 1111tl11111pcd response o l HSI)( )I N)'Nlll lll to the f'o1ci11g
function shown in Fig . l'J. (1 (1 .
p

l 1111hl11111 11

67

f1 1) , l>ltlw L' Xptcss l1111s for lht undumpLd ,csponsc of a SDOF system to the paral>ulil' g1ou11d 11ccclcrntio11 i/H a,(1 1 /1,/ given in Fig. P2.6-9.

iig

Figure P2.6-6

2.6-7. Let the ground acceleration shown in Fig. P2.6-7 be imposed on a SDOF
system, and find expressions for its undamped response.

Figure P2.6-9

iig

o,

--,,...------------

l , I 1. For u step function of magnitude P = Pi , calculate the undamped response of

ti SDOF system, using a recursive procedure with 10 equal time steps of duration
t.1 T/ 10.
I I l, Assume that a ramp function P = P, t/t, is applied to an undamped SDOF
system . With a step-by-step procedure, find the response of the system for 10
l'qual lime steps of duration !::.t = T /10 = t 1.
, 1 ,. ( 'on firm the approximate results of Example 2.9 in Table 2.1.
~. f 4, ( 'onflrm the results of Example 2.10 in Table 2.2.
, f.lfi. Using 20 equaJ time steps with !::.t = T/20, determine the undamped response
of u SDOF system to the forcing function shown in Fig. P2.7-5.

Figure P2.6-7

.l.(,-H. For the ground acceleration shown in Fig. P2.6-8 , determine expressions for the
undumped response of a SDOF system.

iig
P,

a,
Qi-_ _ __L_ _ _

__;ll.----~----~---t,o

--------~~ ~------(

,.,

--------- -

1,

1,11111111 p.1.7-~

""J..7.(1.

Systems w ith One Degree of Freedom

Chap. 2

Calculate the undamped response of a SDOF system to the forcing function

shown in Fig. P2.7-6, using 20 equal time steps with l:!.t = T/20 .

Chap. 2

2.7-8. Determine the undamped response of a SDOF system to the parabolic forcing
function P = P,t 2 / do shown in Fig. P2.7-8, using 10 equal time steps of
duration At = T/30.

P,

69

Problems

lo---------.

P,

QP--~~~~~~~--t~~~~~~~~--r~~~~

t,o

P,

t20

o .....""""'~~~~~~~~--''--~~~~~

_JI--------~

t,o

Figure P2.7 -8
Figure P2.7-6

p
p

Apex
P,

t,o

Figure P2.7-7
Figure P2.7-9

J..7 7. J)i vldc lhl' 1rl1111gulnr i111pulSl' shown in Fig. P2.7 7 iulo 10 cquul time steps

At

'/'/ 10 , 1111d ll11d llw

syshl111 .

1111cl11111pN I ws pollMl' wh\111 it Is Hpplicd to n SDOF

J..7-1). The parabolic forcing function shown in Fig. P2.7-9 has the formula

P 1( 1 12 /do), With 10 equal time steps of duration l:!.t = T/25 , obtain

llw u11d111111wd 1lNpo11sc of' 11 SDOF system.

Systems with One Degree of Freedom

10

Chap. 2

2.7-10. Find the undamped response of a SDOF system to the parabolic forcing function
P = P1[1 - (t - t1 0)2 / do] shown in Fig. P2. 7-10, using 10 equal time steps of
duration 6.t = T /25 .

Chap.2

71

Problems
p

Apex
P,

Figure P2.8-2

2.8-3. Repeat Prob. 2.8-1 for the forcing function shown in Fig. P2.8-3. (See Prob.
2.6-6 for response formulas.)
0

Figure P2.7-10

2.H-1. For the forcing function shown in Fig. P2. 8-1 , plot the response spectrum um/ Us,
and the time for maximum response t,,,/ t, against the time ratio ti/T (See Prob.
2.6-4 for response formulas.)
p

QL----~~.....1-~~~~~--'L-~~~t,
2t,

P,-------

Figure P2.8-3

2.8-4. Repeat Prob. 2.8-1 for the forcing function shown in Fig. P2.8-4. (See Prob.
2. 6-7 for response formulas.)
p

P,
P, '-- -

__,......__ _ _ _ __,

l<'lgml' P2.ll 1

l,H2. lfop111t P, oh ) .H I 101 th, lw d 111'

) '1 "I Im 11~pn11N1' 111111111111H )

i111ll't lo11

Nhow11 111 Fl1,1. I' ) .H l. Ukc l'mh.

72

Systems with One Degree of Freedom

Chap. 2

2.8-5. Repeat Prob. 2.8- 1 for the forcing function shown in Fig. P2.8-5. (See Prob.
2.6-8 for response formulas.)
p

2P,

Finite Elements
and Vibrational Analysis

P,

t,

Figure P2.8-S

2.8-6. Repeat Prob. 2.8-1 for the forcing function shown in Fig. P2.8-6. (See Prob.
2.6-9 for response formulas.)
p

3.1 INTRODUCTION
P,

t,

Figure P2.8-6

In this book we use the method of.finite elements [1-3] to discretize solids and
structures for dynamic analysis. The basic concept is to divide a continuum into
subregions having simpler geometries than the original problem. Each subregion
(or finite element) is of finite size (not infinitesimal) and has a number of key
points, called nodes, that control the behavior of the element. By making the
displacements or stresses at any point in an element dependent on those at the
nodes, we need only write a finite number of differential equations of motion for
such nodes. This approach enables us to convert a problem with an infinite
number of degrees of freedom to one with a finite number, thereby simplifying
the sol ution process . For good accuracy in the solution, the number of nodal
degrees of freedom usually must be fairl y large; and the details of element
lormulations are rather complicated. Therefore, it becomes necessary to program this method on a digital computer.
Figure 3. I shows various examples of solids and structures that are disrrclized by finite elements , with dots indicating the nodes. In Fig. 3. l (a) we see
11 rnntinuous beam that is divided into several flexural elements of the type to
l>L described in Sec. 3.4. The space frame with curved members in Fig. 3. l (b)
1111, uxi nl , flexural, and tors ional deformations in each of its subdivided memhL1s . Figure 3.1 (c) depicts a two-dimensional slice of unit thickness, represent111~ tht rnns lunt slutc known us plnnc strain on the cross section of a long,
p11s11111lk solid . On lht olht1 hand, llw disnclizcd general solid in Fig. 3. l(d)
h111, 11111,1wlt 1l'~liil'l11111 II lltt lhiu pl111t in Fiu. I l(t) lt11s lortts npp lil:d in its

74

Finite Elements and Vibrational Analysis

Chap. 3

(a)

/
/

(c)

(d)

(e)
(/)

Jt'lgure 3.1

75

1. Divide the continuum into a .finite number__of subregions (or elements) of

simple geometry, such as lines, quadrilaterals, or hexahedra.
2. Select key points on the elements to serve as nodes, where conditions of
dynamic equilibrium and compatibility with other elements are to be enforced.
3. Assume displacement shape_functions within each_element so that displacements, velocities, and accelerations at any point are dependent on
nodal values.
4. Satisfy strain-displacement and stress-strain relationships within a typical
element for a specific type of problem.
5. Determine equivalent stiffnesses, masses, and nodal loads for each finite
element, using a work or energy principle.
6. Develop differential equations of motion for the nodes of the discretized
continuum by assembling the_finite-element contributions.

Stresses and Strains

infinitesimal size. The shape functions make generic displacements at any point
completely dependent upon nodal displacements. Similarly, the local velocities
and accelerations are also dependent on the nodal values. With these dependencies in mind, we can devise a procedure for writing differential equations
of motion, as follows:

Sec. 3.2

From the homogeneous form of the equations of motion, we can perform

a vibrational analysis for any linearly elastic structure . This type of analysis
consists of finding undamped frequencies and corresponding mode shapesJ or
the discretized analytical model. Such information is often useful by itself, and
it is essential for the normal-mode method of dynamic analysis described in
Chapter 4.
In the present chapter we develop one-dimensional elements that are to be
used in subsequent work (especially Chapter 6) for analyzing framed structures.
Other discretized continua will be discussed in Chapters 7 and 8, where the
applications include two- and three-dimensional solids, plates in bending, and
shell structures .

Structures modeled by finite elements: (a) beam; (b) space frame; (c)

plunc strain; (d) solid; (e) plate; (f) shell.

own pltme, it experiences a condition known as plane stress. But if the forces
nrc normal to the plane, it is in a state of flexure, or bending. Finally, a general
shell of the type shown in Fig. 3.1 (f) can resist uny kind of loading. All of the
discretized structures in Fig. 3. 1 have mult.iple d1g1tl'N of freedom and will be
1llc11'l'd lo us MOOF .l'y.1t,111s.
Tht liniw l' k11w111 111llhod to be uNl'd in lhlNhook 111v11lws thl ussumption
ol r/i,1'/ lfrll't'l/lt'III .1'1111/11' /tllll'll//11,\' withi11 l'lll'h 1h1111 11I ' ('htw lt111tlio11s giVl'
11pp111x111rnlr 111111II N whr11 1111' 1lt111111t 111 111 1111111 - 111 111111 1x11l'I 11111111~ 111

3 .2 STRESSES AND STRAINS

lt1 I his book we assume that the continuum to be analyzed consists of a linearly
l'lustic material with small strains and small displacements. In any case, strains
1111d their corresponding stresses may be expressed with respect to some rightl11111d orthogonal coordinate system. For example, in a (rectangular) Cartesian
11111, llu; coordinates would be x, y, and z. On the other hand, in a cylindrical
1 omdlnntl systl'rn , the symbols r, 0, and z would serve as the coordinates.
Figun \ , , shows 1111 inrl11itcsi111ul l'ic1mnt in Cartesian coordinates, where
llw 1d)'1'H 1111 ol ltHHths r/1 . di', ,11ut d. Normal nnd s/11r11/111{ .11r<sse.1 arc

76

Finite Elements and Vibrational Analysis

Chap. 3

Sec. 3.2

77

St resses and Strains

For convenience, the six independent stresses and the corresponding

strains usually will be represented as column matrices (or vectors). Thus,

erx
ery
u,
Txy

0-1

I
I
I
I

dy

U4

ri,
)----- -

er6

T,,

-----

EX

indicuted by arrows on the faces of the element . The normal stresses are labeled
,rv, and er,, whereas the shearing stresses are named rxy , Ty,, and so on. From
l'q11ilibrium of the element, the following relationships are known:

Verx

1
E
(- verx

E, =

11 "

'l'hus, only three independent components of shearing stresses need be con11idl1cd.

Corresponding to the stresses shown in Fig. 3 .2 are normal and shearing
,1/ro/11.1. Normal strains Ex , Ey, and E, are defined as

ax

E
y

ay

(2)

o, and ware translations in the x, y, and z directions. Shearing strains,

Y, 1., 'Y111 un<.I so on , are given by
'Y.<y

av

= ay + ax = Yyx

av + -aw
yy: =CIZ
ily
<'. IV ! i)11
dI
ii '

I 1!1111 ,

011l y (h111 111

lh1

~l11111 IIIH ~(1111111, Ill

1111l1111111l111t

E4
Es

Yxy
Yy,

E6

'Yzx

- very

er,)

Yyz

Ty,

=G

1'zx
'Yzx = G

2(1

+ v)

In these expressions Eis Young' s modulus of elasticity, G the shearing modulus,

and vis Poisson's ratio. With matrix format, the relationships in Eqs. (5) may
be written as
e=C<J'
in which
1

-v

-v

-v -v
E 0
0
0
0

C = l_

y,

+ ery - Ver,)

Whl1l'l' 11 ,

au

E,

where

(l)

aw
E =z
az

E3

Ey = - ( -

Stresses on an infinitesimal element.

av
-

Ey

l,W

au
Ex= -

E2

= -E (CT - vay - ver,)

1
E

dx

Figure 3.2

Ex

where the boldfaced symbols denote the vectors shown.

Strain-stress relationships for an isotropic material are drawn from the
theory of elasticity [4], as follows:

f-'"

Tyz
Tzx

ers

! )-.,.
y,v

(J'

er2
er3

E1

-v
-v
1

0
0
0

0
0
0
0

0
0
0
2(1

+ v)
0
0

2(1

+ v)
0

0
0
0
0
0
2( 1 + v)

Matrix C is an operator that relates the strain vector E to the stress vector CJ'. By
the process of inversion (or simultaneous solution), we can also obtain stressstmin rt'lotio11ships from Eq. (7) , ns follows:
(9)

78

Finite Elements and Vibrational Analysis

Chap. 3

where

= c-

(1

E
v)(l - 2v)

J)

J)

J)

l-v

J)

J)

1-

0
0
0

0
0
0

0
0
0

1 - 2v
2

J)

--

ooct.e.s.,
q (t)

D..Q)

--

1 - 2v

--

3 .3 EQUATIONS OF MOTION FOR FINITE ELEMENTS

Wu shall no w introduce definitions and notations that pertain to the finite eleinunts to be studied throughout the book. By using the principle of virtual work ,
we can develop equations of motion for any finite element. Such equations
inc lude energy-equivalent stiffnesses, masses, and nodal loads for a typical
d cmcnl. T hese terms are treated in detail for one-dimensional elements in the
next section .
Assume that a three-dimensional finite element with zero damping exists
in Cartesian coordinates x, y, and z. Let the time-varying generic displacements
11(1) at <Illy point within the element be expressed as the column vector

= {u, v , w}

(1)

whcru u, D, and w are translations in the x, y , and z directions, respectively .*

If the cle ment is subjected to time-varying body forces, such fo rces may
lw placed into a vector b(t) , as follo ws:
(2)

= l, 2, ... , n n)
0

(3)

11111 l11N1tl hy h1 1111N (

l ,111111

wilh

(4)

However, other types of displacements, such as small rotations (av/ iJx, and so
on) and curvatures (a 2v/ ax 2, and so on) will be used later.
Similarly, time-varying nodal actions p(t) will temporarily be taken as
only forces in the x , y , and z directions at the nodes. That is,

p (t)

{p; (t)}

(i

= 1, 2, ... , n.n)

(5)

in which
p;(t)

= {px; , py;, p,;}

(6)

Time functions for Pxi , py;, and p,; at each node may be independent and arbitrary .
Other types of nodal actions, such as moments, and so on, will be considered
later.
For the type of finite-element method to which this book is devoted, certain
assumed disvlacement shave functions relate generic displacements to nodal
displacements, as follows:

u (t)

f g (t)

ru

ln this expression the symbol f denotes a rectangular matrix containing the

functions that make u(t) completely dependent on q(t).
Strain-displacement relationships are obtained by differentiation of the
generic displacements. This process may be expressed by forming a matrix d,
called a linear differential operator. and applying it with the rules of matrix

multiplication. Thus,

E(t) = d u (t)
In this equation the operator d expresses the time-varying strain vector e(t) in
terms of generic displacements in the vector u(t) [see Eqs. (3 .2-2) and (3.2-3)].
Substitution of Eq. (7) into Eq. (8) yields

I lc rl' the symbols b, , h,., and b: represent co mponents of force (per unit of
volu111u, arua, or length) acting in the ru forunce direl'lions nt n gcnuri c point. The
lll W

(i

q;(t) = {qx;, qy;, q,;} = {u; , V; , w;}

Matrix E is an operator that relates the stress vector CT to the strain vector e .
For the elements in this chapter, we will not need the 6 X 6 stress-strain
matrix given by Eq. (10). With one-dimensional elements, only one term , such
us E or G, is required. Later, when we deal with two- and three-dimensional
t:lcments , larger matrices, such as that in Eq. (10) , will be needed.

*'1'11 NIIVl' Np ill'I', rnl1111111 Vl'l ' IIIIN 11111y lw wt1 11,111 111 11
rUIIIIIIIIN ~11p11111l lt1f1 lh1 111111~

= {q;(t)}

where
1 - 2v

u (t)

79

Equations of Motion for Finite Elements

time variation for each component of body force is assumed to be the same
throughout the element. That is , we may have one time function for bx, another
for by, and a third for b,.
Time-varying nodal displacements g(t) will at first be considered as only
translations in the x, y, and z directions. Thus , if n. 0 = number of element

1-

Sec. 3.3

e (t)

= B q.(f2_

whurc

dr

110

I l11llu I l11111u1111111111 I Vll11n11111111I A1111ly1ln

Chop. 3

I q1111llm 111 ul Motlo11 fw 11111111 I 1111111111111

m ,1.,1

81

Muttix B gives struins 111 uny pol111 within the clemcnl d11t lo 1111i( vnlucs of' nodal
displacements.
From Eq. (3.2-9) we have the matrix form of' .1rre.1.1-.l'lrai11 relationships.
That is,
o(t) = E e(t)
( l I)
b,(t) dV

where E is a matrix relating time-varying stresses in the vector o(t) to strains in

E(t). Substitution of Eq. (9) into Eq. (11) produces

o(t)

= E B g (t)

1,odv LJdv

(12).

in which the matrix product E B gives stresses at a generic point due to unit
values_of nodal displacements.
Virtual Work Principle: If a general structure in dynamic equilibrium is
,l'ubjected to a system of small virtual displacements within a compatible state of
dt:f'ormation, the virtual work of external actions is equal to the virtual strain
l'lll'l'RY of internal stresses. When applying this principle to a finite element, we
huve

J--/

dx

)-x"'

z, w,W

(14)

Figure 3.3 Applied and inertial body forces.

(15)

Substitution of Eqs. (17) and (18) into Eq. (13) produces

llsinp lhc strain-displacement relationships in Eq. (9), we obtain

ae = B aq

aeTo(t) dV = aqTp(t)

(16)

Now lhe infernal virtual strain energy can be written as

8uTb(t) dV -

where integration is over the volume of the element.

For the <'Xterna/ virtual work we turn to Fig. 3.3, which shows an infiniltsi 11111I clement with components of applied body forces bx(t) dV, by(t) dV, and
!,, (I) dV. The figure also indicates inertial body forces pu cl\/, pv dV, and pw dV
d11t lo lhc accelerations u, i5, and w. The symbol pin these expressions repreNt11ts !he 1110.1.1 density of the material, which is defined as the inertial force per
111111 HL'L'clcration per unit volume. Note that the incl'liul forces act in directions
lh111 me opposite to the positive senses of' lhc utnllf'11lio11s . Thus, we add the
t'Xlt111111I vlrtunl work of' nodul nnd clislribu(Ld hody loll't'N us follows:
h11 'p (I) I ( Im 'h (I) 111'

Iv

" r'ln 11111 rlV

8uTpii dV

(19)

Now assume that

ii= f

l>W,.

piidV - - - - - -

'l'ht111 lhe resulting virtual generic displacements become [see Eq. (7)]
au= f aq

b,(t) dV

y, V, ii

1, 2, . . . , nen)

//

I b,(t) dV

when: fJU, is the virtual strain energy of internal stresses and aw. is the virtual
wmk of external actions on the element. To develop both of these quantities in
dtlllil , we assume a vector aq of small virtual displacements. Thus,

--

au.= aw.

(i

dy

, ( 18)

'l'ltcn we can substitute Eqs. (12) and (20) into Eq. (19) and use the transposes
nf Eqs. ( 15) and (16) to obtain
J>ql'

LBTE

B dV q = 8qTp(t)

+ aqT

fTb(t) dV - SqT

pfTf dV q
(21)

<'1111ccllution of aq rand rearrangement of the resulting equations ofmotion gives

Mq+ Kq

= p (t) +

Pb(t)

Wltl' ll'

r II

14:

n <IV

Finite Elements and Vibrational Analysis

82

Chap. 3

Sec. 3.4

83

One-Dimensional Elements
u

and

q,

M =

2 q,
------,!.------------x,u

Iv pfT f dV

Also,
Ph(t) =

Iv er b (t) dV

(a)

Matrix K in Eq. (23) is the element stiffness matrix, which contains stiffness
coefficients that are fictitious actions at nodes due to unit values oL nodal displacements. Equation (24) gives the form of the consistent-mass matrix. in

1 ~
(b)

which the terms are energy-equivalent actions at nodes due to unit- values of
nodal accelerations. Finally, the vector pb(t) in Eq. (25) consists of equivalent
nodal loads due to body forcesi n the vector b(t). Other equivalent nodal loads
due to initial strains (or stresses) could be derived [l], but analyses for such
influences are considered to be statics problems.

(c)

Figure 3.4

3.4 ONE-DIMENSIONAL ELEMENTS

In this section we develop properties of one-dimensional elements subjected to

n,dul, torsional , and flexural deformations, starting with the axial element in Fig.
1.4(11). The figure indicates a single generic translation u in the direction of x.
'l'hus, from Eq . (3.3-1) we have
u(t) = u

Axial element.

The first function diminishes linearly from 1 to 0, whereas the second increases
linearly from O to 1.
From Fig. 3.5 we see that the single strain-displacement relationship
du / dx for the axial element is constant on the cross section. Thus, Eqs. (3.3-8),
(3.3-9), and (3.3-10) yield
du
df
E(t) = Ex= - = -q(t) = B q(t)
dx
dx
Therefore, we have

'l'hc corresponding body force is a single component bx (force per unit length) ,
11cting in the x direction. Therefore, Eq. (3.3-2) gives

B = d f = df = .!. [- 1
dx
L

1]

b(t) = bx
Nodul displacements q 1 and q2 consist of translations in the x direction at nodes
I nnd 2 Isee Fig. 3.4(a)]. Hence, Eq. (3.3-3) becomes
q(t)
( ' oi responding

= {q1, q2} = {u1,

u2}

nodal forces at points 1 and 2 are given by Eq. (3. 3-5) as

P (t)

= {pi, P2} = {Pxi, pxi}

Figure 3.4(b) and (c) show linear displacement shape functions/1 and,f; that we
ussumc for this element. That is, Eq. (3.3-7) gives

t' q (t)

wlwn
I'

(I)
Flw1111 I,!!

i\~ i11l

1h'lt11t111tlilll1N

(2)

84

Finite Elements and Vibrational Analysis

Chap. 3

Sec. 3.4

which expresses the strain Ex.in terms of the nodal displacements. Similarly, the
single stress-strain relationship [see Eqs. (3 .3-11) and (3 .3-12)] becomes merely

O'(t)

CTx

85

One-Dimensional Elements

q, 1
-0.
2
q,
- - - - - -.......- - - - - - - - - - - - - X

= E E(t) = EEx = EB q (t)

Hence,

E=E

and

EB=

(a)

f [-1 1]

(3)
1

The latter expression gives the stress crx in terms of the nodal displacements.
The element stiffness matrix K can now be evaluated from Eq. (3.3-23),
as follows:

Iv BT EB dV = i

= EA [
L

1
-1

[-!

Jr-1

-!J

(c)

(4)

assuming that the cross-sectional area A is constant. Similarly, the consistentmass matrix M is found from Eq. (3.3-24) to be

= Jv pfT f dV = { 2
=

pAL[2
6

f L[

L : x ]EL - x

x] dA dx

l]

(5)

1 2

fT bx dx

if[

L :

J dx = bx ~ [ !J

(6)

which shows that the equivalent nodal loads are equal forces at the two ends.
Turning now to the torsional element in Fig. 3.6(a), we use a single
generic displacement Ox, which is a small rotation about the x axis (indicated by
a double-headed arrow). Thus ,
u (t)

Figure 3.6 Torsional element.

which is a moment per unit length acting in the positive x sense. Nodal
displacements in the figure consist of small axial rotations at nodes 1 and 2.
llence,

q(t)

assuming also that the mass density p is constant.

We see that the stiffness matrix K and the consistent-mass matrix M are
unique for a prismatic axial element of uniform mass density . However, an
infinite number of equivalent nodal load vectors Pb (t) may be derived, depending
on the distribution of body forces. For the simplest case, we assume that a
uniformly distributed axial load bx (force per unit length) is suddenly applied to
the axial element. Then Eq. (3.3-25) produces

Pb (t)

(b)

f),

= {q1, q2} = {()xi, Ox2}

In addition, the corresponding nodal actions at points 1 and 2 are

p(t)

= {p ,, pz} = {Mx1, Mxi}

which are moments (or torques) acting in the x direction. As for the axial
clement, we assume the linear displacement shape functions /1 and !2 shown in
Fig. 3.6(b) and (c). Therefore,

ex = f q (t)
in which the matrix f is again given by Eq. (1).
Strain-displacement relationships can be inferred for a torsional element
with a circular cross section by examining Fig. 3. 7. Assuming that radii remain
11lraight during torsional deformation, we conclude that the shearing strain 'Y
vurics linearly with the radial distance r, as follows:
'Y

wlwrc

1/1

d()x

= r - = rijJ
dx

is the twist , or rate of change of angular displacement. Thus,

Corresponding to this displacement is u single body action

b(t)

Ill,

dO,
1/1

ffi

86

Finite Elements and Vibrational Analysis

Chap. 3

Sec. 3.4

87

One-Dimensional Elements

1]

= GJ[
L

(L r21r (R

Jo Jo Jo (r 2 )r dr dO dx

1
-l

(12)

where GJ is constant. The polar moment of inertia J for a circular cross section
is defined as
4
21T
11R
J =
r 3 dr dO = 0
0
2

Figure 3.7

Torsional deformations.

Equation (7) shows that the maximum value of the shearing strain occurs at the
surface. That is,
'Ymax

= Rt/;

where R is the radius of the cross section (see Fig. 3. 7). Also, we see from Eq.
(7) that the linear differential operator d relating 'Y to Ox is

df

dx

= L[-1

1]

(10)

which is the same as for the axial element, except for the presence of r.
Shearing stress r (see Fig. 3.7) is related to shearing strain in a torsional
element by
r

= Gy

where the symbol G denotes the shearing modulus of the material. Hence,

and

If the cross section of a torsional element is not circular, it will warp. Such
warping is mosLsevere for elements of open cross sections, such as channel or
wide-flanged sections. For most practical cases, the theory of uniform torsion
described here may be used by substituting the appropriate torsion constant [5]
for J. If a more precise analysis is desired, the theory of nonuniform torsion [6]
may be applied.
To obtain the consistent-mass matrix M for a torsional element, we will
first integrate over the cross section and then over the length. Due to the small
rotation Ox, the translation of a point on the cross section at distance r from the
center is rOx. Also, the acceleration of the same point is r8x. By integration over
the cross section, we find the inertial momentper unit length to be -p 18,., where
J is again the polar moment of inertia given in Eq. (13). Use of this inertial
moment in conjunction with the corresponding virtual rotation Mx leads to

= r-

Thus, the strain-displacement matrix B becomes

LR

GU

Grl
/,

11

( 11 b)

Thcsl' 1l'111tionships llll' 111111Ioous lo Eqs . ( 1) 101 till' 11x111 l l'll1rn111 ,

W, 1111,y 110w 1111d th, 1<11si111111l stll l11111N 11111111, K hy 11pply111 l\q.
( I I I I), 11~ lollow~

which is a specialized version of Eq. (3.3-24). Integration of Eq. (14) over the
length yields

M = pl2 (L [L - x ][L - x
L Jo
x

pJL
6

x ] dx

[2 1]

(15)

1 2

'f'his array is the consistent-mass matrix for a torsional element of constant cross
Hcction and uniform mass density.
The simplest case of a body force applied to a torsional element consists
of u uniformly distributed axial torque (or moment) mx per unit length. For this
l1111ding Gq . (3 .3-25) gives

(' f"'m, dx

Jn

1'11,'lt' l'q111v11l,111 1111d11I londs

llll'

m1 ('

,, Ju

I1L -, xl dx = m)::.2 [1]

1

(16)

l'q1111l 1111111w11111 nl lhl' two ends or the clement.

Hnlto E:lomontt1 nnd Vll1111tl111111I A1111lynl11

88

Chup. 3

. F_igure 3.8(a) show~ a ~traii!U}l~

elert1J.'III , for whlt:h the x-y plane is
a principal ~Ian~ of bending. Indicated in the figure is a single generic displacc111cnt v, which 1s a translation in they direction . Thus,

=V

U (t)

The corresponding body force is a single component by (force per unit length),

nc

I i1

01111

89

lll11111111lm111I I 111111111111

1111111-1 111 tht y direction . I loncc,

b(t)

= by

I 1111dc I !sec Fig. 3.8(a)] the two nodal displacements q1 and q2 are a transl,1111111 111 the y direction and a small rotation in the z sense. The former is
h11lk 11ttd by a single-headed arrow, while the latter is shown as a double-headed
111111w. Similarly , at node 2 the displacements numbered 3 and 4 are a translation
,111d II small rotation, respectively. Therefore, the vector of nodal displacements
111 l'tllllCS

111 which
fJ
zl

dv,
dx

fJ z
'

= dv2
dx

I ht'Sl' derivatives (or slopes) may be considered to be small rotations even

though they are actually rates of changes of translations at the nodes. Corre-111111ding nodal actions at points 1 and 2 are

P (t)

lb)

,-

f,

y=:==-- ----- =__.~

= {p1, Pi, p3 , p4} = {py1, M,1, Py2 ,

I ht terms Py1 and py2 denote forces in they direction at nodes 1 and 2, and the
-y111hols M, 1 and M,2 represent moments in the z sense at those points .
For the flexural element we assume cubic displacement shape functions in
111111ri x f, as follows:

l/1 Ji f3 J:i]
I
/}

- [2x 3

3x 2L + L 3 x 3L - 2x 2 L2 + xL 3 - 2x 3 + 3x 2L

(17)

x 3 L-x 2 L 2]

(c)

Id)

M,2}

l'hl'sc four shape functions appear in Fig. 3.8(b)-(e). They represent the vari11t1011s of v along the length due to unit values of the four nodal displacements
,,, through q4.
Strain-displacement relationships can be developed for the flexural ele1m111t by assuming that plane sections remain plane during deformation, as
lll11strutccl in Fig. 3.9. The translation u in the x direction at any point on the
1 rnss section is
dv
u = - ydx
1llling thi s relationship , we obtain the following expression for flexural strain:

(11)

y</>

(19)

90

f In Ito Elomont11 nncl Vlhr ntl1111nl A1111lyuls

Cha. 3

111,

3.4

One-Dimensional Elements

91

" L

BTE B dV

y,v

ri- J Ey6

-ry

~--

........

Jo

+---\

~ :t

2
[ ~:;

l 2x

+ 6L

[12x - 6L

. . .

6xL - 2L 2 ] dA dx

6xL - 2L 2

-_;;;-""--~--

Mu ltiplication and integration (with El constant) yields

......-"

I M, /

10

\_,,

_ dv
dx

I '.,........J----

\
\

+---

----------1~-v,....

K = 2El [3:
L3 -6
3L

-:;,I

................

-dx~

[=

~~2

-3L
L2

:3:
6
-3L

~~

]
-3L
2L2

(24)

y dA

Figure 3.9 Flexural defonnations.

in which </> represents the curvature:

d 2v

<I>

= dx2

(20)

I:rom Eq ( 19) we see that the linear differential operator d relating

Ex

to v is

d2

= -y dx2

(21)

rrptl'Scnts the moment of inertia (second moment of area) of the cross section
wi th respect to the neutral axis. Additional contributions to matrix K due to
lll'11ring deformations are given in Ref. 5.
The consistent-mass matrix M for a flexural element will be developed in
1wo parts. A typical cross section of this type of member translates in the y
dlr1l'ti on, as indicated in Fig. 3 .8(a). However, the section also rotates about its
mutru l axis, as shown in Fig. 3. 9. The translational inertia terms are much more
lrupmtant than the rotational tenns, so they will be considered first. Using matrix
I lwm Eq. (17) in Eq. (3.3-24) , we find that

M, = fv pfTf dV =

The n Eq. (3.3-10) gives the strain-displacement matrix B as

d f

= _ 1'._ [l 2x L3

6L

6xL - 4L 2

-l2x

+ 6L

[ 156
22L
420
54
- 13L

6xL - 2L 2]

= pAL

(22)
In addition, flexural stress

er.. in Fig. 3.9 is related to flexural strain Ex simply by

(23a)

l lcnce,

E =E

pAfTf dx

22L
4L 2
13L
-3L 2

EB

IW

(23b)

Element stiffnesses muy now be ohlnirwd 1111111 1111 (,, l .,l, 21, l ) ' us, ,.() ows:
11

-13L
-32

- 22L
4L 2

(26)

whkh is the consistent-mass matrix for translational inertia in a prismatic beam.

Notational inertia (or rotary inertia) terms for a beam can be deduced from
I IH I . 9, where the translation u in the x direction of a point on the cross section
111

and

54
I3L
156
-22L

II

111 thi11 1xprl1Hsion ,

- y f)z

92

Finite Elements and Vibrational Analysis

(), = V,x = f,xq (t)

Chap. 3

(28)

where the symbols v,x and f,x represent differentiation with respect to x. Similarly, the acceleration of the same point in the x direction is

u = -yo,
0, = V,x = f,xQ (t)

(30)

By integrating the moment of inertial force over the cross section, we find the
inertial moment per unit length to be - pl 0,, where I is again the moment of
inertia given in Eq. (25). Use of this inertial moment in conjunction with the
corresponding virtual rotation 80, leads to the formula

93

One-Dimensional Elements

and flexural elements. Let us reconsider the torsional element in Fig. 3.7 and
integrate the moment of the shearing stress T about the x axis. Thus, we generate
the torque Mx , as follows:

Mx

(29)

where

M,

Sec. 3.4

fr

dr d()

Substitution of the stress-strain and strain-displacement relationships from Eqs.

(l la) and (7) produces

Mx

= Gt{,

fr
1r

r 3 dr d()

= GJ t{,

(36)

lf we take Mx as generalized (or integrated) stress and as generalized strain,

the generalized stress-strain (or torque-twist) operator G becomes
G

p!f;xf,x dx

Tr

1T

= GJ

(37)

which is the torsional rigidity of the cross section . Hence, from Eq. (36) we have

which is a modified version of Eq. (3.3-24). Differentiating matrix f [see Eq.

( 17)] with respect to x, we find that

Mx = Gt{,

(38)

(32)

By this method the operator d in ~. (9) does not include .the mu.ltiplier r.
Furthermore, the generalized matrix B [Eq. (10) devoid of r] is used mstead of
matrix B. That is,
(39)
B = rB

Suhslilulion of this matrix into Eq. (31), followed by integration over the length,
pmduccs

Prom this point we can conclude that evaluations of th~ terms in the stiffness
matrix K do not require integrations over the cross section. Therefore,

f,1

= ~3 [6(x

3x L - 4xL + L

xL)

M,

36
pl
3L
30L -36
[
3L

3L
4L2
-3L
-L2

-6(x

-36
-3L
36
-3L

xL)

-3L
L2

-3L
42

3x L - 2xL

2
]

K = JoL BT G

which is the consistent-mass matrix for rotational inertia in a prismatic beam.

Additional contributions to matrix M due to shearing deformations have also
been developed and are given in Ref. 7.
We now consider the simple loading case of a uniformly distributed body
force by (per unit length) applied to a flexural element. Equivalent nodal loads
ul points I and 2 [see Fig. 3.8(a)] may be calculated from Eq. (3.3-25) as

( - Jorl f

p,, t) -

by dx -- by L { 6, L, 6, - L}

12

B dx

(33)

(34)

For this integration the displacement shape functions / 1 through [4 were drawn
frolll Eq . ( 17).
By using f./.l't11'mll: l'd ,\tr,.1.w.1 wul .1trol11.1, wt c1111 uvoid n.ipctitious inlc
111110111, ow, llll' nrn,1, "l'l'tio11s o l om d1111l11si111111l llt111l' nl s Allhou~h this
rom,pt 1, 111thl'I t11vi11l l111 111111x111l lk11u111, 11 l'llll hr 1111111 II Mtlul Im 1t11 s101111I

This expression for K is equivalent to Eq. (3 ._3-23) used previously.

Turning now to the flexural element in Fig. 3.9, we integrate the moment
of the normal stress r:r., about the neutral axis to obtain M ,, as follows:
M,

~ U xY

dA

Then substitute the stress-strain and strain-displacement relationships from Eqs.

(23a) and ( 19) to find
2
(42)
M , = E</> y dA = EI</>

hu this clement we can take M, as generalized (or integrated) stress and </> as
l' l'nc,tilizcd strain . Then the generalized stress-strain (or moment-curvature)
lljll'l'll(OI' /: is
(43)
F

94

Finite Elements and Vibrational Analysis

Chap. 3

which is the flexural rigidity of the cross section. Thus, from Eq. (42) we have

M,

= E</>

(45)

Thus,
Then the shearing stress at any point on the cross section can be found with Eq.
(49).
For a flexural element, the relationship between the nonnal stress CTx and
lhe bending moment M, is
CTx

which is analogous to Eq. (40).

In Sec. 3.5 we will see how to assemble finite elements, and in later
chapters we shall learn methods for calculating the dynamic response q (t) of the
nodes . After those steps, we can find the time-varying stresses C1(t) within each
clement, using the equation

(3.3-12)

This ex pression may be converted to a special fonnula for generalized stress,

depending on the application. For example, an axial element has the simple
.~lrcss force relationship

l'x
A

= -

(47)

where <r., is the normal stress on the cross section and l'x is the axial force. If we
substitute Eq. (47) and E = E into Eq. (3.3-12), the result is
A

.EB q(t)

(48)

where E=EA is the axial rigidity of the cross section. After the axial force has
been determined from Eq. (48), we can find the normal stress using Eq. (47).
A torsional element with a circular cross section bears the following
relutionship between the shearing stress T and the torque M.,:
T

M,,
.I

-My
-' =
I

-yEBq(t)

I knee,

M,

= EB q(t)

(52)

I'quation (52) is a formula for the bending moment at any point along the length
nexural element in tenns of the nodal displacements q (t). After this calcululion, we can find the normal stress at any point on the cross section from
lq (51).
If we wish to find actions only at the ends of a member, we can simply
111111iiply its stiffness matrix__K and the vector of nodal displacements q(t). While
1h1H calculation is valid for framed structures, it has no such physical meaning
liu two- and three-dimensional finite elements, where the resulting actions are
th lllious.

111 11

11 /oni/ axes fo r a finite element are not parallel to global axes for the whole
~,, m lure, rotation-of-axes transformations must be used for nodal loads, dis-

Therefore,

Suhslilulin lhi s 1l1111 1" wl'II m, I~

,(S TRANSFORMATION AND ASSEMBLAGE OF ELEMENTS

l'x = EB q(t)
l'x

= _ M,y

'l'his term as well as E = E and Eq. (45) can be substituted into Eq. (3.3-12),
yie lding

K= fiJTEBdx

(T
X

(50)

Mx =GB q(t)

Then integration over the cross section for terms in matrix K becomes unnecessary. Hence,

C1(t) = E B q (t)

95

Transformation and Assemblage of Elements

(44)

With this approach the operator din Eq. (21) is devoid of the multiplier -y. In
addition, the generalized matrix B[Eq. (22) without the factor -y] may be used
in place of matrix B. That is,
B = -yB

Sec. 3.5

M,r
.I

(42.}

<; 111ul l\q ( l'J) 11110 l\q . (J .J 12) prnduc:es

11l11l"l'll\Cnts~ accelerations, stiffnesses, and consistent masses. Thus, when the

, I, 111l'llts are assembled, the resulting equations of motion will pertain to the
1,1 l11h11I directions at each node.
The concept of rotation of axes applies to a force. a moment, a translation,
11 , 111ull rntatio.n.. velocit~celerations, orthogonal coordinates, and so on.
I 1v.1111 \ . IO(a) shows a two-dimensional force vector F and its components E1111111~ 111 the I and y directions. The figure also gives the components E- and F;,
111 1111 d11cl'tions of inclined axes x' and y'. The scalar values of the components
111 11111, 1u111s of the primed axes can be computed from those for the unprimed
IKI 11, IIN follows:
(1a)
/:,
(11:
10 I ' A11 I: A11 f ~

I\

( I:

I.;,) .I '

Aq / ;

A11 I:

( I h)

96

Finite Elements and Vibrational Analysis

Transformation and Assemblage of Elements

Chap. 3

97

(2)

y'

or

.F' =RF
The 2 x 2 matrix R is called a rotation matrix, consisting of direction cosines
of the primed axes (with respect to the unprimed axes) listed row-wise. Hence,
it is orthogonal, and the reverse transformation becomes

(a)

y
y'

Similarly, a three-dimensional force vector F appears in Fig. 3. 1O(b) . Also

shown in the figure are the force components E, Fy, and F, in the x , y, and z
directions. Components E, F;,, and F,, in the directions of inclined axes x', y ',
nnd z' are given as well. For this case Eqs . (1) can be expanded, as follows:

f'x = (E + Fy + F,) i' = A1,f'x + A12F, + Al3F,

F, = (E + Fy + F,) . j = A2, f'x + A22F, + A23F,
F,, = (E + F;, + E) k' = A31f'x + A32f', + A33F,
I

(5a)
(5b)
(5c)

where k' is a unit vector in the z' direction. In matrix form, Eqs. (5) become
A,2 Al3]
A22 A23
A32 A33

[F,]
F,

(6)

F,

for which the additional direction cosines are associated with axes z and z'.
nquation (6) may be stated more concisely as in Eq. (3), but the rotation matrix
R is of size 3 x 3. In addition , the reverse transformation given by Eq. (4)
pertains to the three-dimensional case as well.
Simultaneous transformation of a force vector F and a moment vector M
lllay be accomplished by

A'= RA= [: :][~]

"
111 this expression Risa
rotation-of-axes transformation matrix containing two
Identical rotation matrices in diagonal positions. The reverse transformation is

z'
(b)

Figure 3.10 Rotation of axes: (a) two dimensions; (b) three dimensions.

In_th1cse cxp;cs~ions the symbols I' and J' denote unit \ll'ctor.1 in the directions
of -~ nndy . h .1rllwrn1ort., llw cmf'fkit.nls A 11 , A 1J, und so on, ore dir,ction
'.'(/,\' /lit',\ '
thl' p1 llllt'd IIXl'S wil lt ll'N Pl'l't lo lhl' 1111p1 lllll'd IIXl'N, ( Fo1 l'Xlllllple , i\
1
IN lhl dlll'rl1011 lo.~ 1m ol 11~11, 1 willt ll'~ Pl'l' I 10 1i,us 1, 11 111 1 ~11 011 ) ' I ht 11111 1r/~
1111111 of h1, t I J 1,

of

0 J[F'J
A= R'A'=RTA'=[RT
0 RT M'
As menlioned before, displacements and other types of vectors also can be
l11111 Nl111111td to and from local and global di rections. Therefore, we shall convert
111, 1q11111lo11s or motion for a finite clement from local axes to global axes. For
111111 p111111>Ml', llI us rtwdtl' Eq . (1 . l ?2) rm incl ined axes, as follows:

M 'c1' I K'c1'

p '( t) I p /,(t)

(9)

Finite Elements and Vibrational Analysis

98

Chap. 3

Sec. 3.5

[DF] + [SFF
SRF

In acc~rdance with Eqs. (4) and (8), premultiply Eq. (9) with RT and substitute
I
..
"
"
'
q = R q and q' = R q to obtain

M q + K q = p (t) + Pb (t)

(3.3-22)

in which

K = RTK'R

(10)

MFF MFR]
[ MRF MRR DR

[DFJ = [AF(t)J
AR (t)

SFR]
SRR DR

parts gives

..

..

..

..

MFFDF + MFRDR + SFFDF + SFRDR

(11)

and

MRFDF + MRRDR + SRFDF + SRRDR

Also,

p(t)

= RTp'(t)

(12)

MFFDF + SFFDF

= R_TpHt)

(13)

= AF(t)

(18a)

= AR(t)

(18b)

If support motions (at restraints) are zero, Eqs. (18) simplify to

and

Pb(t)

= AF(t)

(19a)

and

I !ere the matrix Rcontains rotation submatrices for all the nodes of the element.
After stiffnesses, masses, and nodal.Joads for 'ndividual elements have
h~~~ transformed to global directions, we can assemble them by the direct
sqf./ness method [5]. With this approach we need only add the contributions from
all the elements to obtain stiffnesses, masses, and nodal loads for the whole
structure. Thus , by summation* we have

(19b)
These equations will be used in subsequent work for calculating free displacements DE and support reactions A 8 (t).
In many problems it is sufficiently accurate merely to lump tributary
masses at the nodes of a discretized continuum [8]. When doing so , we form the
lumped mass matrix Me for the whole structure as

lie

Ms= LM;

(14)

i= I

M,
0

0
M2

0
0

0
0

MJ

Mn,,

and
Me=

n,

As(t)

= L p;(t)
i=I

(15)

(20)

i=I

where n, is the n~mber of elements. In Eqs . (14) the symbols Ss and Ms represent
l!~e ~tructural stif.(ness matrix and the structural mass matrix for all the nodes.
S11rnlarly , the action vectors As(t) and Asb(t) inEqs. (15) are actual and equivalent nodal loads for the whole structure. Then the undamped equations of
, notion for the assembled structure become

MsDs + SsDs = A,(t) + A .,b(t)

(16)

i~ wh!ch Ds an~ Ds are vectors of structural displacements and accelerations.

Equation ( 16) gives the structural equations of motion for all nodal displacements, regardless of whether they are free or restrained .
.
In pre paration fo r solving Eq . (16) , we can rearrange and partition it, as
lollows:

(17)

In this equation actual and equivalent nodal loads have been combined into a
single action vector. The subscript F refers to free nodal displacements, while
the subscript R denotes restrained nodal displacements. Writing Eq. (17) in two

and

M = RTM'R

99

Transformation and Assemblage of Elements

flm till' 111w111tlonNIll 111,~. ( Iii ) 111111 (, , i. 1111' 11111!1 IX 111 Vl'l'llll 1111 lh1 1l11ht lllll Nl Ill' 1x.p11111kd
,
With /l'IIIN Ill 111111'1 II 1111 Hllllll' Nill IIM till' llllllii~ 111 Vt'I 1111 1111 1111 h II

where n,, is the number of nodes. The typical submatrix M1 in Eq. (20) signifies
11 small diagonal array defined to be
(21)

lu this expression Mi is the tributary mass lumped at node;, and Io is an identit~

11111lrix with l replaced by O wherever a nontranslational displacement occurs.
' l'hus, the lumped-mass approach has the advantage that the mass matrix Me is
11lw11ys diagonal, although not always positive-definite.
As an b.umpid of transformation and assemblage of element properties, we
11111111 l'Onsider the plane truss in Fig. 3.1 l(a). For this structure let us find the
11ltll11L'l'IS matrix S, and the consiste nt-mass matrix Ms in rearranged and part ll Hllll'd fo1111s. Ass11llle th111 the cross sLclional area of member 1 is equal to A,
111111 ol ,11L111lwl' J is l'q111d lo O,Cl/\, 1111d th11l ol' tlll!lllber 3 is equal to 0.8A.

Finite Elements and Vibrational Analysis

100

Chap. 3

--2

}2

(22)

2}

Those for a typical joint j are illustrated at the left of Fig. 3.1 l(a). For this
structure we have three degrees of freedom (numbered 1, 3, and 4) and three
restrained displacements (numbered 2, 5, and 6). Note that the arrows for
restrained displacements have small diagonal cross lines to distinguish them
from free displacements .
Figure 3. ll(b) shows a typical member i in the plane truss. Its ends are
numbered j and k, and local axes x ' and y ' have their origin at point j. Axis x'
lies along the member, and its positive direction is from j to k, whereas the
direction of y' is perpendicular to the member. These axes are inclined at the
angle y from the global axes x and y. Using the length of the member as

j2

101

Transformation and Assemblage of Elements

jl = 2} - 1

j O - j1

Sec. 3.5

0.8L

= \/x;k + Yjk

(23)

we find the direction cosines of the inclined axes to be

(a)

x'
q~

y'

qk2

\tY

x,,

----- ---

cos )'

A 21

-sin )'

L = Cx

Xjk

\
/\J2 -

= -cy

A22

sm '}' -

cos )'

TABLE 3.1

Plane truss: (a) structure; (b) member.

An arbitrary system for numbering the members and joints of the truss
nppcurs in Fig . 3. 11 (a). Member numbers are enclosed in c ircles adjacent to the
mc1ubcrs, and joinl numbers arc placed adjacent to the joints. The numbering
Hystr111 for joint displuccm<.: nt s is shown hy nu111bcrld arrows that depict the
posi ti w dirl't'lio11s of the poss lhk displ11l'l'IIIL'llts, ind11di111~ thoi,c ut restruincd
polut11. Thl'Sl' displlll'l'llll'lll 111111thl' l'S llll' ol>tul11Ld l'n1111 llll' joint 1111111bcrs, us
lollows :

Cy

(24)
Cx

Direction Cosines
Member
i

Joint
j

Joint
k

1
1
2

2
3
3

(b)

L-

Member Information for Truss of Fig. 3.ll(a)

~qk1

YJk -

In these expressions the symbols Cx and Cy denote direction cosines of the

member axis itself, with respect to axes x and y. Table 3.1 summarizes member
information for the truss , including the arbitrarily chosen joint numbers} and k,
the cross-sectional areas, the lengths, and the direction cosines cx and Cy.

Figure 3. U

A11

Length

Area

c,
L
0.6L
0.8L

A
0 .6A

0.8A

0.6
1.0
0

Cy

0.8
0
- 1.0

By considering two nodal displacements in the x' and y' directions at each
end of the member, we can write the element stiffness matrix K' for local axes
as follows:

K'

= EA
L

~~

-l
0

O
0

-~

~i

1 0
0 0

(25)

This matrix is the same as that for the nxial clement in Eq. (3.4-4), except that
tht zuros arl' insl'llld for st iffmssl'H 111 the y' direction. Thus, the size of the

102

11111111 f- hm11111t11 111111 Vll11 nll1111nl A1111ly1l1

Cho p.

matrix is expanded from 2 x 2 to 4 x 4. Equation ( I 0) l'nnhks us to trunsl'or111

the stiffness matrix from local to global axes. !Jenee,

-c;
K

= RTK'R

= EA
L

[c:!Y
-c;

(26)

[~

Noll' that tht\SC indexes uppeur us subscripts for the .t and y translations at joints
Fig. J. I l(h). As un aid in the transferring process, the numerical values
111 1 t through k.2 arc listed down the right-hand side and across the bottom of
r11d1 mutrix in Eqs. (29). After assembling the structural stiffness matrix using
1111' first of Eqs. ( 14), we obtain
/ 1111d kin

R is

Ss

(27)

~]

where the submatrix R can be written as

R

=[

l'x

-cy

Cy]

(28)

Cx

Ki

0.48
0.64
-0 .48
-0.64

-0.36
-0.48
0.36
0.48

-0.48J
-0.64
0.48
0.64

1.36
0.48
= EA - 0.36
L -0.48
-1
0

0.48
0.64
-0.48
-0.64
0
0

-0.36
-0.48
0.36
0.48
0
0

-0.48
-0.64
0.48
1.64
0
-1

Ss =

[SFF
SRF

SFR
SRR

EA

I
0
-1
0

K, -EA[~
L

0
0
3

00
0
0

-]0
1
0

0
1 0
0 0
-1 0
4

00 J
0
0

(3 1a)

5
6

0
1.36 -0.36 -0.48 ' 0.48 -1
'
-0.36
0.36
0.48 :-0.48
0
0
EA -0.48
0.48
1.64 : -0.64
0 -1
= - -------------------------~-----------------L
0
0
0.48 - 0.48 -0.64 : 0.64
'
-1
1
0
0
' 0
0
-1
0
0
0
0
1

(31b)

5
6

wltl'rc the indexes down the right side and across the bottom show the rear11111gcment.
As with stiffnesses, we can generate the consistent-mass matrix M ' for
lrnal axes of a typical member to be

M'

-1OJ
0

AJ.

2
O
6 1
[
0

= pAL

0
2
0
1

1 OJ
0 1
2 0
0 2

(32)

111 lhis instance, the terms in Eq. (3.4-5) for the axial element are repeated for
1hr 11' direction, because accelerations in that sense also give rise to inertial

The terms in each of these matrices may be transferred to the appropriate

locations in the structural stiffness matrix S, by culculating the following joint
displacement indexes:
)2
2/
JI

J. I

(29)

0
0
0
-1
0

-1
0
0
0
1
0

lfr11rranging and partitioning this matrix in the form shown by Eq. (17) produces

The latter matrix contains only the direction cosines of the member axis.
When Eq. (26) and the data from Table 3 .1 are applied to each of the
members, we obtain the following stiffness matrices for global axes:
0.36
EA
0.48
L -0.36
[
-0.48

103

l 111111,l1111n11!11111 111111 A11111nhln1111 111 I ln1111111lM

-c;

For this type of element the 4 x 4 rotation-of-axes transformation matrix

R=

.u,

s.

-cxCy

-cxCy

nl'

1.J.

,111ro11s. Thus, the consistent-mass matrix for inclined axes is of size 4 x 4. It

11111y he lrnnsformcd to global axes using Eq. (11). However, when that equation
1- 11pplil.'d, we !ind the resu lting matrix M to be the same as matrix M ' in

ht ( \l). Thc1cforc , wt rnnl'luclt that for II plnne truss

1111u111

111111tix ii!

i11v111 lunl

element the consistent-

with rotation of uxcs.

/\pply1111:, Eq . ( \l) 1111d Oil' <111111 ltu111 Tuhlt \. I, we find the consistent-

104

I lnltn I lnmn1111 nnd Vlhr1tl11r111I A11nlyI

Chnp. J

mass matrices for the members to be

M1

=M

M2

M = pAL
s
6

0
0.36
0

M,

0
1
2.72
0
0
3.28
1
0
0
0.64
0.36
0
2

O.MM

(JJ)

0
1
0
3.28
0
0.64

0.36
0
0.64
0
2
0

0
0.36
0
0.64
0
2

Vlhrnllonnl A11nly1l1

105

to the diagonaLten:n_s.

j11 111lw1 M, o, M ,.
Al this point we shou ld observe that considering the masses for the three-

1111111lll'1 truss lo be associated only with jQint translations is a poor approxi11111t1011 to the truth . In fcw.L. there often will be flexure in the membeis thl!.t
hl~11if11.1111tly influences the dynamic characteristics of the structure. For examl'k, ti II time-varying force is applied in the x direction at the middle of member
I l'lt't' Jlig. 3. 1 l(a)I, the flexural deformation in that member will dominate the
11 ~p1111se. We shall deal with this type of problem for trusses by the component11111111 method in Sec. 10.6. In the meantime, we will use Ms or Me in the forms
l111w11 here, even though the numbers of members and joints are small.

(34a)

It VIBRATIONAL ANALYSIS

111 ~kc. 3.5 we derived undamped equations of motion for free nodal displace111r11l'> of a discretized structure [see Eq. (3.5-19a)]. If there are no applied
,1111011s, these equations can be written in homogeneous form as

Rearranging and partitioning this matrix as shown in Eq. (17) gives

M,

111

1111'11 111 lh1. Joi111~ of 11 1111ss, tlwy 111ny IK mld~d directly

= 0 .36M

In these expressions the matrix M is the same as M' in Eq. (32). Terms in
matrices M 1, M 2, and M3 may be transferred to the consistent-mass malrix M,
for the whole structure using again the displacement indexes j 1 through k2 .
Thus , when M, is assembled in accordance with the second of Eqs. ( 14), we find
that
2.72
0

11

= [::: ::]

MD+SD=O

(1)

2.72
1
1
3.28
pAL O
O
- 6
0
0
0.36 0.64

0
3

0 :
0 :
3.28 :

0
0

I
I

1
1 :I 2.72

0 : 0
0.64: 0.36
I

0.36
0.64
0
0
2
0

0
0
0.64
0.36
0
2

I lt1l' the subscript F (for free displacements) is omitted to simplify the notation.
Equation (1) has a known solution [9] that may be stated as follows:

3
4

I
I

'

(i

= I, 2, ... , n)

(2)

wh11c II is the number of degrees of freedom. In this harmonic expression , 4>;

1 11 vector of nodal amplitudes (the mode shape) for the ith mode of vibration.
I h1 symbol w1 represents the angular frequency of mode i, and a; denotes the
11/t,111 a11gle. By differentiating Eq. (2) twice with respect to time t, we also find

where the indexes at the right and bottom again show the pattern of rearrangement.
Alternatively, the consistent mass matrix M, in Eq. (34b) could be replaced by a lumped-mass matrix Me, as follows:
0.68
0
0
0
0
0
0
0.82
0
0
0
0
0
0.82: 0
0
0
0
Me= pAL ------------ -- -----L----------------0
0
0
0 : 0.68 0
0
0
0 ' 0
0.5 0
'
0
0
0 '' 0
0 0.5

D; = 4>; sin (w;t + a;)

(34b)

111111

D;

= - wT4>; sin (w;t +

a 1)

(3)

'111h1,litution of Eqs. (2) and (3) into Eq. (1) allows cancellation of the term
111 (1,,,1 f- a1), which leaves

(4)
3
4

(35)

5
6

I Ills ruanipulation has the effect of separating the variable time from those of
,pun , und we are left with a set of n homogeneous algebraic equations.
Ll4uution (4) hus thc..form oLthe algebraic eigenvalue problem From the
1hr111 y of homogeneous equations 1101, nontrivial solutions exist only if the
d1'111111i11unt of the coefficient mutrix is equal to zero. Thus,

JS

"

For L'IIL'h muss matrix M, and Ml', the totul muss of llll' trt1ss is 2p/\/, for hoth
llw I and 11 dirL'rtions, us it shou ld hL . ll 11011til1ur1u1.il nHIL'L'Utrntcd 11rnss1.:,; also

cu f MJ

(5)

I ,p1111:-;i1111 nl this d1ll'll11i1111111 yi1 ld~ 11 pol y11n111111l of order II called the charac'lh1 1111111l~ ,o: ol llm poly1111111i11 l 1111 thl' 1ht11'<111,ri.1fic values,

1,1/1ft111111,1f11111

11111111 ~ ln111nnt1 nnd Vll11 nll1111nl AnnlyI

100

or eigenvalues. Subslilulion of lhese rools (one at n tillll') 11110 lhc homogeneous

equations LEq. (4)J produces the characteristic v,cwrs. or l'i>:envectors <1>1,
within arbitrary constants. Alternatively [9), each eigenvector may be found as
any column of the adjoint matrix Hf of the characteristic matrix H;, obtained
from Eq. (4), as follows:
H;<I>; = 0
(6)
where
(7)

The methods implied by Eqs. (5), (6), and (7) are conducive to hand
calculations for problems having small numbers of degrees of freedom. Examples at the end of this section demonstrate such calculations. However, a structure with a large number of degrees of freedom must be handled by a computer
program, as described in Sec. 3.8. Such a program would include a subprogram
(or subroutine) for calculating eigenvalues and eigenvectors. The most efficient
type of subroutine for structural vibrations accepts the eigenvalue problem only
in the following standard. symmetric form:
(A - A;l )X;

=0

!11

Ch11p. 3

(8)

in which A is a symmetric matrix and I is an identity matrix. The symbol Ai

dcnoles the ith eigenvalue, and X; is the corresponding eigenvector for a new
syslcm of n homogeneous equations. We can put Eq. (4) into the form of Eq. (8)
hy fm;toring either matrix Sor matrix M, using the Cholesky square-root method
11 , 51. We choose to factor S for an important reason that will soon be apparent.
Thu.~,

111

107

Vll11111l111111I A11nly I

I 'Iii ( I I) I, wl11ch is gunruntccd to be symmetric. In addition, we see that the

, 1p111v11l11t' Ai is cquul lo lhe reciprocal of the square of the angular frequency.
l ht~ 11, tlw t'(.>ll~<:quencq of ~h<ill!.M ..JQ factor matrix S and is numerically
11dv1111t11gl'ous because the highest eigenvalue (corresponding to the lowest fre1111r11tyJ has the greatest accuracy. The eigenvector <l>u; in Eq. (10) is related to
I, hy lhc last of Eqs. ( 11 ). This expression constitutes a change of coordinates
111 11 IIL'W set where the stiffness matrix is equal to I. After the eigenvalues and
1,.111vcctors have been found from Eq. (10), the angular frequencies and mode
h11pts (i n the original coordinates) can be determined as
1

W = - -

'

VA;

Because the matrix Mu in the new coordinates is symmetric, all of its

1lpr11vectors are linearly independent [10]. In addition, two eigenvectors <I>~,;
111111c1,,IJ corresponding to distinct eigenvalues A; and A1 will be orthogonal with
ll ~Pl'L't lo each other. Thus,

(i

* i)

l luwcvcr. the back-transformation giyeilJl.Uhe second expression_in Eq. (12)

,loc~ 1101 necessarily preserve orthogonality among the eigenvectors, Instead, the
, l~lnvcctors <I>; and <1>1 in the original coordinates are orthogonal with respect
IU umltix...S.. as follows:
<l>TS <l>j = <l>b;u- Ts

u-11>uj = <l>~;<l>uj = 0

lli2

111 ~L'C. 4.2 we will show that the eigenvectors are also orthogonal with respect
111111rix M.
From vibrational analysis of structures, we may find repeated frequencies
1111 r,peated eigenvalues) as roots of the characteristic equation. If an eigenvalue
I tl'peated m times, it is said to be of multiplicity m. We can find m linearly
1111llpendent eigenvectors corresponding to the repeated eigenvalue, but such
\'l'l'tors are not unique. A new set of eigenvectors can always be formed as linear
111111binations of the original set and still satisfy the eigenvalue problem. That is ,
111

(9)
where lhe factor U is an upper triangular matrix. Substitute Eq. (9) into Eq. (4)
to obtain

(UTU - w7M)<I>;

=0

Then premultiply this equation by u-T and insert I = u - 1 U after matrix M,

which yields

i=I

u - T(UTU - w7M u - u)<I>; = 0

1

lkwriting terms in reverse order, we find that

Wlll'll'

M11

U 'l'M U

,,,1

l llJ

Eq11ntio11 ( 10) is now in thl' sln11d111d , ~y 111111lt1 ll' 1111111 of till' l'igl11val11l'
p111hll 111 g1vt11 hy 11q (H) TIil' 1111111h A i'4 ll'Pll'~t11ttd hy M11 ltlll' lu st of
0

(15)

i= I

whtic ci is a scalar multiplier of the ith modal vector and Am is the repeated
lu111vulue. ll is always advantageous t0 form a new set of eigenvectors that are
111 thogonal with respect to each other. They will automatically be orthogonal to
1 l l,ll 11vectors corresponding to distinct eigenvalues. But there is still an infinity
111 d1oil'es for nonunique orthogonal vcclors corresponding to a repeated eigen\'llh11. The (iram Sd1111idt 1>1tlrow malizutio11 procedure [ 11] is a formal mathe11111l 1t11 l 11pproad1 rn111111011ly usL'd lo l'o11st111c1 1111 orthogonal set of eigenvectors
1111111 11 l11ll'11rly i1ukp111dt11t HtI
II llll' 'tllflt1l'"" 11ml11~ ts s1:111tlkhnill', tt rn1111ol hl' fnl'torl'd as dcslTibcd
1

(Mu - A;I)<l>u1 = 0

M u L c;<l>u; = AmLC;<l>u;

1111110 1:-lomontN nnd Vlh111ll1111nl Annlyh

108

Chnp. 3

Soc. 3.0

109

Vlbrntlu1111I A1111ly11I

hen:. bccausc ut h:ust one rig,id-body modG (with <v1 0) is p,c::.cnt. In such 11
case ii would be po.s.siblc toJactor the mass matrix in~tcud, using Lhe reduction
technique in Sec. 6.6 if necessary to achieve a positive-definite matrix . If the
mass matrix is di agonal (as in the lumped-mass method) , its factorization be
comes simply
(16)

where M 1l 2 contains diagonal terms equal to the square roots of those in M. In

this case the transformation to standard, symmetric form yields
(17)

where
A;=

WT

(18)

In the first of Eqs. ( 18) the matrix M- 112 contains the reciprocals of the diagonal
terms in M 1l 2 After finding the eigenvalues and eigenvectors from Eq. (17), we
can obtain the angular frequencies and mode shapes from
W;

= \IA;

(19)

Note that the numerical advantage mentioned before is not present in this
)!Pproach.
In general , the process of directly extracting the roots of the characteristic
equation for Eq. (8)' must be done iteratively and is not efficient for large
problems. If the eigenvalues and eigenvectors of all the modes are to be found,
it is best to use Householder transformations (see Sec. B.2) and convert matrix
A to tridiagonal form. Then the final values of A; and the vectors X; can be
determined by iteration with the QR algorithm. On the other hand, if only a few
modes are desired, the method of inverse iteration with spectral shifting is more
efficient (see Sec. B. l).
Example 3 .1
Figure 3. l 2(a) shows a plane truss with two degrees of freedom at joint 1. Assume that
the cross-sectional areas of members 1 and 2 are equal to 0.8A and A. Using the
consistent-mass approach, find the angular frequencies and mode shapes for thi s
structure.
The 2 x 2 stiffness matrix S for the free displacements in this problem is

S = EA [ 0.36
L - 0.48

- 0.48]
1.64

11 Wl' llI .I'

!:/\ / /, 1111d Ill

\,} HriA I

(a)

I
I
I

I
I
I
I

(b)

pAl, , .1 .2H O
(1
0
1,2 H

/ 11 , lhl l111111111l' lll' IHI N l' lllllllillllN in

0.6L----

(a)

and the consistent-mass matrix has the diago nal form

1..,__ _ _

(b)

1-'.q. ('1) hl l' lH1W

1

Figure 3.12

(c)

(a) Plane truss; (b) mode l; (c) mode 2.

Finite Elements and Vibrational Analysis

110

0. 36s - mw'f
[

-0.48s

-0.48s
1.64s - mw'f

Chap. 3

111

Vibrational Ana lysis

J[ <l>1;J [OJ
<P2;

Sec. 3.6

(c)

in which the 2 x 2 array of coefficients is the characteristic matrix H;. Setting the
determinant of H; equal to zero in accordance with Eq. (5) produces the characteristic
equation, as follows:

m 2 w1 - 2mswt + 0.36s 2 = 0

G)

(d)

= 0.2m

1.s.:.
m

(e)

from which we find the angular frequencies to be

Wi

(oTs =
\j-;;-

0.6049
L

@.

\jp

_ 1.815
--

lo
-

'

1~

The roots of this quadratic equation are

wf

o/.l

L
(a)

(f)

S ubstitution of these values (one at a time) into the homogeneous equations (c) produces
the mode shapes
(g)

q>I = [~]

(b)

which arc scaled arbitrarily. These shapes are depicted in Fig. 3. l 2(b) and (c) .

!001.uiue 3..Z
A tnntilc vcr beam consisting of one fle xural element appears in Fig. 3.13(a). It is fixed
nl nodc I but has two degrees of freedom at node 2 . Assuming that the beam is prismatic,
dlltllrn1i ne the angular frequencies and mode shapes, using translational consistent mass
ll'IIIIS.

From

Eq. (3 .4-24) , we find the 2 x 2 stiffness matrix for the free displacements

to he

(c)

S = 2EI [ 6
L 3 -3L

- 3L]
2L2

Figure 3.13

(h)

wr =

Also, Eq. (3.4-26) gives the consistent-mass matrix as

pAL [ 78
M, = 210 - Ill

Lettings

lll]

(i)

2L2

_ 3.533
W1 -

uhlul n the ho mogeneous equations

6(s - 13mw;)

L (3.1 - I lmwr)

-L(3s - t tmwn ] [ <1>11] =

2L2(s mwr)
<1>21

[OJ
0

(j)

whc1l' Ilic 2 x 2 cocfli cicnt matrix is the c hnructeristic matri x 11,. As indicated by
l\q. (~). Wl' Sl't tlw dctc1111in11nt ol' 11, cqunl to Zl'l'O , produci ng
( k)

whil'h IK 1111' 1 l1111 m hiiNll1

1q1111111111

'lh,11 1hr 1111 ,tM111 l

q (~ ) 111 1 lrn1111l to lw

s
m

0.02971 -

W2

s
m

= 2 .885 -

(C)

' uhstituting the known values of sand m and taking square roots , we find the angular
li1quc ncics, as follows:

2EI /C and m = pAL /2 10, we substitute matrices S and M, into Eq . (4) to

(a) Cantilever beam; (b) mode l; (c) mode 2.

/El
\/pA

Wz

= 34.81

L2

/El

\jpi4.

(m)

Wl11111 these formulas arc compared with exact values [9] , the errors are found to be
I 0.48% and e 2 = +58%. Thus , the first-mode frequency is a good approxi111111 hu1, hut the second-mode frequency is very poor.
In this cxumplc we obtain the mode shapes by using the first column of the adjoint

, 1

11111t ilx

II ,', which is

II '/,

.
I

21.2(.1

/, ( \,\

lll(rJl )
I I 1111,,f )

(n)

f lnltu I l11111nn1 11 nncl Vih111tl111111I A11nly1l1

112

Substitution of cu ~ and

wl from
(1)1 =

Chnp. 3

113

!iy111111111tlr 111111 A11t l1ym111nltlt: Mod111

Eqs . (f) into thi s column yil'ld~

L ]
[ 1.378

(())

1- sym.

Of course, the second column of Hf would serve equally well. Figures 3. l 3(b) and (<:)
show the mode shapes in Eqs. (o), which are both scaled so that the translation iN
numerically equal to L.

3. 7 SYMMETRIC AND ANTISYMMETRIC MODES

Figure 3.14 shows two examples of symmetric structures. The plane frame in
part (a) of the figure has one plane of symmetry, while the discretized plate in
part (b) has two such planes , as indicated by the centerlines. When a structure
has one or more planes of symmetry, the natural mode shapes for vibrations all
will be either symmetric or antisymmetric with respect to those planes [12]. In
problems of this type we need only analyze a portion of the original structure .
The reduction to a smaller sized problem may be accomplished by introducing
artificial restraints at joints located on planes of symmetry. In addition, the
properties of members that lie in those planes must be altered. These changes
may be incorporated into the structural data for a computer program and do not
require any additional coding. If there is one plane of symmetry [as in Fig .
3.14(a)], only half of the structure need be analyzed. If two planes of symmetry
exist [as in Fig. 3.14(b)], only a quarter need be analyzed, and so on.
When a vibrational mode is symmetric with respect to a plane of structural
symmetry. the nodal displacements. strains. stresses. and reactions will also be
symmetric with respect to the same plane. Therefore, nodes located on a plane
of symmetry must be restrained in such a manner that the structure deforms
symmetrically with respect to that plane. Figure 3 .15(a) illustrates schematically
a typical node j located on a plane of symmetry that is normal to the x axis. The
figure also shows nodes k and k' that are symmetrically located on opposite sides
of the plane. Displacement vectors at each of these nodes indicate a symmetric
pattern of deformation. Note that translations in the y and z directions and
rotations in the x sense are all in positive directions at both points k and k' .
Therefore, we conclude that the same displacements on the plane of symmetry
must be free to occur. These displacement vectors at point j are labeled j 2, j 3,
and }4. On the other hand , the translations in th~ x direction and the rotations
in they and z senses are in opposite directions at points k and k'. Thus , the same
displacements on the plane of symmetry must be set equal to zero. Hence, the
vectors labeled} 1,} 5 , and j6 at point} need to be restrained, as indicated by the
small slashes on their arrows. In general , the component of nodal translati<)Jl
normal to a plane of__symmetry and the C<)111po11l'nts of rotnt ion in the plane 11111:,t
he prcvc11t1.:d in order to cnforcL' a i:;ymnwt1 k p11ltt1n of distortion.

7,i 77, -

X 7} 7,7,

7,i 7,7,

7,i 77,

7} 7,7,

(a)
/

/--sym.

(bl

1,11111n1 J.14 Sym111c1ric i,11 u~tl11~s: (u) frumc; (b) plate.

114

Finite Elements end Vlbrntlonol Analysis

Chap.3

/A

Plane of
/
symmetry~

k--

//

/ +5

/ -t.1
/
( , t
r----~-~------+---

j2

Y,k,

I/

z1,,

Jl'J3
/.

I/

/4----x

,,-4

-1

i1i4)
/

i6

//

z1

/L:. ---y'

I
I
I /I
1/ I
I
/ Y;,

I /
I /

/,1

-/i

I
I Yw

(/

t i5

tT

j2

~ - ---+----,- -- ~
i4)
'/

r-----+--r-----~--. -

/.i1
3

I ,/ 1
I , j6

7~/

t
t

I
I
I
k-I / I
1/ I
I
I Y;,

//
/

7/:.~
/

I ,

115

Pl'CI to

the plane.

If a member of a framed structure or an element in a discretized continuum

ltt 111 a plane of symmetry. we must divide its rigidities by two in order to cut
llll ~tn1cture into egllill.parts, In the case where a member lies in two P-lanes of
\ 111111ctry, we need to divide its rigidities bs...fuut, and so on Jf a member ar
11 ll111h: element is normal to and bisected by a plane of symmetry, we must
,llvldt it into two equal parts and introduce new nodes on the bisecting plane th.at
llll H'Strained as described aboYe.,

(a)

Plane of ---. /
symmetry
7

Symmetric and Antisymmetric Modes

If a vibrational mode is antisymmetric with respect to a plane of structural

\111111ctry, the nodal displacements, strains, stresses, and reactions will also be
111111symmetric with respect to the same plane. For this case Fig. 3.15(b) depicts
ill!iplucements at points k and k' that represent an antisymmetric pattern of
,Ii l111111ation. That is, the translations in the x direction and the rotations in the
1 1111d z senses are all in positive directions at both points k and k'. From this we
, 1111dude that the same displacements on the plane of symmetry must be allowed
h I nl'nir freely. They are the displacements labeled j I, j 5, and j 6 at point j. On
1111 olhcr hand , the translations in they and z directions and the rotations in the
, 111sc are in opposite directions at point k and k'. Therefore, the same displace111r 11ts on the plane of symmetry need to be set equal to zero. This may be
,11111111plished by introducing restraints corresponding to }2, }3, and }4, as
111d1tntcd by the slashes on their vectors. In summary, the components of nodal
1111111,lation in a glane of sxmmetry and..Jbe com12.onent of rotation normal to the
11li111t must be prevented to give a pattern of distortion that is antisymmetric with
I\

,/

k'i/

n, 3.7

z1,

I , 11111plc 3.3
I 111111 \. I 6(a) shows a simply supported beam composed of two flexural elements. This
111 11111 hus four degrees of freedom and is symmetric with respect to its centerline, as
1111111 nttd in the figure. To take advantage of symmetry, we shall analyze only the
111t111 l111nd half, using restraints at node 2 for symmetric and antisymmetric deformations.
I 111111 1. 16(b) illustrates the symmetric case, for which the rotation D 3 in the plane of
111111111ty is restrained. In addition, the antisymmetric case is given in Fig. 3.16(c),
"111 11 lhc translation D 2 in the plane of symmetry is restrained. For each of these cases
", lluw only two degrees of freedom instead of the four degrees of freedom in the
1111.,111111 problem. Now let us find two angular frequencies and mode shapes from each
111 1111 NUhsidiary problems, using translational consistent mass terms.
Fm lhc symmetric case in Fig. 3. 16(b), the 2 x 2 stiffness matrix for the free
1ll11l11nmcnts (D2 and D4) is

1//
v
(b)

21 [ 6 3e
3e u 2

s=7
\\hlo h INtl111w11 from Eq. (3.4-24).

Figure 3.15 Restraints on planes of sym111ct1 y: (11) syn11m1ric modes; (b) anti
symmetric mmks.

M,

" 11v, 11 hy hi ( I I

(a)

Also, the 2 x 2 consistent-mass matrix becomes

r>A

fr

420

156

I H'

)(!) Thl' l h11111l 'h11NI 11 1111111 ix

l.3el
4f2
II, tnr 1hi~

(b)
L'IISL' hns

1lw I'm m

Finite Elements and Vibrational Analysis

116

Chap. 3

117

Symmetric and Antisymmetric Modes

Sec. 3.7

WT =

0.007305 !_

W3

(e)

= 0.9026 m

Substituting the known values of s and m and taking square roots , we find that
I

,
D'/4

Ja,

CD

/Ji-,

Wt

DJ/~
e--------~

(ii

= 9.909

.
4/ ~ -

\JpA

l2

= 110.}

l2

(ii

\JpA

(f)

where L = 2. When these formulas for the angular frequencies are compared with exact
values [9], the errors are found to bee, = +0.40% and e3 = +24%.
We obtain mode shapes corresponding to w, and W 3 using the first column of the
adjoint matrix Hf, as follows:

--x

e~ , j

2 (s - 2mwT)

H1 = [
'
-(3s + 13mwf)

(a)

Substitution of

(g)

wr and w5 from Eqs. (e) into this column produces

(h)

Mode 1

These mode shapes appear in the left-hand portion of Fig. 3.16(b) . Of course, each of
them represents half of a symmetric mode shape for the whole beam.
Considering now the antisymmetric case in Fig. 3.16(c) , we form the stiffness
11mtrix for the free displacements D3 and D4 as

S 2/e [2 1]
=

lb)

Mode 3

(i)

1111d the consistent mass matrix is

3

_ pAC
M, - 420
Mode 2

I hl.ln the characteristic matrix becomes

2

H, =

[2(s - 2mwf)
s + 3mw f

Mode 4

where .1
giws

2/i//

1m 2 wt - 22mswf

(c)

(a) Symmetric beam; (b) symmetric modes; (c) antisymmetric modes.

HI =

s-

e nnd 111
1

w, M, =

6(s - 26mwn
e(3s + l 3111c,l )

e(3s + I 3mwl)
2
(s 2mwn

ze

11111 11

W2

111m 'l'l

(k)

1 h111 111 h il Mih 1q 1111tl1111

,ll ,11111111; I I.I ~

1111 11111(~ 111 l'q

(ti) 111r

+ 3s 2

()

= 0

1 s

= - -

1m

W4

s
3-

(m)

ding us before, we find the angular frequencies to be

,,, 1

I" 1111

which the roots are

(l' )

pA (' / 420 . /'k'lling thl dctc, mi111111t or 11111tri x 11 1 equal to ~.l'1t1

l"l ',111 ' 111,'

Whh h

s + 3mw f
2(s - 2mwf)

whlre s and m are the same as before. Expanding the determinant of H, and setting it
1 qunl to zero gives the characteristic equation

-.....::_:::;..
Figure 3.16

(j)

-3

(d)
lt11 w li1d1

th1 111111 ~

111r

43. 82
/,J

I 11 'Y.,

J/i/
11A

111111

I J. / 17,,

(1) ,1

2001.8
I,

J!JI
11/\

(n)

118

Finite Elem ents and Vibrational Analysis

Mode shapes are given by the first column of H 1, which is

a

Hi; -

2(s - 2mwt)
-(s + 3mwt)

119

Program VIB for Vibrational Analysis

Chap. 3

Flowchart 3.1

Main program for VIB*

IOI

(o)

SDATt

I. Read and write

structural data.

Substituting w~ and w~ from Eqs. (m) into this vector yields

2. Generate structural
stiffness matrix.

(p)

These mode shapes arc displayed in the left-hand part of Fig. 3. 16(c), where each of them
depicts half of an antisymmetric mode for the whole beam.
The error calculated for the first angular frequency w1 is acceptable, while those
for the other modes are not. Better accuracy for these modes could be obtained by using
more finite elements with more nodal degrees of freedom.

3. Generate structural
consistent mass matrix.

4. Convert eigenvalue problem

to standard, symmetric fonn.

3.8 PROGRAM VIB FOR VIBRATIONAL ANALYSIS

In this section we discuss a computer program named VIB for vibrational

analysis of any type of linearly elastic framed structure or discretized continuum.
Steps in the main program appear in Flowchart 3. 1, which calls seven subprograms indicated by the names in double boxes. Subprogram SDATreads and
writes input data for a particular type of structure and calculates nodal displacement indexes. Subprogram STIF generates the structural stiffness matrix (for
free nodal displacements only) by assembling contributions from element stiffnesses, as indicated by the first of Eqs. (3.5-14). Next, the consistent mass
matrix for free nodal accelerations in the structure is assembled by Subprogram
CMAS, using contributions from individual elements [see the second of Eqs.
(3.5-14)] . The subprogram named STASYM then converts the eigenvalue problem to standard, symmetric form by factoring the structural stiffness matrix, as
shown in Eq. (3.6-9). If the stiffness matrix is found not to be positive definite ,
the mass matrix is factored instead. However, if the mass matrix is also found
not to be positive definite, an error message is written and calculations stop.
Otherwise, the subprogram EIGEN2 solves the eigenvalue problem using
Householder transformations and the QR algorithm (see Sec. B.2). Then the
eigenvectors are transformed back to the original coordinates with Subprogram
TRAVEC, using the second of Eqs. (3.6-12). Finally, the subprogram named
RESl writes the resulting angular frequencies and mode shapes obtained from
solution of the eigenvalue problem . As shown by the flowchart, several structures of the same type may be processed in one run of the program .
Program VIB may be specialized to become VIBCB for continuo us beams,
VIBPT for plane trusses, and so on. The main program for each specialization
has four subprograms that are different for each type of structure, as indicated
by the second footnote helow Flowt:harl .1. I . For l'Xa111plc, the suhpmgru111
named SDAT hl't'O llll'' SDATC'B fill II l'1111ti1111011s lw11111 , Sl>AT(Y(' fo1 n plalll'

5. Solve eigenvalue problem

by Sec. B.2.

6. Transfom1 eigenvectors
to original coordinates.

7. Write angular frequencies

and mode shapes.

Go to IOI and process

another structure.

END

*Applies to any type of linearly elastic structure.

t Subprograms that differ for every type of structure.

tt nd so on. Notatio n for this and other programs is given as Part 5 in the
II 1 111 11otatio n near the end of the book. Detailed steps in the logic for various
ulipwgrnms are shown in the flowchart for Program DYNAPT, which appears
111 Applndix C.
'1'11hlc 1. 2 shows preparatio n of structural data for plane trusses. In the
11111111 ltm of the table arc the number of nodes NN, the number of elements
NI , lhl 1111111bcr of restrained nodes NRN , the modulus of elasticity E, and the
11111 11 d111sily RIIO. Enl'h linl' of till' duta for nodal coordinates (NN lines total)
\ 1111111111, 11 nmk 1111111lw1 J, tlw I l'Oo11li111ttl' X (J ) of tlw node, nnd they coordinate

1111 N,

Flnlto El omonts 1111tl Vlhr 11tlo1111I Arw lysls

120

121

l'royrnm Vll3 lo, Vlbrn1lo1111I A11ulysl11

Chop. 3

t4
,_T

TABLE 3.2 Structural Data for Phmc Trusses

No. of
Lines

Type of Data
Problem identification
Structural parameters
Plane
(a)
(b)
(c)

Items on Data Lines

Descriptive title
NN ,NE, NRN, E, RHO

I
I

truss data
Nodal coordinates
Element information
Nodal restraints

O.Bl

J, X(J), Y(J)
I, JN(T) , KN(I), AX(l)
J, NRL(2J-l), NRL(2J)

NN
NE
NRN

Y(J). The element information (NE lines) consists of the element number I, the
j node JN(I) at one end , the k node KN(I) at the other end , and the cross-sectional
area AX(I) .
Each of the NRN lines in the last block of data contains a node number J
and two code numbers which indicate the conditions of restraint at that node. ThL
symbol NRL(2J-1) denotes the condition of restraint against translation in the \
direction at node J , and the term NRL(2J) gives the restraint against translation
in the y direction. The convention adopted in this book is that if the restraint
exists, the integer 1 is assigned as the value of NRL; but if there is no restraint,
a value of zero is assigned. Of course, the vector NRL must initially contain on ly
zeros (by clearing it) before the restraint information is read.
If desired, we could include data for external masses that are idealized to
be concentrated at the nodes of the structure. Such extra masses may be con,
veniently added to the data lines for nodal coordinates. When the program
assembles the consistent mass matrix for the structure , these concentrated
masses would be added to diagonal terms for translational accelerations . Such
a procedure for handling superimposed masses could be applied to any type of
framed structure or discretized continuum.

~r

t J - - -- - - - - - 0 ~

~l~

\- ~ 0.6l~

(b)

(a)

_l_

Example 3.4
We will now use the specialized program VIBPT to find frequencies and mode shapes
for the plane truss with three members examined previously in Sec. 3.5. This truss is
reproduced in Fig . 3 .17(a), where the free and restrained nodal translations are indicated
by numbered arrows. Structural data for this problem are listed in Table 3.3, in wh ich
the following numerical values are assigned to the parameters E, p, L, and A:
E = 3.0

x 104 k/ in. 2

L = 250 in .

p = 7.35 x 10

k-s2 / in.4

10 in. 2

where the material is steel and US units arc used. (Sec Appendi x A fo r a discussion o1
systems of units and 11111tcrinl properties.)
Tobie \.'I contuins the rnmputer n:sult ~ lor rlus l'x11111pll'. In the rir st pnrt of tlw
lnhle, Wl' Sl'l' 1111 "l'l'itn" pr 1111 o l lht d11111 1l'lld hy lhl' l'0111p11lt1 AIN11 co111p11ll'd 1111d prinll'd
lll l' till' t'lt1t \l'III l11111th11 ll 1111' d11 11 l11 111 l m1 111 ( x 111111 ( 'Y, 1111 11111111w, 111 \lt'j,\ll'\'S 111

(c)

l'l1111rc 3. 17

(d)

(a) Plane truss; (b) mode l ; (c) mode 2; (d) mode 3.

11,, 111 1111 Nl>I, 11 nd the number of nodal restraints NNR. This is followed by the angular
111 q,11 Ill il'~ 11 11d 111odc shupes for cuch of the natural modes of vib.ration, which may be
, 111tll1 i11rd hy h1111d c11k11h.11ions. The angul ar lhqucnc:7 for mode .1 1s w.'... = 420.~;ad/sec
'"' ) ). 111111 1111 IIHHk]. IS (t/J
11 68 s 1 '1111d th111 tor mode 3 IS W3 - 1862 s . Each
11u11lnl vi- 1 1111 11 11 ~ hlt11 ,1111 1111tli:nd with H'N(ll'l I 111 its l11rgcs1 term, and the mode shapes
11 d1 ph lrd Ill ltp \ I l(h) (d )

Finite Elements and Vibrati,Jnal Ana lysis

122

TABLE 3.4
TABLE 3.3

MODE
2
ANGULAR FREQUENCY
NODE
DJl
1 8 . 6725E-01
2 -l.7149E-01
3 O.OOOOE+OO

Alphanumerical Values

Problem identification
Structural parameters

Example 3.4: Three-member plane truss

300(().0
0.000000735
3
3
2

Plane truss data

(a) Nodal coordinates

0.0
150.0
150.0

2
3

I
I

I
3

(b) Element information

(c) Nodal restraints

TABLE 3.4

(Continued)

Structural Data for Example 3.4

Type of Data

123

Program VIB for Vibrational Analysis

Chap.3

MODE
3
ANGULAR FREQUENCY l.8618E+03
DJ2
NODE
DJl
1 1. OOOOE+OO O.OOOOE+OO
2 -6.0504E-01 -6.1068E-01
3 O.OOOOE+OO O.OOOOE+OO

0.0
200.0
0.0

JO.O

3
3

6.0

8.0

1.1677E+03
DJ2
O.OOOOE+OO
l.OOOOE+OO
O.OOOOE+OO

K11.11mple 3.5
1<1 urc 3.18(a) shows a plane truss with a larger number of members. The cross-sectional

11

I
I

111 s
1 11

Computer Output for Example 3.4

PROGRAM VIBPT

of diagonal members are equal to I .SA , and those of other members are equal to
\ This truss happens to be symmetric with respect to its centerline, so we need only
11111lyzc half the structure. For this purpose, we impose restraints in the plane of sym1
1111,11y for symmetric and antisymmetric deformations, as indicated in Figs. 3.18(b) and
ti ), In addition, the cross-sectional area of member 12 is divided by 2 . Assuming that
y

*** EXAMPLE 3.4: THREE-MEMBER PLANE TRUSS***

STRUCTURAL PARAMETERS
NN
NE NRN
E
3
3
2 3.0000E+04
NODAL COORDINATES
X
NODE
l
0.000
150.000
2
150.000
3

RHO
7,3500E-07
y

0 . 000
200.000
0.000

ELElMENT INFORMATION
AX
J
K
ELEM.
10 . 0000
1
2
1
6.0000
1
3
2
8 ..oo 00
3
2
3

ex
CY
0.6000 0.8000
1 .0000 0.0000
0. 0000 -1. 0000

EL
250.0000
150 .0000
200.0000

(al

NODAL RESTRAINTS
NODE NRl NR2
1
0
1
3
1
1
NUMBER OF DEGREES OF FREEDOM:
NUMBER OF NODAL RESTRAINTS:
MODE
1
ANGU~AR FRIDQUENCY
NODIC
I
2
I

OJ l

2.J137lt 01
I OOOOICi 00

0,000!llCtOO

NDF
NNR

3
3

4.1995m~o2
D,12

o. 0000,c, oo
2, 472 21C 0 1
o. ooomc, no

1"1)l1111 ,\. Ill

(11) 1'111111 t111NN; (h) 111111!1- 1 (Nyn1111cttlc); (c) mode 2 (antisymmetric).

124

11111111 I l11mn11t N1111d Vlh1ntl111111I A1111IYNh1

126

I 11ol!ln111a

t hnp I

l lrrirniltl,iko , S. t>., Yo11111-1, I>. 11. , 111111 Wl' IIVN, W, , J1., , Vl/m 1tion Problems in
1,:11Hl1w1rl11H, 4 th ed ., Wil ey, New York , 1974 .
.
r
W
Jr
Mlllri.
A
Algebra
f
or
Engmeers,
2nd ed.'
.
I
W
It) <lill', .J. M., nnt
cave , .,
llrooks /Co lc, Monterey, Calif., 1983.
.
Y k 1973
11 1lohn ' E E. ' W eme11tary Matrix Algebra , 3rd ed .' Macm1llan, New or .'
.
'I
k
'
.
p
G
"Symmetry
in
Structural
Mechanics,"
ASCE,
J
.
Struct.
Dzv.,
Vol.
I l ( t oc 110 1, , ,
1111 , No . ST! , 1973, pp. 71- 89.
11

-- ~-""
l

ll " \.
", I

---~--

(c)

Figure 3.18 (cont.)

the truss is aluminum , we give parameters the following numerical values:

E = 6.9 X 107 kPa

p = 2.62 Mg/ m

l=5m

A = 6

PROBLEMS

F
P3 4-1 shows an axial element with a linearly distribut~d load (fo~ce per
u~~~~; ngth) given by the formula bx = b_i : (h2 - bi)x/ L. Fmd the eqmvalent
f' {. 1; [ ,-{ {_
nodal loads Pb(t) = {pbi' Pb2} due to this mfluence.

10- 3 m2

f" [

where SI units are implied (again, see Appendix A).

To process this truss with Program VIBPT, we must analyze half the structufl
twice. In the first analysis , restraints on the plane of symmetry allow only symmctrk
modes of vibration; and the second analysis uses restraint data for only antisymmetrir
modes. Figures 3.18(b) and (c) illustrate the mode shapes corresponding to the first and
second angular frequencies w, = 79.55 s - 1 and w2 = 168.9 s - 1 . We see that the first
mode is symmetric with respect to the plane of symmetry, while the second mode is
antisymmetric.

. f,~

~ b ,

xJ

' ; [ 1/ '<"'L' ri.:.,

- I--------1 rt [ rh,.
Pb,

p,,,

l\,f)"

b,

.,

, .... 2.

I. Weaver, W. , Jr. , and Johnston, P. R. , Finite Elements for Structural Analysi.1,

Prentice-Hall, Englewood Cliffs, N.J., 1984.

1,~'(

=-1 ).

Figure P3.4-1

REFERENCES

, L-X

(kb,)

L [(L-X)

"

l \,,+ li,._ -~')-;:-

~ ( 1,i-1-( b, - L,) I:_

A parabolical!y distributed load (force per unit length) has . th~ f~rm~~
bx = bi(x / L )2, as illustrated in Fig. P3.4-2. Determme the equiva en no
loads Pb(t) = {pb,, Pb2} resulting from this body force.

2. Zienkiewicz , 0. C., The Finite Element Method, 4th ed. , McGraw-Hill, Maidenhead, Berkshire , England, 1987.
3. Cook, R. D., Concepts and Applications of Finite Element Analy sis, 2nd ed., Wiley,
New York, 1981.

5. Weaver, W., Jr., and Gere, J.M., Matrix Analysis of Framed Structures, 2nd ed .,
Van Nostrand Reinhold , New York, 1980.
6. Oden, J. T., Mechanics of Elastic Structures, McGraw-Hill , New York, 1967.
7. Archer, J . S., "Consistent Matrix Formulations for Structural Analysis Using FiniteElement Techniques," AJAA J., Vol. 3, No. 10, 1965 , pp. 1910- 19 18.
8. Clough , R . W. , "Analysis of Structural Vibrations and Dynamic Response," Rel'.
Adv. Mat. M ethodsStruct . Anal. Dts., ed . R. 11. Gullugher, Y. Ynrnadu, a nd J. 1:
Oden, University of Alabama Press, llunt svillc, Alu ., 197 1, pp. 25 45.

~b~--x,u

p:--i.-,,-~---.1

4. Timoshenko, S. P., and Goodier, J. N., Theory of Elasticity, 3rd ed., McGraw-Hi ll ,
New York , 1970.

.,.....,.

p.,

Figure P3.4-2
" ss ume that an axial clement has three nodes , as shown in Fig. Pd?.41-3(a). I
"te 'rms
of the coordinate x measure d from node 2 ' the quadratic
.
- (2x1sp
- acemer
L)x/ L;
.I
.
pirt (b) (d) of the figure are. f1 ' "
2
0 cnvc thc 3 x 3 stiffness matri
Jishape(I} uncti4.,ons
i )/Im
}, and.{,
(2.i I /,) ,1 / / '..
K l'or thi s l'I Cllll'llt ii' the nxinl d /\id ity /\'; \ is constant along the length.

128

1111110 I lorrntrUN 1111d Vlltrnll111111I A1111fy11IH

--

I . ~q,

q,

/:

./
L

,K

1111hl11111 11

127

----x, u

Q3

~ ~ - - - - --x,IJ,

.I

/1 111

P.,

, - m,
- - - - -

'

L------+l

(a)

Figure P3.4-6

Suppose that a concentrated moment Mx is applied to the torsional element at the

distanccx from node 1, as shown in Fig. P3.4-7. Determine the equivalent nodal
loads Pb (t) = {pb1, Pb2} caused by this moment.

~----____::::;="
(b)

P11,

-M.

-----+--------- -

------

X,

IJ,

Figure P3.4-7
(c)

I ,,,It

The flexural element shown in Fig. P3.4-8 is subjected to a triangular load

by = b2x/L (force per unit length). Derive the equivalent nodal loads
Pb (t) = {pb1, Pb2, Pb3, PM} indicated at points 1 and 2.

(d)

Figure P3.4-3

3.4-4. For the axial element with three nodes [Fig. P3.4-3(a)], derive the 3 x 3
consistent-mass matrix M, assuming that p and A are constant along the length .

3.4-5. Let a uniformly distributed load b., (force per unit length) be applied to the axial

Figure P3.4-8

element with three nodes [Fig. P3.4-3(a)]. Find the equivalent nodal loads
Pb(t) = {Pb1, P1,2, Pb3} due to this body force .

3.4-6. For the torsional clement shown in Fig. P3.4-6, obtain the equivalent nodal
loads p1, (t)
{P111, p1,2} caused by a puruholicully dislribulecl moment (per uni I
length) given ns 111,
111,d I (l'/ 1.)2 J,

1.4-1), Figure P3.4-9 depicts a concentrated force Py and a concentrated moment M,

nppl icd to II flexural elcmcnl at the distance x rrom node 1. Obtain the equivalent
nodtil londs 1>1o(I)

{P1o 1, J>1oi, JJ1,1 , p1,.1) for cuch of these actions.

128

1lnltu I lon10nta1 1111d Vlh1nll1111nl A1111ly1ltt

Chop. :l

y, V

P",

;/1:

pr

I,
PoJ

129

l 111>l>lt1111"

M,/,1
L

P-~,

II II

CD

CD
y

l,

'/

I
3.4-10. In Fig. P3 .4-10 a linearly distributed load by = b1 + (b 2 - b1)x/L (force per
unit length) acts on a flexural element. Find the equivalent nodal loads
Ph (t) = {Phi, Pb2, P&3, Pb4} due to this influence .

0.8LI

0.6L

Figure P3.4-9

I,

0.8L

Figure P3.5-2

Figure P3.5-1

Repeat Prob. 3.5-1 for the plane truss shown in Fig. P3 .5-2, assuming that
cross-sectional areas of members l and 2 are 0.8A and A.
For the plane truss shown in Fig. P3.5-3, repeat Prob. 3 .5-1. In this case assume
that the cross-sectional areas of members 1 through 4 are equal to A and that the
area for member 5 is V2 A.

y,v

r
L

Figure P3.4-10

3.4-11 . Rederive the 2 x 2 stiffness matrix K for a torsional element, using moment M,
and twist i/1 as generalized stress and strain. Assume that the torsional rigidity
GJ is constant along the length.
3.4-12. Derive again the 4 x 4 stiffness matrix K for a flexural element, with moment
M, and curvature rp as generalized stress and strain . Let the flexural rigidity El
be constant along the length.
3.S-1.

The plane truss shown in Fig. P3.5- I has cross-sectional areas of 0.6A and A for
members I and 2. For thi s structu re !incl the stiffness mutri x S., and the consistent
mass matrix M, in rourranged und pnrtit ioncd f'onns.

I,

CD

0
3

0
L

Figure P3.5-3

l.~-4.

Ropcut Prob. 3.5- 1 for the plane truss shown in Fig. P3.5-4. Assume that the
cmss sectional arcus of mc1nhers I 11nd 2 arc 0.8A, those of members 3 and 4
111'l' 0 .6/\. und th11t or 111c111lwr ~ Is l'qunl to/\ .

130

1lnltn ~ lnt11n11t nnd Vlhr11tl1111nl A11n ly1l11

l'rohlnm

Glinp. 3

l'1gtlll' 11 I..~ I NhowNn two d ct11lllll ht1111 1 Im whk'h lhc pnrnmctcrs E, I , A, and
11111c l'OIINtn11l 11ln11g the length . Asscnthle the stiffness matrix S., and the consisllnt-11111ss mutrix M, (for trunslutionul inertias) in rearranged and partitioned
forms .

CD

o----_..;-----o 2

0.6L

131

,.

,/

Figure P3.5-4

4.

~l- ---L~----+------

.1

Figure P3.5-7

3.5-5.

For the plane truss shown in Fig. P3 .5-5, repeat Prob. 3.5-1. Assume that the
cross-sectional areas of members 1 and 2 are 0.8A, those for members 3 and 4
are 0.6A, and those for members 5 and 6 are equal to A.

Repeat Prob. 3.5-7 for the two-element beam shown in Fig . P3.5-8 .
y

y
y

/,

,I

0.6L

- -x

j.

O.BL

Figure P3.5-5

3.5-6.

G)
1~

CD
z

/ ~L

~3

L~

Figure P3.5-8

For the two-element beam shown in Fig. P3.5-9, repeat Prob. 3.5-7.
y

- -x

3~L

Figure P3.S-6

Repeat Prob. 3.5-1 for the plane truss shown in Fig. P3.5-6. In this case let the
cross-sectional areas for members I through 4 be equal to A, while those for
members 5 and 6 are equal to V2 A.

G)

4
l

I
FIMlll'C PJ.5-9

~3

l~

132

Finl to rlum1111t" 111111 Vll11 nllorml A1111ly11l11

Rltwnt Prnh. \.6 2 for the pl1tlll' 11 II NN show11 In Fig. P3 .6-4 , but let the crossNl'Clionul ureas of mcmbcrs I and 2 hc l'qunl to 0.8A and A.

3.5-10. Repeat Prob. 3 .5-7 for the two-element beum show11 In Fig. P3.5- 10.

,~

1~

G)

,/~,

l~

Figure P3.5-10

J1'l1111rc P3.6-4
3.6-1.

3.6-2.

Assuming that the mass matrix M is positive-definite and not diagonal, factor
it into M = VTV. Then transform Eq. (3.6-4) to the standard, symmetric form
of Eq. (3.6-8). Also, show the back-transformation of eigenvectors to the
original coordinates.
Figure P3.6-2 shows a plane truss with two degrees of freedom at joint I.
Assume that the cross-sectional areas of members l and 2 are equal to A and
A V2. By the consistent-mass method, find the angular frequencies and mode
shapes for this structure.

1.,,-s.

The beam shown in Fig. P3 .6-5 consists of two prismatic flexural elements with
two degrees of freedom at node 2. Member 1 has moment of inertia a~d
cross-sectional area equal to I and A, but member 2 has 2/ and 2A for its
properties. Determine the angular frequencies and mode shapes for this beam
using translational consistent-mass terms.
y

[,

.-- L--I
Figure P3.6-2

3.6-3.

1
i-,,, - - --

0.8L - ---....i
Figure P3.6-3

Repeat Prob . 3.6-2 for Lhe plane truss shown in Fig. P3 .6-3, buL let the cross
sectional areas of mcmhors I und 2 lw l'q111al to /\ 1111d O.M .

0.8L- - --

"l
L~ I

1~

r
L~_,.

133

l'whlntn

Chnp. 3

CD
/, A

~L

.1.

0
2/, 2A

~3

L~

Figure P3.6-5

.U-6.

Repeat Prob. 3.6-5 for the continuous beam shown in Fig. P3.6-6. In this.case
the beam has constant values of I and A along its length and has two rotational
degrees of freedom (at points 2 and 3).
y

11111111 1111111111111111111 Vll11 11ll1111nl A1111ly1l1

134

Ch11p J

Repeat Prob. 3.6 5 for the two clement co11tin11011Nlw11111 shown in Fig. PU, l
Cross-sectional properties I and A arc constunt , 1111d the free displacc111cnh 111
joints l and 2 are both rotational .

3.6-7.

"'''
\, I,

1:

l'rnhl"'"'

hu thl Ny1111mtric rnnti11uo11s IK11111 shown in Fig. P3.7-3, find the angul.
tll'(llll' lldcs and mode shapes for (11) symmetric and (b) antisymmetric di

hu lions. Use onl y half the structure, assuming that each of the four elements h,
tht snn1c v11 lucs of I und A.

, _____;:CD:::;.____2_ _ _ _0.;:;2:;.__ _ _~ _3 - - x
z

#i

~l,1--~~~[~~~----,-i-~~~-L---------+-j
Figure P3.6-7

3.7-1.

The fixed-end beam shown in Fig. P3. 7- l consists of two prismatic flexurnl
elements having the same values of I and A. Using only half the beam, find thl'
angular frequencies for the (a) symmetric and (b) antisymmetric modes.
y

1 7 4.
I

i-Sym.
2

CD

1~

Figure P3. 7 -3

/1

3~

.,

The symmetric plane truss shown in Fig. P3.7-4 has cross-sectional areas fo
members I and 2 equal to V2 A , whereas those for members 3, 4, and 5 arc
equal to A. Using only half the structure, calculate the angular frequencies anc
the (a) symmetric and (b) antisymmetric mode shapes.

r-Sym.
I

L= 2e

-1

Figure P3.7-l

3.7-2.

Figure P3.7-2 shows a prismatic continuous beam composeq of four flexural

elements. Determine the angular frequencies and mode shapes for (a) symmetric
and (b) antisymmetric deformations, using only half the beam.

,-:2@

(Ds_l_

/J ' > o - - - - - . = - - - ~ - - - - o ( ' v ' .

r-Sym.

,~

CD

I
z

l-e

,I,

31

e+e
Figure P3. 7-2

~ L - 1 4 L~

~
,J,

~s

Figure P3.7-4

\. - x
e--j

L- u

1.7-5. For the symmetric plane truss shown in Fig. P3.7-5, determine the angular
frequencies and the (n) symmetric and (b) antisymmetric mode shapes. Use only
half the structure, and assume thnt the cross-sectional areas of members I and
2 111c O.<,A , thot for mcmhcr 1 is 0 .8A, nnd those for members 4 and 5 are equal
lo/\ ,

136

Fl11 lto l: lomo11t11 1111cl Vll>111l lo1 111I A11uly11lo

r

Chop. 3

13,

l 1111 l>lt11 1111

y(Sym .)

"ro6L-j-o6L--::i
CD

, 0

'l

O.SL

_j~x
Figure P3.7-5

3.7-6.

Figure P3 .7-6 shows a plane truss having cross-sectional areas equal to A for
members 1 and 2, 0.6A for member 3, and 0.8A for members 4 and 5. Find the
angular frequencies and the (a) symmetric and (b) antisymmetric mode shapes,
using only half the truss.

(a)

rSym.
1

0
l.---=----0---=--------G

- I

0.6L

____.__l- x

- o.sL- -....,, j,_.--o.sL------j

Figure P3.7-6

(cl

(b)

Figure P3.7-7

3.7-7. The symmetric plane truss shown in Fig. P3.7-7(a) has cross-sectional areas of
all members equal to A. Using only half the structure, calcul ate the angular
frequencies and the (a) symmetric and (b) antisymmetric mode shapes. For thiH
purpose, part (b) of the figure shows symmetry restraints, und part (c) showN
antisymmetry restraints. (Note: For the antisymmetric case, nodes 2 a11d I
translate equall y in the x direction; so llll' ll1iw 1 '\ mny h.: trenlLd as 111 igid body.)

4
Normal-Mode Method
of Dynamic Analysis

4.1 INTRODUCTION

In Chapter 3 we discussed the formulation of action equations of motion for a

MDOF structure using finite elements. Then we showed how a vibrational
analysis can be performed by solving the algebraic eigenvalue problem associated with the homogeneous equations. Extracting the angular frequencies and
mode shapes in this manner sets the stage for the normal-mode method of
dynamic analysis [l], described in the present chapter.
A structu.re subjected to impulsive (time-varying) loads responds with a
combination of rigid-body and flexible-body motions. If a structure is restrained
(immobile), the response will involve only flexible-body motions. However, if
the structure is unrestrained or partially restrained, certain rigid-body motions
can occur as well. Within the linear theory of the noanal-mo.de..Jru:.th.od,_such
motions may consist of small or large translations but only .smaluQtations . Thus,
problems involving large rigid-body rotations are beyond the scope of this
theory.
Regarding the flexible-body part of the response, a linearly elastic solid or
structure has an infinite number of degrees of freedom and an infinite number of
natural modes of vibration. If the structure were analyzed as an elastic continuum, its flexible-body response to dynamic loads would consist of the sum of
an infinite number of vibrational motions . However, if the structure is dis
cretized by the finite-element method, the resulting analytical model will have
only a finite number of nodal degrees of freedom and a finite number of nnturul

138

n, , 11.:l

139

l'r lrwlpnl n11d Nor mnl C:111111tl11ntn11

111rnl(S

of vihmtion . 'l'hcrcfme, such a model has only a finite number of vil111111011ul motions contributing to ils dynamic response.
Systems that are subjected to arbitrary dynamic loads be~ome extremely
dlt lll'ult to analyze rigorously in their original p~ysic~l coordmates: We can
nvold these difficulties by using natural modes of vibration as gener~hzed coordl11111cs . When this path is followed, the equations of undamped m~~on bec?me
llllt'tlllplcd . In these coordinates each equation may be so~v~d as 1f it pertamed
hr II system with only one degree of freedo~. Superposition of. ~ese SDO~
1, "ults is accomplished through a transformation back t~ the ongmal .coordiuules. By this means we can evaluate time-vW?'ing nodal displacements, mtemal
lilHsscs, and support reactions for the analyttcal model.
.
.
In this chapter we develop the normal-mode method and apply it to simple
ltuctures such as beams and plane trusses. In later chapters we will use it ~or
111111 c co~plicated framed structures and other continua. discreti.zed b.y fimte
l'lr 111 cnts. Undamped systems are treated first, and special considerations req111rcd for damped systems are discussed in the latter parts of.the chapter.
An important advantage of the normal-mode me~od is t~at onl)'....lh.e
M
l11,11ifa:ant modal responses need be included in a dynamic analysis.. The oth~r
11 u,dnl responses may often be omitted without much loss of accuracy. This
trthnique, known as modal truncation, can make the normal-i:node.method more
dllcicnt than the numerical integration methods to be descnbed m Chapter 5.

4,2 PRINCIPAL AND NORMAL COORDINATES

111 order to study relationships among the natural modes o~ vibratio~, let us
11 msider modes i and j of the eigenvalue problem for action equat10ns, as
lollows:
(la)
<I>; = WT M <I>;

S <1>1

wJ M

(lb)

<1>1

l'lwsc expressions are modified versions of the algebraic eigenvalue problem

lwn previously in Eq. (3.6-4). Premultiplication ~f Eq. (la) by <l>J and post11111ltiplication of the transpose of Eq; (lb) by <I>; yields

= wf<l>JM <I>;

(2a)

~S~=~~M~

<l>JS <I>;

I hl lcfl hand sides of Eqs. (2) are equal, so that subtraction of the second
1q1111lion from the first produces the relationship

(wr - wJ) (l>T M <I>,

1111

the other hand , if Wl' divide holh silks of' Eq . (2a) by

(3)

wr and both sides of

140

Normal-Mode Method of Dynamic Analysis

Eq. (2b) by

Chap. 4

<I> TS <I>1 = 0

(4)

To satisfy Eqs. (3) and (4) when i j and the eigenvalues are distinct
(wr wJ), the following relationships must hold:

141

Principal and Normal Coordinates

Premultiplication of this equation by <f>T and insertion of I = <I> <1>- 1 before D

and D produces

wJ, the right-hand sides become equal. Then subtraction gives

- _l)
( _!_
Wf WJ

Sec. 4.2

<l>JM <I>; = <l>;M <l>i = 0

(5)

<l>JS <I>; = <l>;S <l>i = O

(6)

and

These expressions represent orthogonality relationships between the modal vectors <I>; and <l>i. From Eq. (5) we see that the eigenvectors are orthogonal with
respect to the mass matrix M. Equation (6) also shows that they are orthogonal
with respect to the stiffness matrix S, as demonstrated in Sec. 3.6.
For the case when i = j, Eqs. (3) and (4) yield

which can be restated as

(13)
By virtue of Eqs. (10) and (11) , the generalized mass and stiffness matrices in
Eq. (13) are both diagonal. Also , the displacement and acceleration vectors in
the latter equation are defined to be

DP

= <1>- 1 D

(8)

in which Mp; and Sp; are constants that depend on how the eigenvector <I>; is
normalized.
For operational efficiency, we place all of the eigenvectors columnwise
into an n x n modal matrix of the form
<I>,.]

(9)

where n is the number of degrees of freedom. Then we can state Eqs. (5) and
(7) collectively as

..

MDISI>

in which the modal matrix <I> is given by Eq. (9). The symbol w 2 in Eq. (16)
represents a diagonal matrix with values of wf in diagonal positions, as follows:

wf
0
0

w2 =

(11)

( 12)

(15)

Here we see that the generalized displacements in vector DP operate as multipliers of the modal columns in <I> to produce values of the actual displacements
in vector D. Thus, the shape functions for the principal coordinates of a MDOF
system are its natural modes of vibration.
We now restate the eigenvalue problem in Eq. (la) more comprehensively as
(16)
S <I> = M <I> w 2

in which MP is a diagonal array that will be referred to as a ,vrinciDal mass

matrix. Similarly , Eqs. (6) and (8) are combined into

where SP is another diagonal array that will be called a principal stiffness matrix.
Equations (10) and (11) represent diagonalization of matrices Mand S. If either
of them is already diagonal, the operations merely scale the values on the
diagonal.
To take advantage of the diagonalization process, let us reconsider the
action equations of motion for free vibrations of an undamped MDOF system,
as follows:

..

D=<I> DP

(10)

<f>T s <I> = Sp

(14)

Ihe generalized displacements De given by the first of Eqs. (14) are called
Jl.rincipal coordinates, for which the equations of motion [Eq. (13)] have neither
inertial nor elasticity coupling.. From Eqs. (14) we find that the displacements
and accelerations in the original coordinates are related to those in principal
coordinates as follows:

(7)
and

Op = <1>- 1 I)

(J)~

W5

(J)~

Ll1l

This array, sometimes

caHed the spectral matrix, will be referred to as the

eigenvalue matrix, or matrix of characteristic values. It postmultiplies the matrix <I> in Eq. ( 16), so that a typical modal column <I>; is scaled by the correspondPremultiplying Eq. (16) by <l>T and using the relationships
ing eigenvalue
( 10) and (11) , we obtain
(18a)

wt.

llcncc,
( 18b)

Normal-Mode Method of Dynamic Analysis

142

Chap. 4

Thus, in principal coordinates the ith principal stiffness is equal to the ith
principal mass multiplied by the ith eigenvalue.
Because the modal vectors may be scaled arbitrarily, the principal coordinates are not unique. In fact, there is an infinite number of sets of such generalized displacements, but the most common choice is that for which the mass
matrix is transformed to the identity matrix. We state this condition by specifying that Mp; in Eq. (7) must be equal to unity, as follows:
(19)
where
(20)
Under this condition, the scaled eigenvector <l>N; is said to be normalized with
respect to the mass matrix. The constant C; in Eq. (20) is computed as

C;

i <l>j;(i Mjk<I>)1

V<l>TM <I>; =

j= I

(21a)

k= I

Sec. 4.2

143

Principal and Normal Coordinates

normal coordinates. From Eqs. (14) , these vectors are related to their counterparts in physical coordinates by

DN = <l>N1D

DN= <l>N D
1

(25)

and the reverse transformations are

(26)
The inverse of the normalized modal matrix required in ~~s. (25) ~ay be
easily found . We need only postmultiply Eq. (22) by <l>N to obtam the
relationship
(27)
Thus, the desired inverse may always be calculated by this simple ~atrix
multiplication. If we wish to include only a selected number of modes mm our
analysis, the modal matrix will be of size n x m. Then Eqs . (25) must be
restated as
(28)
This technique-. for using only a limited number of. modes s called modal

If the mass matrix is diagonal, this expression simplifies to

truncation.

II

C;

L Mj <l>J;

(21b)

Example 4.1

j= I

When all of the vectors in the modal matrix are normalized in this manner,
we use the subscript N and revise Eq. ( 10) to become

<1>iM <l>N = MN = I

(22)

To demonstrate the use of normal coordinates for action equations, let us reconsider the
three-member plane truss in Fig. 4. l(a), which we studied prev!ously in Cha~ter_3. Fro~
Hq. (3.5-3lb) the 3 x 3 stiffness matrix for the free nodal displacements md1cated m

Pig. 4. l(a) is

Thus, the principal mass matrix is now the identity matrix. Furthermore, the
principal stiffness matrix, from Eqs. (11) and (18a), is seen to be
(23a)

Or, for the ith mode,

- 0.48]
0.48
1.64

(a)

In addition the consistent mass matrix from Eq. (3.5-34b) is

(23b)

Therefore, when the eigenvectors are normalized with respect to M, the stiffnesses in principal coordinates are equal to the eigenvalues. This particular set
of principal coordinates is known as normal coordinates. Rewriting Eq. (13) in
normal coordinates, we have

'

M =

2 2
;

pAL[
6

~J

3.~8
O

(b)

3.28

hu this structure, three eigenvectors were calculated in Example 3.4 (see Table 3.4).
I hl'NC vectors become the columns of the modal matrix, as follows:

(24a)
or

-0.36
0 .36
0.48

1.36
S= -0.36
L [ -0.48

EA

<I>
111 1111111111 11,l thii. 111 111y wtth

0.2314
1.0000

0.8673
0 . 171 5

l.OOOOJ
- 0 .6050

O.2472

I.()()()()

0 .6 107

ll'S \ll' l' l

to thr

11111HH

111nt11x,

Wl'

(c)

compute from Eq. (2 1a) the

144

Ll111p. 11

Nn111111I Mnd11 M1111t111l 11f 11y11111 11lc l\1111lynlt1

l'1l11nlpnl 111111 N,11111111

< 11111dlt111l n

146

, 1111~llllllN

C2
F,,

-,

,.

I ll vldl11g the columns o f'

)(

0.9384

1. 2114

- 0.1856

\/pAL - 0.2995

1.0820

SN = <l>;!::S <l>N

/
/

F,,

-J I--

(a)

lb)

F,,

/r ----+-1I

-'.j+I

i1 r ,.

F 32

r,

I
I
I

/
/
I

I
I

/
/

'/

-0.7542

(e)

0.2701
0
O

0
2.088
O

0 ]
0
5.308

= w2

(f)

wi, and w~on the diagonal. Their square roots are the
the angular frequencies given in Table 3.4. Of course, the eigenvalues in Eq.
11) were already available; so the advantages of transforming stiffnesses to normal
, 1u11 di11atcs are not obvious. These advantages will become apparent for the response
, ,iii 11lutions in later sections.
111111 11s

,
- 0.6 L - - -

1.2350]
-0.7472

" l1 k h contains the values of wr,

2.__io----~~~
-

:2 [

(d)

111hNlilution of Eqs. (a) and (e) into Eq. (23a) yields the result

0.2803

1
[

C3 = 0.8097\/pAL

by these values produces

<I>N = -

(J>

0.9242\/pAL

If a structure is statically determinate and immobile [2J, flexibility

11w/ficients are not difficult to calculate. In such a case, the action equations of
1111111011 expressed by Eq. (12) may be replaced by displacement equations oJ.
11tr>tiulL, as follows:

FMD + D=O

(29)

111 whi ch

F =

s-1

(30)

I lit.flexibility matrix F contains values of free nodal displacements due to unit

v,tlucs of the corresponding actions. We tr.ansforrn Eq. (29) to principal coordi1111lts by substituting Eqs. (15) for D and D. Then premultiplication by <1>- 1 and
hlMt' I tion of the identity matrix I = 1>- Tcl>T before M produces

1>- I F 1>- Tcl>T M cl> Op + Op = 0

whk h can be rewritten as
(31)

II

(c)

l'.'lgurc 4.1

l10n 111

I.

(a) Plane truss; (b) condition A,

Id)

I; (c)

condit ion A2 = J; (d) condi-

1lw symbol Fp in Eq. (31) represents & orincioal flexibility maJ.Jix., correspondl11p to S,,, and is defined as

Fp

= <I>

'F<I>

= Sr'

(32)

111 ,m11sc, this definition applies only when S (nnd hence S,,) is positive-definite.
l<'u,thl.!nllol'c, th<. cxpflndcd frn111 ol llw t'i)'l:nvuluc problem in Eq. (16) is

Normal-Mode Method of Dynamic Analysis

146

Chap. 4

F M<I>

= <I> A

(33)

The eigenvalue matrix A in this expression consists of a diagonal array containing values of .Ai = 1/ in diagonal positions, as follows:

wt

A1
0
0

.A2

0
0

.A3

0
0
0

An

= (1) - 2

4.3 NORMAL-MODE RESPONSE TO INITIAL CONDITIONS

(35)

For a MDOF system, suppose that w~ know the initial conditions (at time t
uf displacements Do and velocities Do, as follows:

Do=

(36)
Thus, lhe flexibility matrix in normal coordinates becomes the eigenvalue matrix
~. which is also equal to the inverse of w 2 From this we conclude that Eq. (24b)

Hives the equations of motion in normal coordinates, regardless of the method

ol formulation in the original coordinates.

wmJe U

/\s 1111 cx11mplc of the use of normal coordinates for displacement equations, we again
rn11sidcr the three-member truss in Fig. 4.1 (a). Because this structure is statically deter111l1111l c nnd immobile, we can find terms in the flexibility matrix by the unit-load method
Il l, Figures 4 . l(b)- (d) show the flexibilities obtained from applying unit values of A,,
/\ J, and A.1 (all forces). The resulting flexibility matrix for this structure is

F =

:A

l. 000
1.000
[
0

1. 000
5.556
- 1.333

0
-1.333
1.000

(g)

<l)N

= <1>i M = VpAf,

0.3290
0.3945
[

0.4353

0.7089
0.05494
- 0.2026

- 0.1637]
0.5915
- 0.41 23

(h)

I,., I J1' l N1

,,, /['70) ()
I

()

ti I /IN

tl

ll

II I HH I

Do=

Doi
Do2
D03

(1)

111 accordance with Eqs. (4.2-25), these initial values may be transformed to
11mmal coordinates by premultiplying them with the inverse of the normalized
1110<.lal matrix. Thus, we have

DNo

= <l>,v 1 Do

DNo

= <l>,v' Do

(2)

I he second relationship in Eqs. (2) is obtained by differentiation of the first with

11'Hpcct to time. The forms of the displacement and velocity vectors in normal
111ordinates are the same as those in physical coordinates given by Eqs. (1).
From Eq. (4.2-24b) we see that a typical equation of motion for undamped
l11t vibrations in normal coordinates is

wt DNi = 0

(i = 1, 2, ... , n)

(3)

I 11l'h equation of this type is uncoupled from all of the others , and we will tr~at
1111 l'Xpression as if it pertained to a SDOF system.[se~ Eq. (2.2-4)]. Knowmg
1111' rnnditions of initial displacementDNo, and velocity DNo;, we find the response
111 11 ll' i th normal coordinate as

(4)

1111'4 tx p1'l' Ssio11 is dl'awn from Eq . (2.2 9) for un undamped one-degree system.

w,

(i)

= 0)

Don

(i = l, 2, ... , n)

With this matri x und the nex ibility matri x from Eq . (g) , we find FN using Eq . (36), as
l11llow11:

Ii',.,

Doi
Do2
Do3
Don

DNi

S11hslituti11g <1,i and M from Example 4.1 into Eq. (27), we obtain the inverse of<IJN as

and .\3 = 1/wt as

At this point we should remember that working with displacement equations and flexibilities is feasible only for structures that are statically determinate
nod restrained against mobilities. Upon reflection, we realize that this is a very
limited class of structures that would be encountered rather infrequently. In
~cneral, the approach using action equations and stiffnesses is much more
Nuitable for dynamic analysis.

When the modal matrix is normalized with respect to the mass matrix, the
principal flexibility matrix from Eqs. (32) and (35) takes the form

11'

= 1/ wf, A2 = l I wt

(34)

Transforming Eq. (33) to principal coordinates as before, we find that

FpMp = A

147

Normal-Mode Response to Initial Conditions

This diagonal array contains the values of A,

confirmed by Eq. (f) in Example 4.1.

replaced by

A. =

Sec. 4.3

11u111ppl y Hq. (II) rqwlitiwl y to rnlrnl11tl' till' terms in the. vector of normal111111lr tl1 hplmt111l'1Ils l)N \nN,l 'l'hr N1 1t1nilt 'I 111t thl'II tn111Hlornwd hack to the
111lp11111I 111111d11111tr 'I, ll 'l i11p 1hr 11p1 111111111 p1vrn hy th1 1111,I ol 11.qs, (11. 2 Ui) .

Normal-Mode Method of Dynamic Analysis

148

Sec. 4.3

Thus,

DNo

<l>NDN

...,, _ ,
= 'YN
Do = vr-;-;pAL {4.486,

=0

= DNo; + DNo;t

(7)

D =

,-1

o,~

- 0.1407

1.257
=

PoL
EA

COS W1 t -

5.434 cos w,t

[ -1.343

COS Wit -

1111d
' - - - 0 . 6L - l

The.initi td displacements of the joints caused by the force may be calculated as tlw
produet of lli nnd the second column of the flex ibility matri x F in Eq. (4.2-g). Thus,

(u l

AI No, Ill 111111 I

() tilt 111ltl11l w lodtkN llll' i'>11
0 IJN it11-1 tlu tli NI ol l\qN: (l). Wl'
1111 111111111 t1111pli1111111' llt ~ ht 1111111111l 111111dl1111h N, IIN lnllnw" :

111111Ml111111

= pAL
420

Plane truss with initial load.

1:.'.~'

0.09600

w2 t

COS W 2t

0.1738

COS W3t

+ 0.1051 cos w3 t
+ 0.1061 COS W 3t

(d)

Assume that the flexural element in Fig. 4.3(a) is initially at rest when node 2 is struck
In such a manner that it suddenly acquires a velocity D03 . Determine the smalldisplacement response of the element to this impact.
Nodes of the flexural element have no restraints against either translations in the
I' direction or rotations in the z sense. Therefore, the stiffness and mass matrices are

l.:lJ.1 )

COS W2t -

In this example the response of the first mode of vibration is about one order of magnitude
)lrcater than each of the other two.

- 6

-3L

L2
3L

-6

3L
2L2
-3L

- 3L

3L

L2

- 3L

2L2

K-2Et:

( 1.000, 'i. 'i:'i6,

0.08325

+ 0.01647 cos

0.8L

Ou

COS W3t

11;xample 4.4

Po

Figure 4.2

(c)

~:

<l>NDN

Example 4..3
Suppose that the plane truss in Examples 4.1 and 4.2 is subjected to an initial force R
in the x direction at joint 2 , as illustrated in Fig. 4.2. Let us find the free-vibrationa~
response of this structure due to suddenly releasing the load.
3

4.486 cos w, t
- 0.08872 cos W2t

Then with the normalized modal matrix <l>N from Eq. (4.2-e), we transform these results
back to the original coordinates, using Eq. (5) as follows:

This f~rmula is used in_ place ofEq. (4) to evaluate the response of a rigid-body
mode m normal coordmates, assuming that rotations are small.

'VpAf,

DN =

(6)

Integration of this equation twice with respect to time yields

DN;

(b)

EA

For this purpose the inverse of the normalized matrix <l>,v is available from Eq. (4.2-h).
According to Eq. (4), we find the terms in the vector of normal mode displacements
to be

wr

DN;

} PoL

- 0.08872, - 0.1407 1

This _sequence o~ operatio?s is the same regardless of whether the original

equat10ns of motion are wntten as action equations or displacement equations.
Howeve~, _for the action-equation approach . there exists the possibility of
one or more ngid-body modes. For such a principal mode the eigenvalue is
zem..._and Eq. (3) becomes

..

149

Normal-Mode Response to Initial Conditions

22L

54

156
22L

4L2

13L

54

13L

156

- l3L

- 3L2

-22L

-13l2
- 3L

(e)

1
(f)

-22L
4L2

llh itiven by Bqs. (3.4-24) and (3.4-26). In this case the stiffness matrix is positive"1'11iidefinite, so we should expect lo find two repeated zero roots when solving the
1 l1wnvnhu.: prnb lern . Thus, so luti on or Eq. (4.2- l a) yields the e igenvalues

,,,,'

'

'" '

I io fo.'/

11Af

8400 El
I

pAJ,''

(g)

160

No rmal-M odo M othod ul l)yn11mlc Analysis

Chap.4

Sec. 4.3

Normal-Mode Response to Initial Conditions

151

l'11cse eigenvalues and eigenvectors were found using a computer program called VIBCB
Im vibrational analysis of continuous beams (see Sec. 3.8). Each of the four column

D,/

t'

CD

,/

Vl'Ctors in Eq. (h) is scaled so that the first translational displacement is numerically equal
tu L. Figures 4.3(b) and (c) depict the rigid-body modes, and Figs. 4.3(d) and (e) show
the vibrational modes. For the purpose of normalizing columns in the modal matrix with
l\'Spect to M , we use Eq. (4.2-2la) to find the constants

tb03
D3L

o./.1

C1 =

LVm

C2 =

L~

C3 =

L~

C4 =

Li

(i)

111 which m = pAL. Dividing the columns of Eq. (h) by these values, we obtain the
11ormalized modal matrix as

(a)

l>N= - l [~

LVm L

(b)

L\/3
-2\/3
- L\/3
-2\/3

LVS

LY?]

- 6V5

-12V7

LVS

- LY?

6V5

- l2V7

U)

Initial velocities in normal coordinates now may be calculated using the second
in Eqs. (2). Hence,

1 1qlrcssion

L ~
(c)~ - L

.
DNo =

L~

l>N Do = 1>iM ~OJ D 03

1

~---~

~L

(k)

(d)

I q1111tions (4) and (7) give the vibrational and rigid-body responses in normal coordiL~

1111ll'S

,,----.____

~ " 1-L

as

ce)

(e)

Figure 4.3

N111t that there is no response of the third normal mode, which has a symmetric vil1111timuil shape. Finally, with Eq. (5) we transform the results in Eq. (e) back to the
1111pi1111l coordinates to obtain

(a) Flexural element with initial velocity; (b) mode I; (c) mode 2;

(d) mode 3; (e) mode 4.

us wl ll as the modal matri x

<I>

r~

()

- tL + (L sin W4t)/ w4]

12t - ( 12 sin W4t)/ w4 [)03
I ltl
(L sin W4t )/w4 lOL
[
121 ( 12 Sill W,il)/ W,1

/,

/,

"

~21
/,

I)

(h)

1111'~1' ll'Npt!llNl'H Ill\' v11 )1d 011ly Im

111111111

111-dd hotly 111t111ltHIN.

(m)

152

Normal-Mode Method of Dynamic Ana lysis

Chap. 4

4.4 NORMAL-MODE RESPONSE TO APPLIED ACTIONS

Now we shall consider the case of an undamped MDOF system that is subjected
to applied actions corresponding to the nodal displacements. For this situation,
the action equations of motion are

MD+SD=A

(1)

Sec. 4.4

the terms in the vector of normal-mode displacements DN = {DN;}. Then the

results are transformed back to the original coordinates using Eq. (4.3-5).
For a normal mode corresponding to a rigid-body motion , the eigenvalue
11Jt is zero. Then Eq. (6) becomes
(8)
In this instance, the normal-mode response (with the system initially at rest) is

where the symbol A denotes a column vector of time-varying applied actions,

as follows:

(2)
and n is the number of degrees of freedom. Equation (1) is transformed to normal
coordinates,.by premultiplying both sides with <1>i and substituting Eqs. (4.2-26)
for D and D to produce

<1>iM<1>NriN

+ <1>is

<I>NDN

= <1>iA

This equation may also be written as

DN+ w 2 DN = AN

(3)

The symbol AN on the right-hand side of Eq. (3) represents a vector of applied
act ions in normal coordinates, computed by the operation
AN= <1>iA

(4)

153

Normal-Mode Response to Applied Actions

DN; =

J:

(9)

AN; dt" dt'

This equation replaces Eq. (7) whenever a rigid-body mode is encountered.

In summary, we calculate the dynamic response of an undamped MDO~
~ystem to applied actions by first transforming those actions to norm~! coord111ates using Eq. (4). Then the response of each vibrational mode is ob~amed from
tire integral in Eq. (7), and that for each rigid-body mode is de~enmned by t~e
double integral in Eq. (9). Finally, the values of the actual displacements m
physical coordinates are found with the back-transformation. operation of Eq.
(I\ .3-5). If applied actions do not correspond to the nodal displacements, the
11ppropriate equivalent nodal loads discussed in Sec. 3.3 can always be found as
11 preliminary step (see Example 4.6 at the end of this section).
At this point, let us examine the effect of a dynamic load A1(t), correspondIng to the jth nodal displacement, on the response of the kth nodal displacement
" k From Eq. (5) the ith normal-mode load due to A1 is

In expanded form, the results of this multiplication are

AN!]
AN2

[ ~N;, =

[<l>N11A1 + <l>N21A2 + , , , +
<PN12A1

+ <PN22A2 + , , , +

(10)

<I>NnlAnl
<I>Nn2An

;;ln~I -~ ;N~n~~ ~- ~ ;;,.,.~n

(5)

If the system has only vibrational modes, the response of the ith mode is found
lr om Eq. (7) to be
W;

= 1, 2, .. . , n)

(6)

where the normal-mode load AN; is taken from the ith row in Eq. (5).
Each of the normal-mode equations of motion given by Eq. (6) is uncoupled from all of the others and has the same form as a SDOF system with unit
mass. Therefore, we can calculate the response of a typical vibrational motion
using Du/w111el' s integral, as follows:
t') rlt'

Jo

(11)

dt'

'l'nrnsformation of this result back to the original coordinates by Eq. (4.3-5)

yidds the response of the kth nodal displacement as

The i th equation of motion in normal coordinates is

(i

= <l>Nji 1 A1 sin W;(t - t')

DNi

(7)

TIIIH t'Xjlll'HHio11 Is d111w11 lm111 Hq. () <> <,) 1111d w11s dlriwd for 1111 u11d111111wd
cu11 dl'l-' ll't' 'lYNll' lll 111111 ,~ l11i1l11lly 111 11~1 W1 11pply II H' ptliliwly lo c11ll'lll1111

(Dk)A ,

[<l>Nki<l>Nji rAj sin W;(t - t') dt']

i= l

W;

Jo

(12)

Sl111ilarly, the response of thejth nodal displacement caused by a dynamic load

, \ 1{(), corresponding to the kth nodal displacement, may be written

(0j)A1 -_
11 /\,
1 q1111tt

~ [<_l)Njl cpNki

L,
I

W/

lt (t _ ') d ']
Ak

Sill W;

(13)

/\ 4, the right hand sides of' Eqs. (12) and (13) are equal, and we may
llw kfl 111111d sides to ohtai 11

U>1h,

(/l1 ),,

(14)

154

Normal-Mode Method of Dynamic Analysis

Chap. 4

This relationship constitutes a reciprocal theorem.for dynamic Loads [3] that is

similar to Maxwell's reciprocal theorem for static loads [2]. It states that the
dynamic response of the kth nodal displacement due to any time-varying action
corresponding to the ith displacement is equal to the response of the ith disP-lacement due to the same action applied at the kth displacement. The theorem bolds
.for systems with rigid-body modes as well as vibrational modes, as can be seen
by using Eq. (9) in place of Eq. (7) in Eq. (11).

we obtain the vector of normal-mode displacements, as follows:

1.2114(1 DN =

COS W1 t)/ Wi]

- 0.1856(1-cosw2t) / w!
[

-0.7472(1 -

COS W3t)/w3

4.485(1 - cos w,t)

A= {O, Pi, O}

DN =

(b)

- 0.1408(1 -

P2pL2

1.000 - 1.257c,
D

= <l>NDN =

COS W3t)
+ 0.08341 c2 + 0. l 739c3]

AL

5.556 - 5.433c 1 - 0.01650c2 - 0.1052c3 EA

[

-1.333

(e)

+ 1.343c, + 0.09618c2 - 0.1062c3

where c 1 = cos w 1t, c2 = cos w 2t, and c3 = cos W3t . Inspection of these results shows
1hat the joints of the truss vibrate about the displaced positions:

-0.7472

P2L
D .. = {l.000, 5.556, -1.333} EA

where m = pAL. For this purpose the normalized modal matrix <l>N is available from Eq .
(4.2-e). From the result of the Duhamel integral for a step function, given as Eq. (2.6-8),

(d)

Transforming this solution back to the original coordinates using Eq. (4 .3-5), we find that

As required by Eq. (4), we transform vector A to normal coordinates with the operation:

AN= ((>NA=

(c)

JEVm

- 0.08889(1 - cos Wi t)
[

(a)

1.2114] P2
-0 .1856 Vm

2
Substitutionofwi = 0.2701E/ pL2,w~ = 2 .088E/ pL , andw5 = 5.308E/ pL2intoEq.
(c) yields the simpler form

Example 4.5
We shall consider again the plane truss of Examples 4.1, 4.2, and 4.3. Figure 4.4(a)
shows a step force of magnitude P = A applied in the x direction at joint 2. Let us
determine the response of the structure to this suddenly applied load, starting from rest.
The vector of applied actions for this case is

155

Normal-Mode Response to Applied Actions

Sec. 4.4

(f)

due to the force Pi applied statically.

Proceeding in a similar manner, we can also calculate the response of the truss to
11 step force of magnitude P = P, applied in they direction at joint 2, as indicated in Fig.
I.4(b). In this case, the results are

+ 0.3108c, - 0.4863c2 + O. l 755c3J

AL
- 1.333 + l.343c 1 + 0 .09618c2 - 0. l062c3 EA
0

P = P2

(g)

1.000 - 0.3321c, - 0.5607c2 - 0.1072c3

P.quation (g) shows that the joints of the truss vibrate about the displaced positions:

AL
Ds, = {O, - L333, 1.000} EA

(h)

due to applying the force P3 statically. If we let the step force P2 be equal to ?3, th~n the
11sponsc D 3 caused by P2 in Eq. (e) will be equal to the response D2 caused by P3 m Eq.
(11). This equality confirms the reciprocal theorem for dynamic loads in Eq. (14).
1'

ANsumc that the unrestrained flexural element in Example 4.4 is subjected to a ramp force

(b)

~ 0.6 l ~

/'

P,t/ t, applied in they direction at its center, as indicated in Fig. 4.5. We shall
111ll11late the response at the nodes due to this influence, beginning with the element at
11NI

(n)
ll ll(lll'l' " "

l'llllll'

IIIIN~ (11) Wllh Nh' Jl 11111 I'

mpl~4.6

,.,. (h)

wllh ~hp lmn ,.,

lJs111p I
I /'2 111 ll w displnl'l'llll'lll sh11pl' lunctions f from Eq. (3.4-17) and
npply, 1114 Fq ( I I )~) without inlrp,1111011 , w1 hntl th1 1quiv11knt nmlul actions to be

Normal-Mode Method of Dynamic Analysis

156

Chap.4

Sec. 4.5

157

Normal-Mode Response to Support Motions

Lt 3

ISL~ - ~ sin

90~ - ~ sin
3

ISL~ - ~ sin
3

- 90~ - ~ sin
3

W3t) / w5
W3~/

w~

(n)

w3t) / w5
W3t) /

w~
3

llere we see that the first rigid-body mode contributes translations equal to P1t / 6mt1 at
hoth nodes l and 2.
Figure 4.5 Flexural element with ramp force.
AP -- f T P --

4 , L, 4 , - L }Pit
8t1

(i)

which consists of a force and a moment at each end . Premultiplication of this vector with
the transpose of the normalized modal matrix from Eq. (4. 3-j) yields
T

AN = <l>NAP =

Ir

2, 0, -vs, 0

Pit
1
2t1 vm

(j )

wlwre 111 = pAL. The vector in Eq. (j) contains normal-mode loads of types 1 and 3 for
th is example. Now we integrate the first term with Eq. (9) to obtain

P1t 3
6t1Vm

(k)

DNI = - -

whil'h is a symmetric rigid-body motion [see Fig. 4.3(b)]. Similarly, evaluation of the
thinl term in Eq. (j) in accordance with Eq. (2.6-c) for a ramp function produces
DN3 =

P1VS

2 1
2t1W3V m

(t

-1
W3

.
sm

W 3t

ce)

4.5 NORMAL-MODE RESPONSE TO SUPPORT MOTIONS

We are often interested in the response of structures to support motions instead

of applied actions. In this section we discuss problems in which either rigid-body
ground displacements or accelerations are specified. In addition, independent
111otions of multiple restraints will be treated.
Figure 4.6 shows six possible displacement components Dg,, Dg2, . . . ,
nH 6 for a point g that is assumed to be a reference point on ground. The figure
1tlso depicts a typical joint (or node) j on a structure that is connected to ground.
The six possible displacement components at pointj are labeled Dj1, D j2 , . . . ,
n1c,. A location vector r8 j is directed from point g to point j and has scalar
,wnponents x 8 j, Y gj , and Zgj We may calculate the displacements at j due to
t lf,!,id-body displacements of the ground at point g using the concept of translotion of axes [2, 4]. For this purpose, the rotational components of the ground
tliNplacements must be small. Under this condition, the displacements at j in
ti, ms of those at g are
(1)

whldt is a sy mmetric vibrational response [see Fig. 4.3(d)]. The antisymmetric rigidhody and flexural modes 2 and 4 do not respond to the centrally-placed load. Altogether,

till' vector of normal-mode responses is

where

(2)
0

-3vs( -~
1

sin

(m)

(3)

w,,) / w5

0
1'1 1111Nlrn11111tlo11 o l thl'Nl' dlsplttL'l'lllL'lllN bnLk to phy11k-11t L'ttonll1111tL'N using llq . (4 .J ~)
fLIV!'N

1111d

I II Hq. ( I ) the transformalion matrix is

l,
0

(4)

Normal-Mode Method of Dynamic Analysis

168

Chap. 4

to,~
.to,,
/ o,6

tD
1~

o,~,I

This matrix equation has the same form as Eq. (2.3-11) and can be restated

MD + SD

--D,,

.:;\bSo \ ..t.Q..

Coo-rd,.,,..fe.i (9)
(10)

Thus, the vector A on the right-hand side of Eq. (9) contains equivalent nodal
8
actions due to rigid-body ground displacements .
Similarly, the accelerations at a typical joint j may be expressed in terms
of rigid-body ground accelerations at point g , as follows:
(11)

I
:

:~
-----------r---------,,---X

/
/

= A8

in which

D14

95

159

Normal-Mode Response to Support Motions

To include the effects of ground displacements in the action equations of

motion [see Eq. (4.2-12)), we write them as
(8)
M D + S (D - ..:1g) = 0

I .

Sec. 4.5

o.,

Dg3

1I

Dg6

--- - -- . __ _____ __J/

(12)

/:
/

where

//
91

llnd
(13)

X gj

Then the vector of accelerations

/
I

..:i8

at all free joints becomes

..

..:18

Figure 4.6 Rigid-body ground displacements.

(14)

T 8 D8

In order to use this vector in the equations of motion, we must change to the
111 which 13 is an identity matrix of order 3 and

cL =

-c8i

relative coordinates:

~i

= C18 = [-~, 1
Y g;

Xgj

-:.;J

D*
(5)

This skcw-sy~metric submatrix contains positive and negative values of the

l'0111poncnls of vector r81 These components are arranged in a manner that
p, od 11tcs Ihe cross product of a small rotation vector at g and the location vector
l'NJ Then the time-varying displacements ..:18 at all free joints may be written as

(6)

D - ..:18

D* = o - ..:i

(15)

In these expressions the symbot D* denotes a vector of displacements relative

lo the ground, and the vector D~. contains the corresponding relative acceler11tions. Substituting D - ..:18 and D from Eqs. (15) into Eq. (8) and rearranging,
wc find that
MD*

+ SD*

= A[

re.t"-t, ve

Cao'<~;

nt1.ti>{ l6)

111 which
(17)

Whl' H'

TT _
/I -

T
TI,1
[
T

112

,- , 1'

I NII/

1111d 111 I~

IIH

11111111 11'1

ol NIil ii 10111111

(7)

'l'lll'rcforc, the vector At on the right-hand side of Eq. (16) consists of equivalent
11od11I ac tions caused by ri gid-body ground accelerations. If we compare Eq. (16)
111 illl SDOF counl crpurt in Eq . (2.3- 18), we see that both are of the same form.
Afkr tlw l'q11 ivule11t nmlul actions AHor At have been found, the response
l llil'11l 11tions in nhsolull' m 1tl11l iVl' l'OOl'di1111ll'S proceed as described in Sec. 4.4
1111 11pplild mlillllH 11111 IIH' 1111111 l 11sr, lht 11h~11l11tt disph1l'l' 1Hl' nl s at free nodes

160

Normal-Mode Method of Dynamic Analysis

Chap.4

may be calculated from the first of Eqs. (15) as

D

= D* + 4g

(18)

Sec. 4.5

The rigid-body rotation 8z, centered at point 2, causes the followi ng step-translations at
point 1:

48

which is the sum of the relative displacements and the effects of the rigid-body
ground displacements.
Example 4.7
Figure 4. 7 shows the plane truss used previously in Example 3 .1. In this truss the
cross-sectional areas of members I and 2 are equal to 0.8A and A. Now let us calculate
the response of the structure to a small rigid-body step-rotation D86 = 8, about point 2.
From Example 3 .1, the stiffness and mass matrices for the free displacements at
point I are

S=s [

0.36
-0.48

-0.48]
1.64

(a)

w~ = 1.8..::.

= TJD8 =

A8 = S 4 8 = [

1 [3

8,

(c)

-0.288]
sL8,
0.384

(d)

which result from applying Eq. (10). From this point we may proceed to calculate the
response by the method shown in Sec. 4.4. Thus, Eq. (4.4-4) gives the normal-mode
loads as

A= <l>T A= [-1]
0.48sL8,
'\!'To,;;
3
N

<l>N= ~ 1

[-0~8L]

as given by Eq. (6). The 2 x 1 operator TJ in Eq. (c) contains only terms from the first
and second rows and the sixth column of the general operator in Eq . (4). Equivalent nodal
loads (forces) at joint 1 are

in which s = EA/Land m = 3.28 pAL/6 . We also know that

w, = 0.2-

161

Normal-Mode Response to Support Motions

(b)

(e)

Then the normal-mode responses to these step loads become

_[-(1 - cos w,t) / wf] 0.48sUJz

3(1 - W2t)/w~ '\!'To,;;

(f)

DN -

COS

These expressions may be simplified by substituting wf and w~ from Eqs. (b), as follows:

_[-3(1l -- cos
cos w,t) ] 0.8mL8,
'\!'To,;;

DN -

(g)

W2f

Transformation of this vector back to physical coordinates with Eq. (4.3-5) produces

_
_ [-10
D - <l>NDN -

+ 9c1 +

c2 ]0.8L(J,

3c1 - 3c2

-10

(h)

where c, = cos w 1 t and c 2 = cos w 2 t . Here we see that the truss vibrates about the
displaced position given by Eq. (c).
0.Bl

1, 11mple 4.8
4.8 . We
M
hnll determine the steady-state response of node 2 caused by a rigid-body rotational
ll('tClcration Dg6 =
sin Ot of ground at node 1.
Stiffness and mass matrices for node 2 are

'l'hc prismatic cantilever beam analyzed in Example 3.2 is shown again in Fig.

e,

S =
/

Dno l _
/

0.6l

s[

- 3L

-3L]
2L

M=m [

78

-lll]

- llL

2L2

(i)

whllll' .1
2/:'/ / 1,1 and m = pAL/2 l 0. Normalization of the modal matrix (from Examph 1,2) with respect lo the mass mulrix yields

(l,N

I
I.

Ill

0. 1,\()4/,

0. 1943/, l

O I ()21

I .181

(j)

Normal-Mode Method of Dynamic Analysis

162

Chap.4

Sec. 4.5

MFF MFR][~F] + [SFF SFRJ[DFJ

[ MRF MRR DR
SRF SRR DR

Dg

(3.5-17)

AR

produces
(3.5-18a)

<I

[ AF]

As in Sec. 3.5, the subscript F denotes free displacements , while the subscript
R pertains to restraint displacements. Writing these equations in two parts

JD,

CD

1~

163

Normal-Mode Response to Support Motions

and
(3.5-18b)

MRFDF + MRRDR + SRFDF + SRRDR = AR

Rearranging Eq. (3.5-18a) gives

Figure 4.8

Cantilever beam with rigid-body ground acceleration.

MFFDF + SFFDF = AF - SFRDR - MFR DR

In this form we can see that the terms

Due to the ground acceleration at node 1, the accelerations at node 2 are

. [L] 0,

"18 = T;'D8 =

AFR
(k)

sin fit

ns given by Eq. (14). In this case, the 2 x 1 operator T; in Eq. (k) has only terms from
rows 2 and 6 and column 6 of the general operator in Eq. (4). Equivalent nodal loads (a
force and a moment) at node 2 become

Ai

.. = - [ 67 ] mLO, sin fit

-M .1
8
-9L

(C)

* <l>NAg
T
*= AN=

7.613 ] L ,v
-0.3109

..
.
me,
sm

fit

(m)

F,om these actions we find the steady-state normal-mode responses to be

D,t

= -

7.613{3,/wf] . , .
[ - 0. 3109/32/ W22 L v m (}, sm fit

* -_ <I>N D N* -_

Example 4.9
We shall now reconsider the plane truss in Example 4.7 , which appears in Fig. 4.7. Let
the support point 3 have a sudden independent step translation DR 3 = din the x direction.
Then determine the response of the free displacements at joint 1.
By including terms for the restraint displacement at point 3, we can extend the
Nliffness matrix in Eqs. (a) to become

(},

0.36

SFR
s = [sRF sRJ = s [ ~~:::
SFF

Sill fit

(o)

--~:::

i-~:::

3.28 0 i lJ
M [M . MRR] m, [- ~1_ _ }_:_~l~
i2
=

When u stru cture has multiple connections to ground , it is also possible to

l'nku lalc the respon se to i11rl<'fJ<'tult11t 11101io11s t?f'support rl'.11rai11t.1 by generntinp
lltl' 11ppropri11tL Ntiff11n 1s 1111d 11111ss l'ocffkiL'llts I 1, 'i I, 111 such II t'llsc, thl' rcl11tiVl'
diMpli1cl'llll' llls 111 thl' ~11pp111 h 11111st Ill' s 11111II i11 111dl1 to Ht11i11 liiwur lwlwvlrn
I 1t II~ 11w1it1 1111 11rnl 111 11p1d l'q11111lo11~ ol 1111,111111 1111 1111 prn11, tl>l1 11od11l displ11n

-0.48 i -0.36J
(p)

Slil'fnesses for unmoving restraints are omitted from Eq. (p). Similarly, the mass matrix
11 0111 Eqs. (a) is extended to
MFF

llll" III N Ill 1111 11111111d 1111t l pill titlo111d 1111111 , 11~ 111111 ,w~

(20)

DE

(n)

)L] . .

[ (l.06 lb , - 0.0604lb2
l.46 lb1 - 0.4604b2

-SFR DR - MFR DR

on the right-hand side of Eq. (19) are equivalent nodal loads. They are caused
by the independent restrai1:.t displacements in the vector DR and their ~orresponding accelerations in DR .. After the displacements DE and the accelerations
are found b y normal-mode analysis , the reactions Ae at su1mort goints may
be obtained from Eg. (3.5-18b) if desired. Of course, .@Y..._restraints without
motions may be represented with zeros in vectors D,uind D.R (or omitted al-

llpon transforming this vector back to physical coordinates, we find the relative displacelltLlllls:

Logether).

whkh is dictated by Eq. (17). Then the normal-mode loads are

1

(19)

MFR

(q)

RI

,,M,/6. Thon l'rom Eq . (20), the equivalent nodal loads due to the step
dl~p(lll'l ' llll' lli J>R I
r/ lll l '
w hiil' 111

() \/, I
O IH

rl

0(),,tH
.361,I'(/

(r)

164

Normal-Mode Method of Dynamic Analysis

Chap. 4

Next, the normal-mode loads may be calculated as

T
AN= cf>NAFR
=

0.6] -sd- 1.8 V10m

(s)

From these actions, we find the normal-mode responses to be

t)]

- [
DN-

0.6(1 -cosw,t) / wT] -sd

- - [3(I - cosw 1
- 1.8(1 - cos W2t)/ w~ V10m - - (1 - cos w 2t)
Finally, the transformation back to physical coordinates yields
D

= cf>NDN =

[10 -- 9c,
J!!...
3c1 +
10
c2

3c2

md
(t)

V10m

Sec. 4.6

Damping i n MDOF Systems

This matrix contains damping coefficients that are defined as actions required for
unit velocities. That is, any term Cik in an array of viscous damping coefficients
is an action of type j equilibrating damping actions associated with a unit velocity
of type k. This definition is similar to those for stiffness and mass terms and
implies that the damping matrix is also symmetric.
To form the damping matrix, we consider first the systems for which this
urray is assumed to be linearly related to the mass and stiffness matrices. That
is , we take

C
(u)

In this case, joint 1 vibrates about the displaced position:

(v)

165

= aM + bS

(3)

where a and bare constants. The formula in Eq. (3), attributed to Rayleigh [3],
is called proportional damping because the matrix C is proportional to a .linear
combination of Sand M. In such a case the equations of motion [Eq. (1)] are
uncoupled by the same transformation as that for the undamped system. Thus,
in principal coordinates we have
(4)

obtained by static analysis.

where
(5)

4.6 DAMPING IN MDOF SYSTEMS

Damping in solids and structures is not understood as well as stiffness and mass
properties . Often the effects of damping upon the response of a vibratory system
cun be ignored, as has been done in Secs. 4.2 through 4.6. For example , the
iulluencc of a small amount of damping on the response of a structure during an
cxcilution of short duration is not likely to be significant. In addition, damping
plays a minor role in the steady-state response of a system to a periodic forcing
function when the frequency of the excitation is not near a resonance. However,
l'or u periodic function with a frequency at or near a natural frequency, damping
is of primary importance and must be taken into account. Because its effects arc
usually not known in advance, damping should ordinarily be included in a
vibrational analysis until its importance is ascertained.
When a discretized solid or structure is assumed to have viscous damping,
lhc uction equations of motion may be written as

MD+Ci>+SD = A

(' .

C2,
( ' 11

C,2

(6)

'l'hc diagonal matrix w in this expression contains the characteristic values w?

for the undamped case [see Eq. (4.2-17)]. Therefore, the ith equation of motion
in normal coordinates is
2

..

DN;

' + w;2DN;
(a+ bw;2)DNi

c,,,

C22

Cn
C21

C211

(' IJ

(' II

( '111

(i = l, 2, ... , n)

AN;

C Ni

n;

= 2n; = a + bwr

'}'; = -

111 these relationships the term CN; = 2n; is defined as the modal damping
,1111sta111 for the ith normal mode, and '}'; represents the corresponding modal
rlfl111pi11R ratio. Using the first of these definitions in Eq. (7), we obtain

= 1, 2,

11 1

( '11

"'

',,,,

. .. , n)

(9)

lmlt ol th1 11 l'q1111tio11 s 1l' iWl'Sl'lllt'd by lhis l"Xprcss ion is uncoupled from all of
1111' 111111'1 ~ 1'11l'11lo11, Wl' l'IIII d 1ti1111i111 lli1 ll'M iH>IINI' ol' Jill' /th IIH >(k in llw sum1..
0

(8)

W;

(i

(2)

(7)

To make this expression analogous to that for a SDOF system (see Chapter 2) ,
we introduce the notations

( I)

whkh upplics only to free nod~! displacements. The damping matrix C pre
111ultiplying lhe velocity vector Din Eq. (l) has the general form

c,,

'l'he symbol CP represents a diagonal array that will be referred to as a principal

damping matrix, and it consists of a linear combination of MP and Sp. When the
modal matrix is normalized with respect to M, the damping matrix in normal
rnordinates becomes

1111\lllll' I II N 111111 1111 II

SI)( II ~yH tt-111 Wllh Vl ~l 11111, d11111pi11p

166

Normal-Mode Method of Dynamic Analysis

Chap. 4

. From the definitions in Eqs. (8), we may express the modal damping ratio
Y1 m terms of the constants a and b, as follows:
Yi=

a + bw7
2wi

(10)

This. formula is useful for studying the effects upon the modal damping of
varymg the constants a and b in Eq. (3). For example, setting the constant a
equal to zero (while bis nonzero) implies that the damping matrix is proportional
to the stiffness matrix. This type of damping is sometimes refeJTed to as relative
damping because it is associated with relative velocities of displacement coordinates. Thus, under the condition that a = 0, Eq. (10) becomes

bwi
'Yi = 2

(11)

which means that the damping ratio in each principal mode is proportional to the
u~1da mped angular frequency of that mode. Therefore, the responses of the
hi gher modes of a system will be damped out more quickly than those of the
lower modes .
O n ~he othe~ h~nd, setting b equal to zero (while a is nonzero) implies that
!he cla~pmg matnx 1s proportional to the mass matrix. This type of damping is
Homct1mes called absolute damping because it is associated with absolute velocities of displacement coordinates. In this case Eq. (10) simplifies to
a
'Yi=-

2wi

(12)

.~o thul the damping ratio in each mode is inversely proportional to the undamped
1111gular frequency. Under this condition the lower modes of a system will be
suppressed more strongly than the higher modes.
.. .It has been shown by Caughy [6] that the criterion given by Eq. (3) is
s11 l llc 1c nt but not necessary for the existence of principal modes in damped
11ystems . .The ess~ntial condition is that the transformation which diagonalizes
t hl' dump111g matn x also uncouples the equations of motion . This criterion is less
H'Nltictivc tha n that in Eq. (3) and encompasses more possibilities.
llowcver, in the most general case, the damping coefficients in matrix ('
llll' suc h !hut the damping matrix canno t be diagonalized simultaneously with thl'
1111,ss and stiffness matrices. In this instance, the natural modes that do exist haw
phust re lar io~ships that complica~e the analysis. The eigenvalues for this type o l
Nysll'lll arc ei ther real and negative or complex w ith negative real parts. Tlw
l'otllpltx c igcnval.ues ~ccur as conj ugate pairs, and the correspondin g cigcn
Vl'l'lms uls.o to~1s1st of complex conju gate puirs. In hi ghly damped systems,
wh1:1l' lhl 1nwg 111111 y ll'1111s dm to di ssipnl iw f'otTL'S nn sig nifit-11111 , (he 111elhod
ol h>HNI / I 11u1y lw llSl'd 'l'hi~ 11ppn111l'11 l11v11lVl's l1 1111s l0111111li1111 of' llw II Sl'eond
111d11 l' <llllllll!II~ ol IIH!ltOll lillo >11 IIIH'llll)lhd ilt,1 111d11 1q1111IIOIIH

Sec. 4.6

167

Damping in MDOF Systems

Lightly damped structures need not be treated in such a complicated

manner, especially in view of the fact that the nature of damping in physical
systems is not well understood. The simplest approach consists of assuming that
the equations of motion are uncoupled by use of the modal matrix obtained for
the structure without damping. In other words, the matrix cl> is assumed to be
orthogonal with respect to not only M and S but also C, as follows:

ct>J C cl>; =

cl>; C cl>j = O

(i -:/= j)

(13)

This expression implies that any off-diagonal terms resulting from the operation
CP = cf>T C cl> are small and can be neglected. In addition, it is more convenient
to obtain experimentally (or to assume) the damping ratio y; for the natural
modes of vibration than to determine the damping coefficients in matrix C
directly. We can usually find the damping ratio y 1 for the first mode of vibration
by field testing a structure or by previous experience. As mentioned in Sec. 2.4,
the range of this constant for metal structures is approximately 0.01 to 0.05,
while that for reinforced concrete is about 0.05 to 0.10. With the value of Yi on
hand, we can extrapolate to other values of y; using the approximate formula:
'Yi

W)e1

= 'YI ( W;

(0.5

::5

e1

::5

(14)

0. 7)

This expression suppresses the higher modes in accordance with damping experiments [8], but not as severely as in Eq. (11) . Alternatively, we can simply
determine y 1 and then let 'Yi = y 1 for all other modes.
Now we rewrite Eq. (9) in terms of 'Yi as
(i

= 1, 2, ... ,

n)

(15)

where CN; = 2yiwi. In order that this equation may pertain to a lightly damped
structure, let us also specify that O ::5 Yi ::5 0.20 for all modes. The type of
damping associated with this set of assumptions is of great practical value, and
it will be referred to simply as modal damping. It should be remembered that this
concept is based on the normal coordinates for the undamped system and that
dumping ratios are specified in those coordinates.
When modal damping is assumed in the normal coordinates for a structure,
ii may also be of interest to determine the damping matrix C in the original (or
physical) coordinates. This array can be found by means of the reverse
trnnsformation
(16)
lustcacl of attempting to invert cl>N, ho wever, we use the relationship
1sec Eq. (4.2-27)) and rewrite Eq. (16) as

I>/ <l>iM

(17)

l'hii; f'o nu of' lhl' ll'll11Hlo111111lio11 is tspl'd11 ll y appropri ate when not all of the
11111d1, 1111 ind11dld 111 till' 1111ul y1. is (111od11 I t111t1l'1ttio11).

11111111 111

168

Normal-Mode Method of Dynamic Ana lysis

Chap. 4

Example 4.10

Sec. 4.7

Damped Response to Periodic Forcin g Functions

169

where

As an example of modal and physical damping, we shall reconsider the 3-DOF plane
truss analyzed in Examples 4 .1 and 4.2. For this case, let us assume that the structure
is made of steel and that the damping ratio for the first mode of vibration is y 1 = 0.02.
Prom Eq. (4.2-f) the three angular frequencies are
wi =

0.5197c1

w2

= I.445c1

W3 =

2 .304c1

(a)

1.445

,'2

= 0.02 ( 0.5197

'}'3

2.304 )
= 0.02 ( 0.5197

0 6

.
2n;DN;

2
W; DN; -

= 0.04887

DNi

a
(,....
= Pm;
2 JJi cos ~ Lt

O;)

1
(c)

{3; = Y[ l - fl/w;) 2] 2 + (2y;fl/ w;)2

and the phase angle O; is

Using these values in Eq. (16) along with <l>iv 1 from Eq. (4.2-h), we obtain

O; - tan

C = <l>ivTCN<l>,v 1

1, 2, ... , n)

(3)

(4)

The magnification factor {3; in this expression is

= 2(0.04887)(2.304)c1 = 0.2252c 1

= A YpE

W;

= 2(0.02)(0.5197)c1 = 0.02079c 1

0.06154

(i

Pmi cos flt

(b)

CN2 = 2(0.03694)(1.445)c1 = O. I068c1

CN3

(2)

in which p 111; is a constant. This equation has the same form as Eq . (2.4-23), so
we can take the damped steady-state response of the ith mode to be

= 0.03694

Then the normal-mode damping constants are

CN1

= {P1 , P2, A, ... , Pn}

In Eq. (1) the terms in Pact as scale factors on the function cos flt. Transformation of the action equations of motion to normal coordinates produces the
typical modal equation
..
DN;

where C1 = (VE/p)/L. Applying Eq. (14) with e1 = 0.6., we find that

0 6
)

-0.01270

- 0.01 270
0.02002

-0 .01662

0.01987

- 0.01662]
0.01987

(d)

0.07620

wllich is the symmetrical damping matrix in physical coordinates.

4 .7 DAMPED RESPONSE TO PERIODIC FORCING FUNCTIONS

- I[

2y;fl/ W;

1 _ (fl/ w;)2

(5)

(6)

Equations (4) , (5) , and (6) are drawn from Eqs . (2.4-31),.(2.4-32), and (2.4-33) ,
respectively. The response given by Eq. (4) may then be transformed back to the
original coordinates in the usual manner, using Eq. (4.3-5).
To determine the response of the mode having its angular frequency w;
closest to the impressed angular frequency, we need only use the modal column
<l>N; in the transformations to and from normal coordinates . That is, Eq. (4.4-4)
is specialized to
(7)
Pm;= <1>1;P
and the back-transformation in Eq. (4.3-5) becomes

As mentioned in Sec. 4.6, damping is of greatest importance when a periodic

1xcitalion has a frequency that is close to one of the natural frequencies of a
MDOf! system . In this section we consider the normal-mode approach for
l'llit'uluting steady-state responses of discretized structures to periodic forcing
l1111l'tions. Knowing the imposed frequency of such a function and the natural
lilqucncies or the system, we can obtain in a direct manner the steady-state
ll'Sponses of the modes having frequencies in the vicinity of the imposed frc
'(lll'llt'Y . Both simple harmonic and general periodic forcing functions will bl'
di.~l11 ssed, and modal damping will be assumed, as described in Sec . 4.6.
Ir II lightly damped structure is subjected to a set of actions that uru all
p1op111tio11ul to tlw siinpk harmonic fun<:tion <.'OS
tlw action vector A lllny
h1 WI iltl'II HS
,

n,.

!' ON

01

(Il

= <l>N;DN;

(8)

If desired, this process can be repeated for other modes with frequencies in the
vicinity of fl.
Now we shall consider a lightly damped structure subjected to a set of
actions that are all proportional to the general periodic functionf(t). In this case
the applied action vector A may be written as
A = F(t)

= Pf(t)

(9)

where the vector Pis given by Eq . (2). Proceeding as described in Sec. 2.5 , we
i:xpn:ss f(r) in the form of' a Fourier series, as follows:
,.

/'(I)

<In I ~~ ("1 rnNjUi I h1 sin jHI)

I I

( 10)

Normal-Mode Method of Dynamic Analysis

1'70

Chap.4

Sec. 4.7

which is drawn from Eq. (2.5-1) with i replaced by j. The coefficients a1, b1 , and
in Eq. (10) may be evaluated as indicated by Eqs. (2.5-3).
Transformation of the action equations of motion to normal coordinates
produces the typical modal equation

(i

where p,,.; is again the constant given by Eq. (7). From the solution of Prob.
2..5 5, we talce the damped steady-state response of the ith mode to be
DN; =

Pm;{ao +
w,

{3u[a1 cos (}flt - Ou) + b1 sin (}flt -

j= l

in which the magnification factor

8;J]1

- VpiL

The magnification factor for the first mode is obtained from Eq. (5) as

~1=--;========:=~=====(~=:=:='
0 .25 )
0.25 7)
4

(12)

( l - 0 .2701

f3u = Y[l - (jfl/w;)2] + (2y;Jfl/w;)2

DNI =

(13)

und lhe phase angle Ou is

=
- I[

(d)

11.94

+ (0.0 ) 0.2701

From Eq. (4) we find the damped steady-state response of the first mode to be
2

~J
(c)

f3u is
1

eij - tan

vk[0.2803 1.2114 -0. 2995{

l.2114Pi

(11)

1, 2, ... , n)

p., - "'" p

110

171

Damped Response to Periodic Forcing Functions

2y;jfl/ W;

l _ (}fl/ w;)2

(14)

Because a multiplicity of terms contribute to the ith mode in Eq. (12), the
possibility of resonance (JD, ""' w;) is much greater for a general periodic func1i1111 than for a simple harmonic function . Therefore, it becomes more difficult
lo predict in advance which of the natural modes will be strongly affected.
l lt1wl~vcr, after the forcing function has been expressed as a Fourier series, each
of the JO frequencies can be compared with thew; frequencies for the purpose
ol p1'ldicting large-amplitude forced vibrations.

2114
1.
~~(11.94) sin (flt - 81)
0.2701 V pAL
2
53 .55P2L p . CA
8)

E\,lpAf,

where
81

tan

Sill ut -

(e)

- 1

(0.04)(0.5) ]
(0.5197)
= 270 21'
_
0 25
l - 0.2701

(f)

as given by Eq. (6). Transfonning the response of the first mode back to the original
coordinates with Eq. (8), we obtain
D = <l>NIDNI = {15.01, 64.87, -16.04}~~ sin (flt - (Ji)

I~ nnmk 4.1 t

(g)

Proceeding in a similar manner, we can determine the response contributed by the

I t'1 tlw plane truss in Fig. 4.l(a) be subjected to a simple harmonic forcing function
/', Ni ll Ht, applied in the x direction at joint 2 (corresponding to displacement D 2). The
11111111l11r frequency of the forcing function is n = (0.5/L)VE/p. Calculate the steadyHinh ll'Nponse of the structure, assuming that the modal damping ratios are y 1 = 0.02,
Yi 0.0.15, und 'Y1 = 0.05 (see Example 4.10).
Till' squurc of the impressed angular frequency is

n2

= 0.25
L2p

/,,

,,

(h)

D = {-0.09478, 0 .01875, -0.1093}~~ sin (flt - 82)

1111d that due to the third mode is found to be

(i)

D = {- 0.1824, 0.1104, 0 .1114}~~ sin (flt - 83)

(a)

The amplitudes in both of these vectors are small compared to those i~ ~q. (g). Further111ore, the innucnce of damping is significant for the first mode but neghg1ble for the other

'l'hlN v11h1l' is t'nirly close to the square of the first angular frequency
0 .270 IH

second mode as

IWO.

(h)

11H 1'iw11111 l'11hlt I I 'I h1Hlo11, w1 Nho11ld 1xp1111h1 tl1 NI 11111111 ol 1hl' s1n1vtl11l' 10.lw
1111 1111111111 y lCllllllh111111 hi !Ill' 11~pn11Nl' IJ~111p Jlq (/) ,WI d1,1r1111h11 lh1 lliN1 IIOlll\111
llllllh lo11d NI 1111 1111 llll In h1

11:,mm1>lt 4.12
lli\tlll' 'I .') Mhows 11 1wriodic forci11A 1'1111l'tio11 in lhe sh'.1pc..of' a square wave. lf this
lluatlmi ls 11 ppllid 111 llw I tlh111io11111 jnl111 I ol 1111' IIIINN 11111~. 4. l(a), find the damped
Hh'illl y Hlilh' ll'H Jlllll/'11' 1111 l 'lll hot lh1 IICllllllll 111111hH
0

172

Normal-Mode Method of Dynamic Analysis

Chap.4

2'!T

'lT

- P,

3'!T

TI

4'lT

D Ni

t--

Expanding the square wave as a Fourier series (see Prob. 2.5-1), we obtain
4 1
p (sin fit+~3 sin 3fit
7T

+ . . .)

(j)

Trunsformation of the load vector to normal coordinates produces

AN

= <I>J.; A = <l>J,;

F(t)]
0
[

[cf>N11]
<I>Nl2 F(t)

(k)

<l>m 3

111 which the load corresponds to displacement D 1 . In accordance with Eq. (12) the
nw mul-modc responses become
'

3
<I>m1~11 sin (fit - 811) + /3~ sin (30t - 8 13) + .. -}/ wf
l)N = 4P1
m 2{/321 sin (fit - 821) + /323 sin (3fit - 823 ) + . . .} /w~
7T
'l'N1
3
<I>N13~ 31 sin (fit - 831) +
Wlil'tC

13;3 sin (3fit -

833 ) + . . -} /

-11/(v
I

NOi

cos

Wd; t

+ DNOi +

n;DNOi

.
sm

WJ;t

(1)

Wdi

which is drawn from Eq. (2.4-6). The angular frequency of damped vibration in
l~q. ( 1) is
(2)
WJ; = V wJ - nT = w; V 1 - YT

Figure 4.9 Periodic forcing function.

F(t) = Pif(t) =

173

Damped Response to Arbitrary Forcing Functions

nre relatively high ( y; > 0.05) , the effects of damping could have some conNc4uence. Therefore, we shall now modify the formulations in Secs. 4.3, 4.4 ,
1111d 4.5 to account for the influence of damping on vibrational response in
normal coordinates. Action equations and modal damping will be assumed
th roughout the discussion.
In Sec. 4.3 we formulated the normal-mode responses of a MDOF struc111re to initial conditions of displacement and velocity at time t = 0. In the
presence of damping, the free-vibrational response of the ith mode, given by Eq.
(1. 3-4), must be changed to

p1

Soc. 4.8

wl

_the magnification factors and phase angles are given by Eqs. (13) and ( 14),

I l'N IWt"t1vc ly,

4 .8 DAMPED RESPONSE TO ARBITRARY FORCING FUNCTIONS

1>11111ping sl_1ou_l~ be included in transient response calculation s whenever ii

111l1,1l~t Ill' s1gn1licnnt. For tx111nplc, if the durntinn or u forc ing function is
ll' l11t1v1ly 10111 to111p1111.d to illl 1111turnl pt1'iods ol II st111 ctun, dninpiug rn\ild lw
1111po1 l1111t. Also , wlll'11 1111 t11111 of i11t11tsl ii.. i..hrn I h111 till' 111rnlnl d11111pi11g ,nti wi

111 which w; is the undamped angular frequency. Transformation of the initialvectors D 0 and D0 to normal coordinates remains the same as in Eqs.
(U -2), and back-transformation of the response is still given by Eq. (4 .3-5).
Similarly, the calculation of normal-mode responses to applied actions, as
d1scribed in Sec . 4.4, requires only a few modifications associated with modal
dumping. Transformation of applied actions to normal coordinates is the same
11N in Eq. (4.4-4), but Duhamel's integral in Eq. (4.4-7) must now be written as

1 lllldition

DN;

e - n;r
=WJ;

lt e '

tH'

'

AN;

sm

WJ;(t -

t ) dt

(3)

which is taken from Eq. (2.6-4).

Normal-mode responses to support motions, covered in Sec. 4.5 , may also
It, ultered to include the effects of modal damping. For rigid-body ground
,111l'lcr~tions, there is neither displacement coupling nor velocity coupling be1w11n the masses and the ground in relative coordinates. There exists only
111111tiul coupling with ground , which is the same as that for the structure without
il11111ping. To determine nodal responses relative to the ground, we first calculate
th1 1quivalent nodal actions in the vector At, as given by Eq. (4.5-17). Transli11111111ion of these actions to normal coordinates yields the equivalent modal
11111\ lh
(4)
1 luu the relative response D~1 in each normal coordinate is obtained from Eq.
1 \ J. wi th Ai 1 replacing ;\ Nf, Finally , these displacements are back-transformed
11 l11p
(5)

174

Normal-Mode Method of Dynamic Analysis

Chap.4

Sec. 4.9

175

Step-by-Step Response Calculations

(e)

which_ gives the relative responses in physical coordinates. As before, the absolute d1splace~ents at free nodes may be found with Eq. (4.5- 18).
.
In ~ertam cases where support displacements are specified, there is velocco_uplmg bet~een fre~ displ~~ement coordinates and support restraints. This
s1tu~t1on can ar1~e for e1th~r ng1d-body ground displacements or independent
mot1~ns o~ multiple restraints. Methods for handling such circumstances are
described m Ref. 1.

4.9 STEP-BY-STEP RESPONSE CALCULATIONS

lt:xnmple 4.13

111 Sec . 2.7 we examined step-by-step solutions for SDOF structures, where the

We s~all now repeat the first part of Example 4.5 , including the effects of modal
dt11pmg. Rec~II that ~e three-member truss in Fig. 4.4(a) is subjected to a step force
l i, correspondmg to d isplacement D 2
Symbolically transforming the vector of applied actions to normal coordinates
pwduces

lorcing functions are not necessarily analytical expressions. The basic approach
111 that section was to approximate the forcing function (or data points) using
1>iecewise-linear interpolation and then to use the Duhamel integral within small
lime steps. We shall now incorporate this technique into the normal-mode
111cthod for calculating transient responses of MDOF structures. As in the pre1cding sections, modal damping will be assumed throughout. Because of the
1xtensive calculations required, it is implied that the method of this section is to
he program.med for a digital computer. Such a program is described in Sec. 4.10,
where numerical examples are also presented.
Let us consider again the piecewise-linear type of interpolation illustrated
hy Fig. 2.18. Without loss of generality, only one such forcing functionfe(At1)
will be handled at a time, and the piecewise-linear action vector Ae, (or M;) may
Ill expressed as

AN=

~U =

(a)

~ { ; ] = [ :: }

I >uc lo the step function, the damped normal-mode responses are

9}/

<l>N21{1 - e-"ir(cos Wd1 t + ;~ sin wd1

1
DN

= P2 <l>N2+ -

e- "2 ' (cos wd2t

<l>N2+ -

e- " '(cos W113t

+ ;: sin wd2t)} /

9}/

+ ;: sin wd3

wf
w1

where i = 1, 2, 3. The expression in Eq. (e) is a modified version of Eq. (2.7-2c).

(b)

w5

'lhlsc expressions are drawn from Eq. (2.7-2b).

11:x11m1>tc 4.14
S11.p1iosc l!lat t~e gr:~und in Fig. 4.4(a) accelerates in they direction in accordance with
lhl 111111p_fun ct1on D82 = a2t/t2. Formulate the damped responses of the normal modes
Nl111t111g from rest.
'
this problem we work in relative coordinates, where the equivalent nodal toad

1:m

(j + 1 = 1, 2, ... , n1)

where At represents a small but finite time step, and n1 is the number of steps.
1
III this form the values of P act as scale factors on the common function f e(At1).
11 more than one such function is applied simultaneously, the responses for each
111 them handled separately can be superimposed.
Transformation of the action equations of motion to normal coordinates
pt oduces the typical modal equation
t'
(2)
DN, + 2n, DN, + w,2 DNI = ANI,) + M NI ,).At
J

Vl'l'IOI IS

(i=l,2, ... ,n

Mm~:,- mM<:'
IIN Jt lVl'n

(<:)

(1)

and

+ 1 = I, 2, ... , n1)

wh11e t' = t - t . The symbol AN;,J in Eq. (2) represents the ith normal-mode
1
11111d at time t1. Thus,
(3)

hy Eq . (4.5- 17). Then the cqu ivulcnt nonnnl mode loads arc
//, I

/1111

I
hll 1(11 lllllljl ltllll llllll, 1(11 d11lll(II

ii

(d)

111 11dditio11, we have the change in the ith modal load during the time step At1,
,11 li111d 11s follow s:
(4)

II

1ltlllli11' llitllh II ~(11111~1 I, (111 lllllt

wl1111

lh1

11y111l111I

A,,, 111 dl11111111

ihl' m ltnll

111

111111 1111

176

Normal-Mode Method of Dynamic Analysis

Chap. 4

In a manner similar to that in Sec. 2. 7, we can express the response of the

ith mode at the end of the jth time step as the sum of three parts, which are
= (DNI

D Ni,j+I

DN2

+ DNJ)i,j+ I

(5)

The first part of the response consists of the ith free-vibrational motion caused
by the conditions of displacement and velocity at time tj (the beginning of the
interval). Therefore , we have
(D NI) i,j+ I =

e -ntJ.t(D
I
Ni ,j cos
I

Wd;utj

+ DNi,j +

n;DNi,j

.
sm

Wd;Utj

Wd;

(6a)

= -ANi,j[
2- 1
W;

- e - ntJ.t(
' 1 cos

wd;Atj

.
+ -n; sm

wd;Atj)

Wd;

whkh is drawn from Eq. (2.7-Sc).

We can also write the velocity of the ith mode at the end of the }th tinw
Ntlp in three parts, as follows:
(DNI

D N2

(7)

DNJ)i,j + I

'l'lu:sc lhrec contributions may be formed by extending the notation in Eqs.

(J. .7 6) to obtain
(/) NI ) /,/1 I

= e

ntJ.t[
1 1

(v

Ni,jWd;

+ n;l)Ni,j +Wdin;DNi,j) sm
.

Wd;Utj

(811)

DN;,J

cos w,11 AtJ

(8h)

AA.Nl,/1I
'"/ At1

,.

111 11

t 1(cos

w,11 At1 I

111

'",11

DNi,j+I = DNi,j

/)Ni,j Atj

1(ANi,j

1AAN;)(AtY

(9)

DNi,j+ 1

'
= DNi,j
+

(ANi,J

A
2I AANi,j ) ut1

(10)

11.quations (9) and (10) pertain to rigid-body motions with no absolute damping.

4.10 PROGRAM NOMO FOR NORMAL-MODE RESPONSE

(6c)

pmvide the initial conditions of displacement and velocity at the beginning of

Htep j + 1. These expressions may be applied repetitively to obtain the time
history of response for each of the normal modes. Then the results for each time
11tution are transformed back to the original coordinates in the usual manner.
If the ith mode of a structure is a rigid-body motion, appropriate expres~ions for rigid-body response must be used instead of the recurrence formulas
~iven above. That is , the displacements in Eqs. (5) and (6) are replaced by

(6b)

which is taken from Eq. (2. 7-5b); and that associated with the triangular portion
becomes

l)Ni,j+ l

177

Program NOMO for Normal-Mode Response

1111d the velocities in Eqs. (7) and (8) are supplanted by

This fo rmula represents an extension of Eq. (2.7-5a).

The other two parts of the response in Eq. (5) are due to the linear forcing
function within the time step. The rectangular portion of this impulse yields
(D N2 ) i,j+ 1

'foe. 4.10

sin w,11 6.1)

(He)

l\q1111tlo11s (.,) 1h11111)'h (H) trn1 stil11lt ll'l'lll ll'lll't' 111111111111.~ lon11kul11ti~11 th1
d111111wd ll'N p1111 ~1 ol 111d11111111111l 11111tl1 1111h1 111d 1111111 /lh lii11t sltp 'l'lwy 11bo

l'hc normal-mode method for calculating dynamic responses of structures will

110w be applied in a program named NOMO. This program can be used to
1111nlyze any type of linearly elastic framed structure or discretized continuum.
l'he main program for NOMO calls six subprograms, as shown by the double
boxes in Flowchart 4 .1. The first subprogram appearing in the flow chart is VIB,
which is the program for vibrational analysis described previously in Sec. 3.~.
I lerc it is treated as a subprogram that calls the seven other subprograms m
llowchart 3.1. However, as the last step in Subprogram RESl, the eigenvectors
111 c normalized with respect to the mass matrix. Then the number of loadin_g
111,1tems NLS is read, the loading number LN is initialized to zero, and LN 1s
liK-rcmented by one. Next, Subprogram DYLO reads and writes dynamic load
d11(a, and the output includes the loading number LN as well as the n~mber of
luutling systems NLS. This is followed by Subprogram TRANOR, which tran~lmms initial conditions and actual or equivalent nodal loads to normal coord11111h.:s, using Eqs. (4.3-2) and (4.4-4). At its beginning , TRA~OR reads and
w1itcs the number of modes NMODES for the purpose of usmg modal trun1111 ion.
Subprogram TIHIST calculates time histories of normal-mode displace1m111ts and velocities with the step-by-step method described in .sec. 4.9. Be11111se uniform time steps are to be used, the coefficients of DN;,J , DN;,J, ANi,j, and
,\ANt,J in Eqs. (4.9-5) through (4.9-10) becom~ consta~ts t~at need ~e deter11il1ll'd only once at the beginning of the analysis. For v1brat10nal motions, the
11Nponscs in Eqs . (4.9-5) through (4 .9 -8) can be written in eight parts , as
lt1llows:
n N1.i1 1
/,NI/II

nNI.,

t (' , 1>N1. 1 t c,A.N1. 1 -1 C4 AAN1 .1

(I )

( \ /)NI,/

I ( ',/>NII I (' 1A.N/,f I ('K/J.A.N/,/

(2)

(',

178

Nurmnl Moctn M111h111l 11l l>y1111111h Annlyala

179

Pro,1rem NOMO for Normal-Mode Response

I h1 l'Ollstunt coefficients C 1 through C8 appearing in Eqs. (1) and (2) have the
iii ll111tions
Flowchart 4.1 Main program for NOMO

n~(

C1 = e ' cos

IOI

l. Prngram VIB from Sec. 3.8,

plus nom1aliza1ion of eigenvcciors
with respcc1 to mass mairix.

C2

A +-~
wd;ut

= - 1- e -n;t!.t sin wd;!:J.t

Wd;

sm

A~
wd;ut

Wd;

C3

= -(1 - C1)

Wf

(3)

Read number of loading

systems.

Initialize loading number to

zero.

Increment loading number by

one.

2. Read and write dynamic

load data.

3. Transfomi initial conditions and

loads to nonnal coordinaics.

4. Calculate lime histories of

nonnal-modc responses.

5. Transfomi displacements back

10 physical coordinates.

6. Wrilc ancVor plot rcsulls of

response calculations.

Check for las1 loading sys1c111.

Go 10

101

,truclurc
lND

Applies lO uny type of lincnrly ch1Ntlc

Nlluctu11,

IS11hp1ow11111~111111 dllkr for cvc1y ty1w ol Hl1m1111,

and rrocc,, 111101hcr

1111 Nl' coefficients, as well as simpler constants for rigid-body motions [see Eqs.
and (4.9-10)], are coded in Subprogram TIHIST.
After the response calculations have been completed, the normal-mode
1ll~pl11cements are transformed back to physical coordinates by Subprogram
I 11/\BAC, using Eq. (4.3-5) . The last subprogram, named RES2, optionally
'" 111s and/or plots resulting time histories of nodal displacements and axial
l111ns in members. After writing, the maximum and minimum values of these
11111111tities and the times of occurrence are written as well. At the end of the
lh1wchart the test of LN against NLS determines whether to return for another
h111dl11g system or another structure.
The logic in Program NOMO implies that the time histories of nodal
1IINplacements are stored in a matrix of size NDF x NTS, where NDF is the
1111111bcr of degrees of freedom and NTS is the number of time steps. Although
1111, procedure is conducive to plotting and calculation of internal actions or
l11sses, the use of such a large block of storage is not efficient. We could
111111sfcr blocks of these displacements to auxiliary storage if desired. However,
lhl~ upproach would require more data, more intricate logic, and more detailed
1 ,pl11nutions.
Program NOMO may be specialized to become NOMOCB for continuous
11111111s, NOMOPT for plane trusses, and so on. The main program for each
p1'l' inlization has certain subprograms that are different for each type of struc11111, us indicated by the second footnote below Flowchart 4.1. That is, the
~11hp1ogram named DYLO becomes DYLOCB for continuous beams, DYLOPT
1111 plune trusses, and so on. As for Program VIB, notation for Program NOMO
,1ppt11rn in Part 5 of the list of notation near the back of the book. Also, the
lh1wch11rt for Program DYNAPT in Appendix C contains detailed steps for the
1,,.,1c in the subprogrums of Program NOMO.
Tnhk 4 . I shows prcpurution of d1111m11i< load data for plane trusses. In the
ll1 11t lt1w of' thl' tuhll 111c thc n11111h11 lll 1i111t st1ps NTS, the duration of the
111111111111 liit1l' sll'P In: und llw d11111p111p 1ut111 I >AMPR. pcrtaining to nll modes.
1 I 1l lJ)

Normal-Mode Method of Dynamic Ana lysis

180
TABLE 4.1

Program NOMO for Normal-Mode Response

181

Dynamic Load Data for Plane Trusses

Type of Data
Dynamic parameters
Initial
(a)
(b)
(c)

Sec. 4.10

conditions
Condition parameters
Displacements
Velocities

Applied actions
(a) Load parameters
(b) Nodal loads
(c) Line loads
Ground accelerations
(a) Acceleration parameter
(b) Acceleration factors
Forcing function
(a) Function parameter
(b) Function ordinates

No. of
Lines

Items on Data Lines

NTS, DT, DAMPR

NNID, NNIV
1, D0(2J-l) , D0(2J)
J , V0(2J-l), V0(2J)

NNID
NNIV

NLN
NEL

NLN,NEL
J, AS(2J-l ), AS(2J)
I, BL!, BL2, BL3, BIA

1
1

IGA
GAX,GAY

NFO
K, T(K), FO(K)

NFO

"Omit when IGA = 0.

Nl'Xt, the initial-condition parameters NNID and NNIV give the number ol
nodes with initial displacements and initial velocities, respectively. Each line ol
llw data for initial displacements (NNID lines total) contains a node number J,
the I component of displacement D0(2J-l), and they-component D0(2J). Sim
ilul'ly, the NNIV lines of data for initial velocities indicate terms that are anal
opous lo those for initial displacements.
Applied actions and ground accelerations carry appropriate dimensions
11ml me to be multiplied by a dimensionless forcing function given at the end ol
tlw l11blc. In the data for applied actions, the load parameters are the number ol
lo11dL'd nodes NLN and the number of elements with line loads NEL. Data fo1
11od11I loads (NLN lines) consist of a node number J and scale factors for llw
1 rnmponcnl of force AS(2J-1) and they-component AS(2J). Figure 4.10(11)
dlpit1s lhcse components of the applied force acting at node j of a plane truss.
Also, Fig. 4.1 O(b) shows linearly varying line loads (force per unit length)
upplied in the x and y directions along the length of a typical plane-truss elemt'lll
1. 1>11111 for this condition of loading appear as part (c) under applied actions,
wht1l' L't1c:h of NEL lines contains an clement number I and scale factors for lh1
load i11tt11sitics 13LI throuh BI A shown in Fig . 4. IO(h) . ll can easil y he show11
lhul 1h1 tq11i v11k11l 11od11I loud~ i11 slruclunil dinctiom1 al joints j Hnd k 111'L'

(b)

Figure 4.10 Loads on plane truss: (a) nodal loads; (b) line loads on element.

AS(Jl) = (2BL1 + BL3)L/6

AS(J2) = (2BL2 + BL4)L/6
AS(Kl)

= (BLl +

AS(K2)

= (BL2 + 2BL4)L/6

(4)

2BL3)L/6

where subscripts JI through K2 are obtained from Eqs . (3.5-30). Of course, an

infinity of other load sets could be applied to the element and their equivalent
uodu I londs dcri vcd .

182

Normal-Mode Method of Dynamic Analysis

Chap.4

Data for ground accelerations consist of two lines, the first of which gives
11 parameter IGA for ground accelerations. If this number is nonzero, ground
uccelerations exist; otherwise, a zero indicates nonexistence. On the second line
arc scale factors GAX and GAY for ground accelerations in the x and y directions. The possibility of rotational ground acceleration is omitted from this
program.
For either applied actions or ground accelerations, the load data must
include information defining a dimensionless piecewise-linear forcing function.
Such a function is given as the last block of data in Table 4.1. On line (a) we
have the number of function ordinates NFO. Ordinates of the forcing function
11rc given in NFO lines, each of which contains a subscript K, the time T(K)
when the function ordinate occurs, and the value of the function ordinate FO(K)
al that time. For simplicity, we restrict the time T(K) to be equal to an even
number of time steps DT. If the forcing function has a discontinuity at time T(K),
lwo lines are required to define FO(K) on both sides of the discontinuity (for the
sume time). Note again that the function ordinates will receive dimensions only
when they are multiplied within the program by either applied-action or ground11ecclcration scale factors, thus creating time-varying proportional loads.
ln all computer programs for dynamic analysis we write and plot seleclivcly to limit the volume of output. Table 4.2 shows our method for selecting
nodal displacements and element stresses for writing and/or plotting in Subpmgrnm RES2. Line (a) of the table contains output parameters that have the
lollowing meanings:
IWR
IPL
NNO

= indicator for writing (0 or 1)

= indicator for plotting* (0 or 1)
= number of nodes for output

Output Selection
(11) Output pnrnmctcrs
(h) Nodnl displnccmcnts

(l') l\ltlll('III Ntll'SN('N

we calculate and write only the axial force at the k end, which has the
M
11me sign as the axial force in an unloaded member. On the other hand, with
lwo- and three-dimensional finite elements, the stresses are calculated at sampling points for numerical integration (see Chapters 7 and 8). If desired, ~e
,ould specialize the variables JNO( ) and IEO( ) to have a second subscnpt
cli,noting a particular type of nodal displacement or element stress.
l~xnmple 4.15
l'o illustrate using Program NOMOPT, we shall analyze the three-member pla~e truss
IINCd previously in Example 3. 4. Dynamic responses of this structure will be obtamed for
the following influences:
I. Initial displacements of 0.1 in. at all nodal degrees of freedom, with
DAMPR = 0.0
2. Piecewise-linear force in x direction at node 2, with DAMPR = 0.0
'- Same as case 1, but with DAMPR = 0.1
4. Same as case 2, but with DAMPR = 0.1

11, 11 11 cases we take the values of E, p, L, and A as given in Example 3.4, where the
11111tcrial is steel and US units apply.
Table 4.3 contains a partial listing of the output from Program NOMOPT for the
111111, analyses considered. With the formats coded in the program, this output becomes
TABLE 4.3

Computer Output for Example 4.15

STRUCTURAL PARAMETERS
NN
NE NRN
E
3
3
2 3.0000E+04
NODAL COORDINATES
X
NODE
0.000
1
150.000
2
150.000
3

RHO
7.3500E-07
y

0.000
200.000
0.000

ELEMENT INFORMATION
AX
J
K
ELEM.
10.0000
1
2
1
6 . 0000
1
3
2
8 . 0000
2
3
3

Output Selection Data

No. of
Lines

11 usses

PROGRAM NOMOPT

lt1 line (b) the node numbers JNO( ) for output of displacements are listed, and
tkmcnt numbers IEO( ) for output of stresses appear in line (c) of the table. J-101
l 111111l'd struclures, this output usually consists of generalized (or integrated)
Nlttsses al the j and k ends of members (see Chapter 6). However, for plant

Type of' Data

Items on Data Lines

NNO

IWR, IPL, NNO, NEO

JNO(l), JN0(2) , .. , , JNO(NNO)

NEO

1110 ( I ) , 11 \0(2) , , , , , IEO(NEO)

+1111111m plilllillp 111p11hlll1 y v11d1N

111 1111 I 111. II "' . ,,,"

lllllllllll, Pllll'IIIIII II Nl' I " ,

11!1 p1111um11, 1 11'1.

11111y

ex
CY
0.6000 0.8000
1.0000 0.0000
0. 0000 -1. 0000

EL
250.0000
150.0000
200.0000

NODAL RESTRAINTS
NODE NRl NR 2
1
3

0
l

l
l

NlJMl\lOH 01r PMOH!ilVlS 0 1r J<'Rl'l llJDOM I

NUMIIMII OJI NOO/\L lll9!l'rltl\ I N'l'fl I
I q1111I 111 11111

183

Program NOMO for Normal-Mode Response

*** EXAMPLE 4.15: THREE-MEMBER PLANE TRUSS***

NEO = number of elements for output

TABLE 4.2

Soc. 4.10

ht "'''
111r1rN1r.OI

Ml\'1'111 X IJl1t'OMl'O dr.11

NOF
NNH

3
3

184

Normal-Mode Method of Dynamic Analysis

TABLE 4.3

(Continued)

MODE
1
ANGULAR FREQUENCY 4.1995E+02
NODE
DJl
DJ2
1 2 . 3137E-Ol O.OOOOE+OO
2 l.OOOOE+OO -2 .4 722E-Ol
3 O.OOOOE+OO O.OO OOE+OO
MODE
2
ANGULAR FREQUENCY
NODE
DJl
1 8, 6725E-Ol
2 -l.7149E-Ol
3 O.OOOOE+OO

1.1677E+03
DJ2
O.OOOOE+OO
l.OOOOE+OO
O.OOOOE+OO

MODE
3
ANGULAR FREQUENCY l.86i8E+03
NODE
DJl
DJ2
1 l.OOOOE+OO O.OOOOE+OO
2 -6.0504E-Ol -6 .10 68E-Ol
3 O.OOOOE+OO O.OOOOE+OO

*** LOADING NUMBER 1 OF 4 ***

DYNAMIC PARAMETERS
NTS
OT
DAMPR
20 l.OOOOE-03 O.OOOOE+OO
INITIAL CONDITIONS
NNID NNIV
2
0
INITIAL DISPLACEMENTS
NODE
001
D02
1 l.OOOOE-01 O. OOOOE+OO
2 l.OOOOE-01 l .OOOOE-01
APPLIED ACTIONS
NLN NEL
0
0
GROUND ACCELERATIONS
IGA
0
NORMAL MODE SOLUTION
NMODES
3
OUTPUT SELECTION
IWR IPL NNO NEO
1
1
2
1
NODES:

ELEMENTS:

DISPLACEMENT TIME HISTORY FOR NODE

1
STEP
TIMID
DJ1
DJ2
0 O.OOOOE+OO l.000010 01 o.oooomoo
1 1 . OOOOIC 03 6 , 70b lM Oil O.OOOOJlltOO
1. 1,000CJM O'I !, 'J. ., ltil~ () i
CJ, 00QCJJ~ I QQ
I
I, OCJO(IIC !l I
1.uo11m 01 () (l{l(l()lq I ()(1
4 4, OOOOIC O I I !\Ii 1111f1l (I ,I o. 11110111q, no

185

Program NOMO for Normal-Mode Response

TABLE 4.3

5
6
7
8
9
10

5.0000E-03
6.0000E-03
7 . 0000E-03
8 . 0000E-03
9 . 0000E-03
1. OOOOE-02
l,lOOOE-02
l.2000E-02
l.3000E-02
l.4000E-02
l.SOOOE-02
l.6000E-02
l.7000E-02
l.SOOOE-02
1. 9000E-02
2.0000E-02

(Continued)

9.7853E-02
4.9458E-02
-7 . 4544E-02
-l.0603E-Ol
-5.4252E-02
2.7924E-02
9.2962E-02
4.0849E-02
-8.1017E-02
-6 . 9040E-02
6.8279E-02
l.1979E-Ol
4 . 7138E-02
-3 . 6778E-02
-8.3692E-02
-5 , 2695E-02

O. OOOOE+OO
O. OOOOE+OO
O.OOOOE+OO
O. OOOOE+OO
O.OOOOE+OO
O. OOOOE+OO
O. OOOOE+OO
O.OOOOE+OO
O. OOOOE+OO
O. OOOOE+OO
O.OOOOE+OO
O. OOOOE+OO
O.OOOOE+OO
O.OOOOE+OO
O.OOOOE+OO
O.OOOOE+OO

MAXIMUM
TIME OF MAXIMUM
MINIMUM
TIME OF MI NI MUM

l.1979E-Ol
l . 6000E-02
-l.0603E-Ol
8 . 0000E-03

O.OOOOE+OO
2.0000E-02
O.OOOOE+OO
2.0000E-02

DI SPLACEMENT TIME
STEP
TIME
0 O.OOOOE+OO
1 1.0000E-03
2 2.0000E-03
3 3.0000E-03
4 4 . 0000E-03
5 5.0000E-03
6 6.0000E-03
7 7.0000E-03
8 8 . 0000E-03
9 9.0000E-03
10 l . OOOOE-02
11 l.lOOOE-02
12 l.2000E-02
13 l.JOOOE-02
14 l.4000E-02
15 1. SOOOE-02
16 1. 6000E-02
17 l.7000E-02
18 1.SOOOE-02
19 l.9000E-02
20 2.0000E-02

HISTORY FOR
DJl
1. OOOOE-01
8 . 5266E-02
7.2848E-02
6.0750E-02
-5.4078E-03
-8 . 4207E-02
-9 . 8212E-02
-8 . 5678E-02
-9.3358E-02
-8.2739E-02
-5 , 1067E-02
-2.9195E-02
l.9009E-02
9.6665E-02
l.2087E-Ol
8 . 872BE-02
7.6180E-02
7 . 2053E-02
3.4921E-02
-3 . 4530E-03
-3.8910E-02

MAXIMUM
TIME OF MAXIMUM
MINIMUM
TIME OF MINIMUM

l.2087E-Ol l . 0992E-Ol
l.4000E-02 l . lOOOE-02
-9.8212E-02 -l,1650E-Ol
6.0000E-03 l.9000E-02

11

12
13
14

15
16
17
18
19
20

MEMBER
STEP
0
1
2
3
4

b
h

NODE

2
DJ2
1. OOOOE-01
l.6390E-02
-l.0676E-0 1
-1. 0298E-Ol
3 . 5295E-03
1.0144E-Ol
l.0809E-Ol
2.3188E-03
-9.5989E-02
-3.8360E-02
9 . 6986E-02
l.0992E-Ol
-7,1293E-03
-l.0699E-Ol
-l.0634E-Ol
-l.2480E-02
8.6587E-02
5.6728E-02
-7 . 7855E-02
-l . 1650E-Ol
2 . 4600E-03

FORCE TIME HISTORY FOR ELEMENT

TIME
AMl
O.OOOOE+OO 9.6000E+Ol
1. OOOOE-03 2 .8848E+Ol
2.0000E-03 -2.6 47 2E+O l
3.0000E- 03 l. 7503E+Ol
4.0000ID 03 1. O<,bbIC101
b. oooorn o 1 I, I'/0410 t OI
I, , OOOOltl ll I 'J. hht1 1 m, no

188

No111111J Mocln M11tl1111I 11l l>y1111111 lo An11lyel11

Program NOMO for Normal- Mode Response

TAJJLI~ 4.J (C'or11J11111dJ

7
8
9
10
11
12
13
14
15
16
17
18
19
20

7 , 0000E - 03 -5 .790 21!l+ OO

B.OOOOE-03 -8. 3 0 2 7E+Ol
9,0000E-03 -5 . 7337E +Ol
l.OOOOE-02 3.6233E+Ol
l . lOOOE-'02 l . 7570E+Ol
l.2000E-02 -2.2569E+Ol
l . 3000E-02 2.5220E+Ol
l.4000E-02 3.4642E+Ol
l. 5000E-02 2 . 7426E+OO
l.6000E-02 5,l 722E+Ol
l.7000E-02 7.2398E+Ol
l.BOOOE-02 -2. 3117E+Ol
l. 9000E-02 -S.4068E+Ol
2.0000E-02 l . 2287E+Ol

MAXIMUM
TIME OF MAXIMUM
MINIMUM
TIME OF MINIMUM

***

9 , 6000E+Ol
O.OOOOE+OO
-B.3027E+Ol
8.0000E-03

LOADING NUMBER

2 OF 4 ***
DYNAMIC PARAMETERS
NTS
OT
DAMPR
20 l , OOOOE-03 O. OOOOE+OO

INITIAL CONDITIONS
NNIO NNIV

APPLIED ACTIONS
NLN NEL
l
0
NODAL LOADS
NODE
AJl
2 2.0000E+Ol

AJ2
0.0000E+OO

GROUND ACCELERATIONS
IGA
0

FORCING FUNCTION
NFO
7

FUNCTION ORDINATES
K
TIME
1 0 . 0000E+OO
2 l . OOOOE - 03
3 3.0000E- 0 3
4 4.0000E-03
5 5. 0000E-03
6 7.0000E- 0 3
7 B.OOOOE- 0 3

FACTOR
O. OOOOE+OO
1. 5000E-Ol
8 . 5000E- Ol
1. OOOOE+OO
8 , 5000E- Ol
l . 5000E- Ol
0.0000E+ OO
NORMAL MODE SOLUTION
NMODES
3
OUTPUT SMl ,11CTl ON
I WU
I Pl, NNO Nl!O
I
I
I
I

187

TABLE 4.3 (Continued)

NODES :

ELEMENTS:

DISPLACEMENT TIME
STEP
TIME
0 O.OOOOE+OO
1 1.0000E-03
2 2.0000E-03
3 3.0000E-03
4 4.0000E-03
5 5 , 0000E-03
6 6.0000E-03
7 7 . 0000E-03
8 8 . 0000E-03
9 9 . 0000E-03
10 1. OOOOE- 02
11 l.lOOOE-02
12 l.2000E-02
13 l . 3000E-02
14 1.4000E-02
15 l.5000E-02
16 l . 6000E-02
17 l.7000E-02
18 1.BOOOE-02
19 1.9000E-02
20 2.0000E-02

HISTORY FOR
DJl
O.OOOOE+OO
5.3221E-04
4.4538E-03
l,5965E-02
3.7705E-02
6 . 4468E-02
8.7960E-02
9 , 9425E-02
9.4028E-02
7 , 1941E-02
3.7520E-02
-i . 2319E-03
-4 . 4130E-02
-7 . 7419E-02
- 9.6296E-02
-9 . 8483E-02
-8.4354E-02
-5 . 5403E-02
-l.6520E-02
2.4923E-02
6 . 2330E-02

MAXIMUM
TIME OF MAXIMUM
MINIMUM
TIME OF MINIMUM

9 . 9425E-0 2 2 . 4408E-02
7 . 0000E-03 l . 4000E-02
-9 . 8483E-02 -2 . 3223E-02
l . SOOOE-02 7.0000E-03

NODE

2
DJ2
O.OOOOE+OO
-2,0013E-05
-5.1199E-04
-3.0147E-03
-8.9069E-03
-l.6887E-02
-2.2416E-02
-2 . 3223E-02
-2 . 1405E-02
-l,8146E-02
-l.1247E-02
-2 . 5762E-05
l . 1875E-02
2.0566E-02
2 . 4408E-02
2 . 3290E-02
l . 9033E-02
1. 3650E_-p2
6 . 0714E-03
-4.9980E-03
-l . 6433E-02

MEMBER FORCE TIME HISTORY FOR ELEMENT

STEP
TIME
AMl
0 O. OOOOE+OO O. OOOOE+OO
1 l.OOOOE-03 4.8184E-01
2
2 . 0000E-03 3 . 2844E+OO
3 3.0000E-03 9 . 0341E+OO
4 4 . 0000E-03 l . 647BE+Ol
5 5 . 0000E-03 2.1947E+Ol
6 6 . 0000E-03 2 .7 113E+Ol
7 7 . 0000E-03 3 . 1667E+Ol
8 8 . 0000E-03 3 . 08l4E+Ol
9 9 . 0000E-03 2 . 2442E+Ol
1 0 l . OOOOE-02 l.0842E+Ol
11 l.lOOOE-02 -7.8519E-01
12 l .2 000E-02 - l . 3479E+Ol
1 3 l .3000E-02 -2 . 4655E+Ol
14 l . 4000E-0 2 - 3. 0138E+Ol
1 5 l .5000E - 02 -3 . 0917E+Ol
16 l. 6 000E-0 2 -2. 7 5 91E+Ol
17 l .7 000E-02 - l .77 88 E+O l
1 8 l. BOOOE - 0 2 - 4 . 09 57 E+O O
19 l.9 000 E- 0 2 8 .l366E+OO
2 0 2. ooo om- 02 1. oee1m+o1
MMC IMUM
'I' I MIQ 0 1~ MAX I MUM
MINIMUM
'l' IMI~ OI' MINIMIIM

l, l (1t1 7 1QI 01
I, 00 0 01~ 0~
I. II 111 /If, I O I
I . hfllllll~ ll ~

188

Normal-Mode Method of Dynamic Analysis

Chap.4

self-explanatory. Note that the first page of the listing repeats the vibrational analysis
given previously in Table 3 .4. For each load case we use number of time steps NTS = 20
and duration of time steps DT = 1 ms. Partial results for the four load cases are also
plotted (by computer) in Figs. 4. 11 and 4. 12. Figure 4.11 shows that the initial displacements excite all three modes of vibration. However, the responses in Fig. 4. l 2(b) due
to the applied force in Fig. 4.12(a) demonstrate mostly first-mode contributions.

20

-"

0.

0
D

0.10

z
......

o.os

-,

15

10

20

(ms)

<: -I 0

~
z

10

(\J

DAMPR-o.o
-

189

Program NOMO for Normal-Mode Respon se

Sec. 4.10

<:

-20

o.oo

,_ -o.os

(a)

<:

, - 0 . 10 0

DAMPR-0-0

(a)

0. 10

DAMP'<=O.O
-

0 . 10

.: o.os

o.os
(\J

~
0
z

o.oo

o.oo

z
......

.....
-o .os
<:

<:

-o.os

-,
0

-0. 10

( b)
( b)

DAMPR-o.o
-

0 . 10

Figure 4.12 Three -member plane truss: (a) applied force; (b) responses.

- o.os
(\J

11.,mmplc 4.16

o.oo

<

N
~

0 . 05
0 , 10

.
NOMOPT let us reconsider the symmetnc p ane
AN n ~ccond example usmg Programs). For this.structure dynamic responses due to the
1111KN trom Sec . 3.8 (see Example 3.
lnlluwing innucnccs arc desired:

Stcp force or magnitude I 00 kN appI'icd m the y direction at node 3, with

I >AMPR
O.O

with DAMPR = 0 0
Jlkttwi1H: li1w11r ~round nccl.llurntlon In the Y direction ,
.
S111111 IIH \' llH~ ) , h11t with ))t\M l'I{
(), I

190

Normal-Mode Method of Dynamic Analysis

Chap.4

For these cases we use the values of E, p, L, and A stated in Example 3.5 , where the
material is aluminum and units are SI.
To take advantage of symmetry, we must decompose the load for case 1 into
symmetric and antisymmetric components and then analyze half the structure twice.
Thus, half the force must be applied using symmetric restraints on the plane of symmetry;
nnd the other half of the force must be applied with antisymmetric restraints. Of course,
the results of these two analyses must be added to find the total solution for the left-hand
pa rt of the truss.
Figure 4 .13 consists of computer plots showing the total step force of 100 kN in
purl (a), symmetric responses to half the force in part (b), and antisymmetric responses

z 100

Program NOMO for Normal-Mode Response

Sec. 4.10

to half the force in part (c). For each load case the number of time steps NTS "'.' 40, and
lhc duration of time steps DT = 5 ms. The responses in Fig. 4.13(~) consist of the
1,.translation at node 6 for all symmetric modes as well as truncation to mode. l .
Similarly, Fig. 4.13(c) shows the responses from all antisymmetric modes ~nd truncat10n
lo mode 2 (for the same nodal translation). We see that the first (symmetric) and second
(untisymmetric) modes produce the major con~ributi_ons _to the responses.
. .
Figure 4.14(a) is a computer plot of the p1ecew1se~h~ear ~round acceleration m the
1, direction. Because this influence induces symmetnc mertial loads, we_ need only
,nnsider symmetric responses of the structure. Time histories ~f the resultmg y-transliuion at node 10 appear in Fig. 4.14(b), with and without dampmg. Evidently, the first
11 1ode is the primary contributor to both responses.

80
,.i.,,

60

40

0
0

z o.so

< 20

I-

<

et:

~ 0

so

100

ISO

200

(a)

(ms

w o.oo
.....J
w
(.)
(.)

<
0

ALL MODES

ee
<J)

I. 00

'E

I-

-o.so

=i

et:
<:!>

-1 . 00

>-

(a)

'<

~ 0
200

.oo

(ms)

..=

0. so

::!5

0. 00

0
7

..r - 0 ,50
('J

-.,

~ 0

1 00

l1'l111m 4.11

(b)

l'l1111r IIIINN (II) 11pplltd lor n : (h)

hy11111111rl1 111~p1111 \ 1 -

NY lllllllltl1

191

IIINjlOII Nl'N; (t' ) 111111

1!'11111111 4, 14

1'111111" IIII NN 111) ~11 1111111 111 l't h111lio11; (h) ll'NJ1UII Nt'N

192

Normal-Mode Method of Dynamic Analysis

Chap.4

REFERENCES
I. Tim?she~ko, S. P. , Young, D. H. , and Weaver, W. , Jr. , Vibration Problems in
Engineering, 4th ed., Wiley, New York, 1974 .
2 Weaver, W., Jr., and Gere, J. M. , Matrix Analysis of Framed Structures 2nd ed
'
.,
Yan Nostrand Reinhold , New York, 1980.

3. Rayleigh , J. W. S., Theory of Sound, Vol. I, Dover, New York , 1945.

4
;ea~er, W., Jr., and Johnston, P. R. , Finite Elements for Structural Analysis
rent1ce-Hall , Englewood Cliffs, N.J. , 1984.
'
5. Weaver'. W. , Jr., "Dynamics of Discrete-Parameter Structures," in Developments in
Theoretical and Applied Mechanics Vol 2 ed w A Sh
p
p
York, 1965, pp. 629- 651.
'
. ' . . . aw, crgamon ress, New
6. Caughy , T K , "Class1caI N ormaJ M odes m
Damped Linear Systems " J A /
; ech., Yo!. 27, 1960, pp. 269-271; also Vol. 32, 1965, pp. 583-588'. PP

?ss, K. A., '_'Coordinates which Uncouple the Equations of Motion of D

d
Lm~ar Dynamic Systems," J. Appl. Mech., Vol. 25, 1958, pp. 361-364. ampe

7.

8. Loui~,. J. J. ~- , "~amping in Structures-a Review," Engineer thesis, Department

of Civil Engmeenng, Stanford University, June 1976.

PROBLEMS
4.J-1.

4.J-2.

~~sumc that the pla~e truss_ i~ H g. 3.12(a). is at rest when joint 2 is sudden!
~t1 uck_s~ ~at 1t acquires an m1tial velocity D 0 1 in the x direction. Determine th~
'.rcc-v1biat1onal ~esponsc of the structure due to this impact. The properties
angular frequencies, and mode shapes for this problem are all given in Exam~
pie 3. 1.
Calc_ulate the free-vibration~! response of the cantilever beam in Fig. 3.13(a) to
t~c sudden release of a static force Po in the y direction at node 2 Exa I 3 2
~1vcs the properties, angular frequencies , and mode shapes for this pr:e~.
lo: t~e pla~c truss in Pro?. 3.6-2, find the response caused by initial displace,
mcnts Doi - D 02 = d at Jomt 1.

4 ..1-4. Obtain the response of the plane truss in Prob 3 6-3 t

l I
the y direction at joint I.
. .
o an m1t1a ve oc1ty Do2 in

4..1-5.

J,'t t~corcc~'.anePc, m_trussthe innegative

Pro?. 3.6-~, determine the response to the sudden release
o
direction at joint I.

4..1-6.

Su1~~'.l~~ that ~ode 2 of the beam in Prob. 3.6-5 has an initial velocity Doi in tlw
v d11 cct1on. Find the response of the structure due to this influence.

,t

4..1-7. Let the bca'.11 in Prob. 3 .6-6 initially have a small positive rotation < at node 2
and a negative rotation O,o at node 3. Calculate the response of the s(/ucturc th1tl

,esults.

41-H. AsN111ne. that

1111

initial posi1iv1; mClmcnt M O ut node I of the beum in Proh ..U,

1
Oh1111n till' l l'hJ)llllNl' o l till' ~II Ul lllll' l'ltll\l'd h y this l'Oll<li
'

Ih h<Iddutl y Hk11,ld

lion

Chap. 4

Problems

193

4.3-9. If joint 1 of the symmetric plane truss in Prob. 3.7-7 has an initial velocity Doi
in the x direction, determine the antisymmetric response of the structure.
4.4-1. (a) Calculate the response of the plane truss in Fig. 3.12(a) caused by a step force
of magnitude P = Pi applied in the x direction at joint 1. (b) Confirm the
reciprocal theorem for dynamic loads by obtaining the response due to the same
fo rcing function applied in the y direction at joint 1. Example 3. l gives the
properties, angular frequencies, and mode shapes for this truss.
4.4-2. Assume that a hannonically-varying force P = Pi sin Ot acts in the y direction
at node 2 of the cantilever beam in Fig. 3. l 3(a). (a) Find the steady-state response
at the free end of the beam due to this loading. (b) Confirm the reciprocal theorem
for dynamic loads by calculating the response caused by a moment M =
M 2 sin Ot in the z sense at node 2, where M 2 is numerically equal to Pi. The
properties, angular frequencies and mode shapes for this problem are all given in
Example 3.2.
4.4-3. For the plane truss in Prob. 3.6-2, determine the response caused by a ramp force
P = Pit/ ti applied in the negative y direction at joint 1.
4.4-4. Obtain the response of the plane truss in Prob. 3.6-3 to a step force of magnitude
P = Pi applied in the y direction at the middle of element 1.
4.4-5. Assume that a hannonically-varying force P = P, cos Ot is applied in the x
direction at joint l of the plane truss in Prob. 3.6-4. Calculate the steady-state
response of the structure due to this influence.
4.4-6. Apply a step moment of magnitude M = M 2 in the z sense at node 2 of the beam
in Prob. 3.6-5, and find the response of the structure.
4.4-7. Suppose that a hannonically-varying force P = Pi sin Ot is applied in the y
direction at the midpoint of element 2 in Prob. 3.6-6. Determine the steady-state
nodal responses of the beam caused by this loading.
1.4-8. Let the beam in Prob. 3 .6-7 be subjected to a ramp moment M = M2t/t1 in the
z sense at node 2, and calculate the response of the structure.
t4-9. For the symmetric plane truss in Prob. 3.7-7, obtain the antisymmetric response
to a ramp force P = Pit/ t I applied in the negative x direction at joint 1.

4,.!l 1. For the plane truss in Fig. 4. 7, calculate the relative response of joint 1 to a
rigid-body ramp acceleration Dg1 = a 1t/ t1 of ground in the x direction.
,t ~-2. Considering the cantilever beam in Fig. 4.8, let the ground at point 1 have a
sudden translation D 82 = din the y direction. Find the response at node 2 due to
this influence.
,I. ~.J. Suppose that the ground at point 3 in Prob. 3.6-2 has a small rigid-body step
rotation DR6 = 8, . Determine the response of joint 1 in the plane truss caused by
thi s motion.
,I,~ 4. For the plane truss in Prob. 3.6-3, evaluate the relative steady-state response at
joint I due to a rigid-body ground acceleration Dg2 = a cos Ot in they direction.
,I,~-~. Ass11n1t th11t the support ut point 3 of the plane truss in Prob. 3.6-4 translates
111dl'Pl'lldln1ly in tlw y dircltion lll'l'mdin8 to the fun ction D R3 = d sin Ot. Calculnh thr 1(~11ltln Ntl1tdy st11h H'~p1111s1 111 joint I.

184

N111111111

Mu 111 M11ll111d "' I>v1111111lc. A1111ly11h1

Chnp

4.!i '' fl lht ~round in Prob I c , 1

(} in lhc ncgai ivc z ,sc."1s',, ,/;'.~1i':11~1;11i:: "t'1111ll 1l8fdhody step rotulion " n
t ~-7. For the continuous b, .
' 111 llu iw1pons0 of the beam ut 11od1
cam 111 l'roh I c, c, f 1
,
1 .. relative steady ~1111
rcspons0 due to a rigid-body , t I'
I
i ti 1111111.(lt.l tie
node 3.
I() U 10 1111 IIC
'('t'f11111lio11 l)N(> = {), Sin fit Of grOUlltf Ill

4.~tt Let lhe support at point 3 in Prob 3 6 7 h

OH1
0, si n Ot in the z sense. Calc~lat~ t~e ave a small harmonic rotu11t111
and 2 of the beam caused by thi . fl
steady-state responses at nod1- I
s m uence.

Direct
Numerical Integration
Methods

11.1 INTRODUCTION
111 this chapter we shall discuss various numerical integration methods for calculnling dynamic responses of structures. These techniques are usually more
cI fl cient than the normal-mode method, unless modal truncation is used. Thus,
1i1hcr approach involves approximations of different types. While the normal11wdc method applies only to linearly elastic structures, direct numerical integralIon can be used for either linear or nonlinear systems.
Any method for direct numerical integration of second-order differential
l'quations of motion may be visualized as some type of finite-difference formul"tion. Although we use finite elements for discretization in space, we find that
the discretization in time is more conveniently handled by finite differences. This
upproach is most apparent in the next section, where a second central difference
Im rnula is converted to an explicit extrapolator for dynamic response . Although
II is not always evident, the other approximation formulas used in subsequent
Nl'Clions are also various types of finite-difference expressions.
For a SDOF analytical model, the damped equation of motion is

mu + cu + ku = P (t)

(1 )

in which the terms were all defined previously in Chapter 2. On the other hand,
the damped equations of motion for a MDOF structure have the matrix form

Mfi +c 6 + SD = A00

ill

<li.:Sl'Jihcd in C'lrnpll'I' 4 . To beg in nny numerical integration procedure, we

111111,t ht1Vl' 11v11il11hll' till' 111itiul condi1io11s of displnccmcnts, velocities, and loads

11H

196

10/

f><l1np11l11ll1111 with I . ,,11, 1111111111111111

1111 1111 N11111n11t nl ltllM11fnllo11 M111hod11

1U8

ti111c 1 110111 pHviow, rcsponscH by 11si11g 11pprox.imulion formu las.

11 1
,1il1111H111uthmls for doing this tll'i.) tk:sct'ib1.:d in thll fo llowing sections for both

I 111111Hl' 11!

11t SI
t11m
0. Fmn1
II
>OFlo system
as these qunnlllll II w, 11111v, 11h 11l11tl lhc luitiul act:clllrnli1111 l11

111 ,1111nd
..

uo

J (

= -;

/{l

kllu

,,io)

MDOF problems .

I I)

:l XTRAPOLATION WITH EXPLICIT FORMULAS

is obtained from Eq . (1) at f ime to. Similarly'

.
(wltil'h
l ) yilllds
for a MDOF structure ' I q

Do= M-1(Ao - S Do - C Do)

( I1
If
till'
Vl' i.'
lor.muss matrix Mis singular ' the vec tor OO .m Eq. (4) may be taken as u 111111
.
iigure 5. 1 shows a graph of the nu
. 1
S l)OF system. This plot is represent d menca solution for the response ol n
Ihough it may actually have slight
ast. a s.n:iooth curve in the u-t plane, i.'Vl II

d.:

i con mu1t1es. The symbols Uo

denote values of u at the time stations t t t ' u1' u2, . . .
/11 1, . . . ' and so on . The time interval !:,,.t b
. o, 1, 2, . . . 'lJ ,, r,.
t11kl.ln to be of uniform duration t:,,.r altho jh etween tim~s !j ~d tr: 1 is usuul lv
HI ii y. 'l'hll objective of the
'. 1 . ug su~h a specialization is not ne,1"
numenca integration process is to calculate tlu

"1

i,

"1, UJ' i,

11 1xplicil
111111 1
in

extrapolation procedure consists of expressing the displacement at

terms of the displacements, velocities, or accelerations at previous
11 1
111111' stotions.
Referring to Fig. 5.1 , we can write the second central difference
111 llll' displacement at time ti as the acceleration:
1

Uj = (At)/ui- 1 - 2uj

(1)
Ui+ 1)

hklt implies a SDOF system with a unifonn time step At. Similarly, the

...

l11l'ily at time

tj

may be approximated as
.

Uj

1
!:,,.t (ui+l -

(2)
Uj-1)

',11hstituting iij and Uj from Eqs. (1) and (2) into Eq. (5 .1-1) and rearranging the
/I

11

~ult produces

",, ,
P'(t;)

t'

(3)

~ P(I;) - G- !~}, - L(:,)' - i'i.},-,

(4)

l1q11L1tion (3) is known as the central-difference predictor, which can be applied

1rpctitively to obtain ui+ 1 for each time step . Then the acceleration ili and the
w locity ui at time ti may also be found using Eqs. (1) and (2) if desired.
The central-difference predictor in Eq. (3) is a two-step formula, so it
1 1111not be applied directly in the first time step. In order to derive a starting
111 ocedure, we apply Eqs. (1) and (2) at time t0 and solve for U - i , as follows:
U- 1

= uo - uo Lit + ! iio(At) 2

(5)

l'his result can be used in Eq. (4) to evaluate P*(t1) for the first time step.
For a MDOF structure, the expressions given above can be generalized in
11 1nulrix format. Thus, Eqs. (1) through (4) become
()

t,

..

t,

11'(111111 ~. I

D1 ""' (iit) 2 (Di- , - 2Di + Di+1)

S11l111lt111 hy tllil'l'I 111111 111111111 l11h' f''11 1l1111

lllr!I I N111111111t nl lt1h1111ntlon Muthoda

198

I(~r)2 M
A*(r1) =

111 th is exu111ple we c 1111 cxp1li1111 tlw

t

2 ~( cJ01, 1 A+(I/)

l(A

A(tJ - [s - (;t)2 MJ01

)2
1

'

M - 2~rc]Dj,

D- , = Do - Do At + 2 Do(At)

D;,i+ I

"'

(i

M; A;' (ti)

=2 =~
w,,

= }5_
=
w2

k(.!_)
21r

(a)

o, so the initial acceleration is

Po

P,

(b)

iio=-=-

1, 2, ... , n)

- ) with R =Pi.To start the solution, we apply Eq. (5)

hh h Is drawn from Eq. (5 . 1 3
o
1o, 11h111in
2
2
2
(c)
1
Pi ( T ) _ Pi (21r)
= O 04935~
U - 1 = 2iio(A.t)2 = 2m 20
- 2k T
20
.
k

(!_)

.
. Eq ( ) ves the displacement at time
1111 II the central-difference predictor m . 3 g1

which is simply a multiplication of the effective action Af (ti) by the scul111

(At) 2/ M;. Therefore, high-speed computer core storage need only contain info,
mation for one displacement (or more conveniently, one node) at a time.
The central-difference predictor is probably the most widely used explkll
formula for solving structural dynamics problems. However, all expressions ol
this type have a critical time ste[l.., above which the solution becomes numerically
unstable and diverges [l]. Nevertheless, among all of the known second-order
pied ictors, the central-difference method has the largest stable time step [2]. The
value of the critical time step for this technique is

(At)c,

_ Li 0 , we have uo - o -

which may be used to start the procedure. If there is no damping and the n111 NN
matrix is diagonal , a so-called nodewise solution may be devised. That is, lho
solution for a particular displacement D;,i+ I in Eq. (8) becomes

(At)2

. , r ti , sti t'fness k and the period

f\l1llls o 1e
2

/I.I 111111 10
l ..

111111111111 111

In addition , the generalized form of Eq. (5) is

199

I x 1111pol11tlu11 wllh I xplh It I 111111111111

Chnp

u, =

(t:,.: ~1 -

ltxuma,ll 5.1

as

U -I

(d)

.
. Eq (c) A second application of the predictor at time t2 yields
"hkh 1s the same as U- 1 m . .

u2 =

C~

2f, -[k - c: 2Ju, - o}

= [ (; )2(2;)2o

(12)

SuppoNl' thut an undnmpcd SDOF linear system is suhj1.1.t1.<111> 11 step force of mugnit11<1,
/'1, Ht111 ll11g fw111 nst. l.11 us rnkulute the 11pp1oxi 111111l 11sponse using lhl' l'1.111rnl
1lllhHm, p11dii101 with 1w,11ty unifrnm 1i11w ~hpN 111 d111111in11 t\/
'/'/'20

Pi
P1
= (0.09870 - 0.04935)k = 0.04935k

7T

The symbol w,, in this expression denotes the largest angular frequency in the
analytical model, and T,, is the smallest period.
Key and Beisinger [3] applied the central-difference predictor to the linem
dynamic analysis of thin shells. In addition, Krieg and Key [4] showed that us in~
11 diagonal mass matrix improves the accuracy of the procedure. Successful use
of this method in nonlinear analysis has also been reported by Key [5], who USl'd
111ti!lcial damping to control the inherent instability. Morino et al. [61, searchld
Im the optimal predictor for systems of second-order differential equations and
1.onl'l uded that the central-difference formula is best. The main disadvantage ol
this method is that for a fine network of e lements a very small time step i11
11. qu i1c<l to obtain stable results without damping. The topics of numericnl
Ntahility un<l accuracy will be discussed more thoroughly in Sec. 5.6.

0- (:)2U-1] (;0)2(2;)2 ~ -

ti

_ 0.04935) + c2)(0.04935)]~
(e)

P,
0 . 1925-k

.
(3) is a lied in the same manner, and Eqs. (1) and
lu iuch subsequent time step, Eq.
p~
I ..
11
I
d to find the accelerations and ve oc1t1es as we .
t 'J rnuld be emp .oye
f
b . db this procedure are listed in Table 5.1,
The approximate values o u! .o ta1~e y
ws lots of the approximate results
,1l11ng with the exact values. In add1.t1on, ~1g. i'2 ~~ 1 PFor the scale used in Fig. 5.2,
1111 both zero damping and a dampmg ratio o r.
..
1111 plotted curves are indistinguishable from exact responses.

I' uauplc 5 2

.
Exam le 3 1 Recall that the
l'luurc 5.3(a) shows the plane truss analyzed prev1ous~\m O 8A ~nd
Now we shall
' i 11Ns sectional areas of members l an~ , 2 tare t~!u:o ;e force P (t), applied in the x
,h1fl11ll1i1te the undamped response of this s rue .
. F' 5 3(b) Let us
1
1llH'l'llOII Ill jnint I. The time histordy ~; th~.d~n:::~ ~~~~~1~Xrr~=:~~e ~~; di~tor ~ith 20
, 11k11l11tr the displacements I), an
i IISlllK
,
.
,
t:,. 1
'/'1/ J..O st111 tlllij lrnm rest.
1
1111ltoi 111 111111 Ht0ps ol durnt on . .
'
ll'l'H for the free displacements at
1111111 E x11111pll \ I llw st1llmss 11111 1 11111N~ II 1111 i

A..

200

Direct Numerica l Integration Methods

Chap. 8

oc.

5.2

Extra polation w ith Explicit Formulas

TABLE 5.1 Response for Example 5.1

Using Central-Difference Predictor"
y

Approx. u

Exact u

Approx. u

Exact u

1
2
3
4
5
6
7
8
9
JO

0.04935
0.1925
0.4154
0.6960
1.007
1.317
1.595
1.815
1.955
2.000

0.04894
0.1910
0.4 122
0.69 10
1.000
1.309
1.588
1.809
J.951
2.000

11
12
13
14
15
16
17
18
19
20

1.947
1.800
1.574
1.292
0.9804
0.6712
0. 3944
0.1774
0.04 157
0.00034

1.951
1.809
1.588
1.309
1.000
0.6910
0.4 122
0. 1910
0.04894
0

P(t)

"Tabulated values to be multiplied by P1/ k.

i------

- -0.6L
(a)

P,

I' I
k

1.----------,

Ol__---- - -- -+:-----~--Tt.t2:o-- t,o

/ '
P,
(b)

()

Fi~urc 5 .3
I

(a) Plane trnss; (b) applied force.

201

202

Direct Numerical Integration Methods

Chap. 5

Sec. 5.3

TABLE 5.2 Responses for Example 5.2 Using Central-Difference Predictor

joint 1 are

S=

s[ - 0.36
0.48

M = m[ol 01]

-0.48]
1.64

(f)

in which s = EA/Land m = 3.28 pAL/6. In addition, we have

wf

0.2 .:_
m

w~ = 1.8 .:.

(g)

Using the first expression in Eqs. (g), we can relate the mass constant m to the stiffness
'
constant s and the fundamental period 7; , as follows:
2

m = 0.2s ( 7i71")

(h)

~.l tim~ to = 0 the initial displace~ents an~ :~locities are (D 1) 0 = (D2) 0 = (D 1) 0 =

( >2)0 0. Therefore, Eq. (5.1-4) gives the m1t1al accelerations as

Do=

M - 'Ao = ;[~

UJ ~

~][~] =

(i)

Before beginning the step-by-step procedure, we apply Eq. (10) to find

D- 1 =

I2 Do(At)

I[ ~ (1i)

1
]
2 0 m 20

~ [~] /;s (2;)2(~~)2 =

[ o.2;67]

D, = (~
=

[~

~]{[~]

(:;)2 [~

{[o.~34] _ [ o.2;67]}

(j)
ti

produces

~Jn-J

[ o.2;67]

(kl

[1 O]{[P']
O 1
0

_ s[-0.48
0.36

.. (2!.20.) _!!_
( 271") [1 - 0.08883]
0 .2s Ti
O + 0.1184
2

o] }

-0.48]
~ [1
1.64 D, + (Llt) 2 o I D,

+ 2[0 .2467] ~
o

0.4496 -I [0.4935]} ~ = ro.9431 p 1

, 0.05845
o
s
Io.os845 -;
{'

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

Approx. D1

0.2467
0.943 1
1.979
3.227
4.574
5.928
7.188
8.223
8.898
9.107
7.839
5.330
1.952
- 1.929
-6.024
-10.06
-13.70
- 16.50
- 18.07
-18. 15

Exact D ,

Approx. D2

0.2432
0.9322
1.963
3.210
4.556
5.901
7.148
8.179
8.868
9.111
8.382
6.315
3.221
- 0.51 87
- 4.556
- 8.592
-12.33
- 15.43
-17.49
-18.22

0
0.05845
0.2930
0.7593
1.375
1.962
2.365
2.557
2.627
2.679
2.720
2.416
1.420
-0.2636
- 2.191
- 3.772
-4.683
-5.049
- 5.238
-5.467

Exact D2

0.0047 1
0.06831
0.2932
0.7350
1.333
1.932
2.374
2.598
2.662
2.667
2.653
2.462
1.787
0.4618
-1.333
- 3.128
-4.454
-5 .128
-5.319
- 5.333

exact values are also listed. Figure 5.4 depicts plots of these responses for both the
11pproximate and the exact analyses. Here we see that the curves for the approximate
responses deviate slightly from those for the exact responses.

6.3 ITERATION WITH IMPLICIT FORMULAS

The technique to be explained in this section is called the predictor-corrector

method [6, 7]. In each time step an explicit formula (a predictor) is first used to
estimate the response at the end of the step. This is followed by one or more
applications of an implicit formula (a corrector) to improve the results. Although
such an iterative procedure is not required for linear analysis, it is included here
l'o r use in nonlinear problems, where physical properties can change in each

which is lhc same as D- , in Eq. (j). Applying Eq. (8) again at time t2 yields

(Llt)
m

Tabulated values to be multiplied by P, / s.

lJsl ng this result in the matrix form of the predictor [Eq. (8)] at time

02_

203

Iteration w ith Implicit Formulas

({')

In .suhsl~ru.e nt ~n lculutions Wl' upply Eq . (8) rcpctili vllly 10 find rhc response at cnch ol
ihl tl111111111111g 11ml' st111lo11s. Noll' ih nl lhc 11111ss n1111il x 111 thi s l'x11111pll is diugonnl so 11q
(K) dms not 1tq11i1 t sol vl11i.: sl 11111l11111lt111 s l'q1 111th111s
' ,
1<1~1111~ 1111111 lhl~ 11pp1 wd111111t 111111lys is l111 /l 111111I /l , 111r 11 rv,11 111 'f'nhk' 'i l, wlwtl'

eycle of iteratio n.
By an approach that we shall refer to as the average-acceleration method,
lhc veloc ity u11 1 for a SDOF system at time tj+ t is approximated as
(1)

in whic h 1i1 is the ve lt)cily Ht the preceding time station 11 (see Fig. 5 .1). This
lrn nnlla , know n lo nu,rnri rnl 111111lysts as till' t1'<tf>l'zoidul rttfr, implies that the

205
204

Direct Numerical Integration Methods

Chap.

Sec. 5.3

Iteration with Implicit Formulas

acceleration in the stepJ s taken to be the average of ii1 and ii1+,. Similarly , the
displacement u + at the end of the step is approximated by the trapezoidal rule

D,

10 ~s

with the expression

Approx.
Exact

(2)

where the velocity in the step is taken to be the average of u1 and u1+1 Substitution of Eq. (1) into Eq. (2) yields

5~
s

(3)

When applying this method , we do not use Eq. (3) directly; but Eqs. (1) and (2)
are used in succession. Because the value of ii1+ 1 is not known in advance, the
approximation is said to be implicit, so the solution must be iterative within each
step. The following recurrence equations represent the ith iteration of the Jth
step:
- 5~
s

10 ~s

(i > 1)

(4)

= R1 + Hu1+i); At1

(i 2: 1)

(5)

l
= -(~+1

(i 2: 1)

(6)

(u1+ 1);

= Q1 +

(u1+1);
(ii1+1);

Hu1+1);- 1At1

- ku1+1 - cu1+1);

where

(7)

15 ~s

and
R1 =
20 "1
s
(a)

Uj

+ ! Uj

(8)

At1

This iterative procedure is not self-starting because it requires a supplementary formula for determining the first estimate of u1+ 1 in each time step. After
evaluating the initial acceleration from Eq. (5.1 -3) , we may start the iteration for
the first step by approximating 1 with Euler's extrapolation formula, as follows:

(j

0; i

= 1)

(9)

Then the first approximations for u 1 and ii 1 are obtained from Eqs. (5) and (6).
All subsequent iterations for the first time step involve the repetitive use of Eqs.
(4), (5), and (6).
To start the iteration in the Jth time step, we may again apply Euler's
formula to determine a first estimate of

u1+ 1 as
(i = 1)

fl'l1411n1~.I
(h) /l ,

(10)

Hoth Eqs. (9) and ( I 0) imply constant values of the accelerations within the
steps . To improve the accuracy of the rcsu lts fo r the first iteration of the jth step,
we l:HI\ 11se till' slil'htl y IIHll'l' cl,,borntc fo1 mul11 thul is valid only for uniform time

206

Direct Numerical Integration Methods

111

6.3

207

Iteration with Implicit I ormul tJII

I q1111tion (16) is the same as Eq_. ( l) of the aver.age-acceleration method, but Eq.

steps:

1 l /) is slightly different_from its clount~erpart

(i = 1)
This expression spans two equal time steps from tj- I to tj+i (see Fig. 5.1) and
utilizes the midpoint acceleration at time tj.
Equations ( 10) and ( 11) are called explicit predictors because they provido
estimates of Uj+I in terms of previous values of u and u. On the other hand, Eq
(1) is referred to as an implicit corrector that yields an improved value of u11 1
after an estimation of uj+I has been obtained. The method described here in
volves one application of a predictor, followed by repetitive applications of thl'
corrector.
An iterative type of solution requires some criterion for stopping or chani,t
ing the step size, such as a limit on the number of iterations. A convenient
method for measuring the rate of convergence is to control the number of
significant figures in uj+ 1, as follows:

I(uj+1); -

(uj+1);-1 I <

(12)

where 11 is some small number selected by the analyst. For example, an accuracy
of approximately four digits may be specified by talcing ,, = 0.0001. That lewl
of accuracy is used in the numerical examples of this section. For a MDOI
structure, we use the length of a translational or rotational vector in Eq. (12),
which is equal to the square root of the sum of the squares of its components .
Another implicit approach for approximating responses is known as tht
fl11ear-acceleration method. As indicated by its name, this technique has thl'
ussumption that the acceleration varies linearly within each time step. Thus , nn
expression for a during the step f:.tj may be written
a (t ')

aj

(iij+ 1 -

uj) f:.t.

. + Ujt
.. + (.,Uj+ I

= U;

(u1+1);

= Rj + 'i

*=

Uj

+ ("Uj+ 1

1i1 I

1Ui1

(20)

. the recurrence equatio.ns for

F MDOF structures, we can genera1ize
.
II rntio~rinto matrix formats. Considering first the average-acceleration method,
1w replace Eqs. (4) through (8) by
(21)
(i > 1)
(Dj+1); = Q1 + Hi>1+1);- 1 D.tj
(22)
(i ~ 1)
(D1+1); = Rj + HDj+l)i D.tj
.
(23)
(i ~ 1)
(D1+1); = M - l(Aj+I - s Dj+l - C D1+1);
Wlll'l'C

(24)

I\IHo, Eqs. (9), (10), and (11) are supplanted by the matrix expressions
(D1)1 = Do + Do f:.to

o" A
(D;+1)1 = D1 + i utj

") (t')J
f:.tj

Uj

( 15)

I il1,,)At1

(j = O; i = 1)
(i = 1)

(26)

(i = l)

(28)

(27)

'
'lli 11 IIurl y' for the linear-acceleration
method' Eqs. ( 19) and (20) generarize to
* 1 (D. ) 1::,.
(i > 1)
(29)
(D;+1)1 = R1 + !! 1+1 ; t1
-

( 1(1)

Rj

1111d

( I '/l

(25)

= Dj + !Dj D.tj

Rj

( 14)

At the end of' the step the velocity and displacement become
,;, ,,

(19)

1111 t the iteration in each step.

(t')2
= Uj +I+"
Ujl
Uj
2

as:

A
+ 32 Uj ut;
+ oI Uj.. ( uA(j )2

(D+ 1)1 = Dj-1

')

U1+1 ;

u1+1

" lll'l'C

und
U (t

,1 1111

.. ) (t f:.t
')

Uj

-gu1

l lt11s, we form the recurrence equation for the ith iteration of

* I ( ) D,.t
(i ~ 1)

t1

CIJ >

where t' is measured from the beginning of the step (see Fig. 5.1). If till'
uccelcration varies linearly, the corresponding velocity and displacement wil I
vary quadratically and cubically with time. Therefore,

( ')

= u1 + H2u1 + Uj+I

U t

U1+1

I Ill' formulas given earlier [see Eqs. (9) and (10) or (11)] may be used agam to

E,,I (uj+1)d

:~o~f~

~3!anner analogous to that

We will apply the lmear-acce era ion me
.
E (l)
.
h Because Eq ( 16) 1s the same as q
,
li11 the average-acceleration appro~c :
.
. . is the same as that in Eq.
1111 recurrence expression for the ith iteration of u;+ i .
I for ii in
1 Il To obtain a direct relationship between U1+ J an~ Uj+ '.' we so ve
J+,
I q ( 16) and substitute the result into Eq. (17), which yields
.
. ) /::,. + I .. (f:.t )2
( 18)

I II

(' IHll'Ht',

D1 I

~ I>,

A.11 I

~ i>1 (Atj}2

(30)

.
( I) (l\) ( tl) ('JC,) (J'f), uml (2K) apply to both methods.
I.qH. 2 t t 2 t
t

208

1>1111 I N1111111rl1 11l l111n111n1lu 11 Mulhod1

1111111111111 wi th lmpllc It

II is we ll known thar !ht 11111111 11111 I 111111111 111tlhod is somewhur 1111

uccurure than the average-accch.-111111111111r 1I111d I / l llowever, it has been shu
181that the former technique is onl y , ,111,/11/1111,tl/\ .1tahle. Therefore, as in th
central-difference procedure, the solut ion diwqit11 if the time step is too 111111
On the other hand, the average-accelcnition nicthod is unconditionally .1tuli/
although less accurate. As with the central-difference predictor, a nodtwiM
iterati ve solution of a MDOF problem is feasible if the mass matrix is diago1111I

1:q. (I ):

201

I 1111111,I H

/ '1

Ill

,., '/'

( Ii iJ I

(I

(111 ),1

= 0 .04816 ~
0 I 0 .!MHXO
T
k
k ( ' '120/' ' ) ( 21r)

() I/ ~ I H)

I()

O Oi! HHO

Ill

Eq . (5):
J',q.

(11 1)4 = ~
m ( I - 0 .048 16) = 0.95 18 ~

(6).

Example 5.3
We shall now repeat Example 5. 1 using the iteration methods described in this sect,1111
Recall that m = k(T / 211')2, as given by Eq . (5.2-a), and that the in itial acceleration 111
ti me to = 0 becomes u0 = P1/ m, in accordance with Eq . (5 .2-b) .
To apply the average-acceleration method, we start the first iteration in the fir I
time step us ing Eq . (9) to estimate the velocity at time t 1 = At= T/ 20, as followh
(u1)1

= uo + f1o tlt = 0 + (~)

(:0) =

nse has converged to within fo ur significa nt

111111 lh iterat ion we see that the re~po . T ble 5 3 along with the number of
II~ II l(csults for 20 time steps are given m a
. '
II i,1!11111~
required in each step .

It 1111

"1

TABLE 5.3

Average-Acceleration Method

0 .05 P~T

Then the displacement at time t 1 is fou nd from Eq. (5) to be

(u1) 1 = Ro + I (u. 1)1 At= 0

0.05
+ 2k

2
)

(Pi T (2T1r)
20

= 0. 04935 P1

wh1d(6)
1 isasthe same as u1 in Eq . (5 .2-d). Next, we obtain the acceleration at time t1 from
11q,

(ui)1

1
= -(P1
-

ku1)1

Pi
= -(I
m

- 0 .04935)

= 0 .9507 -Pi
m

Por the second iteration in the first time step, Eqs. (4), (5), and (6) yield

'>
11

'

H
1/
10

II

1:2
Eq. (4):
Eq. (5):
Eq. (6):

(ii1h

(I

P1T
+ 0.9507)P1T
- = 0 .04877-

40m

( ) =O+
Ui
2

Pi .
(u1)i = -(I
m

0 .04877 (Pi T
2k
20

2
)

( 21r)
T

= O.O48 l 3 ~k

Pi
0 .95 19 -

Third iteration :

Eq. (4):
Eq. (5):
Eq . (6) :

lo1111h lll' rn lio n:

11

1,1
l:i
JI,
)/

IH

- 0.04813) =

Respons e for Example S 3 Using Iteration Methodsa

11)
)0

Linear-Acceleration Method

II;

Approx. u

n;

Approx . u

4
4
4
4
4
4
4
3
3
3
3
3
3
4
4
4
4
4
4
5

0.04816
0. 1880
0.4061
0.6813
0.9873
1.294
1.573
1.797
1.944
2.000
1.959
1.827
1.6 14
1.343
1.038
0.7300
0.4478
0.2 187
0.06497
0.00 126

4
4
4
4
3
3
3
3
3
3
3
3
3
3
3
4
4
4
4
5

0.04855
0.1895
0.409 1
0.6861
0. 9936
1.302
1.581
1. 803
1.948
2.000
1.955
l.8 18
1.60 1
1.326
1.0 19
0.7 105
0.4299
0.2047
0. 05665
0.00025

l'uhulatcd values to be multiplied by P1/k.

(ti1 h

P1T
P1T
= ( 1 + 0.95 19) 40
m = 0 .04880-;-

(u1 )1 = 0

0.04880(P1T2)(21r)2 = 0.048 16~

2k
20
T
k
0 .048 16)

0,9 5 18 Pi
Ill

.
.
~thod using Eqs. (19) and (20) in place
Nex t, we apply the lmear-accelerat~on: res o~ses are calculated somewhat more
111 1 q~. (5) and (8). In this c~se th~ appr~:1:~s fo/this second analysis also appear in
,111 1111111.: ly and with few er iterations.
1,,hlt '> . L

I \llllljlh. 5.4
truss in Example 5 .2 which was shown
1, 1 11~ 111111ly1.c hy_ itcra~ive methods thct ancthe relationships m = 0 .2s(Ti / 21r )2 and
111 l'lp '> . 1(a ). h ,r this prohl':m we av~
5 2-)
1111 ! l , 0}/ '1/ 111 , as giw n hy l:qs. (5 .2-h) and (. t .

210

Direct Numerical Integration Methods

nc. 5.4

Direct Linear Extrapolation

t b For t_he a~erage-acceleration method, we begin the first iteration in the first timo
s ep Y estimatmg the velocities at time t 1 with Eq. (26), as follows:

(D,),

=Do +

D0 At= o +

[~](:) (~) =

TABLE 5.4 Responses for Example 5.4 Using Iteration Methods

[ois] P;

=Ro+ i(D,)

At=

;;_ic(~;;

O+ [0.05]
= [0.2467] [l
0 0.PiT..2s (T..)(27T)
40 T..
O
s
(cl

i::: same as D1 in Eq. (5.2-k). Then the acceleration at time t, is calculated from

(f)

For the second iteration in the first time step, Eqs. (21) , (22), and (23) produtt

ico

At={[~]+ [o.9;12]} ~ (~t)

[1.911 2] !j_ (Ti) = [0.04778 JP, I;

(D,h = Q0 +

1) 1

0 . 11 84 m 40

Eq. (22):

Eq . (23):

(D,)2

0.002960

=Ro+ ~(D,h
At= 0 +[0.04778
J P,Ti (Ti)(21r)2
2
0.002960 0.2s 40
Ti
= [0.2358 J!j_
0.01461 s

(D,h =

M- 1(A1 -

= _!_

[1

D;

Approx. D,

Approx. D2

6
6
5
4
4
4
4
3
3
3
4
4
4
6
4
4
4
4
3
3

0.2369
0.9121
l.932
3.173
4.515
5.851
7.081
8. 102
8.807
9.101
8.691
7.150
4.451
0.9388
-3.005
-7.028
-10.82
-14.09
-16.53
- 17.87

0.01168
0.08378
0.2961
0.7005
1.263
1.864
2.357
2.648
2.734
2.693
2.596
2.409
1.950
1.014
-0.4398
-2.193
-3.841
-5.009
-5.537
-5.533

5
5
4
4
3
3
3
3
3
3
4
3
3
5
3
4
4
3
3
3

0.2400
0.9220
1.948
3.192
4.535
5.875
7.112
8.139
8.838
9.110
8.742
7.146
4.394
0.8498
-3.106
-7 .142
- 10.97
-14.27
- 16.73
- 18.02

0.00834
0.0761 8
0.2941
0. 7165
1.298
1.902
2.373
2.629
2.697
2.670
2.618
2.482
2.031
1.024
- 0.5508
-2.386
- 4.002
-5.034
- 5.425
-5.395

"Tabulated values to be multiplied by

P, / s.

t \ti)

O]{[P']
_ s[ -0.48
0.36
0

m O 1

0.9221 JPi
-_ [ 0.08922
;;;

- 0.48] [0.2358
1.64 0.01461

J!j_}s

11 ,4 DIRECT LINEAR EXTRAPOLATION

Such ,7ul~ul~tions arc re~eated u_ntil convergence of displacements is obtained to within

lour s1gn1f1cant figures (m the sixth iteration). At that stage the values of the displ ..
111t11ts cnlculatcd from Eq. (22) are
au

'l'11hil "i.il t:011t11i11s tilt displ11t1111<'11ts /) 1

1111111lw1 ol 1tt111t101111 11, 101 1111 h ~tip

Approx. D2

As a second analysis, we apply the linear-acceleration method, using Eqs. (29) and
in place of Eqs. (22) and (24) . The results for this case are also listed in Table 5.4.
c 11111parison of these responses with the exact results in Table 5.2 shows that they are
1111111 llCCurate than those for the average-acceleration method.

s D,h

(D i)t> _

Approx. D 1

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

- 0.48] [0.2467] !j_}

1.64
0
s

Eq. (21):

D;

I
2

Linear-Acceleration Method

Average-Acceleration Method

Next, we find the displacements at time t 1 from Eq. (22) to be

(D,),

211

0.2369

0.01 168

1111<11) 1

P1

,\'

Im lwiuty linw sttps , 11s wl'II Its

(g)
fill'

11 lht equations of motion for a MDOF structure are linear, it is possible to avoid
lti 11t1ion of implicit formulas for numerical solutions. Instead, direct linear
, l111polation procedures may be devised for both the average- and linear11111ltr11tion methods. We can formulate either total-response algorithms for
l1t1r111 systems or incremental-response methods for nonlinear systems. Because
1h, 11111nhcr of nrithmetic operations is about the same for either approach, we
1 h111111l' lo dlvl'lop lhl' i1
wn.111lnt11l ttchniqm , which upplies to both linear and
111111l1rn111 p1ohlt1n~ F111 lht1111111t, 11 ,., ft11s1hlt to stl up simultuncous equations

213

Direct Numerical Integration Methods

212

!loc. 5.4

for incremental accelerations, velocities, or displacements. Here we will use

incremental displacements as unknowns and solve a pseudostatic problem for
each time step. To save space, only MDOF structures will be considered; and
any SDOF system becomes merely a special case.
At time ti (see Fig. 5.1), the damped equations of motion for a MDOI
linearly elastic structure are

M Di + C Di + S Di = Ai

..
!::,.Di

Subtraction of Eq. (1) from Eq. (2) produces the incremental equations of
motion as

t::,.D
Q
i- i

(11)

(12)

Now substitute Eqs. (11) and (12) into Eq. (3) to obtain

_ 4_ 6.D- _
M [ (!::,,.tj)2
J

Similarly, at time ti+ 1

(2 l

= (Cl.ti )2

6.Dj = 6.t 6.Dj - Rj

( I)

= ti + D.tj, the equations of motion become

M (Dj + D.Dj) + C (i>j + D.Dj) + S (Dj + 6.Dj) = Aj + D.Aj

Direct Linear Extrapolation

t 'ollecting

QljJ + C [~
!::,,.D- l:::,,.tj

R} + S 6.Di

;J

= Cl.Ai

terms, we rewrite this equation in the form

S 6.Di = 6.Ai
111 which

(14)

(3)
These equations will be used for both the average- and linear-acceleration
algorithms developed in the following discussion.
For the average-acceleration method, the incremental velocities obtained
by the trapezoidal rule at the end of time step D.ti are

6.Di

= 2 (Di + Di+1) D.ti = Di D.ti + 2 6.Di D.ti

Similarly, the incremental displacements at the end of the step become

_

D.Di -

2 (Di

'

+ Di+ 1) !::,.ti - Di D.ti + 2 6.Di !::,.ti

(5)

Substitution of Eq. (4) into Eq. (5) yields

D.Di

1 ..

..

= Di b.ti + 2 Di(b.ti) + 4 b.Di (!::,.ti)

(6)

= (!::,,.ti )2 !::,.Di -

(7)

Substituting this expression into Eq. (4) produces

.
2
.
!::,,.D = - tJ.D- - 2D1
Lltj
J
J

(8)

Now we define vectors Qi and R1 that contain only combinations of i>i and

~
R1

4
-

u.11

D1+ 2D1

(15)

thus, the pseudostatic equations represented by Eq. (1~) are to be solved fo~ the
lnl'l'emental displacements 6.Di in each step. Then the mcrement~l accelerations
.\i) and velociti~s ai>i mar. be found using Eqs. (11) and (12) . Finally, the total
1
vulues of Di+1, Dj+l, and Dj+l are

D1+1 = Di + 6.Di

(16)

i>j+1 = i>j + !::,,.l)j

(17)

= i>j + anj

(18)

Dj+l

i>1

Ulrcct Linear Extrapolation by the Average-Acceleration Method

Determine Qi and Ri from Eqs. (9) and (10) .
i 1:ind Sand !::,.Ai using Eqs. (14) and (15) .
Solve Eq. (13) for t::,.D1.
Ca lculate 6.Di and t::,.D1 with Eqs. (11) and (12).
.
\ Add the incremental displacements , velocities, and accelerations to prending values using Eqs. (16) , (17), and (18).

('))

2f>1

11~1111-1 llw~t 1kf1111tio11~. Wl' ll'wiitl' Eqi. (7) 1111d (K)

6.Ai = 6.Ai + M Qi + C R1

11 ucture consists of the following calculations m each time step:

4 .
..
!::,.ti Di - 2Di

as follows:

1111d

In summary , the procedure for obtaining ~ynamic ~esponses of a MDOF

Solving for the incremental accelerations in Eq. (6) gives

..
!::,.Di

(13)

( IOI
il !j

'l'lll lling now to the linear-acceleration ~ethod , we may again derive the
1111 1ri 111 111ol vc lod ty vector froni the trnpc1.01dal rule to be
6.1>, 1>1 6.11 I ; 6.1>1 6.11
(4)

215
110. 5.4

Direct Numerical Integration Methods

214

llowever, the incremental displacements in the step must be found f

(5 .3- 17), as follows:

E
rom

ADj

+ H2Dj + Dj+1)(Atj)2
i>j Atj + !Dj(AtY + !ADj(AtJ2

Dj

ll

Atj

( 19)

Solving for the incremental accelerations in Eq. (19) gives

..

ADj

= (Atj) 2

Atj

(20)

Dj - 3Dj

Now substitute Eq. (20) into Eq . (4) to obtain

AD. = -3 1.1AD
1
1

1.1

Then define

* = A6

Qj

1.1tj

R j*

--

30

..

Dj + 3Dj

(22)

2Atj Dj..

(23)

Qo

4.

(b)

+ -4m- = k +

= (Atj )2 ADj - Q/
J

Eq. (15):

Mo

Eq. (13):

Auo =

Eq. (11):

( 0.04816.. = - 4 - Auo - -Qo = 4 (20)

P ) - 2A
Auo
(At)2
T
k
m

(At)2

Mo
T
=

S* AD1 = AA1*

(26)

6 - M+-C
3
S* = S + (Atj)2
Atj

(27)

(d)

2A

A) -

tt

= (
=

= 4 -20) -k ( 0.04816( 27T m

k

Auo =

Pi = - 0.048152m
m

(f)

ti..uo - Ro = 2( ~) (o.04816~) - 0

Pi) (2T )

40) (,
\0.04816;;;

7T

(g)

AT
= 0.04880--;;

Eq. (16):

U1

Eq. (17):

u, = uo + Auo = 0 + 0.04880m

Uo

(e)

_ k = 0.04816,;
41 53
2

Nl'Xl, we substitute Eqs. (24) and (25) into Eq. (3) and collect terms to find

= 41.53k

= Mo + mQo = 0 + 2P, = 2Pi

(25)

(c)

Eq. (14):

(24)

AD = l_ AD - R*
At

4k( - T ) (20)
21r
T

(a)

-k = k

Eq. (12):
J

Pi

P,
m

+ 2- = 2-

= 0

= 2uo = o

1111d rewrite Eqs. (20) and (21) as

..
ADj

..

= -At Uo + 2uo

(2 1I

Atj

Eq. (9):

Eq. (10): Ro

..
3D 1 - -At1 D
1

11 ~nmple 5.5
o\fLuin , we shall repeat Example 5.1, using the direct linear extrapolation methods
1h1tlved in this section. For that purpose, we need the relationships m = k(T/211-)2 and
rl11 P1/m from Eqs. (5.2-a) and (5.2-b).
Starting with the average-acceleration method, we apply Eqs. (9) through (18) in
1111 first time step. When doing so, we use notation for the undamped SDOF system, as
l11llows:

6
ADj -

Direct Linear Extrapolation

+ Auo = 0 + 0.04816,; = 0.04816,;

(h)

i11 wllkh

1111d

AA/

= AAi +

Q/

Eq. (18):

+ CR/

(28)

l \quutll)IIS (~6), (27), and (28) are analogous to Eqs. ( 13) (14) and 15
dt11wd previou sly for the average-acceleration method Thus 'th
,
I ( 1'
Ii . I' . . .
.
.

, , c proceu,rc
tu
, th ,

1 Ht 1 111~ 111 cxtrnpolat 1on hy the linear accclcralion method foll

H(tp H HS h I'll' ,
,
I

OWS
C SH llll
,
t t c, tXl'cpt t 1111 hqs . (9) through (I~) nrc rc1
, lnccd by Fqs ("'JJ
lh1011,lt ( )8) .

' '.

PiT

AT
m

= 0 .04880-

p
p
p
ii., = ii. +Mo=_!. - 0.04815_!. = 0.9518_!.
0

(i)

(j)

t~11h lhllt the values in Eqs. (h), (i), and (j) are the same as those obtained in Example

~ I hy the method of iteration. This procedure is repeated for each of 20 time steps.

V11lucs of the response from direct linear cx.trapolation by the average-acceleration

11111hod 111l' listld in Table 5.5. Also given in the tublc arc slightly more accurate values
, 11111p11tld hy tlw 1111~111 lll'l'l' krntion ,m,thml.

1111 1 I N1111111h ul l11t11urntlo11 Motl1111l1

lln

I
l.q , ( 11 ),

t.\1>11

3
4

5
6
7

8
9
10

0.048 16
0. 1880
0.4061
0.68 13
0.9873
1.295
J.573
1.797
1.944
2.000

0 .04855
0 .1895
0.4091

11

l ,IJ;W

12
13

0.6861
0 .9936

14
15

1.827
1.614
1.343
1.038

1.302
1.581

16
17

0.7300
0.4478

1.803
1.947
2.000

18
19
20

0.2188
0.06500

1.019
0 .7105
0.4299
0.2047
0.05673

0.00130

0.00033

Av11 Atwl.

1.955
1.81 8

l
'/' I

~ = :/' +

Eq. (10):

Ro = 2D0 = O

13q. (14):

S=

Eq . (12):

= s[ 8.466
- 0.48

J!2 _[ 2J!l

0.01167 s

O m
(p)

2
(40)(0 .2369 JP,
ilDo = -Lit ilDo - Ro= -Ti
0.01167 -s - 0
2

I 0.04801 JP, Ti
L0.002365 m

(q)

0.2369 ) P,
Eq. ( 16): D, =Do+ L\Do = 0 + [ 0.01167 -s
=

[1

~mo

-0.48] + (4)(20)
1.64

Eq. (17):

Eq. (18):

. .

)Ar..

(s)

. [l]A

D1 = Do + ilDo =

0 -m +

( -0.07976JA
0.09459 -m

}!l

= f 0.9202
0.09459 m

)2

(t)

1lill' the values in Eqs. (r), (s) , and (t) are the same as those found previously by iteration

0 ~Jo.2s({~

- 0.48 J
9.746

= [0.04801
0.002365

(kl

(r)

.
.
.
(0.04801 J PiTi
D, = Do + ilDo = 0 + 0.002365 - m

~]

2[1

ro.2369 }!l
L0.01161

(/)

S + _ 4_ M = S + (4)(20)
0.36
- 0.48

= (40)f0.2369 J0.2Pi(!l)
T,
0.01167 m 27T

20, = 0+ 2[:J; = [~];

CM2

J'-'

0.09459 m

1.601
1.326

For the average-acceleration method in mat . f

shown. Thus, for the first time step we have nx orm, we use Eqs. (9) through ( l Kl

Eq . (9):

() ,() 11 67

= [ - 0.07976J!l

11:xnmplc 5.6
Now we will calculate the responses of the Ian
.
txtrnpolution methods. From Eqs (5 2 h) pd (; ~ss m Example 5.2 by direct lim111
nnd l>o {l, O}Pi/m.
. . . an
.2-1) we know that m = 0.2s(7; /211 )

Eq . ( 15):

(,\)( l())J\ (),) \<,I)

()11

(27T)2m

"Tabulated values to be multiphed by P,/k.

s[

t.\1>11
1

Lin . AtttI

== (4)(20) (0.2s) l0.2369

Li n.-Acccl.

Avg.-Acccl.

llN

(t.\/)

(Ill)

LlAo = ilAo + M Qo

1111\,rnmple 5.4. Following this time step, 19 other sets of calculations are performed in
1111 ~111nc manner, and Table 5 .6 gives the results.
ror the linear-acceleration method, we apply Eqs. (22) through (28) in place of
I q~. (9) through ( 15). Responses for this second analysis are also listed in Table 5. 6.
N1111 lhat all of the values in the table are practically the same as those obtained by
111111tion in Table 5.4.

= 0 + m[~ ~] [~]~ = [~JP,

l!q . ( 13):

L\Do =

S , L\Ao =

_ l_[9.746

82.28 0.48

y,

,0 .2'.\69
().() 1167 ,\'

(11)

0.48 J [ 2JA
8.466 0 ~
(o)

NEWMARK'S GENERALIZED ACCELERATION METHOD

111 this section we study a family of methods that were described by Nathan M.
NLwmark in a renowned ASCE paper. Two other variants, derived by Wilson
, I 111 . tmd Hilber et al., will also be discussed.

219

2.11

N11w111n1k'1 Un11n1nll1111I /1,111111111111011 Mnlhoil

lllr1 I N111111111t I l111nu,nllol1 Mnthocl

rhl' p11111111t'll'I /I 1111'11 I 'l 1111111111, 1hr vc1111111011 of acceleration within the
hp For this 1cu~o11 !Ill' ll'l h11h1111 IM,l'lccced to as Newmark's generalized
, , lr1111 ion method (or Ncw111111 k' s {J 111cthod). For example, if we take f3 = 0 ,

1111

Avcragc-Accclc111111111
Method

Approx. D,
I

0.2369
0.9121

3
4

5
6
7
8
9
10
11
12
13

16
17
18
19
20

1.932
3.173

0.00835
0.0762 1
0.2941
0.7 164

4.515
5.851

1.264
1.865

4.536
5.875

1.297
1.900

7.080
8.101
8.806

2.359
2.649

7. 113
8.139

2.370
2.627

2.737
2.697
2.600
2.409
1.948

8.838
9.108
8.741
7.145
4.397

2.699
2.675
2.623
2.482
2.022

10.89
14.09
16.53
17.86

I I

Approx. D2

0 .2400
0.9220
1.948
3.192

0.9392
- 3.005
-7.028
-

ApjllOX. D,

0.01167
0.08374
0.2961
0.7006

9.099
8.689
7.149
4.452

14
15

Approx . /Ji

I 1111111 Al'CCICnllion
Mr lhod

1.013
-0.4404
- 2.193
- 3.843

0.8543
-3.104
-7.143
-10.97

-5.011
-5.541

- 14.28
- 16.73

1.011
-0.5576
-2.379
-3.988
- 5.024
-5.428

- 5.544

- 18.01

5.412

u111

(3)

f3 = L

Eq. (2) yields

(4)

lhlMa prcssion is the same as that in Eq. (5.3-3) for the average-acceleration
m, r/111rl . When we take f3 = i, Eq. (2) produces
2
Uj+I = Uj + Uj Di.tj + H2iij + U1+1)(Di.tj)
(5)
1111111s case the formula is identical to Eq. (5.3-17) for the linear-acceleration
'"' 1!,11d
. let us consider a MDOF structure and cast the Newark-/3 method into
Now
111,,tii x l'ormat. In each time step we shall solve a pseudostatic problem for

11111,111cntal displacements by direct linear extrapolation , as in Sec. 5.4. For this

p111pnsc, Eq. (1) is restated in incremental matrix form, as follows:
t:i.i>j = [(l - y)Dj + yDj+ tJ t:i.tj

= oj t:i.tj +

(6)

'Y t:i.Dj t:i.tj

111 11ddition, Eq. (2) is restated in incremental matrix fonn as

1
"
"
2
Di.Di = Di !::i.tj + [( i - {3)Dj + /3Dj+1](Atj)
(7)

1 "
2
..
2
= Dj
Di.tj + 2 D/!::i.tj) + f3 t:i.Dj(t:i.tj)

In hi s 1959 paper Newmark [8]

.
intcgrut ion methods that h' ad b
.
generalized certain direct numcrirnl
een m use up to that t'
H
111 npproximaling the velocity and d' I
ime. e presented equatio1111
I fo llows
IIN
isp
acement
of
a
SDOF
system at time tJI I
"

+ [(l - y)uj + 'Yii1+ 1] !::.tj

= u1 + u1 t:i.tj + [0 - f3)u + 13 ... ](A )2
Uj

(Il

u, + t ut1
( 'l
. (or algor 'th . ) d
11 ll p11ru111etcr 'Yin Eq . (I) produ ces numerical
:t t1111c step !::.11, If 'Y is taken to be less than ,
. l .mic ar~1pi11g withi n
u 11ult1-1. On the other hand r . .
i , an arttfic1al negative dampin):
.
, 1 'Y is g1eater than ! such th
..
uvoll 1 n111mncul damping 1llogethcr ti
I
'. .
'mpmg ,s pos1t1vc. To
(I) l>tto111ts lhc trnpczoid~d rule. , 1c vu uc ol 'Y must he equal to j; und Hq

'l'I

becomes

II pl11l'cmcnt.
II we let

Newmerk-/J Method

1)

I hi l111111ula is known as the constant-acceleration method, because the accel1,1111111 fl at the beginning of the time step Di.ti is taken to be constant within the
I
h I' 11.quation
(3) also corresponds to a truncated Taylor series that results from
11 1111' Huler's formula [see Eq. (5.3-10)] for velocity and the trapezoidal rule for

"Tabulated values to be multiplied by P,/ s.

Ujt I

'

'-i11lving for Di.Di in Eq. (7) produces

..

ADj

= f3 (Ati )2 t:i.Dj - .f3

',111l11tilution of Eq. (8) into Eq. (6) yields

t,:01 = ~ ~

11

6.01 -

7i 01 -

1 .
1 ..
Ati Dj - 2/3 Di

u~ -

(8)

1) t.t f>

1 1

(9)

hn rnnvcnicncc, we define the vectors

(10)

221

Direct Numerical Integration Methods

220

Rj =

Sec. 5.5

j 1\ + ( 2~ - 1) Ati f>i

Newmark's Generalized Acceleration Method

ii

Now rewrite Eqs. (8) and (9) in the forms

ADi =

/3 (Ati )2 ADi -

'
ADi =

/3

(1 2)

Qi

'Y
Ati ADi - Ri
A

(13)

Then substitute Eqs. (12) and (13) into the incremental equations of motion,
given previously as Eq. (5.4-3); and collect terms to obtain

S ADi =

AA.i

(14)

in which
1

= S + /3 (At)2 M + /3

LM,_JI;+, r

'Y

(15)

Ati C

~M,~

and
(16)

We solve the pseudostatic problem in Eq. (14) for the incremental displacements
ADi and substitute them into Eqs. p2) and (13) to find the incremental acceler
ations and velocities ADi and ADi. Then the total values of displacements,
velocities , and accelerations at time ti+ I are determined using Eqs. (5.4-16),
(5.4-17), and (5.4-18).

Figure 5.5

We solve Eq. (18) for AD 8 to obtain

..
6
6 . - .. .
AD = - A D e - - Di 3D1
e (At8 )2
Ate
.

ADe = -

Wilson et al. [9] , extended the linear-acceleration method in a manner that

makes it numerically stable. The basic assumption of the Wilson-0 approach is
that the acceleration ii varies linearly over an extended time step Ate = 0 Atj, as
depicted in Fig. 5.5. During that time step the incremental acceleration is
Aue = 0 Aui. It has been shown [ 10] that the optimum value of the parameter
0 is 1.420815, which can be rounded to 1.42.
As with the Newmark-,B method, we shall construct the Wilson-0 variant
in an incremental matrix format applicable to a MDOF structure. From Eq.
(5.4-4) for the linear-acceleration technique, we have the incremental velocities
at the end of the extended time step At0 , as follows:
,

..

..

AD o = Di Ato + ~ AD o Ato

( 17)

where the symbol AD 0 denotes a vector of incremental accelerations. Simi larly,

Eq . (5.4- 19) gives the incrlmentul displucellll'llls us

I ..

D1 At11 I 21>1(A111)

I ,.Al>11(A111l

( 18)

(19)

.
. t Eq (17) which results in
l'hen substitute this expression m o
.
'

Wilson-0 Method

Al>11

Linear-acceleration method extended to Wilson-6 method .

Ate

Now define

Q*
e

ADe - 3Di -

= ~ i).1 + 3:0j

(20)

(21)

b.te

.
Rt= 3Di
11 11d

Ate ..
Di

b.te ..
Di

(22)

restate Eqs. (19) and (20) as

6 AD
..
Q*
ADe = (b.te)2
ee

3
AOo - R!
b.to
.
( c; 4 1) nnd collect terms to find
'i11hHt11utl' l~qs. (23) 11ml (24) into 1~q.

st M>,1 AAt

ADo -

(23)
(24)

(25)

'l'J'J.
'l'hl

ll!t It N11111n1 h ,11 l1111111111tlo11 Mnthod11

st11111.! d ll!1111s in this

1 It ll

l''lll1tfhtt1 1111

st

~
,"l

N11111111 le 111

223

t.tnhlllty 111111 /\1, 111 111 y

n NUMERICAL STABILITY ANO ACCURACY

(1
A

(ul,i)

I 11 ~tucly the stability and accuracy of various one-step direct numerical

11111 p,111tio n procedures, we may cast them into

und

U1+1 = A U1 + L
(J. I)

where

AAo =

e AA1

(.'HI

The
problem in
so1ved ior
.: the mcremental
.
AD pseudostatic
Tl
. Eq . (25)
. is
disp1acenw111 II
1en we ca~ determme the mcremental accelerations AD 8 from Eq (2 II

HiltI reduce them lmear1y by the formula

..
ADJ

I 'I (I) represents a column vector containing the three response quantities u1,
,,, 1111d

a1 at

the time station t1. That is,

U1 = {uj,
111111

Next, the incremental velocities AD1. and displacements AD are obt d f

Pei (5 4 4)
d (5
L.l J
ll.me ro11 1
s.
- an
.4-19) for the time step At. As before Eqs (5 4-1()
( '\ .4 17), and (5 .4-18) give total values of the dis I
'

l '
lll'l'l'lcrntions at t'
t
p acements, veloc1t1es, and
1me 1+ 1 .

u1, u1}

(2)

the vector U1+ 1 is similarly defined at time tj+ t to be

U1+1 =
(2'))

(1)

Pj+I

1ltlNl'xpression pertains to a SOOP system that might equally well be considered

, 1111c natural mode of vibration for a MDOF structure. The symbol U1 in

..

= -AD
e o

operator form [12], as follows:

{u1+1, u1+ 1,

(3)

ii1+1}

1111' toefficient matrix A in Eq. (1) is a 3 x 3 array called the amplification

1t1tt11/, that we shall examine to answer questions about stability and accuracy.

I l1111lly, the symbol L denotes a column vector called the load operator, which
I 11111ltiplied by the load lj+ 1 at time t1+ i If there is no loading, Eq. (1) simpli111 H to

(4)
Hllber-a Method

lt11 free-vibrational response.

. To imp~ove control of numerical damping, Hilber et al. [11 ] introduced

p111 amctcr a into the equations of motion at time t
.: IIows..
J+ l , as 10

MD1+1 + C D,+1
. + (1 + a)S D1+1 -

aS D1

= A,+ i

11

(30)

~11btracting similar equations of motion at time t from Eq. (30) prod

ti .
tru.:rnmcntal equations
,
uces 1(

M AD1 + C AD1 + (1 + a)S AD1 - aS AD,- ,= AA1

(3 1)

Ncl>lw substitute Eqs. (12) and (13) from Newmark's method into Eq (31) and
cct terms to find

l'O

To investigate the stability of a numerical algorithm, we apply spectral

,l,,0111position [13] to the amplification matrix A, as follows:
(5)

ltt this equation A is the spectral matrix of A, containing eigenvalues A1, A2 , and
\
In diagonal positions; and <I> is the 3 x 3 modal matrix of A, with
1
1 l~lnvcctors <1> 1, <1>2 , and <1> 3 listed columnwise. If we start at time t0 == 0 and
l ll~ l' 11 time steps using Eq. (4) , we have
1
Vn = AnjUo
(6)
1

wlwrc the vector U0 contains initial conditions, and vector Un gives the response
)

Sa ADJ= AA.a}

(32)

111 which

v11 l11cs at time t,.J' Raising the decomposed form of matrix A in Eq. (5) to the
powe r n1 yields
(7)

(33)
N11w let us define the spectral radius of matrix A as

1111d

,.,

,.,

AA<\i = AA1 + aS AD1

(J4 )

Expressi<~ns for S and A~ were derived previously us Eqs. ( 15) and ( 16) F .
lhl llrst time step, we luke A Di 1 _
in Eq . ( ,,I).
or

(r)A

= max IA,\

(i

= 1, 2, 3)

(8)

I hl'II Eq . (7) shows that we must hnvc

(1 )A

(9)

224

1 11c.

Direct Numerical Integration Methods

in o~d~r to keep the numerical solution from row

.
,
c~nd1_hon is known as the stability criterion for a !ven:!t:~h;t boun~. Th~H
~p~ym~ ~111
~ntenon to the constant acceleration method (see Sec 5 5) .
time step for this conditionally stable approach to be. . ' we n t e cntical

(At)c,

= _!!T.1T = 0.318Tn

(J 0)

This value is the same as that given b E (S 2

.
.
Y q.
-12) for the central-difference
predictor For th r
1
.
e mear-acce erat1on method, the critical time step is

(At)c,

= V3 '7T
[;_ = 0.551T.,,

(JI l

On the other hand the spectral d' f; h

always unity. The;efore it has ra IU~f o~ t_e average-ac~eler~tion method is
dilionally stable.
'
no en ica time step and is said to be uncon
The matter of accuracy of a n
1
.
re lated to that of stability Figure 5 6 u:enc\ mteration procedure is closely
system to an initial dis la~e
. s ows t e un amped response of a SDOJ
Ihose labeled 2 3
~ 4 ment uo. The_ curve labeled 1 is the exact result, and
' ' an
represent various approximations Curve 2 d
Nlrat<.:s an amplitude increase (Al) that r
.
emon
shows no amplitude change and curve
ie~ an unstabl~ algonthm. Curve .l
Utcuuse of the stabTt
-' . .
epicts an amplitude decrease (AD)
1 1

Y en 1enon m Eq (9) only curves of t

3 d

ndlllissiblc approximations Curve 3

b '
.
ypes an 4 arc
uvcrugc-acceleration meth;d h. h hmay e cons1der~d to be _the result for the
siblc algorithms are represen~e: b1yc cu~/;pech~ahl ~adml_s of umty. Other_admis
' w IC imp 1es a spectral radms Jess

i:~

LJ

--.. PE

'

Al

AD

1111(1111

~.<,

S 111h1l11 y 1111d

un111m y 111 11111111,111111 lr11tp1111i1111 llll'lhrnlN

5.7

Program DYNA for Dynamic Response

225

lhnn unity. Thus, one important type of error to be considered is amplitude

11tppression, as exhibited by curve 4.
All of the approximate responses in Fig. 5.6 also show period elongation
l PE), which is a second type of error introduced by any numerical algorithm.
lloth the amplitude suppression and the period elongation may be made negli1,11ble by using sufficiently small time steps. Newmark [8] recommended a time
M
l<.:p of duration equal to! or of Tn, which is the smallest period of a MDOF
Nltucture. However, a more commonly used time step is lit= Tn/10.
At first glance, the average-acceleration method appears to be the best
1 hoice among the implicit approaches, because it has no amplitude suppression
1111d the least period elongation [12]. However, a small amount of amplitude
~uppression is desirable to reduce or eliminate unwanted responses of higher
111odes in an undamped MDOF structure. But if the spectral radius of an
11111plification matrix is too small, the response of the structure will be unduly
~uppressed, as in the Houbolt method [14]. Although the optimized Wilson-8
lmmulation produces a reasonable level of amplitude suppression, it also exlllhits an undesirable tendency to overshoot the true response in the first few time
~1,ps [15]. Thus, we conclude that the best choice of algorithm is the New11111rk-/3 method with the Hilber-a modification, as described in Sec. 5 .5 . Probahl y, the optimum selection of parameters for this approach is to let a = -0.1,
/I 0.3025, and 'Y = 0.6.

,7 PROGRAM DYNA FOR DYNAMIC RESPONSE

Now we shall describe a general-purpose program named DYNA that calculates

dy11umic responses of linearly elastic structures. This program includes vi111111 ional analysis from Chapter 3, normal-mode responses from Chapter 4, and
dl1l'Ct numerical integration from the present chapter. By virtue of a branch in
1111 main program, responses may be obtained using either the normal-mode
11111hod or direct numerical integration. For the latter approach, we select the
Ntwmark-/3 method with the Hilber-a variant discussed in Secs. 5.5 and 5.6.
Flowchart 5.1 shows the main progrnm for DYNA, which calls the five
, 111,programs indicated in double boxes. As in Flowchart 4.1, Subprogram VJB
11111s the seven other subprograms given in Flowchart 3.1, including nor111111izution of the eigenvectors with respect to the mass matrix. Then the program
1, 11tls the number of loading systems NLS and checks it against zero to determine
wlwlh<.:r to process a loading system or another structure. Next, the loading
1111111hcr LN is initialized to zero, and then it is increased by one.
'l'h<.: s<.:cond subprogram to be called is DYLO, which is almost the same
,, 111 Flowe ha, I 4 . I. I lowevcr, th<.: first dynamic parameter now becomes the
l111ll111lor ISOLVE for type of sol11lio11 . The two choices of solution are the
11111111111 111e111l' 1tll'lhod (ISOLVE
0) 1111d dirl'L'l numcricnl integration

M11w1'11111 I 'I, I

1\,111111 1111111111111 t111 U\' NA

227

~. I

IOI
V IIJ1

1'11111111111

VIII

11 11111

S1'1' I H,

JIiii , lllllllllllilllllllll Oi Cl)IC ll Vl'('l(ll N

II 111 1 1,,1w1'I Iii 11111,s IIIHIII X.

R1ud 111111,hcr of loading syslcms.

If NLS = O, go to IOI and process

another structure.

Initialize loading number to zero.

Increment loading number by one.

2. Read and write dynamic load

data, including indicator !SOLVE
for type of solulion.

lf!SOLVE = l ,goto 103.

3 Calculate dynamic responses

by nom1al-mode method.

Go to 104.

4 Calculate dynamic responses by

direct numerical integration.

5 Write and/or plo1 results

of response calculations.

Check lo r las, loading sysle m.

Go to IO I nncl process a11o 1hc,

s 1ruc 111rc.

l !ND
'~lll'lll'S 111 1111y IYl'l' ol l111111oly 1'111M 11 ,,11111'lll ll'
IS11 hp111~111111~ 111111 dilll1 1111 l'Vl'IY ' Yl'l' 111 , 1111l lilll'

I) . SubNt'qlll'lltl y. 1lll' p111g1 11111 l'tllllpt1r0s !SOLVE with unity to

I , hh whk h ly pl! of solulion prnnd111t is to bi! used for dynamic response

,h 11l111ions. II' ISOLYE I- I , Subprngrum NORMOD for the normal-mode

111, 1h,,d in turn calls Subprograms TRANOR , TIHIST, and TRABAC, which are
1111111hl'1\d 3, 4, and 5 in Flowchart 4. I . Alternatively , if ISOLVE = 1, Sub1111111111111 NUMINT computes dynamic responses by direct numerical integra111111 For this purpose , it reads and writes the integration parameters ALPHA,
Ill I A, und GAMMA required in the Hilber-a formulas.
Within Subprogram NUMINT it is also necessary to generate the damping
111,111 Ix. C to be used in the response calculations. Toward this end, we substitute
lh1 d11mping ratio DAMPR and the results from Subprogram VIB into Eq.
1I,, 17) to create the damping matrix in physical coordinates.
The final subprogram called by the main program is RES2, which writes
111111/rn plots selectively the results of response calculations as before. Testing LN
1111111111! NLS at the end of the flowchart determines whether to return for another
li 1111ll11g system or another structure.
Program DYNA may be specialized to become DYNACB for continuous
111 1111111, DYNAPT for plane trusses, and so on. As before, the main program for
, ,11h specialization has subprograms that are different for each type of applicallt 1111 us indicated by the second footnote below Flowchart 5.1. Note that Subp111grnms NORMOD and NUMINT are the same for all types of structures. As
1111 l'mgram NOMO, notation for Program DYNA is included in Part 5 of the
11-1 of' notation near the back of the book. Details of the logic in Subprogram
NI IMINT appear in the flowchart for Program DYNAPT, given in Appendix C.
Thus, in several stages we have devised a program that will handle not only
~ll11111ional analysis but also two types of dynamic response calculations. Vari1111111 of the program apply to all linearly elastic framed structures and discretized
, 1111llnua discussed in this book.
I,111111,tc 5.7
I 11 Nhow how Program DYNAPT is used, we shall repeat Example 4.15 (the three1111111hcr plane truss), using the same number and size of time steps. In this application
~, 1,i11ploy direct numerical integration instead of the normal-mode method by setting the
p111 11111l1tcr ISOLVE equal to unity. Also, the values of the integration parameters read
hy S11hprogram NUMINT are taken to be ALPHA= - 0.1, BETA= 0.3025, and
t 1/\MMA

= 0.6 .

I ,ct us examine again the responses of the structure to the dynamic influences given
11111,.,nmplc 4. J 5. Table 5.7 lists part of the output from Program DYNAPT for these four
,11111lys1s. figures 5.7 and 5.8 (on pages 233 and 234) also show computer plots of the
11 ~1,IIN ohJained . The responses in Fig. 5.7 due to initial conditions follow the same
11111r111I lt'l!nds us those in Fig. 4. 11 , but they differ because the time step is too large.
1111 1hr nlhcr hand, !he smoother responses in Fig. 5 .8(b) caused by the applied force in
111 't .K(u) mc prnctically indistinguishuhlc from those in Fig. 4.12(b). This good
111111l111lon INdm to thl' facl that the ti11w s11p Is short enough to model the slowly varying
II ~plllllll 'N

lll'l'llnlll'ly,

228
Direct Numerical Integration Methods

Program DYNA for Dynamic Response

TABLE S.7 Computer Output for Example S.7

229

TABLE S.7 (Continued)

PROGRAM DYNAPT
*** EXAMPLE 5.7: ~HREE-MEMBER PLANE TRUSS***
STRUCTURAL PARAMETERS
NN
NE NRN
E
3
3
2 3.0000E+04
NODAL COORDINATES
NODE
X
l
0.000
2
150.000
3
150.000

GROUND ACCELERATIONS
IGA
EL
250.0000
150.0000
200.0000

ex

CY
0.6000 0.8000
1.0000 0.0000
0. 0000 -1. 0000

DIRECT NUMERICAL INTEGRATION

ALPHA= -0 . 1000 BETA
0.3025

GAMMA

OUTPUT SELECTION
IWR IPL NNO NEO
1
1
2
1

NUMBER OF DEGREES OF FREEDOM

NUMBER OF NODAL RESTRAINTS:

NDF
NNR

STIFFNESS MATRIX DECOMPOSED

MODE
1
ANGULAR FREQUENCY 4.1995E+02
NODE
DJl
DJ2
l
2.3137E-Ol O.OOOOE +OO
2 1. OOOOE+OO -2.4722E-Ol
3 0.0000E+OO O.OOOOE+OO
l.1677E+03
DJ2
O. OOOOE+OO
l .OOO OE+OO
0.0000E+OO

MODE
3
ANGULAR FREQUENCY l.8 618E+03
NODE
DJl
DJ 2
l
l.OOOOE+OO O.OOOOE+OO
2 -6 .0504E-Ol -6.1068E-Ol
3 O.OOOOE+OO O.OOOOE+OO
LOADING NUMBER

APPLIED ACTIONS
NLN NEL

0.000
200 .0 00
0.000

NODAL RESTRAINTS
NODE NRl NR2
l
O
l
3
l
l

MODE
2
ANGULAR FREQUENCY
NODE
DJ l
1 8.6725E- Ol
2 -L 7149E-Ol
3 0.0000 E+O O

INITIAL DISPLACEMENTS
NODE
DOl
D02
1 l.OOOOE-01 O.OOOOE+OO
2 l.OOOOE-01 l . OOOOE-01

RHO
7.3500E-07

ELEMENT I NFORMATI ON
ELEM.
J
K
AX
l
l
2
10.0000
2
l
3
6.0000
3
2
3
8.0000

INITIAL CONDITIONS
NNID NNIV

l OF

DYNAMIC PARAMETERS
JSOLVE NTS
DT
l
20 l.OOOOE - 03

4 ***

DAMPR
O.OOOOE+OO

3
3

NODES:

ELEMENTS:

DISPLACEMENT TIME
STEP
TIME
0 O.OOOOE+OO
1 l.OOOOE-03
2 2.0000E-03
3 3.0000E-03
4 4.0000E-03
5 5.0000E-03
6 6.0000E-03
7 7.0000E-03
8 8 . 0000E-03
9 9 .000 0E-03
10 l . OOOOE-02
11 l . lOOOE-02
12 l .2000E-02
13 l .3000E-02
14 l.4000E-02
15 l.5000E-02
16 l.6000E-02
17 l.7000E-02
18 l.8000E-02
19 l.9000E-02
20 2.0000E-02

HISTORY FOR NODE

1
DJl
DJ2
l.OOOOE-01 O.O OOOE+OO
6.9151E-02 O.OOOOE+OO
-9. 7574E-03 O. OOOOE+OO
-8. 0921E- 02 O.OOOOE+OO
-6 . 9989E-02 O. OOOOE+OO
2.1718E-02 O. OOOOE+OO
8.8292E-02 O.OOOOE+OO
3.9844E-02 O. OOOOE+OO
-7.5170E-02 O.OOOOE+OO
-l.2500E-Ol O.OOOOE+OO
-5.7660E-02 O.OOOOE+OO
4.7566E-02 O.OOOOE+OO
8 . 9567E-02 O.OOOOE+OO
5.1576E-02 O.OOOOE+OO
-l . 2272E-02 O.OOOOE+OO
-4 . 6102E-Oi O.OOOOE+OO
-2.5622E-02 O.OOOOE+OO
3.6195E-02 O.OOOOE+OO
8.5913E-02 O.OOOOE+OO
6 . 2893E-02 O.OOOOE+OO
-2.9158E-02 O.OOOOE+OO

MAXIMUM
TIME OF MAXIMUM
MINIMUM
TIME OF MINIMUM

l.OOOOE-01
O.OOOOE+OO
-l.2500E-Ol
9.0000E-03

O. OOOOE+OO
2.0000E-02
O.OOOOE+OO
2.0000E-02

0.6000

230

Direct Numerical Integration Methods

TABLE 5.7 (Continued)

TABLE 5.7 (Continued)

DISPLACEMENT TIME
STEP
TIME
0 O.OOOOE+OO
1 l.OOOOE-03
2 2.0000E-03
3 3.0000E-03
4 4.0000E-03
5 5.0000E-03
6 6.0000E-03
7 7 ,0 000E-03
8 8 . 0000E-03
9 9.0000E-03
10 l.OOOOE-02
11 l . lOOOE-02
12 l.2000E-02
13 l . 3000E-02
14 l.4000E-02
15 l,5000E-02
16 l . 6000E-02
17 l.7000E-02
18 l.BOOOE-02
19 l.9000E-02
20 2.0000E-02

HISTORY FOR
DJl
l.OOOOE-01
8.8727E-02
6.8467E-02
4,9368E-02
l.2394E-02
-5,2366E-02
-l,1040E-01
-l,2015E-01
-8,9808E-02
-6,2312E-02
-5.2629E-02
-3,4703E-02
9.6841E-03
6,2708E-02
9 .8915E-02
l.1326E-01
l.0852E-01
8.0000E-02
3.1105E-02
-l,48 6BE-02
-3 ,9867E-02

MAXIMUM
TIME OF MAXIMUM
MINIMUM
TIME OF MINIMUM

l . 1326E-01 l . 1863E-Ol
l.5000E-02 6.0000E-03
-l.2015E-01 -l,3161E-Ol
7.0000E-03 l.5000E-02

MEMBER
STEP
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

HISTORY FOR ELEMENT

AMl
9 . 6000E+Ol
4.6852E+Ol
-2.5370E+Ol
-2,4355E+Ol
1. 6683E+Ol
1. 2893E+Ol
-2,9174E+Ol
-4.2452E+Ol
-2,3751E+Ol
-2.4867E+Ol
-4 .330 1E+Ol
-2. 3119E+Ol
3.4926E+Ol
5.8727E+Ol
2.1256E+Ol
-l.159BE+Ol
l.3378E+Ol
5.3656E+Ol
4.3667E+Ol
-4.5712E+OO
-3. 2732E+Ol

FORCE TIME
TIME
O.OOOOE+OO
l.OOOOE-03
2.0000E-03
3.0000E-03
4,0000E-03
5.0000E-03
6.0000E-03
7.0000E-03
8.0000E-03
9.0000E-03
l.OOOOE-02
l. lOOOE-02
l.2000E-02
l . 3000E-02
l,4000E-02
l . 5000E-02
1. 6000E-02
l,7000E-02
l.BOOOE-02
l.9000E- 02
2.0000E-02

MAXIMUM
TIME OF MAXIMUM
MINIMUM
TIME OF MINIMUM

9.6000E+Ol
O.OOOOE+OO
- 4.3301E+Ol
l.OOOOE - 02

231

Program DYNA for Dynamic Response

***

NODE

2
DJ2
l.OOOOE-01
3.4123E-02
-B.5095E-02
-1. 2309E-Ol
-4.4408E-02
6.8993E-02
l.1863E-01
7.5778E-02
-l,3763E-02
-7 ,2918E-02
-4,8878E-02
3.7619E-02
9.6294E-02
5.2826E-02
-6.1248E-02
-l.3161E-01
-B.6674E-02
2.3038E-02
8.6592E-02
5.3559E-02
-2.6064E-02

LOADING NUMBER

2 OF

DYNAMIC PARAMETERS
!SOLVENTS
DT
1
20 l.OOOOE-03

***

DAMPR
O.OOOOE+OO

INITIAL CONDITIONS
NNID NNIV

APPLIED ACTIONS
NLN NEL
1
0
NODAL LOADS
NODE
AJl
2 2.0000E+Ol

AJ2
O.OOOOE+OO

GROUND ACCELERATIONS
!GA
0

FORCING FUNCTION
NFO
7

FUNCTION ORDINATES
TIME
K
1 O.OOOOE+OO
2 1. OOOOE-03
3 3.0000E-03
4 4,0000E-03
5 5.0000E-03
6 7.0000E-03
7 8.0000E-03

FACTOR
O.OOOOE+OO
1. 5000E-01
8.5000E-01
1. OOOOE+OO
B,5000E-01
l,5000E-01
O,OOOOE+OO

DIRECT NUMERICAL INTEGRATION

ALPHA= -0 .1000 BETA
0,3025

GAMMA

0.6000

OUTPUT SELECTION
!WR IPL NNO NEO
1
1
1
1
NODES:

ELEMENTS:

2
DISPLACEMENT TIME HISTORY FOR NODE
DJl
DJ2
STEP
TIME
0 O.OOOOE+OO O,OOOOE+OO O.OOOOE+OO
1 1. OOOOE-03 B,4579E-04 -l.OOOBE-04
2 2,00 0 0E-03 5.5055E-03 -8.5533E-04
3 3.0000E-03 l,7942E-02 -3,4897E-03
4 4.00 0 0E-03 3.8969E-02 -8,9831E-03
5 5.0000E-03 6,3694E-02 -l,6261E-02
6 6.0000E-03 8.4309E- 02 -2,1959E-02
7 7.0000E-03 9.4437E-02 -2 ,3226E-02
8 O.OOOOE- 03 9.0 37 4E- 02 -2. 0476E-02
9 q,OOOOF. 03 7.0970E 02 l.6052E-02
1,07/ bE 01 1. 0521E-02
10 l.0000~ 02

232

Direct Numerical Integration Methods

TABLE 5.7 (Continued)

11
12
13
14
15
16
17
18
19
20

l.lOOOE-02
l. 2000E-02
l . 3000E-02
l.4000E-02
l.SOOOE-02
l . 6000E-02
l.7000E-02
l. SOOOE-02
l.9000E-02
2 . 0000E-02

9.4437E-02 2 . 3972E-02
7 . 0000E-03 l.4000E-0 2
-9 . 568BE-0 2 -2 . 3226E-02
l . SOOOE- 02 7 . 0000E- 03

MEMBER
STEP
0
1
2
3
4
5
6
7
B
9
10
11
12
13
14
15
16
17
18
19
20

HISTORY FOR ELEMENT

AMl
O. OOOOE+OO
6 . 1168E-Ol
3 . 4981E+OO
9 . 63 72E+OO
l. 7047E+Ol
2 . 2466E+O l
2 . 5649E+Ol
2 . 8511E+Ol
2 . 9596E+ Ol
2.4238E+Ol
l.1821E+Ol
-1. 9340E+OO
-l.2285E+Ol
-2 . 0281E+Ol
-2 . 7599E+Ol
-3 . 1270E+Ol
-2 .7 564E+Ol
-1. 7935E+O l
-6.9027E+OO
3 . 9111E+OO
l.5351E+Ol

FORCE TIME
TIME
O. OOOOE+OO
l. OOO OE-03
2 . 0000E- 03
3.0000E-03
4.0000E-03
5 . 0000E-03
6 . 0000E-03
7 . 000 0E-03
B. OOO OE-03
9 . 0000E-03
l . OOOOE-02
l.lOOOE-02
l . 2000E-02
l . 3000E-02
l . 4000E-02
l . SOOOE-02
l . 6000E-02
l . 7000E-02
l. BOOOE-02
l.9000E-02
2 . 0000E-02

MAXIMUM
TIME OF MAXIMUM
MINIMUM
TIME OF MINIMUM

DA'1PR=O.O

l . 3673E-04 - 2.3257E-03
-3 . 7755E-02 8 . 5393E-03
-6 . 9338E-02 l . 8614E-02
-9 . 0054E- 02 2 . 3972E-02
-9.5688E-02 2. 3788E-02
-8 . 4382E-02 l . 9802E-02
-5 . 8635E-02 l . 354 2E-02
-2.3925E-0 2 5 . 5924E-03
l.4150E-02 - 3. 2004E-03
5 . 0421E-02 -l.1543E-02

MAXIMUM
TIME OF MAXIMUM
MINIMUM
TIME OF MINIMUM

233

Program DYNA for Dynamic Response

Soc. 5.7

0. 10

- o.os
~ o.oo
0

-o . 05

1-

<

, -o.
0

( a)

2.9596E+Ol
8 . 0000E-03
-3.1270E+Ol
l .SOOOE- 02

0.10
C

- o.os
N

~ o.oo
0

I-

<

-o .os

, -o .
0

(b)

DAMPR=O . O
-

0 . 10
C

- o.os
N

~ o.oo

l r.101111l)IC

5.8

N1x1, we repeat Example 4. 16 (the symmetric pl ane truss), with the same number 1111d
time steps. llowcvcr , the first load case in that example will be treated diffcrcnl ly
l11Nll11d ol' using modal truncarion , we shall compare results from the Ncwmark-/3 method
111111h1sl those obtained by the l lilbcr-a 1cchniquc. In the lirsl instance, we h11vc Oil'
i111q~rntion p11r11 1111i tcrs ALP! IA
0, BETA
0.25, und OA MM A
0.5; und in !hr
Nt1t'1>11tl wt t11k1 ALPHA
0 . 1, BETA
0. 102'>, n11d ClAMMA
0.6.
Ni11 of

,_ -o.os
<

ci " 0, 10

<c 1
11'IJ\ U1'l'

~.7 'l'l!i,t

1111, 111 tw1 pl 1111l' 11 u NN: H'Nl)lHl NCN 10

inilinl displacements.

234

z 100

20

80

"'C.
-

235

Program DYNA for Dynamic Response

Sec. 5.7

Direct Numerical Integration Methods

10

w 60
D
0

(\/

0
0

10

15

z
-;

20

<(

20
(ms

I--

<(

40

I--

w O

(\/

~ 0

so

I 00

so

200

(a )

(ms)

(ms

-,
<(

-20

NEIIMARK
~ 4

(a)

DAMPR =O.O
Q. J 0

'::;: I
(\/

-, 0

_': 0 OS

JOO

( b)

200

(\/

~ o. oo
0
z
E 4

I--

< - o.os

..,

lD

NE I/MARK
3

8z 2

- 0. I 0

~l
(\/

( b)

Figure 5.8 Three-member plane truss: (a) applied force; (b) responses.

Figures 5.9(b) and (c) show computer plots of the symmetric and anti-symmcl r rt
l'lllllributions to they translation at node 6 for the Newmark and Hilber methods. Thcs1
H'Nponscs should be the same as those for all modes given previously in Figs. 4. 1J(hl
1111d (l') . They all mulch fairly well , even though the lime step is rather long . Nole th111
lhl 11pprniti111011 rlsponscs in Fig. 5.9(c) huvc II signilirnnl f)Ni()(/ 1'lo111,1ar/011 rclutiw lo
lhut i11 Fjg, 4. 1 l(l') , Also, llw llilhlr plol in Jllg , 'i .')(c) hus II rmlkl11hk (//ll/1llr11tl1
1111i1111.1.1'io11 rd111iv, tu llw N1w111111k plot , 11s it should

Figure 5.9 Plane truss: (a) applied force; (b) symmetric responses; (c) antisymmetric responses.

Responses to the piecewise-linear ground acceleration in Fig. 5.lO~a) .are/~ot!~d

I11 Fi 5 , O(b.) i-or the two cases of DAM PR = 0.0 and 0.1. They were o tame y e
!:I n tll)JWOttch using the same .intcgrn11on

I lilhcr
parameters as before .hThese plots
l are
d

'
,
J"
4
14(1
)
hich
were
found
by
t
e
norma
-mo e
li11IINti11guiNht1hlc from those 111 tg. , 1 , w
11111(hotl ,

23 1
2311

1111 ,11 1 N1111111lt11l li1ln11111tlm1 M11thod11

...

I , y, S W , uml IkiNiiW'' , Z I\ , "' I 11111NH'111 I >y1111111ic Analysis of Thin Shells by

1111 1111111~ 1ih;n w11t Method ," l'11w . ltd C'CII(/ Mal. Methods Struct. Mech., AFIT,
Wlifr,hl Pulh.irson Al<B, Ohio, 197 1, pp. '17'J 5 18 .
I 1h'f\, R. I) ., and Key, S. W., "Transient Shell Response by Numerical Time
l111t fllltlion," In!. J. Numer. Methods E11g ., Vol. 7, No. 3, 1973, pp. 273-286.
I , y, S. W. , "A Finite Element Procedure for the Large Deformation Dynamic
11, Nponse of Axisymmetric Solids," Comp. Methods Appl. Mech. Eng., Vol. 4,
t~11 2, 1974, pp. 195-218.
I, f\1111 !110, L., Leech, J. W , and Witmer, E. A., "Optimal Predictor-Corrector Method
1111 Systems of Second-Order Differential Equations," AJAA J., Vol. 12, No. 10,
111/4, pp. 1343-1347.
' l l111oshcnko, S. P. , Young, D. H ., and Weaver, W., Jr., Vibration Problems in
I 1111i11eering, 4th ed., Wiley, New York, 1974.
N1wmark, N. M. , "A Method of Computation for Structural Dynamics," ASCE J.
/1111. Mech. Div. , Vol. 85, No. EM3, 1959, pp. 67- 94.
'' WI ison, E. L., Farhoomand, I., and Bathe, K. J. , "Nonlinear Dynamic Analysis of
1'0111plcx Structures," Earthquake Eng. Struct. Dyn., Vol. 1, No. 3, 1973, pp.
1,11 252.
111 llnthe, K. J., and Wilson, E . L., "Stability and Accuracy Analysis of Direct Integratl1111 Methods," Earthquake Eng. Struct. Dyn., Vol. 1, No. 3, 1973, pp. 283-291.
llllbcr, H. M., Hughes, T. J. R., and Taylor, R. L. , "Improved Numerical DissiII
pulion for Time Integration Algorithms in Structural Mechanics," Earthquake Eng.
,\'Imel. Dyn. , Vol. 5, No. 3, 1977, pp. 283-292 .
llnthc , K. J., Finite Element Procedures in Engineering Analysis , Prentice-Hall,
I'
l~n~lewood Cliffs, N. J. , 1982.
11 lll1rc, J. M., and Weaver, W., Jr., Matrix Algebra for Engineers, 2nd ed.,
ll1 ooks/Cole, Monterey, Calif., 1983.
I loubolt, J. C. , "A Recurrence Matrix Solution for the Dynamic Response of Elastic
11
Aircraft," J. Aero. Sci. , Vol. 17, 1950, pp. 540- 550.
llllber, H. H., and Hughes, T. J. R., "Collocation, Dissipation, and 'Overshoot' for
'l'l 111e Integration Schemes in Structural Dynamics ," Earthquake Eng. Struct. Dyn. ,
Vol. 6, No. 1, 1978, pp. 99- 117.

I 00

('l

"

-'
l

o. so

....
<C

n::

w o.oo

...J

(.)
(.)

<C
0

-o.so

::i
0

n::

'f'

- 1 . 00

>-

(a )

I . 00

o.so

o.oo

....

< -o.so

..,
N

,,

Cl

- 1 . 00

( b)

PROBLEMS*

Figure 5.10 Pla ne t russ : ( a) ground acceleration; (b) responses.

REFERENCES
I. Dnhlqulst, G. G., "A Special Stabilit Pr
Nord. 'l'lrl.1kr. !11.f. Bl'hw1d/i11g Vol 3 YI 9;1blcm ~2or
). K i
,
'

, pp. 7

rill,\ , I,. I)., " Unconditional Stnbilit N

.I. Ap11!, M1c lt Vo l 110 N11 2 I 117; 111 111111,111111
''

, '/ ., , pp. I I I ,111

L111car Multi step Methods,"

43
,:

1111w ln11;11n1tlon MulhodN"

'

I. I. Confirm the approximate results of Example 5 .1 in Table 5 .1.

12. Repeat Prob. 2.7-2 using the central-difference predictor.

1..\. Repeat Prob. 2.7-3 using the central-difference predictor.
Repeal Prob. 2. 7-4 using the central-difference predictor.
i-~. Ri;peat Prob. 2.7-5 using the central-difference predictor.
l (1, Repeat Pmb . 2.7-6 using the central-difference predictor.
+ Solutions for pmhlc111s in this ch11p1c1 til l' , 111hcr tedious and should be handled using a
jt 1-1111111 Vlllllplltl'I,

238

Direct Numerical Integration Methods

5.2-7.
5.2-8.
5.2-9.
5.2-10.

Repeat Prob.
Repeat Prob.
Repeat Prob.
Repeat Prob.

2.7-7 using the central-difference predictor.

2.7-8 using the central-difference predictor.
2.7-9 using the central-difference predictor.
2.7-10 using the central-difference predictor.

l111p. 5

~.4-5.

5.2-13. Re~at Prob. 4.4-4 (plane truss) for 20 time steps, using the central-drn
.
1
erenlr
predictor with !::.t = Ti./20.

5.3-1. Confi_nn the iterativ~ results of Example 5.3 in Table 5.3 for both the avera 'l'

~.4-6.
~.4-7.

~.4-8.

~.4-9.

and hnear-accelerat1on methods.

5.3-2. Repeat Prob. 2.7-2 using iteration by the average- and linear-acceleratiori
methods.

~.4-10.

5.3-3. Repeat Prob. 2 7 3 us1n g Iterat10n by the average- and linear-acceleration

.4, 11.

5.3-4. Repeat Prob. 2 .7-4 us1ng 1terat1on

by the average- and 1

J
methods.
mear-aece eration

~.4-12.

S.3-5. Repeat Prob. 2.7-5 using iteration by the average- and linear-acceleration

.4-13.

5 ..1-6. Repeat Prob. 2. 7-6 using iteration by the average- and linear-acceleration

,4, 14.

methods.

methods.

methods.
5.3-7. Repeat Prob. 2.7-7
using iteration by the average- and linear-acceleration
methods.
5.J-8. Repeat Prob. 2.7-8 using iteration by the averageand linear-acceleration
methods.
5.3-9. Repeat Prob. 2.7-9 using iteration by the averageand linear-accelerat ion
methods.

5.3-10. Repeat Prob. 2. 7-10 using iteration by the average- and linear-accelerat ion
methods.

,l-15.

8,4- 16.

4 17.

5.3-11. Confirm the iterative results for the plane truss of Example 5 4 T bl 5 4

. bh h
.
.mae . ,
using ot t e average- and Imear-acceleration methods.
!i.J. J2. Repeat Prob. _4.4-3 (plane truss) for 20 time steps, using iteration b thl
average- and ltnear-aceeleration methods with !::.t = Ti./20 = ti.
y

4 Iti.

!i.J-1.l Re~eat Prob. _4.4-4 (plane truss) for 20 time steps , using iteration by !ht
average- and linear-acceleration methods with !::.t = Ti./20.

I 19.

5.4- 1. Confirm the a_pproximate results of Example 5.5 in Table 5.5 ~

b h !ht
or ot

I lO.

!i.4-2. Rlipcat P.rob. 2.7-2 using direct linear extrapolation by the avcrnge- and li tll'llt
Ut'l'tlerut 1011 met hods.

5.4-J ,

Rl' Pl' HI Jlmh . '2.7 l usinR cli1vtt lint11r cxtrnpol111ion hy tlw uwrugc- und lintru
1111'1 hrnJi,

lll'l'l' ll'l llt IOII

239

t 4-4. Repeat Prob. 2.7-4 using direct linear extrapolation by the average- and linear-

5.2-11. Confinn the approximate results for the plane truss of Example 5.2 in Table 5.2
2
5, -12. Rep~at Pro~. 4.4-3 (plane truss) for 20 time steps, using the central-differenlc
predictor with !::.t = Ti /20 = r1

average- and hnear-aeceleration methods.

Problems

,,t,l l .

acceleration methods.
Repeat Prob. 2.7-5 using direct linear extrapolation by the average- and linearacceleration methods.
Repeat Prob . 2.7-6 using direct linear extrapolation by the average- and linearacceleration methods.
Repeat Prob . 2.7-7 using direct linear extrapolation by the average- and linearacceleration methods.
Repeat Prob. 2.7-8 using direct linear extrapolation by the average- and linearacceleration methods.
Repeat Prob. 2.7-9 using direct linear extrapolation by the average- and linearacceleration methods.
Repeat Prob. 2. 7-10 using direct linear extrapolation by the average- and linearacceleration methods.
Confirm the approximate results for the plane truss of Example 5. 6 in Table 5. 6,
using both the average- and linear-acceleration methods.
Repeat Prob. 4.4-3 (plane truss) for 20 time steps, using direct linear extrapolation by the average- and linear-acceleration methods with !::.t = Ti / 20 = t1
Repeat Prob. 4.4-4 (plane truss) for 20 time steps, using direct linear extrapolation by the average- and linear-acceleration methods with !::.t = Ti. / 20.
The equations of motion for the average-acceleration method may be converted
to the form: S Di+ t = Ai+ Derive e.:pressions for the matrices Sand Ai+t to
be used in this approach.
The equations of motion for the average-acceleration method may be converted
to the form: C Di+ 1 = Ai+1 . Derive expressions for the matrices C and Ai+ 1 to
be used in this approach.
The equations of motion for the average-acceleration method may be converted
to the form: M Di+1 = A1+ 1 . Derive expressions for the matrices M and Ai+ 1
to be used in this approach.
The incremental equations of motion for the average-acceleration method may
be converted to the form: C !::.Di = !::.Ai. Derive expressions for the matrices C
and !::.Ai to be used in this approach .
The incremental equations of motion for the average-acceleration method may
be converted to the form: M !::.D1 = !::.Ai Derive expressions for the matrices
M and !::.Ai to be used in this approach.
The equations of motion for the linear-acceleration method may be converted to
the form: S * D1+ 1 = A *1+1 . Derive expressions for the matrices S* and AJ+1 to
be used in this approach.
T he equations of motion for the linear-acceleration method may be converted to
the form: C * D
1 , 1 = Aj, Derive expressions for the matrices C* and Aj+ 1 to
he used in this approach .
'f'hc cqu utinns of motion for the linear-acceleration method may be converted to
lhl' f'w 111: M"' i)1 A 1, 1 , Dt iiw expressions for the matrices M* and AJ+ 1 to
ht IISl'd 111 lhl~ 11ppm11~ h

240

Direct Numerical Integration Methods

5.4-22. The incremental equations of_ motion for the linear-acceleration method may h
. expressions
.
converted
J . Denve
for the matrices (
d AA* to the form:. c * .ADJ = AA*
an

to be used m this approach.

5.4-23. The incremental equations of ~otion for the linear-acceleration method may h
D enve
expressions

converted
to the form
M* d AA*
b
. M*
. AD
. J -- AA*
J
for the matri<.'l'"'
an

to e used m th1s approach.

Framed
Structures

5.5-1. ~he ~rations_of, motion ~or the Ne':mark-/3 method may be converted to 1hr
. orm: D1+ 1 - A 1+ 1. Denve expressions for the matrices S and A- to be " I
1+1
usu
m th1s approach.

5.5-2. iThe equations

. CD.
_of' motion for
. the Newmark-/3 method may be converted to 1hr

. 0 rm:
J+ 1 - A1+ 1. Denve expressions for the matrices Cand A.
t b ,, I
m this approach.
1+1 o e use,

5.5-3. :0~:~1ttons : 1;1otion fo~ the New~ark-/3 method may be, conver,ted to thl'
.

. 1+ 1

A 1+ ,.

used m this approach.

Denve express10ns for the matrices M and A.

1+ 1

t I
o ll

5.5-4. The incremental equ,atio~s of motion for the Newmark-/3 method

b
.
may econ
verted to the form C AD - AA D .
AA'
b
d . . .
J J
enve express10ns for the matrices C u111I
1

to e use m this approach.

5.5-5. The incremental equ,atiol)..S of motion for the Newmark-/3 method

b
may e con
vcrtcd to the form MAD - AA D .
AA' t b
. . .
J J
enve expressions for the matrices M u111I
J

o e used m this approach.

0.1 INTRODUCTION
I\ framed structure consists of members that are relatively long (or slender)
to their cross-sectional dimensions. Points where members intersect,
li ce ends of members, and points of support are called joints of the structure. In
ll11ile-element terminology, we refer to the members as linear (or perhaps curvilinear) elements, and we call the joints nodes. Thus, the matrix analysis of
II nmed structures (1] becomes a subset of the more general theory of finite
11l'ments [2] for discretizing and analyzing continua. However, a framed struc1111 c inherently is divided into elements, unless it becomes necessary to subdivide
11wmbers into smaller elements.
T here are six distinct types of framed structures that designers use to resist
, 01nmensurate sets of loading systems. Six commonly occurring types of force
11wtems are: (a) parallel-coplanar, (b) concurrent-coplanar, (c) general-coplanar,
(ii) parallel in space, (e) concurren.t in space, and (f) general in space. Each of
1h1 rramed structures illustrated in Fig. 6.1 is specifically designed to resist one
111 lhesc load sets at its joints. When the structures are loaded, their joints
1111d<.rgo lranslational displacements (corresponding to forces) and rotational
1tl11placcmenls (corresponding to moments) . In addition, certain types of internal
111111,us (or stress rcsullants) arise in the members, depending on the type of
(111l'111rc . We will bric lly describe the characteristics of each type of framed
t1lll'IIIIL' 11ppcnri11g in Fi!' , () . I , l'VCll lhough beams and plane trusses already
h11w hl'l'll 11~ld 101 1x111Hplli, 1111d pmhll'IIIN iu pnvio11s (' haptcrs .

1 rnnpared

242
Framed Structures

/
z

.t

/ /-x

2 /

(a)

i
(b)

-I

--x

(c)

t2

(d)

. t2
I e -... .....
/,.-;
(e)
(I)

l<'l1t11n 6.1
pl11111

Typl~ ul 1111 11 , 1

11111111; (cl) wlcl

(1) :,1:,:./',',',',lH~ur,1, ~, (11) rn,111i11111111- 11111111, Ch) pl11111 hllNN, (l')

"

" P1111

1111111

,t

oc. 6.1

Introduction

243

Figure 6. l(a) shows a straight continuous beam, subjected to a parallel1oplanar set of forces in thex-y plane, which is a principal plane of bending . Any
11pplied moments must act in the z sense, as implied by the double-headed arrow
111 the right end of the beam . This type of loading system produces at a free
(unrestrained) joint a translation in the y direction and a rotation in the z sen se.
l'hese displacements are indicated in the figure by the arrows numbered 1 and
) at a typical joint j. From flexural theory, internal member actions (generalized
Ntrcsses) within the beam are a shearing force in they direction and a bending
moment in the z sense. If a force applied to a beam has a component in the x
direction , the resulting internal axial stress must be combined with the flexural
~tresses due to bending, even though the analyses are uncoupled .
Similarly , Fig . 6. l(b) depicts a plane truss in the x-y plane with forces
11pplied to its members as well as its hinged joints. As with a beam, if a moment
IN applied to a member, its direction must be parallel to the z axis. Reactions at
lhe pinned ends of members caused by loads become forces at the joints when
lheir signs are reversed . Thus , the joints resist systems of concurrent-coplanar
lorces, which produce axial forces in the members. Due to the accompanying
11xial strains, joint displacements become translations in the x and y directions ,
nN indicated by the numbered arrows in the figure. For a member with loads
11pplied directly to it, the axial stress from the truss analysis must be combined
wi! h the flexural stresses arising from local bending , although the analyses are
uncoupled.
T he rigidly connected plane frame in Fig. 6. l(c) carries a general-coplanar
r( of forces in the x-y plane, while applied moments act in the z sense. Resulting
di splacements at a free joint are translations in the x and y directions and a
111lution in the z sense. Internal actions occurring in the members consist of an
11xiul force, a shearing force , and a bending moment.
Figure 6.l(d) illustrates another type of planar structure, called a grid,
which lies in the x-y plane. This structure usually has rigid joints and is designed
11, resist forces in space that are all normal to its plane (or parallel to the z axis).
11 lollows that any applied moments have their vectors in the x-y plane. Displace111111ts a t a typical free joint consist of rotations in the x and y senses and a
111111slation in the z direction. Internal member actions are a shearing force, a
hl'nding moment, and a torsional moment, or torque.
The space truss shown in Fig. 6. l(e) is similar to a plane truss, except that
th1 llle mbers may have any directions in space. This type of structure carries
l111 11~s at its hinged (or universal) joints that are concurrent in space. The forces
ell t111g on members and joints may be in arbitrary directions, but any moment
uppl kd to a member must have its vector normal to the axis of that member. The
1r11No11 f'or this restricti on is that a truss member is incapable of resisting torque.
I llNpl1111mcnts at u free j o int ure three components o f trans lation in the x, y, and
d111l'lio11s. Dlll: to npplicd louds, u 111c1ulwr of II space tru ss may have local
lln1111 111 two pr111c1p11I plum:, of lwmlinA "" Wl'II "" 1111 uxiul for1c from the truss
11111tly1i1N

245
244

Framed Structures

Chap.

Figure 6.l(f) shows a space frame, which is the most versatile and complicated type of framed structure. Locations of joints and directions of members
are completely arbitrary, and the structure is designed to carry forces that are
general in space. At a free joint there are three components of translation (as in
a space truss) and three components of rotation as well. Internal member actions
consist of an axial force, two shearing forces in principal planes of bending, two
bending moments in the same principal planes, and a torsional moment.
A typical prismatic member in each type of framed structure has stiffnesses, masses, and equivalent nodal loads drawn from those of the axial,
torsional and flexural elements described in Sec. 3.4. Such member properties
are first expressed in local directions and then transformed to global directions
by the rotation-of-axes technique explained in Sec. 3.5. Next, we can assemble
member contributions to form stiffness , mass, and load matrices in the equation!!
of motion for the whole structure [see Eqs. (3.5-14), (3.5-15), and (3.5-16)].
These equations are solved for joint displacements in structural directions , and
the displacements pertaining to individual members are rotated to local clirections. Using such local displacements, we can find internal actions at any point
for axial , torsional, and flexural deformations with Eqs. (3.4-48), (3.4-50), and
(3.4-52). Of course, internal actions due to static influences at time t = 0 must
be added to those associated with dynamic response to obtain the total values.
Support reactions can also be calculated from member end-actions if desired .
In this chapter we deal only with linearly elastic framed structures having
small relative displacements and small absolute rotations due to dynamic loads.
We also assume that there are no interactions among axial, torsional, and
flexural deformations. Because continuous beams and plane trusses have been
discussed before, no separate sections are devoted to them. However, Secs. 6.2
through 6.5 give further information about plane frames , grids, space trusses,
and space frames. Then dynamic analysis programs are discussed in Sec. 6.6 for
these types of framed structures, as well as for continuous beams. Next, Secs.
6. 7 and 6. 8 cover methods for reducing the number of degrees of freedom for
beams, grids , plane frames, and space frames. Finally, Sec. 6.9 describc11
specialized computer programs that use these reduction methods in the dynamic
analyses of plane and space frames.

!Jue. 6.2

Plane Frames

K'

[~t !t]

r,
O
12
I
Sym.
2
O
6L
4L l
-------------~----,,
O
O l r,
o -12 -6L i O 12
2
0
6L
2L2 I O - 6L 4L

y'

Z,

z'

y
k2

x,,

- --- -

--- - -

7
k3

kl

I
IY,.
I

I
i3/

6.2 PLANE FRAMES

Figure 6.2(a) shows a typical member i of a plane frame with local (primed) and
global (unprimed) axes. The three numbered di spl acements in local directions ul
each end consist of a translation in the x' direction , a translation in the 1'
direction, and a rotation in the z ' (or z) Sl' IISl' . Ass11111i11g the memht~r is prh.
matic, we cun write its<, X 6 stiff1wss 11111l11 x Im locul dlll'l't11u1s us

j1

/
lb)

l'IMUl'I'

(i,l

Phuw I111111l'

w, (1t) lmnl

111r111I

dill'l 11 nnN, (h) ~lohnl dh cc1io11s.

(1)

246

Framed Structures

uc. 6.2

which is partitioned in accordance with joints j and k. In matrix K' the dimen
sionless ratio r1 is AL2 / lz, where / , is the second moment of area of the crosH
section with respect to the z axis. We form the stiffness matrix for a plane frame
member by adding contributions from an axial element and a flexural element ,
which were derived in Sec. 3.4 [see Eqs. (3.4-4) and (3.4-24)). Note that the
terms from these two types of elements are uncoupled for local directions.
Similarly, we can form the 6 x 6 consistent-mass matrix in local direc
tions, as follows:
140

0
pAL
420

156

Sym.

22L

4L

70

140

0
0

54
- 13L

13L

O
O

----- - --------, - -----3L

f4

where

= L

lllr

[bx']
by

, _

L dg

df3

d/5

L dg

, _

(4 )

(611)

'l'ht !'unctions in this submatrix for the j end are

df6 = M, df6

/, () 0 I
I /, I,,
()

(7b)

L dg

(8a)

+ 1)

_ g)

=M

(3g2 _ 2g)

= -pM2

(8b)
(8c)
(8d)

111 which the subscript M replaces the subscript b. Note that no integrations are

1111

follows:

(9)
111 which
Cx

= [ -;Y

~l
6 ~J

Cy

(10)

. h
f th' 3 x 3 rotation matrix R consist of direction cosines
I 11111s m t e rows o
is

. S
35
ply matrix
1
..
'
andz' withrespecttox,y, andz. Asm ec. , weap
I I lllWS X ' y '

u 11 11d

its transpose to ob:i:

(6h)
1 ' // , .

(3g2 _ 4g
'

L dg

PM6 - M, dx'

where

io whkh the dimensionless coordinate is ~

_ g)

= M, dfs = _ 6M,(g2

PMs - M, dx'

(g3 - g2)L

= M, df3 = M

(5)

[/1O !20 hOJ

/6 =

3g2

11quired for evaluating the~e formulas.

d l d atrices to global (or strufln order to convert stiffness, mass, an oa m
.
. R
1111111) directions, we compose a 6 x 6 rotation-of-axes transformation matnx '

which contains forces (per unit length) in the x' and y' directions, as indicattd
in Fig. 6.2(a). The 2 x 6 matrix of displacement shape functions in Eq. (3) ha~
the partitioned form

f. =

= M, dfi = 6M, (g2

dfi

PM2 - M, dx'

PM3 - M, dx'

(3)

-2g3

~=~:a~:~::t:~y also be subjected to ~ime-varyF~ng ~o~c(e)n]tr~~~

P.
11
moment M at any pomt [see ig. . a .
d
lnrces Px an y' as we as a
.
:Uanner similar to that for distributed
, uncentrated forces may be han~led m .a
. d On the other hand, for
lurces b , and b . expect that no mtegrat10ns are reqmre .
'
nd
x
Ille moment
vecyt,or M , we need first derivatives with respect to x off2,hJs, a
/ri to determine the following equivalent nodal loads:

(2)

In this expression, the vector of time-varying body forces b'(t) is

b '(t)

Is =

ax:~~::e

156
-22L 4L2

fTb'(t) dt

=g

.
.
n Eqs . (6b) and (7b) are drawn from Eqs. (3.4-1) and (3.4-17)
l1111ct1ons
given
1

Again, this matrix is obtained by adding terms from those in Sec. 3.4 for an axial
clement and a flexural element [see Eqs. (3.4-5) and (3.4-26)).
1n addition, equivalent nodal loads in local directions due to distributed
forces on a plane frame member may be calculated from

Pl (t)

247

Plane Frames

RTK'R = [Kii

Kik]

(ll)

Kk1 Kkk

Also , for the k end we huvr

(711,

(12)

Framed Structures

Chap,

and

Pb(t) = RTpb(t) = [Pb/t)J

Pbk(t)

(Bl

Perfonning the multiplications indicated in Eq

submatnces
(11) yields the stiffness

= L3z

Kkk

= -L3

EI, [r1c;
(r I

12c;
12)CxCy

6Lcy

10c.

6.3

Grids

II' such forces happen to be in structural directions instead, a conversion to

111ember directions must precede use of the formulas in Eqs. (16).
After stiffnesses, masses , and equivalent nodal loads for individual members have been transformed to structural directions , we can assemble them in the
111anner given by Eqs. (3.5-14) and (3.5-15). For this purpose we must calculate
the global indexes j l through k3 shown in Fig. 6.2(b) , as follows :

Sym.J

r1c; + 12c;
6Lcx

kl

(14a)

4L2

El [-r1c;- 12c; -(r1 -12)cxcy 6Lcy ]

-(r 1 - 12)cxcy -r1C Y2 - 12c2x -6LCx
-6Lcy
6Lcx
2L2

Kkj

jl

EI [r1c; + 12c;
Kii = L3z (r1 - 12)cxCy
-6Lcy

(14b)

Sym.J

-r1Cy2 + 12c;
-6Lcx

= pAL [140c_~ + 156c;

M
11

(14c)
4L2

Mkj

420

kk

22Lcx

pAL [70c; + 54c;

420
l6CxCy
I3Lcy

AL [140c; +
= L_

420

+ 156c;

140c;

16cxcy
70c;

54c;

- 13Lcx

(l 5 a)

= 3j

k2

- 1

j3

3k - 1

k3

= 3j
= 3k

(17)

ltxample 6.1
(ligure 6.2(b) includes a unifonnly distributed body force by sin Ot, acting in the y
1lircction. Let us calculate the equivalent nodal loads in structural directions due to this
lnnuence.
First, we transform the body force to member directions, using a 2 X 2 rotation

b' (t)=Rb(t)=[-::

: J[~]bysinOt=[: } ysinOt

(a)

Then we apply Eq. (3) to obtain

4L2
- I3Lcy]
13Lcx
- 3L2

(b)

(15b)
which are forces and moments in local directions. As the last step, Eqs. (16) give

156cY
140c~

Sym.J
156c;

-22Lcx

[wi~~;;~ ~t;,Jrs

Pb (t) = {O, 6, Lex, 0, 6, -Lex}~~ sin Ot

(15c)

4L2

OI course, K;k = K t and Mk = MT bee

.
Fl11ully' Eq. ( l 3) results in ~he foni ' . aus; matnces Kan? Mare symmetric .

j2

-16cxcy
22Lcy

Pbi(t)

111atrix, as follows:

Sym.J

-16cxCy
-22Lcy

= 3j = 3k -

These indexes serve as subscripts for placing terms into the stiffness, mass, and
load matrices for the whole structure.

Also, multiplications given in Eq . (1 2) produce the consistent-mass

.
submatrices

249

248

of equtvalent nodal 1;:::

(c)

lly inspection, we can see that these results are correct.

0 .3 GRIDS

p 1,3

p/,~(\,l

p /14C',

P1ik(t) -

p/,1<\ 1 1,;,,,-.

( 16h)

/1 t,r,

11~11 Wl' i111ply 111111 hud y lon'l'S lll'l' >,:iv111 111 tlw d l111 lh111Nol lm11l 1111111lwr ;llWS

'l'o make the analysis of a grid similar to that of a plane frame, we place the
iltructure in thex-y plane, as illustrated by the typical grid member in Fig. 6.3(a).
'l 'hc three numbered displacements in local directions at each end of the member
111 l H rotation in 1he x' sense, a rotation in they' sense, and a translation in the
1 (rn z) direction. If the memhcr is prismatic, its 6 X 6 stiffness matrix for local
1ll1l'l'tio11 s

lwro11111-1

250
Framed Structures

110. 6.3

Grids

251

K' =

Ki KJk]
K;j K;k

4L2

g_ __Q_ ___:-.f___l]_
L3

-r2L2

0
0

2L2
6L

111 this matrix the dimensionless ratio r 2 is Glx/Ely, where Ix is the torsion
, ,mstant of the cross section and l y is its second moment of area about the y ' axis .
I krc we combine stiffnesses from a torsional element and a flexural element, as
11ivcn by Eqs. (3.4-12) and (3.4-24). In this case the torsional and flexural terms
,ut uncoupled for local directions.
We can also write the 6 x 6 consistent-mass matrix in local directions as

Sym.

Tr~L-2 __ _
I

-6L I 0

4L 2

-12 / O

6L

12

140r~

y'

M'

x,.

[M
MJk ] = pAL
Mkj Mkk
420

-----

4L2
-22L

0
0

Sym.

156 1

---------- ---~------0
- 3L2
- l3L

?Or:

0
0

0 J 140ri
13L I O

4L2

54

22L

(2)
156

ri

= J / A is the radius of gyration squared. Terms in matrix M' come

where
I10 111 consistent mass matrices for torsional and flexural elements [see Eqs.
( 1.4- 15) and (3 .4-26)].
Furthermore, equivalent nodal loads in local directions for a grid member
111ny be obtained using the previously stated expression
(6.2-3)
z, z'
(a)

However, in this case the vector of time-varying body forces is

b' (t)

x,,

------ --

tk2.

k1

= [ ::-]

(3)

1'11is vector contains a distributed axial torque mx (per unit length) in the x'
dli cclion and a force b, (per unit length) in the z direction, as indicated in Fig.
<1 \(a). As before, the 2 x 6 matrix of displacement shape functions in Eq.
111 2 3) has the partitioned form

f = [f;

fd

(6.2-5)

f; are switched, and the sign of

reversed. Similarly, functionsf5 and/;; in submatrix fk are interchanged, and
tlw sign of f 6 is reversed. Thus , the submatrix tj becomes
11111 now the functionsf2 and f 3 in the submatrix

i3/ i

11 - - - - - -- - - - L - - - - -x

/1 ls

(b)

Odd llll'1t1hc r: (11 ) lmul dh n 111111N, Ch) 11luh11l dlt1l'l in11~

~]
(4)

,11111 !he s 11h111n1rix 14 is

21>2

253

I rumod Struoturoe

.I,

log

0.11

Speoo Trusses

(JI
./11 h

I {)

1111,

(8a)

- (f'

g )L
2

O
]
- 2g3 + 3g2

~or a grid member, we may also have cone

(8b)

P, apphed at any point, as shown in Fig. 6 3( ) ;;trated act10ns Mx', Myi, 111111
P, may be treated in the same mann
. a . e moment Mx' and the Im ,
.
er as m , and b but wth
t
.
1
x
"
ou mtegrat11111 ut
the functions in Eqs. (4) and (5 ) H
negatives of first derivatives with re~:~~~~ ;?r the mome~t My'. we need lh
rows of Eqs. (4) and ( ) b
P.
of the functions m the scl'oiul
5

l40r;c; + 4L2 c;
22Lcx

.
' ecause positive rotation . th ,
negative slopes. Thus, we have
s m e Y sense are equnl 111

-M, dfi __ My' d/2

T dg
y dx' PMS=

P M6

-M ,d(-/5)
Y
dx'

6M ,
--tce - g)

(C,h)

= My,

= - M , dfs = Y dx'

df5 2
L dg - My,(3g - 2g)

My, dfs - 6My, ( 2

L dg - T
g - g) = -p;,,,3

- E1y (- r2c; + 2c;)L2

-(r2+2)L2CxCy
[
-6Lcy

kj-u
-

Kkk -

he;+ 4c;)L

ElY

(r2 - 4)L2cxCy

111cmber.

For this example we can integrate in accordance with Eq. (6.2-3), as follows:

Jr[~J #
1

pi,(t) = L

cos Ot

= {0, -L, 6 , 0 , L, 6}

Lb,

l2 cos Ot

(a)

(-r2c;+2c;)L2
6Lcx

6LCy
-6Lcx
-12

S ym.

'l'hcn Eqs. (6.2-16) produce

pb(t) = {Ley, -Lex, 6, -Ley, Lex, 6}

~~ cos Ot

(b)

(7h)

Ciel
- 6Lcy
6L<\
12
S11111 larly, th1.; consistent-mass subrnal . . f
lions
lll'l'

rices or 11 1, Id riwmhcr in slructun1l dinc
,. .

A uniformly distributed body force b, cos Ot acts in the z direction on the grid member
111 Fig. 6.3(b). Find the equivalent nodal loads in structural directions at the ends of the

(711)

12
X

(r2c; + 4c;)L2

156

Subvectors of equivalent nodal loads in structural directions for a grid member

111 c the same as those in Eqs. (6.2-16) for a plane frame member. However, the
lll'tions in each subvector become a moment in the x direction, a moment in the
y direction, and a force in the z direction.
Assemblage of stiffnesses, masses, and equivalent nodal loads follows the
pultem described in Sec. 6. 2 for plane frames. Global indexes j 1 through k 3
used as subscripts for this purpose appear in Fig. 6.3(b), and we may calculate
their values with Eqs. (6.2-17).

Sym.J

-Cri + 2)L2c cy

(8c)

ICxample 6.2
(C,d)

/\gain, these terms ~o ~ot require integrations.

Because the gnd hes in the x
1
the sequence x-y-z, rotation-of-axes-~rfn:~i:d. the nodal displacements art 111
plane frame. The resulting terms in stiff
~hons_ are the same as those fo1 n
become
ness su matnces for structural directio1111

2
Ely[(r2c; + 4c;)L
2
Kg = U (r2 - 4)L cxcy (r2 c; + 4c;)L2
6Lcy
-6Lcx

Sym.]

It is easy to see that these values are correct.

6 .4 SPACE TRUSSES

A typical member of a space truss appears in Fig. 6.4(a), having universal hinges
111 ioints j and k. Due to this idealized type of connection, rotations at the ends

111 the 111c111lwr llt'l' rn1111idcrcd to be immutcrinl for the analysis of the truss . Local

264
f r nmnd Struoturn 1

II 4

1-1111

8p11co I r11u

266

n1

11 and z ', in conjum;tion with axis x ' o f the member , de fine two principal

1ln111H of bending. At each end o f the member we see three numbered arrows
I
11 lnrnl directions, representing translations in the x ' , y ' , and z directions. If
1111 111l rnber is prismatic, its 6 x 6 stiffness matrix for local directions is
'

K' [K
Kkj

1(

0 0
Sym.
I
1
0
0
0
K]k =EA
- --- --- --+- ----- L - 1 0 0 11
K kk
I
0 0 O I' O 0
0 0 olo 0 0

(1)

N111lll:ro terms in this matrix, pertaining to translations 1 and 4 in the x 1

ill11ltion, are drawn from Eq. (3.4-4) for an axial element. However, most of
1111' ltims in matrix K' are zero, because a truss member has no joint stiffnesses
111 dtn:ctio ns perpendicular to its axis .
In a similar manner, we can write the 6 x 6 consistent-mass matrix in
1111111 directions for the member, as follows:

z
(a)

I
I

Sym .

O O 2
-_ 6pAL 1--0-01_
2______
1

M'

(2)

k2

o o

-7'

k3 /

k1

0 10

Io o

I Ir I l' we use terms for the x ' direction that were derived previously for the axial
i1,~11u.:nl in Eq. (3.4-5). Also, due to accelerations in they ' and z' directions, the
11lht1r consistent-mass terms in Eq. (2) are the same as for the x ' direction.
Equivalent nodal loads in local directions caused by distributed body
l 111 ll\S on a space truss member are calculated as

I
I Y;,
I

(6.2- 3)

h11 this type of member, we can accommodate three time-varying components

111 lt11e loads,
z

Figure 6.4

(b)

b'(t)

pace truss member: (a) local directions; (b) "lobal direct

c,

IOl!S .

{bx , by' , b,,}

(3)

I r 1111s in th is column vector are forces (per unit length) in the x', y 1 , and z 1
,lhccti ons, as shown in F ig. 6 .4(a). The matrix of displacement shape functions
111 11.q (6 .2-3) may again be stated as

Iii

f1 I

(6.2-5)

257

256

Framed Structures

Chap. 11

Space Trusses

In these expressions the symbol e represents a unit vector in the direction

However, we now have two 3 x 3 submatrices,

f; - I,!, - [

Sec. 6.4

i !~}

Indicated by its subscripts. In particular, the unit vector ex is

(8a)

(4)
where

und

fk

= l)/4 =

1 01 0OJ f4

[0

(5)

t,

PM2 = M, df,
dx'

= M, , df, = - M,,

PM3 = -My df1

dx'

L dg

=-

(6a)

My df1 - My,
L dg -y

PM6

= M,, di4 = M,,

dx'

-M , df4
Y dx '

df4
L dg

= M,, -

df4 _ _ My' =
L dg L
- PM3

t, )( ,..

(8b)

(8c)

R, we have
(9)

In which the direction cosines of axes x', y', and z' are listed in rows 1, 2, and 3.
For the purpose of transforming stiffnesses and equivalent nodal loads to
structural directions , we form a 6 x 6 rotation-of-axes operator R as in Eq.
(6.2-9). Using matrix R as indicated in Eq. (6.2-11) , we find the member
stiffness matrix in structural coordinates to be
2
Cy

Cxe,
eyCz
-e; - cxcy
2
-exey -ey
-cxe, -eye,

e,
-exc,
-eye,
_
2
e,

I
I
I
I

Sym.
(10)

- - - - - - - - - - - - - ---,-- - - - - -

(6d)

c~
I
i CxCy Cy2
I
I exC, Cye,

c2

Notice that only the direction cosines of axis x' (the axis of the member)
Influence these terms. Applying matrix R in accordance with Eq. (6.2-13)
pmduces the equivalent nodal loads
-

A~

'()

Pb t - R Pb t ()

[RTPbj(t)J
RTPbk(t)

(11)

('/a)

which are now in structural directions . Of course, the detailed forms of the terms
111 1>1i (t) depend on the nature of the loads o n the member and the location of the
1hlrd point p. Finally, the consistent-mass matrix for a space truss member
IN l11vniin11t with rotation of axes, jl1Sl HS it is for ll plane truss member (see

('/I))

S11-. I , 'i ),

11 11d

(',,

ZJ
= ..J._
L

A similar description may be given for the unit vector ejp, using the coordinates
of points j and p. Collecting the three required unit vectors into a rotation matrix

(6c)

As w '.' )1 c?~centrated. forces' these terms need no integrations.

I o (01111 a rotat10n matrix for a space t
point p (in add ition to j and k) to d fi
1:1ss. member, we can use a third
<>A(n) shows such a point lying in\hne a,Pr~nctal plane of bending. FigurtWhtncvcr possible, this point would be\~ -y pane and. ~ot ?n the x' axis .
lor whid, the coordinates in spa
kn en as another JOmt m. the structure ,
n1111ri x may be found using proper~:t~; th ownt. T(hen the terms m the rotation
e vec or or cross) product, as follow~ :
_
Ce X C,,,
Cr - ,,
c,, X ('1,,

(6b)

L - -p M2

= _ My

= Y xJk + yJk + z}k

CxCy
PM5

Yk
_L

und

where Ji = 1 andf4 =
as before.
A space truss member may al b
b
d
.
forces and moments at any point ;~ e su ~~ctef to time-varying concentrated
lhc moments My and M,, included ~: ;fns1st o the forces P,,, Py,, ~nd P,,. and
hccause a truss me b
g. 6 .4(a). (Note that Mx' is omitted
m er 1s mcapable of re t'
1
!ruled forces may be handled in th
s1s;ng ~n axia moment.) The concen
no integrations are required For t:;anner t:nbed for distributed forces, bul
and ./, with respect to x,, b~t for M , ::men z' we ~eed fir~t derivatives of Ji
In Sec1 6 3 Th
Y
must use their negatives, as explained
us,
. . .

259
258

Sec. 6.5

Framed Structures

As with plane frames in Sec. 6.2 and grids in Sec. 6.3, we can assemble
stiffnesses, masses, and equivalent nodal loads by assessing these properties
from individual members. Figure 6.4(b) shows the global indexesjl through k3
used as subscripts in the assembly process, and their values are again calculated
using Eqs. (6.2-17).
At the end of Sec. 3.5, we mentioned that the analytical model for a truss,
as described here, is not really suitable for a structure having only a few
members. For better accuracy in the dynamic analysis of such trusses, we
recommend the component-mode method discussed in Sec. 10.6.

P'(t)

Px] [cos if!]

[-0.6889]
Z,: = sinO if! P (t) = 0.7;49 P(t)
[

(h)

Then the equivalent nodal loads in local directions become

pplt)

= fJP'(t) = PPk(t) = fIP' (t)

(i)

= {-0.3445, 0.3625, O}P(t)

l'or which the subscript p replaces the subscript b. Also, their counterparts in global
directions are

Example 6.3
In Fig. 6.4(b) a time-varying force vector P(t) acts at the midpoint m of member i. This
force is directed toward point p, which is located on they axis. As in Fig. 6.4(a), the
third point p, along with points j and k, defines the x '-y' principal plane of bending.
Dimensions appearing in the figure are
Xj k

Xk -

Xj

= 4

Yik = Yk - Yi = 3

Zjk

Zk -

Zj

= 2

YiP = YP - Yi

Xk

= \!xJk + Y% + zJk = V29 = 5.385

cY

= 0.7428

(b)

Z k

= Yik
L = 0 .5571

= {-0.4365, 0.1091, - 0.2182}P(t)

c, = { = 0.3714

I OJ

(c)

frame which is similar to

. . ' .
~
truss member- but now rotations are included at Joints J and k. As befor~,
II space
'
.
h
, , lane and the x, -z, plane. Six
the principal planes of bendmg are t e x -y p
. t f tr s-

a~~

nu~bered displa~em~nttsh, in~icyat~d e:~~i::~t~nt:\~i~:~;ri~::!:~ ;em~:r,

lat1ons and rotations m ex , ,
f
6 x 6
the 12 X 12 stiffness matrix for local axes is composed of the o11owmg
1mbmatrices:

r1I,
E

0
0

(c)

Sym.

12/,

0
0

12/y

0
0
0

r2 L2ly

-6Lly

6Lfz

0
0

KJ; = L3 0

Applying Eq. (7a), we obtain

_
)
0 8944

1:i ure 6.5(a) depicts a typical member z o a space

(d)

e,, = [ -0.3714 0 0.7428) = [ _

0 4472
V(0.3714) 2 + (0.7428)2

(j)

.
b E (11) Because of the central location of P(t), the equivalent nodal loads
ns given y q.

nrc the same at both ends of the member.

For this example the unit vector eiP is

eip = [O

= RTPPi(t) = PPk(t) = RTPPk(t)

6.5 SPACE FRAMES

and the direction cosines for its axis are

= {

PPi(t)

(a)

=2

Find the equivalent nodal loads for both local and global directions due to the concentrated force.
From the given dimensions, the length of the member is

Cx

Space Frames

(la)

4L2 ly
4L2 I,
0

and Eq . (7b) gives

e y

= [-0.4983

1111d

0.8305

- 0.2491)

(f)

-r1lz

Altogether, the rotation matrix becomes

[
R

0.7428

0.5571

- 0.4983

0.8305

0.4472

0.37141
- 0.2491

K'kJ - LJ

(g)

0 .8944

which contnins thl.! uni! v1.1.1ors c,,, c,,., and c , lis11.d 10w wlst.
llsin11 1111. 1ttomtII y of II innglc jmp, w1. rnn llncl th, nnitll 1/1 lo lw 2 , 11 111d
'l'h1.1l'11111, 1111' 1111111H111111t N111 tlw 1111\' l' wrtm P(I) In 1111111cll111.11111111 1111.

0
0
0
0
{)

111111

0
- 12/,
0
0
0

6U

0
0
0
6Lfz
0
0
0
0
6Lly
0
- 12/y
2
0
0
- r2L ly
0
2
0
2L ly
0
6Uy
2L2 l ,
0
{)
0
0

(lb)

261
260

Framed Structures

Kkk =

r 1 I,
0
0
E
LJ 0
0
0

12/,
0

Chap. 6

Sym.

12/y
0
0
r2L2f y
0
6Lly
0
4L2I y
-6LI,
0
0
0

( le)

Sec. 6.5

Space Frames

Nonzero terms in these submatrices come from Eqs. (3.4-4) , (3.4-12) , and
(3 .4-24) for axial , torsional, and (two) flexural elements. All of the structural
parameters in Eqs. (1) have been defined previously.
Similarly, the 12 x 12 consistent-mass matrix M' for local directions
contains the 6 x 6 submatrices

4L2/,
,

Mji

140
0
pAL 0
420 0
0
0

156
0
0
0
22L

Sym.
156
l40r~
0
4L2
0
-22L
0
0
0

(2a)

4L2

und

M'
kj

(a)

M'

kk

k1
/ , k3
k6

. t i2

k4

I
I

I Y;k
I

70
0
pAL 0
420 0
0
0

140
0
pAL 0
420 0
0
0

0
54
0
0
0
- 13L

0
0
0
0
0
54
70r 2g
0
0
l3L
0
0

156
156
0
0
0
22L
0
0
- 22L

x,. ----- ~ /

11'1111111 ,,. ~

/ z,k

(2b)

Sym.
(2c)

140r}
0
0

4L2
0

4L2

(6.2-3)

0
l3L
0
0
0
-3L2

Nonzero entries in these submatrices are taken from Eqs . (3.4-5), (3.4-15), and
( L4-26), and all of the structural parameters were given earlier.
Distributed body forces applied in local directions to a space frame memhcr cause the equivalent nodal loads

I
/J3 /1 /4 ______ :_I_
/__,_ _ _ x

,1e

0
0
-l3L
0
-3L2

lhnt urc also in local directions. As for the space truss member, the three

111111puncnls of force (per unit length) arc

(6.4-3)

Iii)
l.]' 11111 I

I1111111

1111"111 Illl l (II) I111111

illt11 ll1tt1II,

(It)

11l11h1tl 1ll111IIIIIIN

whk h 11pp11 11 nl 11 1-f,l'111~1k p11l 11I on 1111 1111111hl' I' 111 Fl1 . (1 , 'i(n) . In Eq . (6 .2-3) the

262

Framed Structures

Chap. 8

Sec. 6.5

263

Space Frames

matrix of displacement shape functions is once more

= [~

fd

(6.2-5)

t, - [~

But now the two 3 x 6 submatrices have the forms

~-G
and

ck = [fio~

~J
1 ~ ~ ~ ~J
0

!2

0
0

!2

ls

0
0
0

(
4)

= {f,, P y', P,,}

(6)

w.hich. require no integrations. Furthermore, three moment components in local

<11rcct1ons are

M'(t) = {Mx, My', M,,}

(7)

To cal~ulate e9uivalent nodal loads caused by these moments, we set up the

following matnx-vector multiplication:

= f:,.,M'(t)

(8)

~hi~~ ~ecds no in~egrations . However, Eq. (8) does require the matrix f ,
rnntammg appropriate first derivatives of displacement shape functions wi;,;
respect to x '. The first submatrix in f ,x' is

-1.x' 0
0

/6,x'

Ax

0
0

}J

(10a)

.,

2 I:
.,

(10b)

To convert stiffnesses , masses, and equivalent nodal loads from local to

global directions, we form a 12 x 12 rotation-of-axes transformation matrix R,
us follows:

R- [[

f ; ~]

(11)

The 3 x 3 rotation matrix R , appearing four times in matrix R, is identical to

lhat in Eq. (6.4-9) for the space truss member. Again, the third point p is
ussumed to lie in thex '-y ' principal plane of bending, as indicated in Fig. 6.5(a).
lJsing matrix R, we apply Eqs. (6.2-11), (6.2-12), and (6.2-13) to obtain the
matrices K, M, and pb(t) for structural directions.
As for the other types of framed structures in previous sections, the stiffnesses, masses, and equivalent nodal loads are assembled from member conldbutions. To transfer terms from the member arrays to the structural matrices,
we need the global indexesj 1 through k6 shown in Fig. 6.5(b). Their values are
l'alculated by the formulas

(9u)

wlwrc

dx'

M,,.

h.x

= df6 = 3 J:2 -

Functions / 1 and / 4 appearing in Eqs. (9a) and (1 Oa) are not differentiated with
respect to x ', because they are to be simply multiplied by the torque Mx. On the
other hand, nonzero terms in the second row of each submatrix are the negatives
of first derivatives that multiply the moment My' Finally, nonzero terms in the
third row consist of positive values of first derivatives that multiply the moment

(5)

pf,(t) = fTP'(t)

0 fi.x

- Js,x

I' '
J 6,x

Due to these forces, the equivalent nodal loads become

~.., = [~ ~

(3)

o -A o

Ii

f4

-/3

. Figur~ 6.5(a) also .shows the possibility of six time-varying concentrated

~ctlons ap?he~ at any pomt on the space frame member. Three force components
m local d!fecttons are given in the column vector

pl,(t)

where

0
0

Functions Ji throughA in these arrays were defined previously in Eqs (6 2-6b)

and (6.2-7b).

P'(t)

0
0

j l = 6j - 5

j2 = 6j - 4

j 3 = 6j - 3

j4 = 6j - 2

j5 = 6j - 1

j6 = 6j

kl

= 6k -

k2 = 6k - 4

k3 = 6k - 3

k4

= 6k

- 2

k5

= 6k

- 1

(12)

k6 = 6k

which arc used as subscripts in the assembly process.

h ,,'

d/i

"I

111 rnld1lto11 , 1111 ~1'1'111111 , 111111111111 11 IM

.h ,

l "I

.,A/:
., I I

ll:xum1>lc 6.4
.

(%)

l l~ll ll'

6 'i( h) shows

II

ti11 ll' v111ying unifrn 111ly distributl'd force of intensity b1 (t) , acting

111 tl w 1 di1 t1t 1t111 1111 tht N(ltll 1' I 1111111 1111111h11 I .111111 io11Nol pnints .I, k, nnd p arc the same
111, 1111 thr ~plll'l' IIII NII 1111 111(!1'1 111 I .ll.llllll'k Ii \ ( hllN, lhl' 1ot11t1rn1 1111 111 ix

264
Framed Structures

0.7428 0.5571
R =

- 0.4983 0.8305
-0.4472

Chap

0.3714]
-0.2491

li ons a~sing the rotation matrix R , we can find the components of by(t) in local dirl'l

-+:,J ~ [~::~:},(,)

lul

Then apply Eq. (6.2-3) to obtain

p;,ir)

= L

fJ b'(t) d(

= {0.2786, 0.4153, 0, 0, 0, 0.0692Il}lby{t)

(h)

Hild

pJ,k(t) = l

I >YNAPF for plane frames, DYNAGR for grids , DYNAST for space trusses,
DYNASF for space frames . They represent the dynamic counterparts of the
1111ic analysis programs described in Ref. 1.
First, let us consider the task of generating Program DYNACB for con1l11uous beams , using Program DYNAPT for plane trusses as a guide. Both types
,ii structures have two displacements per node, but those for a beam consist of
H translation in they direction and a rotation in the z sense, as indicated in Fig.
II I(a). Modifications of the structural data for plane trusses (see Table 3.2) are
minimal. The y coordinates of nodes can be left blank, and we need only add
lhe moment of inertia ZI(I) of the cross section to each of the lines for element
l11formation . In addition, the meanings of terms in the nodal restraint list be1111ne: NRL(2J- 1) = restraint against y translation and NRL(2J) = restraint
IIHllinst z rotation.
Dynamic load data for continuous beams may also be specified as
ftlOdifications of those for plane trusses (see Table 4.1). However, the two nodal
11t'lions for a beam are a force in the y direction and a moment in the z sense,
u~ shown in Fig. 6.6(a). Because forces applied to a flexural element act only
111 the y direction, the linearly varying line loading depicted in Fig. 6.6(b)
ttquires only two parameters (BLl and BL2) for its definition. The equivalent

rlb'(t) d(
y

= {0.2786, 0.4153, 0, O, O, -0.0692Il}lby{t)

,,

265

Programs for Framed Structures

,111d

0.8944
still pertains. We shall detennine equivalent n d 1 1
directions caused by the distributed force.
o a oads for both local and gloh11I

b'(t) - R b(t)

nc. 6.6

(l

(A, )2, ,

.h ~~x~~~~r:~~~~i.n (~:~~~~~e: nfio:dal loads in member-oriented directions at joints} uml

t
/ i

OJ

RT
p,,J(t) = [ 0 R T p;,j(t)

= {O, 0.5000, 0,

- x

(A, )2,

-0.03095L, 0, 0.06190l}Lby(t)

(d)

'

(a)

lllld

p,,k(I)

= [ ROT

OJ

R T Pbk(l)

= {O, 0.5000, 0, 0 .03095l, 0, - 0.06 190L}Lb,.(t)

(i)
I 11111N in these vectors consist of equivalent n0 d I I
. .
.
.
tu l1111f1d dlrct:tions.
a oads at Joints J and km struchUl'

8.8 PROGRAMS FOR FRAMED STRUCTURES

In a 11111111111 r11wlopo11s lo lhut fi>r Proorar11 l>YNt\l'T (

d1v I

"
sn
I op w1111n111r ()1C1frn 1111, for llrl' dyr1111111r u1111ly<11 ~ ol rlw
lr111111d ~r,11r r1111~ 'llr1N1 profr11111N1111 n1111wd l>\'Nt\< 'II I

~ . ., 7
., . ) we c1111
li v1 oflll'r I ,
I
.
YPt'\ c,

"'t t ,

or 11u1t111110111, lita rttN,

J_

b,,

ii 11111111111
l
l
11 I IO
_
x
0
k

Jill
I 111111 1111 111111 l11111111N 111111111 (II ) 111111111 l11111IN ,

(h) 11111 11111d

1111 11!111111111

266

Framed Structures

Chap. 9

od;l loads at points j and k may be taken as the solution for Prob 3 4 _10 AI
;~Y.rograthm Df YNACB the only scale_ factor needed for ground ac~eler~tio:<:~
' so e actor GAX must be omitted.
Example 6.5

'loc. 6.6

l'hc tenn x;(t) in this formula is the distance that the load has traveled along element i
111 time t, as illustrated in Fig. 6.7(b) . Because the functions in matrix fare cubic, the
1quivalent nodal loads given by Eq. (a) will also be at least cubic in time, depending on
how x;(t) varies. For a moving load with constant acceleration, the fonnula for x;(t) is

Figure 6. 7(a) shows a simply sup rt d

.
. po e pnsmattc beam, divided into four flexural
elements of equal lengths A
from left to right so that .. t mo~~ng load _P of constant magnitude traverses the span
translational resp~nses oft~:
at various speeds [3].

~~=~ :~ ::2e 1;

;t:~nc~~on _of)time. We will calculate the

llll pomt caused by the load moving

Equivalent nodal loads for this prob)

b i
. .
ment shape functions for a flexural eleme~~r:?Eq~ ~~~~l~t:i~t:,1!tfo~~:;~placcP&;(t) = Pr[x;(t)]

(a)

~-/----x

x;(t) = Vopt

~x(t)----1.,
i + - - - X;(t) ----..i~.

t
t- - -__________
p

1-/ - - 0

___,;,

Pb3

/ /'

(I ))

l'l111m,

,,.7

2-

(b)

(i - 1)

f = 0), and a0p is the value of its constant acceleration. Substitution of Eq. (b) and the
!unctions from Eq. (3.4- 17) into Eq. (a) produces the desired equivalent nodal loadspb1
through p1,4 indicated in Fig. 6. 7(b).
Of course, we need to extend Program DYNACB to accommodate a moving load.
The load parameters NLN and NEL (see Table 4.1) are augmented with a third paramtitcr, IML, which is an indicator for a moving load. If IML = 0, there is no moving load;
hut if IML = 1, a moving load is present. We also need the additional load data:
(d) Moving load (P, VOP, AOP) to be input for the case when IML = 1. Subprogram
I >YLOCB reads and writes this data and calculates nonproportional equivalent nodal
loads in accordance with Eqs. (a), (b), and (3.4-17). We must also modify the loads
processed by either Subprogram NORMOD or Subprogram NUMINT by adding to them
the equivalent nodal loads due to the moving load. If desired, we could further extend
the code to handle more than one moving load simultaneously.
Now let us assume that the beam in Fig. 6.7(a) is reinforced concrete and has the
physical properties:

E = 3.6

103 k/in. 2

L = 4 = 240 in.

P = 10 k

p = 2.25 X 10- k-s / in.

2

/,

= 3375 in.

where the units are US. Figure 6.8 shows computer plots of Di, at node 3, obtained by
the modified version of Program DYNACB and using Subprogram NUMINT for
3
lhe response calculations. The case of constant velocity (VOP = 3.585 x 10 in. /s)
produces a maximum translation of 0.4036 in. at node 3, and that for constant accelerntlon (AOP = 1.071 x 105 in./s2 , with zero initial velocity) gives 0.3625 in. Their
111tios to the static deflection of PL 3/ 48/, = 0.2370 in. (due to the load applied grad1111lly at node 3) are 1.703 and 1.530, respectively. The values ofVOP and AOP selected
In thi s example both produce travel times equal to the fundamental period of the beam,
which is 66.95 ms.

p"'

t2

1unning

(a)

+ aop

In this expression Vop is the velocity of the load when it first contacts the beam (at time

A= 12 x 15 = 180 in.

P., t
/-.

267

Programs for Framed Structures

(11) Sl111ph lw11111 wifh n111vl11n lune! , Ch) rq11lv111,n1 11ucl11J lrnulN

)(

Second, we shall describe Program DYNAPF for plane frames, using

Programs DYNAPT and DYNACB to guide us. As shown in Fig. 6.l(c), a
typical node j in a plane frame has three displacements, which are translations
111 the x and y directions and a rotation in the z sense. Table 6.1 lists the form
of structural data required for plane frames. As with continuous beams, each line
tontuining clement information must include the moment of inertia ZI(I) of the
noss section. Also , a line of nodal restraints has three types, which denote
1tstrnints against x trunslution, y trnnslntion , and z rotation, respectively.
'l'nhk' c, .) shows dyn11111ic lond dntu Im pln,w r,11111l'S. In this cusc, the lines
lor 11ii1l11I dlNplmt111t11t., 1111d w locitlt~ rn11t11i11 tllll't' q1111111i1itf'I i11sltnd or two .

268

Sec. 6.6

Framed Structures

TABLE 6.2 Dynamic Load Data for Plane Frames

VOP CONSr .

0.40
AOP CONST.

No. of
Lines

Items on Data Lines

!SOLVE, NTS, OT, DAMPR

Initial conditions
(a) Condition parameters
(b) Displacements
(c) Velocities

l
NNID
NNIV

NNID, NNIV
J , D0(3J-2), D0(3J-1 ), 00(3J)
J, V0(3J-2), V0(3J- l ), V0(3J)

Applied actions
(a) Load parameters
(b) Nodal loads
(c) Line loads

I
NLN
NEL

NLN,NEL
J, AS(3J-2), AS(3J-l), AS(3J)
I, BLl, BL2, BL3, BI.A

Type of Data
Dynamic parameters

0.30

C:

~ o. 20
0

Ground accelerations
(a) Acceleration parameter
(b) Acceleration factors

a
0.10

1
I

Forcing function
(a) Function parameter
(b) Function ordinates
omit when IGA

o.oo

IGA
GAX,GAY

1
NFO

NFO
K , T(K), FO(K)

= 0.

flexural elements), we find equivalent nodal loads in member directions. Then

the equivalent nodal loads in structural directions can be computed from
Eqs. (6.2- 16).

(ms)

Figure 6.8

269

Programs for Framed Structures

Translational responses at center of beam.

Example 6.6

The plane frame in Fig . 6. 10 consists of three prismatic members and has an initial load
Po applied by a cable connected to node 1. If the cable suddenly breaks, the frame
responds to the initial displacements caused by the load Po applied statically. Such a
response is the sum of free vibrations of the natural modes excited by the initial displacements; and in the presence of damping, they will decay with time.
We shall assume that all elements in the frame are steel W 12 x 85 sections with
the following properties:
E

No. of
Lines

Problem identification
Structurul pnrurnctcrs
l'h111t lrn1m <111111

(11) Nodul rn111cli1111tr N

Ch) 1' 11 11111111 1111111111111111 11

l N111l11I 11~11,11111~

(1

I
I
NN

NI!
Nl(N

106 kN/ m2

A = 1.61 x 10- m

p = 7.85 Mg/m3

l , = 3.01

10- m

l=2m
4

for which the units are SI. From static analysis , initial displacements due to the load
P0 = 10 kN are: (1) at node 1, (Dot , = 0.5697 mm, (Dot2 = -0.2923 mm, and
(D0 )1 3 = - 4.739 x 10- 5 ; (2) at node 2, (Dot , = 0.3642 mm , (Dot2 = 0.1149 mm,
nnd (D0)13 = 6.598 x 10- 5 . We used these values as input data for Program DYNAPF,
with DAMPR = 0.02 and solution by Subprogram NUMINT. Figure 6.1 l(a) shows
computer plots of translational responses Di I and Di2 at node l due to the load release;
nnd in part (b) we have the bending moments AM3 and AM6 for the j and k ends (nodes
I 1111d 2) of clement I . For all of these responses, the maximum (or minimum) values
m 'l ' III' at time I
O und diminish thcrcuftcr bccnusc of damping.

Items on Data Lines

TABLE 6.1 Structural Data for Plane Frames

Type of Data

= 207

Descriptive title
NN,NE,NRN, E, RHO

I, :\CI), Y( I)
I, INCi), KN(I), AXC I). './,1(1) '
I, NIU ( II JI, NIU ( 11 I), NIU .( II)
-

2/0
I 111111111 i,1111111111111

C l111p

I 11011ntl 111111111 111

l'roi,110111 1111

271

(I

--,,

i3

2L

__L_,

(a)

/ ~l- L-L
/

Figure 6.10

Plane frame with load release.

lrn sion constant Xl(I) and the moment of inertia Yl(I) replace the moment of
1t1tr tia ZI(I) on each line containing element information. Furthermore, the

in the nodal restraint list become restraints against rotations in the x and
senses and translation in the z direction.
Similarly, the dynamic load data for grids are symbolically the same as
1111 plane frames in Table 6.2, but the meanings are different. When the structure
111 11 grid, the first two initial displacements and velocities at a joint are rotational
(in x and y directions); and the third is translational (in the z direction). Nodal
l1111ds consist of moments in the x and y senses and a force in the z direction, as
Nlmwn in Fig. 6.12(a). In addition, the linearly varying line load illustrated in
l1g. 6.12(b) acts in the z direction and is defined by only two parameters (BL1
11ml BL2). Equivalent nodal moments at joints j and k due to the line load are
1111,ily converted to structural directions using Eqs. (6.2-16) withpb 1 = pb4 = 0.
I inally, the scale factor GAZ for ground acceleration in the z direction replaces
lht factors GAX and GAY.
lr 11ns

bl1

/ - - - -- x

(b)

lt'lgurc 6.9

Load
I
s on p ane frame: (a) nodal loads; (b) line loads on element.

_Third, let us brie fly examine Pro ram DYN

of ~11ds. Because this type of structu! ha h A~R for the dynamic analysis
111oiirn111 will be very similar to Program D~~l~/tplacements per node, tlw
flw lh1'l'L' types of displacements at a t . I
I .. or plane frames. However ,
111
1 11111! \' senses und ~ lranslatio~ in th : P'c,~ . ~~~ l' J
~ gr!d are rotations in tlw

T 0 I 1111
c z ' l l t l 11011 H~ indicated in F'1 6 I ( I
l ' 1 Al' llw plu1ll' f'rume structural d11111 1 ',

?' ').
wt n1us1 11<ld lht :..ht111111n n1odulw, <, I II ,
n l 11hlt (1 . I lo that for a !!rid ,
'

' o

II ~111111111111 p11ra11tL'ltrs . Also', lht

11,)(umplc 6. 7

I ill grid illustrated in Fig. 6.1 3 is made of brass and has rather small dimensions and
l111uls. The cross section of each prismatic clement is a solid square with side 0 .25 in. ,
,11ul lhc applied actions consist of step nodal und line loads. Physical parameters in this
I

lllllltpJC UI~'

I'

1.5 X 10 1 k/111

(i

()

'i (1 "

101 k/111.'

f'

R.10

10- 7 k-s 2/in .4

212

Framed Structures

'loc. 6.6

0 , 60

e
e

273

Programs for Framed Structures

z

0,40
w

25

0.20

o.oo

(/J

-o . 20
X

I-

<
_,
-o. 40

(/J

z
<

a:: -0,60

I-

(a)

(a)

E
z

AN3
10

-"

I-

z
w
I:
w

s
b ,,

~o

z
(/J

-s

I(.)

<
0

-10

( b)
1

1 1gure 6.11

Responses of plane frame: (a) displacements; (b) bending moments.

(bl

I,

4 in.

,.

0.02 k

A = 6.25 X 10- 2 in. 2

b1..

Ix= 2!y = 6.510

10- 4 in. 4

= 0.004 k/in.

Figure 6.12

Loads on grid: (a) nodal loads; (b) line load on element.

llie member Jscc Fig . 6.3(a)], and the second is a bending mome~t about they ' a~is.
Mu idmum (or minimum) values of the nodal translations are 0 .1 157 m. and 0.08451 m.;
2
3
k-in . and -8. 181 x 10- k-in.
111111 lhosc for the moments arc 9.333 x 10

111

where the units arc US.

~ilh lhis ~al a we ran Program DYNAOR, 11.~ing DA MPR

0.05 and rcspousi
l'llle11lu11011s by Subprogrnm NORM OD . Translulions of' 11rnks 2 ond 3 in the z dirccliwi
111t plolll~l lu Fin , 6. 14(11), 1111d lhc 1110111c nl sAM1 1111dtl~,, 111 1h1 I. cud (nodc4)of'dcnilIII
I 11pp1111 111 p11,1 (h) ol 1hl ll1111 rl'. T lw ll1sl ol lhtNr 11111111r11I N iN111t,1qm nhout llil' 1 ' lllliN

J/o11 1lh, we shull disl'uss Progrum DYNJ\ST for spucc trusses, a?ai.n using
J>YNJ\PF 1111 l'o111p11iiso11s. 111 till s l' IINl', tlw tllrt'l' types o f displace-

11,1w11111

274

Framed Structures

Chap. 6

275

Programs for Framed Structures

Sec. 6.6

NOOE 2

z
0 -10

j3

::; 0. 05

-i2

(f)

/i

~ o.oo
,_

j1

<(
~

(f)

~ -0 ,05

a::
,_
I

-0 . 10

(a)

Figure 6.13 Grid with step loads.

ments at a typical node j are translations in the x, y, and z directions, as shown

in Fig. 6. l(e).
In order to make the plane-frame structural data in Table 6.1 apply to a
space truss, we must add Z(J) to each line of nodal coordinates. On a line of
clement information, the moment of inertia ZI(I) is replaced by an identifier IP
that indicates whether a third point p is necessary for locating principal planes
of bending. If IP = 0 , the x '-y' principal plane in Fig. 6.4(a) is taken to be
parallel to they axis. However, if IP = 1, the next line of data must contain the
coordinates XP, YP, and ZP of the third point p. (This type of data is shown in
Table 6.3 for space frames.) Last, the terms in a line of nodal restraints denote
restraints against x, y, and z translations .
Dynamic load data for space trusses are similar to those for plane frames,
hut the meanings are different. For a space truss the initial displacements and
wlocitics at a joint refer to translations in the x, y, and z directions. The thrcl'
nodal loads at a joint are forces in the x, y, and z directions, as depicted in Fig.
h. I .5(u). Also, the three sets of linearly varying line loads appearing in Fig .
<, I Ci(b) require six parameters (BLl through BL6) for their definitions. Eq ui v
nltnt nodal loads due to these line loads are easily obtained by extending Eqs.
(I . IO 4). To accomodate three components of ground acceleration , we must also
mid the scale factor GAZ to the data table.
1,:xn1111>lt 6.8
ttpllH' C, IC, dtpil ts II spun 11 IINN with nine p11s11111t11 !'lt' llll'IIIN , h11v111g l'(lllll l l ' IONN
hl'I ti111111I llll'IIS 111111 r11111t111s1d 111 hf1-h sti111pt h 111,111111111 I h111 ~11111 tllll' i~ s11hj1rttil to

-;- 0 .10
Q.

, ..:,f,

AMS

0,05

AH4

1-

I:

o.oo

T Cm s l

w
z

UJ -0-05

I-

~ -0 . 10
I

z
w
(b)

Figure 6.14 Responses of grid: (a) displacements; (b) moments.

llu cc sets of double triangular impulses P(t) at locations indicated in the figure. For this

plllhlcm the physical parameters are

2

117 X 106 kN/m

10

111 whtl'h SI units 111t impli1d

p = 4.49

P.,m = 100 kN

Mg/m3

L = lm

276

Framed Structures

Chap.a

Sec. 6.6

Programs for Framed Structures

TABLE 6.3 Structural Data for Space Frames

Type of Data
Problem identification
Structural parameters
Space frame data
(a) Nodal coordinates
(b) Element information
Coordinates of point p
(c) Nodal restraints

a Requued

when IP

No. of
Lines

277
y

Items on Data Lines

1
1

Descriptive title
NN,NE,NRN,E,G,RHO

NN
NE

j2

J, X(J), Y(J), Z(J)

I, JN(I), KN(l), AX(I),
Xl(I), YI(l), ZI(I), IP
XP, YP, ZP
J, NRL(6J-5), NRL(6J-4), NRL(6J-3)
NRL(6J-2), NRL(6J-l), NRL(6J)

I
NRN

/i
j3

"""

= I.

o -i1

~7'---X

(a)

Figure 6.16 Space truss with triangular impulses.

b,2

Fifth, let us build upon all of the previous programs to describe Program
DYNASF for space frames. With this type of structure, the six displacements at
a typical node j consist of three translations and three rotations in the x, y, and
z directions, as illustrated in Fig. 6. l(f).
Table 6.3 contains the structural data required for space frames. As for
grids, the shearing modulus G is added to the line of structural parameters; and
each of the lines for nodal coordinates includes Z(J) , as for space trusses.
Element information includes AX(I), X I(I), YI(I), 7J(T), JP, and (optionally) the
l'Oordinn11H of II lhird poiul ,, 1111 lornli111 ptinclpul pl111ws of'lxnding. In thc lines

----x
I
UI

We used this data in Program DYNAST, with IP = 0 for all elements,

DAMPR = 0.02, and called Subprogram NUMINT for numerical evaluation of responses . Figure 6. l 7(a) shows a computer plot of the applied force P (t); and resulting
axial force-time histories for elements 5, 6 , and 7 are given in Fig. 6. l 7(b). Maximum
(or minimum) values of the axial forces in these three elements are 325.6 kN,
- 325.6 kN, and 175 .l kN, respectively.

{111

278

Framed Structures

! f!C. 6.6

279

Programs for Framed Structures

100

TABLE 6 .4

so

Dynamic Load Data for Space Frames

No. of
Lines

Type of Data

a..
~ 0

c::

Dynamic parameters
10

15

20

I.L.

Cms)

8 -so
_J

a..
a..
<

- 100

(a)

1
NNID
NNIV

NNID, NNIV
J, D0(6J-5), D0(6J-4), ... , D0(6J)
J , V0(6J-5), V0(6J-4), ... , V0(6J)

Applied actions
(a) Load parameters
(b) Nodal loads
(c) Line loads

1
NLN
NEL

NLN,NEL
J , AS(6J-5), AS(6J-4) , . .. , AS(6J)
I, BL!, BL2, .. . , BL6

Forcing function
(a) Function parameter
(b) Function ordinates

200

V)

"Omit when IGA

IJ.J

ISOLVE , NTS, DT, DAMPR

Initial conditions
(a) Condition parameters
(b) Displacements
(c) Velocities

Ground accelerations
(a) Acceleration parameter
(b) Acceleration factors

.:!

Items on Data Lines

I
1
l
NFO

IGA
GAX, GAY, GAZ
NFO
K, T(K), FO(K)

= 0.

,,~ those for a space truss element. However, the equivalent nodal loads for a
~puce frame element must be obtained by first resolving the line loads into
1 umponents that are parallel to its axis and two principal directions. Using these
1111nponents and the solutions from Probs . 3.4-1 and 3.4-10 (for axial and
tl11lural elements), we can find equivalent nodal loads in member directions.
I hen the equivalent nodal loads in structural directions may be computed from
l1q. (6.2-13). Finally, all three scale factors GAX, GAY, and GAZ must be
1,1 1vcn for ground accelerations.

0
IL

_J
<('

~ -200

( b)

Figure 6.17

Space truss: (a) triangular impulses; (b) responses.

y

'i'

nodal ~estraints, the terms indicate restraints against three translational and
1 lltc mtat1o~al displacements in the x, y, and z directions.
Dynamic load data for space frames in Table 6.4 are more extensive lhun
101
Hn,Y, 0th.er type of framed structure. The six types of initial displacements uud
Vl' 10~1t1es rn the table consist of three translational terms, followed b ~hll'l'
1111111,onul terms. For nodal loads we have three force co
dy
. .
.
mponcnts an thrl'l'
1
11ortil 11l crn.n(mncnts 1n the .t , y , and z direc tions, us shown in Fig 6 f 8(i) 'f'l .
fhll'l' sl'ls of l111i:11rly vr11 y i11g Iinl' louds in Fig . (1 I 8(h) ait chosl'n ;0
1111 ::.

,;l' tl;l. ~

fl'l1111n 6. IK

I .oudN 011

N[llll'!'

lru,rn (11) 110<1111 lo11dN; (b) line loads on clement.

281

Programs for Framed Structures

Sec. 6.6

E = 6.5 X 10 3 k /in. 2
L = 10 in .

p = 1.71 x 10- 7 k-s2 /in.

G = 2.4 x 103 k /in.2

Ix = 2/y

A = 0.01 in .2

= 21, = 1.667

10- in.

und the units are US.

Using this data, we ran Program DYNASF, with IP = 0 for all elements,
l>AMPR = 0.05, and Subprogram NORMOD for response calculations. The plots in
tlig. 6.20 represent the forcing function for ground acceleration and time histories of the
lnmslational responses at node 2 in the x, y, and z directions. Maximum (or minimum)
values of Dji, Dj2 , and Dj 3 are 0.07213 in., 0.04685 in. , and -0.09977 in., respectively.

"'II)

l0

'

~----x

D
I-

<(

0::

(b)

Figure 6.18 (cont.)

w
-'
w

u
u

<(

200

400

800

60

l 0 00

l 00

ms

-s

Cl

K11.11mplc 6.9

:::>

The space frame in Fig. 6.19 has three rismatic

.
.
stcli<lns that are solid squares with side!o 1 . A~agnesm~ elements with equal cross
14round acceleration Dg1(t) in the x directi~nmPh ~o stown m the figur~ is a rigid-body
ys1ca constants for this example arc

0
0::
0

-1 0

(a)

0 .1 0

DJ3

('\J

o.os

Cl
D

1<(

0. 0 0

Cf)

z
0

~
I-

-o. os

<

-'
(/)
%
;;5

:11

( b)

,
l' ll,IIIII' II jll

- 0 , IO -

'lp 1111 1111 1111

wli lt

} 111111111 11,, i Ii l lllltlll

DJl

282

Ch

'rnmncl Structurn 1

nI

6.7 GUYAN REDUCTION

The concept of

.
matrix co11de11101/011 11 , I mg the number of unknown d .. . ' I~ II Wl'll known procedure for rrd
~pp~ications no loss of accurac/::sl~1~~:~r011ts 111 ''. stuti_cs problem . With NU
is simply Gaussian elimination of d. I o m the ,~duct1on, because the mcrh
analy

tsp acemcnts in m tri ti
. s1~, a s1mtlar type of condensatio
.
a x onn. For dy11u111
bnngs m ~ ad~itional approximation. n was introduced by Guyan [41, whl
Startmg with static reduction we w .
.
free displacements in the partition~d fo;te act10n equations of equilibrium fi

[:: ::][!;] -[~]

SaADA

Solving for th

U, 'l',w U11

+ SanDo = A8

(7b)

I "' 11 lor static analysis, this relationship is correct only when actions of type A
1111 11111 exist [see Eq. (3)]. However , Eq. (7a) follows the finite-element theme
111 "~lnve" and "master" displacements. Differentiating Eq. (7a) twice with
11 , p1'l'l to time produces

(8)
I 111 tI1e purpose of reducing the equations of motion to a smaller set, we can form
1111 transformation operator

111 which 18 is an identity matrix of the same order as S 88 . Substituting Eqs. (7a)
(8) into Eq. (4.4-1) and premultiplying the latter by Tl gives

111111

(10)

D;1 = S,U (AA - SABDo)

111 this equation the matrices S18 and Al still have the definitions given in Eqs.
1,1 nnd (6) . However, the reduced mass matrix M%8 is found to be
(4)

in which

and

(.,)

Al= Ao -

s-

SBA M 1 A A
From Eq .. (4) we see that Eqs . (2) have been
(ll)
~~ly_the independent displacements in vecto r~uc~ to a smaller set involvi111l
as m Eq . (5) is a modified version oft r .n: e reduced stiffness matri,
reduced action vector A*
in Eq (6)
?e ongma1 submatrix Saa. Also 1/1"
8

th t
contams tenns
d'f
' "
a are considered to be equivalent loads f
mo t ymg t~e subvector A,,
Fu~ennore, Eq. (3) may now be view~ ?pe B due to actions of type A.
required to find vector DA exactly fro
s the back-substitution formulu
T

m vector D
ummg next to dynamic reducti
B
undamped equations of motion for free od~' Jwe recall from Sec. 4.4 that the
tsp acemcnts arc

Ml) -ISD
11

A
(4 .4 I)

lll'W approxi murion lhur 1hr <li1;pl11cl111t 111~ of

(7a)

(9)

Substitute Eq (3) t Eq
.
mo
. (2b) and collect terms to obtain

Then assume m,

ol typ1 II , llli l111l11w11,

= AA

e vector of dependent displacements DA in Eq. (2a) yields

p1 111ll t1l 011 ll1<1Sl'

IIJ

Here the subscript A denotes the d' 1

subsc t B ti
tsp acements that are t be 1
np_ re ers to those that will be r .
o . e immated, and th
of equations, as follows:
etained. Now rewnte Eq. (I) as two ~or

SMDA + SAaD8

213

C111yn11 Unclw 11011

ry pe A

llll'

M%o = TJM Ta= Mso + T1MAB + Mo;1T;10 + T io MMTAB

(11)

A~ mentioned before, all of the condensed matrices in Eq. (10) are approximate.
II clumping is to be included in the equations of motion, a reduced damping
111111rix 0 also can be derived, which has a form analogous to Ml 8 in Eq. (11).
When applying Guyan reduction to framed structures, we usually choose
111t111ions at the joints of beams , plane frames, grids, and space frames as the
,lqitndent set of displacements. However, the method can be used in a much
111111c general manner for various discretized continua. That is, any arbitrarily
-rll'l'lcd set of displacements may be referred to as type A, while the remaining
,li~placements become type B. The trouble with this generality is that a good
, hoice of "slave" and "master" displacements is not always obvious. Even with
l111111ec.J structures there are cases when joint rotations are more important than
t11111slations and should not be eliminated.

c;

l1 11111nple 6.10

1111' fixed beam in Fig. 6.21(a) is divided into three flexural elements, each of which has
1lw ~111ne properties, I , p, and A. By Guyan reduction , we shall eliminate the rotations
111 11mb 2 and 3 and retain the translations. The reduction will be followed by a
,1l111111onnl annlysis that is compared against exact results.
For 1h is cx111upk, the assembled structural stiffness matrix (without consideration

284

Fromod Structures

Sec. 6.7

285

Guyan Reduction

of restraints) is
y

,----------- -------,
/ 6

, 3e

I
/ -6

= 2/

e3

CD

ze2

-3e
e

Sym.

r--------4---- ----I 12

I
I
I
1
2
1
I
,
'2"
I
I
'-t:J
I
I --------+---------1------- ---l----------1
I
I
I
I

3e
0
0

I
I
I
I

4e

-6

-3

12

I
I
I
I

42

I
I
I
I

'3"

-3e

o I, 3e

e2

I - 6
I

5
6

6
-

(a)

I
I
I
I
I

'-:!.I

L------- - ~--- ---- - I

3e

2e2

(a)

L - - - - - - - - - - - - - - - - -- - 1

Similarly, the assembled structural mass matrix becomes

r ------ - ---------- ---~

/ 156
I
I
I
I
I

pM

22

54

13

,-

/- Be

I
I
I

- 3 l
2

CD

I
- - - - -- - - --l- -- - -- - - -I
312
I
1
@

se l
2

Sym.
,

3
4

I
I
I

Ms = 420 L---- - -----f----- ------/--- --- ----f-- --- ---- I

54

13e l

312

/ -13f

-32 /

82

Q)

'----------4----------1
/

/
I

2
o !l___- _____
13e -3e
-22e
_ _ _ ___
_ ___ _42
__ /

54

13

156

/
!

(b)

- 1

7
(c)

8
Figure 6.21

(a) Three-element fixed beam; (b) mode I; (c) mode 2.

(h)

4e

Dushcd boxes in Eqs. (a) and (b) enclose the contributions of elements 1, 2, and 3, which
urc drawn from Eqs. (3.4-24) and (3.4-26). Moreover, the joint displacement indexes for
the problem [sec Fig. 6.2 l(a)] are listed at the right side and below the matrices.
As the first step , we remove the first, second , seventh , and eighth rows und
tolumns from matrices S, and M,,, because displacements I , 2, 7, and 8 arc rcstrai,wd
hy supports. Then the remaining 4 x 4 arrays arc rcarrnngcd to put the rotationul torn1N
heforc the trans lutionul terms, as fol lows:

SAA SAD]

S=

[ SaA

Sall

2EI

Sym.

12

- 3e o i

-6

12

e2

4e2

--------j------0

3e

(c)

lU/

Fromod Struoturoe

186

pA I'
420

IJe I

()

13e

Cl11y1111

ti. /

llcuhmtlo11

"'h11l' L
3e. These values arc in error by I O.h 1% 1111d I 2.6%, respectively; and they
1111111titute upper bounds of the exact angular f'n:qucncics \51. The corresponding mode

Sym.

K('J

Ill ,

h11p1s arc
(k)

312

54

312

5
wh1l'11 appear in Figs. 6.21(b) and (c) .

1111llll

Eq . (c) the inverse of SAA is

4 -1]

e [

- 1 _

30/ -1

AA -

of S from Eq. (c) into Eq. (5) prod

Substituting this array and the other submatrices
.
J2/ [
Ull
6
St11 = &
_
E!
= 6El [
1
2
5e 1 4
5e3 - 11
16
5
<, l

2 -lJ _ [4 l]

16 -11] 3
3

whk h,i:~ the reduced stiffness matrix .

I o reduce the mass matrix we form the tr
.
'I\,, lo l!q. (7b) and substituting it into Eq. (9). ;~~~~rmat10n matrix Tn by evaluati11p
TAB --

s-AA1SAB= -1
5e

1111d

Tn = TAB
[ In

= I_

1
-4

4]
_1

1111 cc degrees of freedom indicated in the figure.

The 3 x 3 stiffness matrix for unit values of D 1, D2, and DJ has the form

l
I

(C)

11ml the accompanying 3 X 3 mass matrix is

4
1

(h)

MAB

5e

111

22L

22L

! 732

[El

\JpA

(n)

Ahm , the corresponding mode shapes are

'"'

(m)

(i)

- 0.5528

Jn

Sym.J

---------l----

l
= 3.201, 15.14, 32.68 L 2

22.~l ,(,l)/1 I I J

8L 2

which M includes pAL for the mass of member 1. With these arrays we can set up
08
Nolve the eigenvalue problem, which yields the angular frequencies

wh11h is the reduced mass matrix.

theEqterms
. .
111 th1 Using
lrn111 or
. (3 .6in4)matrices
Tl . S111I and
. f M*
1111 , ~e ~an set up the eigenvalue problen,
. 1c angu a1 rcquenc,cs found hy this method arc
"11 ,J

420

8L
- 3L 2

1111tl

Wi , 2, 3

319] 3
[1696
319 1696 5
3

= pAL

88

111 this cnsc the submatrix 1/J in the lower artition of

.
.
1111 two rnmaining translational <lisp!
p
(
matnx 'Is is of order 2 because then
M lrom llq. (d) and T from Eq (ha)c~menEqt
s numbe~s 3 and 5). Substitution of matrii<.

mto
. (11) yields
11

Mlo = pAC
2100

ll.x111nple 6.11
l lgurc 6.22(a) shows a rectangular plane frame that has the same properties E, !,, p, and
\1 l'or each of its three members. Let us find approximations to the fundamental angular
lnqucncy for this frame with and without elimination of the rotations at joints 1 and 2.
l'o Himplify the analysis, we shall omit axial strains in the members, leaving only the

(g)

=-~e----0
1

Frequency coefficients /.Li for prismatic beams with various end conditions
1111 summarized in Table 6.5. In each case the beam is modeled by four flexural
1 kmcnts, and the results for the consistent-mass approach (with and without
1 ll111ination of rotations) are compared with those for the lumped-mass method
(with elimination of rotations). The table shows that the consistent-mass model
p1oduces much better accuracy than the lumped-mass model in beam analysis.

'

I)

<I> = f<1>,

I hi Hl111p1 ol

11111dr

<1>2 <J>3) =

- 0 .5528
I.

I I~ ill11 HfrlllCd in Fig . (i.l.2(h).

- 1

-oLJ

(o)

- - ---------,

Frequency C oefficients ; for Pr ism atic Beams Mod eled by Four Elements

TABLE 6.5
Support Conditions

Mode

~
I

f=

% Error

CM- TO

% Error

LM- TO

% Error

9.870

9.872

+0.020

9.873

+0.030

9 .867

-0.030

2
3

II

+0.38

39.76

+0.7 1

88.23

90.45

+2 .5

94.03

+6 .6

83.2 1

- 5. 7

22.37

22.4 1

+0. 18

22.46

+0.40

18.9 1

- 15

62.06

+0.63

63. 12

+2.4

48.00

-22

w =~
L1

ff.
pA

12 1.9

I
I

+0.83

22. 37

22.40

+ 0 . 13

61.67

62. 24

+ 0.92

123.5

122.4
22.4 1

I
I

62.77

+ 2.2

3.5 16

3.5 16

124.8

+0.00

3.5 16

I
I

+ 1. 2
+ 0 . 18

39.19

86.84
22. 30

I
I

-28
- 0 .3 1

+ 1.8

59.25

-3.9

+3 .2

97.40

-19

3.418

+0.00

-2.8

22. 03

22.06

+0. 14

22.09

+0.27

20.09

- 8.8

6 1.70

62. 18

+ 0.83

62. 97

+ 2. 1

53. 20

- 14

15.43

+ 0 .065

15 .40

I
I

49.05

I - 1.8

I 91.53

I -12

I -0.73

39.63

120.9

39.48

.9

J.."IO

CM- TR

~ .67

frurl

Exact (5)

I
I

15.42

I 104.2

49.97

I
I

15.43
50.28

I 106.6 I

+0.065
+0.62

+2.3

50.56
110.5

+ 1.2

I + 6.0

- 0. 13

CM: Consistent Masses

LM: Lumped Masses
TR: Translations and Rotations
TO Trm,.,IM.-,. Onl~

"',,

a;

'X

:....i

-:
~

':

::;

G)

g "Q
::.

-<
I

('";

:::. -0
o;

:,

"':.

aci'

-:

o'
'"'

-:

0=:

O'Q
-

!;

c.
::s

sCl>

~
~

:,
.;"
t

ol
3
0
-

:,:,
a,

:,

..,

- -

"':.

::..

,z

,z

:,..

--

,.
-

(IQ

5-

...,

o
(>

.::__i

.,

00

0\

...,

"O
C:

C.

l..,

,-

-'

8-

r,

C:

$l.

'<

:,

1----

l
N'

.i "'

s:i i

'-,.'1-1>

a.

o'
~

:i

3
0
&
0

l t

tJ

"'

f'

__l

r>

SJ"'~

N~

,-

.\

g,

"'

JC

"'g.
3

,--.
"O
.._,

5.
;,<

290

Framed Structures

= 24EI _

Sto

36E/

Chap. 8

= 84E/
(q)

5L3
5L3
which is just a single term. Next, the arrays TAB and To are found to be
L

= -s;J sAB

TAB

= _ _l_ [ ]
5L 1

(r)

Sec. 6.8

Constraints Against Axial Strains

291

In this equation the symbols D j t and D j 2 represent the x and y translations of joint
j, and Dki and Dk2 are those at joint k. As before, the x and y direction cosines
of the axis of the member are denoted by c., and Cy.
Similarly, we write the axial constraint condition for zero elongation of a
space frame member [see Fig. 6.5(b)J in the following manner:
(Dit - Dk1)cx + (Di2 - Dk2)cy + (Di3 - Dk3)c,

and

(s)

_ ToT M

569
To = pAL
350

(1)

Now the eigenvalue problem gives

]~
w, = 3.215-/
2

(u)

pA

Cornpari_ng this angular frequency against w1 for the unreduced 3-DOF problem we find
the relative error to be +0.16%.
'

6.8 CONSTRAINTS AGAINST AXIAL STRAINS

In Example 6.11 we omitted axial strains without difficulty because members of

the plane fr_ame wer~ pe1:Pe~dicular to each other. However, for frames with
me~be'.s onented ~rb1tranly m_ a plan~ or in space, the matter of neglecting axia l
strams 1s not so simple. In this sect10n we will introduce axial constraints in
plane and space fra?1es, pri~arily for the purpose of reducing the number of
d~grees of freedom m dynamic analysis. The number of constraints introduced
will be equal to the number of members m for which axial strains are to bl'
neglected. By automatically selecti~g m of the joint translations to be dependent
the rest of th~m, we ca? devise a method for reducing the number ol
mdepend~nt_eq~ahons of motion. This reduction method may be combined with
Gu~an e1Jmmat10n of joint rotations, which was described in the preceding
section. 'v!e ~ssume that the frame to be analyzed is underconstrained, so that
no c~mphcat1ons from redundant constraints (or superfluous members) need tw
considered r6J .

?n

The axial cons~'.aint condition for zero elongation of a typical member; in

a plane frame !see f<l g. 6.2(b)I muy be stHted as

(n,,

n11 k,

(n,.

n1 ,l1\

(1l

(2)

I !ere Dj3 and Dk3 are translations of jointsj and kin the z direction, and c, is the
z direction cosine of the member axis.
Assembling these constraint conditions into a matrix format for either a
plane or a space frame gives

Using the latter operator in Eq. (11), we can reduce the mass matrix to the single term

*
M BB

=0

C D1 = 0

(3)

The matrix C in this expression is called the constraint matrix, which contains
only positive and negative values of direction cosines for the constrained members. This array is of size m x n,, where n, is the number of joint translations.
The vector D 1 in Eq. (3) consists of only free joint translations , because no joint
rotations are involved at this stage.
Due to the constraint conditions, some of the joint translations in the vector
1)1 will be linearly dependent on others. To determine which translations are
dependent and which are independent, it is necessary to investigate the rank of
111atrix C in a systematic fashion. The rank rand the basis (or vector space) of
( ' are found using Gauss-Jordan elimination with pivoting [7]. Although the
1ank of a matrix is unique, the basis is not. Therefore, the choice of dependent
translations is arbitrary , and pivoting automatically produces the best selection.
Thus, it becomes possible (in retrospect) to partition the matrices in Eq. (3) as
lollows:

[C 11

C,2{~:]

(4)

111 this expanded form the vector D 1 represents r dependent translations, and D2
rnntains the remaining n; independent translations. Because no redundant con~t1nints are included, the rank r of matrix C will always be equal to the number
ol members m. Therefore, submatrix C 11 in Eq. (4) is a square array of size
111 X m, and submatrix C 12 is of size m x n;.
Multiplying the matrices on the left-hand side of Eq. (4) produces
(5)

l, 1mwing that matrix C 11 is square and nonsingular, we can solve for vector D1
111 tl't ms of vector D2 . Thus,

(6)
111

whil' h llw

OJll'l ator

T1 2 is
(7)

293

292

Framed Structures

Chap.

lluc. 6.8

Constraints Aga inst Axial Strains

TABLE 6.6 Member Information for

Three-Member Plane Frame

During the Gauss-Jordan elimination

h
.
identity matrix; and C12 is replaced b pr~~ss, t e matnx ~11 is replaced by an
to an identity matrix I of order m / 11 ~- If the operations are also applied
this inverse to be the ~perator
' i wi e replaced by C ;-t We now define

T11

= C 11

Joint j

Joint k

Cx

1
2
3

1
3

0.7071
0
0.4472

(8)

and the augmented constraint matrix is

Ci

[C Im]

(())

To confirm ideas regarding the constraint

.
.
frame in Fig. 6.23 for which th
b . & ma~x: let us consider the plane
'
e mem er m1ormat10n is given i T bl 6 6 T
augmented constraint matrix [see Eq (9)] ior
&
.
n a e . ho
th"1s case is
1

c,

= [o.~001

o.101
-1.000
0

-0 .101
0
0.447

-0.101 1.000
0
: 0
-0.894 : 0

o
1.000
0

o
0
1.000

Cy

Member

1
4

0.7071
1.000
-0.8944

~hown in Fig. 6.18. After the Gauss-Jordan procedure is applied, the matrix in

11.q. (a) becomes

1

c;
(a)

-[f

0
1.000
0

-1.500
0
-0.500

l'hcrefore,

The numbers above the columns of C indicate the translational displacements

T,1 = C 1/

1.000
-1.000
0

0 II 1.414
0 II 0
1.000 l 0

[.414
=
~

-1~18]

1.000
-1.000
0

-Lit~]
0

(b)

-1.118

(c)

-1.118

0.5P(t)

P(t)

1111<.I

t2

T12 = -C;-/ C,2 =

t
P(t)

, - - ----t--- - \ - - - - - -1.

[~J

'"
u111111h11 pl11111 11 111111

(10)

1111' llrst part of D consists of a null vector, representing member elongations

2
urc z.cro) ; and the second part contains the independent translations D .

'1111,u

(d)

I whk h

I 11 11l111c the vector D to D,, we write

111111111 t,,1.1

0
0.500

Thus, the dependent translations are found to be 1, 2, and 5, whereas the

l111h.1pcndent translation is automatically chosen to be 4. Note that there is no
1111,d to rearrange the augmented constraint matrix during this procedure.
In preparation for a coordinate transformation associated with axial con~1111 lnts, we shall define a generalized displacement vector D, as follows:

D
3/ _ _ _ _ _ _ __J__ _

[500]

= T,D,

(l la)

(l lb)

294

I nc. 6.8

Framed Structures

In Eq. ( 11 b) the upper part represents the constraint conditions, and the lOWl'f
part merely re~roduces D2. Because the generalized displacements are indepcn
dent and constitute a complete set, there is also an inverse relationship in th~
form

D, = TcD

( I.? I

where
1
1
Tc= T, 1 = [C,1 -C,1 C12J = [T11
0
12
0

Ti2J
12

1
]

(4.4-1)

( 14)

in :Which the subscr~pts rand t denote rotations and translations , respectively

lJs1~g Guyan reduct10n from Sec. 6. 7, we eliminate the rotations in Eq. (14) and
1cta111 the translations, yielding

M 1~D1 + S!Dr

= A;

( I ~)

where

s: = Sa + T~Srr

( I (1)

A t* -- Ar + TTrrA,

( 17)

M!

= Mu

+ T~M,r + M,, T,.r + T~M,, T,.,

( 181

'l'itl' transformation operator T,., in these expressions relates the rotations D to tht
trnnslntions Dr, as follows:
'

D,

= T,.rD,

( Ill I

Whl'l'll

S,, 1 S, 1

s*rr D,-M*
11 D, +

A* + cTQ

(21)

111 which Q is a vector of axial forces in the constrained members.

To perform the transformation, we substitute Eq. (12) and its second
dl.lrivative with respect to time into Eq . (21) and premultiply by Tt to obtain
M

D+ S D =

C Q

(22)

Writing this equation in expanded form, we have

By segregating rotations from translations, we can write this equation in thr

cxpunded form

[M,M,,, M,M,, [D'ti, ] + [S,,S,, S,Su [D'Dr ] = [A']

Ar

~tiffnesses of members from the structural stiffness matrix. Then to satisfy

,quilibrium at the joints, we must revise Eq. (15) to become

( I II

M~t:ix Tc_ is an operator that can be used to transform action equations from thl'
ongmal displacement coordinates to the generalized displacement coordinatl'N
for _axial ~onstraints. Notice that submatrices T11 and T12 are generated auto
mat1cally m the Gauss-Jordan procedure described previously.
.
Let us now restate the action equations of undamped motion for fi'l'l'
displacements at the joints of a plane or space frame as
MD+SD=A

295

Constraints Against Axial Strains

()OJ

Tlw opt111to1 'I~ lstc Eq . ( 11)1 cun now ht IISl'd to trun sforn1 the rcduttd
1q1111tH111s of 11101l1111 i11 llq . ( I Ci) to lht /'t'1wrnll1td dlspl11tt1rn111 rnordinhti1, I)
Ill J1q ( I()) 1!111111~1 11~i11l l'OIISl111int~ JIil' Ill h1 1111pc1s1d , Wl' i,lwll 11111it 11x111l

M 11
[ M 21

~ J+ [~11
S21

M,2] [
M 22 D 2

~12] [ 0
S22 D2

[Ai]
A2

+ [l']Q
0

(23)

Multiplying terms and rearranging the results produces

(24)

1111d

Q = M12D2 + S12D2 - A,

(25)

I he barred matrices in these equations have the following definitions:

+ TT1 sf2
S22 = TL S i1 T12 + TT2 S i2 + Si'i T12 + Si'2

(26)

A1 = TT1Ai

(28)

S12 = TT1 st\ T,2

A2

= TT2Ai +

(29)

Ai

M12 = TT1 M~ T,2 + TT, Mf2

-M22 = T12M11
T * T,2 + TT12 M*12

(27)

(30)

+ M*21 T12 + M*22

(31)

1q1111tion (24) represents a doubly reduced set of equations of motion that can
Ill' Holved for the dynamic response of tJ.ie independent translations D2. Then the
vn tor D2 and its second derivative D 2 may be substituted into Eq. (25) to
,h ll1mine the vector of axial forces Q in the members. Next, dependent transl111 l1111s D I can be obtained from Eq. ( 6), and the rotations D, are found using Eq.
, 11)) . Finally, other internal actions and support reactions may be calculated
l1rn11 known relationships.
If a damping matrix is included in the equations of motion, its reduction
I "i111ilur to that of the consistent mass matrix. Other topics that could be
11111111dcrcd in this section are redundant constraints and nonzero length changes
111 111l' lllbcrs. I lowcvcr, these subjects arc more complicated and of less interest
1111111 thl 111atlur or Hxinl c:onstrnints, us discussed here.
Wht11tvtr 11xi11l st rnins 11n 0111ittld from :urnlyscs of plane or space frames,
,1 h1~11 11l 11l'l'll1'11l'Y ,~ h111111d t11 mt111 'l'hr 11l1 nilll1111l'l' or such discrepancies will

296

Framed Structures

Sec. 6.8

vary from one problem to another. However, for most practical underconstrained frames, the loss of accuracy due to introducing axial constraints is
likely to be negligible, except in the columns of tall buildings [8] and similar
structures. Moreover, the numerical problem of ill conditioning due to combining large axial stiffnesses with small flexural and torsional stiffnesses is
completely avoided. Of course, when the members in a frame are perpendicular
to each other (as in Example 6.11), omission of axial strains is easily accomplished without the formal procedure of this section.

AL

~ ~20

M,,

- 8 288

6:

"

S = El, [
L2

"

3.879
-2.121

2.121

2.121

2.121

3.195

[ 1412
_ El, -2.121

S - -3
"

0.1564

L [

-I

-2.121
-1.585

2.121

- 2.121

2.121

T,, ==

- s;,. 1 s,,

0.6764

2.980

2.121

-2. 121

-1.692

-0.2301
- 0.3890

S,; ==

El,

_ pAL3 [ 8
M,, - 420 -3

M,,

p/\l}

I W

r I 92
6 .444

<) ,

2
15 .56
9. 11) }

(h)

9. 192

\\21

5
9. 192] 3
5.7 12 6

(ll

5
Sym.11

0.808
0.391

0.565

2.552

-1.017

- 0.303

(n)

1.374

(o)

A: == {1, 1, 0.5, O} P (t)

llq. ( 18) gives
2
[ 317 9
-3 1.46
M" - 420
81.91
1.733

to be

-2.189
-1.463

* _ pAL

-3] 3
8 6

(m)

111 11ddition, Eq. (17) yields

,111tl

si

9.940

Si mil arly, consistent mass matrices for the members are drawn from Eqs. (6.2- 15),
nNNllmbled, and rearranged to give

-0.1787 0.2558] 3
-0.6374 0.2650 6

I !Hing Eq. (16), we find the reduced stiffness matrix

2.336

"'

(g)

(f)

t'hlln from Eq. (20) the operator T,, becomes

(f)

Syml

(k)

- 0 .04790]
0.2313

S,, = El, -0.04790

2
3

A, = {l , 1, 0.5, O} P(t)

291.2

For the purpose of applying Guyan reduction , we calculate the inverse of sub11111lrix S,, as

L
(c)

(j)

8 300.8
62 - 1.6

== .!.[-0.7083

1.414] 3
1.414 4.617 6

11

, . p1g .6. 23 , we see that the parts of the rearranged load vector are
I1om

_ El, [ 6.828
s -

Sym

[3~

A, = {O, O}

Example 6.12
For the three-member plane frame in Fig. 6.23, let us first set up the stiffness, mass, and
load matrices for the six degrees of freedom shown. Second, we shall use Guyan
reduction to eliminate the two joint rotations and retain the four translations. Third, by
imposing axial constraints we will eliminate three dependent translations and keep the
best single independent translation, which was found to be displacement number 4. Last,
we shall calculate the response of the reduced system to a particular set of forcin1,1
functions.
From Eq. (6.2-14), we determine member stiffnesses (without axial terms), as
scmble them, and rearrange the results to produce the following submatrices of thl
structural stiffness matrix:

297

Constraints Against Axial Strains

Syml

289.1

(p)

1.152 262.0
62.84

6.302

293.6

Now we shall furth er reduce the stiffness , load, an~ '.11a~s matric~s to accoun~ f?r
mcmhllrs. By rcnrrunpiog 1111d p,1rt1t1omng the stiffness matnx m

,1, 111 111111 strniuts 0 11 the

I q (II) Wt' fi11d thnl

299
Framed Structures

298

*
S11

9.940

(24) takes the form

Sym.] I

0.808

2.552

-1.017

1305pAL ..
D4
420

(q)

1.374

22.41E/,
L3

O
D4 = 2. A

(a')

l'he eigenvalue problem associated with the homogeneous version of Eq. (a') yields the

* = El,
S 21
L3 [ -1.463 0.391

Programs DYPFAC and DYSFAC

We now have all the matrices needed to find the response in Eq. (24) and the axial
lorces in Eq. (25) for any forcing function. If the function P(t) is a step force A, Eq.

El,
= L 3 -2.189

Chap,

Sec. 6.9

- O. 303] 4 = S *1/

(1)

11ngular frequency:
(b')

* = El,
S22
L 3 [0.565] 4

11nd the response to the step force Pi is

P, L 3
EI,

the last of which is just a :ingle term. Similarly, for the load vector in Eq. (o) we haw

A, = {l, 1, O}P(t)

A; = 0.5P(t)

(I)

Also ,

Furthermore, submatrices of the mass matrix in Eq. (p) are

Mi,=

~fa

[~~;::6

l
M*21 -- pAL[
420 81.91

62.84

(c')

..
AL3w2
D4 = 0.08925-COSWt
El,

289.1

1.733

D4 = 0.08925-(1 - cos wt)

Sym.]

293.6

A
pAL

= 0.6437- cos wt

(u)

Substitution of D4, D4 , and the appropriate barred matrices into Eq. (25) produces

1.152 6.302] 4 = Mi2T

Q=

(v)

4
(w)

M;2 = ~fo[262.0] 4

(d')

Then the barred matrices in Eqs. (26) through (31) become

0.4442]
( 0.6454]
1.617 A 0.7370 P, cos wt
( -0.7714
- 0.6638

(e')

which are the time-varying axial forces in the members. To complete the example, we
also compute the dependent joint translations, the joint rotations, other internal
1111;mber actions, and support reactions. However, these tasks are straightforward and

1 ould

will be left as exercises for the reader.

s,

= ~, [

~::~~]

-21.17

A, = P(t) [

4
1.:l ]
- 1.118

4
EI
S22 = e'[22.41] 4

11.9 PROGRAMS DVPFAC AND DVSFAC

(x l

A2 = P(t)[2 .0] 4

(y)

79 1.31 1
574.3 2
799,7
h11

111 this section we briefly discuss programs named DYPFAC and DYSFAC for
dynumic analyses of plane frames and space frames with axial constraints. These
p1 ograms use Guyan reduction (see Sec. 6.7) to eliminate the joint rotations and
lhl'll further reduce the equations of motion by imposing axial constraints (see
'{l'l'. 6 . 8) . Response calculations for the reduced system are carried out using the
11m nrnl-rnodc method , as described in Chapter 4.
Programs DYPFAC and DYSFAC both have the same outline, as follows:

4
pAI,

470 11 105 1 4

(/)

Outline of P10~1nms DYPF AC and DYSF AC

1111'10' 111111 yN 1111' i111h- 11tN I ' ') 11111 I I I Il' llt lh' llll' lllhl'I 1111111!1111~. w h111l'IIH ll1thx ,1

II1(11t'Nl' III ~ 1hr ltllll'( II 111h Ill 1111111 IIIIIINl11ll1111

lfrud 1111d wil h Nl111l tw11I tl11t11

II

S t1111 t11111l p 111 11111t h 1P1

301

:too

I 11111111<1

b. .loi111

m, 1101111111

1111111111111

llYl'I Al

irnit 11V

I (\I

l'111111ll1111t,

C. Joint l'estnli11I N

2.

3.

4.

5.

6.

7.

d. Member 11110111111111111
Porm constraint lransf'o1111111io11 1111111 kc.:s
a. Assemble constrnint 11111td x
b. Find dependent translations
c. Generate transformation matrices
Generate, condense, and transform stiffness and mass matricts
a. Generate member matrices
b. Assemble structural matrices
c. Eliminate rotational displacements
d. Reduce matrices (due to constraints)
Determine frequencies and mode shapes
a. Convert eigenvalue problem to standard, symmetric forn1
b. Calculate eigenvalues and eigenvectors
c. Write natural frequencies (cps)
d. Transform, normalize, and write modal vectors
e. Normalize modal vectors with respect to mass matrix
Read and write dynamic load data
a. Dynamic parameters
b. Initial conditions
c. Applied actions
d. Ground accelerations
e. Forcing function
Calculate time histories of displacements
a. Transform initial conditions and loads to normal coordinatl'N
b. Determine time histories of normal-mode displacements
c. Transform independent translations back to physical coordi,11111 d . Calculate dependent translations and rotations
Write and/or plot results of response calculations
a. Time histories of independent and dependent joint translation
and rotations
b. Maxima and minima of translations and rotations
c. Time histories of axial forces and other member end-actions
d. Maxima and minima of axial forces and end-actions

Both of these programs perform the reductions optionally. Therefore, tw11

mw structural parameters, [RO and IAC , must be added to the structural dut11
Within the logic of the program , if IRO ,f:: 0 the rotations arc eli minated; wul 11
IA( ' I O axial constraints arc imposed . Further details regarding these progn11111,
111t pivcn in Rel'. 9.
1,:,m1111>1t, <,. IJ
I iHIIH'

(1

J,I(11) II l11Ht rnhN 1111 1111tll1co11s1rnl111d pl11111 I 1111m with I wllw cl1H1l'l' Nof I 111

I'/.

1' (,)

G)

3L

12

/.

-l---..+----X
2L~

(al

P(t)

P,

t,

0
(b)

ix-member plane frame; (b) dynamic load.

Figure 6.24 (a) S

aoa
I rru1111d St1uotur 1111

.102

1'1111!1111111

c;h111
1111 1 tll'h

rn11111l~l 11I I ti tldl .

..
.
.
dom ( 12 DOF) . 'l'I1i.l. I1111111
.
I Y i onn~ctcd pr1s mt1t1c me111h1:1N' 1111
which have the sa me vah1us ol / ' I
111 thu .1 di rection at jo ints 2 u11d ,'1' , '
~ I >y111111 tl c lorces p (1) and 2P (1) !Ill' nppll
Assuming th at the fram; is steel , '".11 t II t 1111 v111111 tion of P (1) appears in Fig. 6, 'ltl 1
' Wl ww ll ll' ll1ll11wlng numerical values ,0
r 1 pan1111r11 1

I ';"''

E = 30

104 k/'m. 1

A = 30 in. 2
L = 50 in.

fJ

7 .35

t2

110111

10- 1 k-s 2 /in. 4

I, '"' 1.0 X 103 in.

10 k

of the th1 cc l'IISl'S W1 Hl I 1\1111 1111 II MpOIIHl' C\IIVC for Case (b) is indistinguishthat fo r Cus1: (11) . 111 11dd l1h111, till' H'HJlllt1Se curve for Case (c) differs by only
11 11111 umount from thu olhl'IH (111nt Is 11pprnxl mate ly - 3.3% at the first peak). Thus,
1111 hvcl o f accuracy shown hy 1hls L'xumple seems sufficient for practical purposes.
1lih

REFERENCES

= 2t1 =

Weaver, W. , Jr. , and Gere, J. M. , Matrix Analysis of Framed Structures, 2nd ed.,
35 ms

whure US units are implied.

.
T hi s structure was analyzed b Pro ra
.
wtlhout reduction (12 DOF) (b) tyh 1 g_ m_ DYPFAC three times, as foll ows t111
Ii . .
'
w1 e immat10n of r t t'
(8
'
~ m1nut1on of rotations, followed b reduc .
o_a 10ns
DOF); and (c) wllh
6.25 shows time histories of respofse t
due to axial constraints (2 DOF). Fip111
or e independent translation D4, as comptth ii

Vnn Nostrand Reinhold , New York , 1980.

Weaver , W., Jr., and Johnston , P. R., Finite Elements for Structural Analysis,

Prentice-Hall, Englewood Cliffs, N. J., 1984.

Yoshida, D . M. , and Weaver, W. , Jr. , "Finite-Element Analysis of Beams and
Plates with Moving Loads," Int. Assoc. Bridge Struct. Eng., Vol. 31, 1971, pp.

:~on_

179- 195 .
<luyan, R. J., "Reduction of Stiffness and Mass Matrices," AIAA J., Vol. 3, No.

D,(i n.)

0.2

G) and@

l, 1965, p. 380.
, 'l'imoshenko , S. P., Young, D. H ., and Weaver , W., Jr , Vibration Problems in
11

/1'11gineering, 4th ed. , Wiley, New York, 1974.

Weaver , W. , Jr. , and Eisenberger, M. , "Dynamics of Frames with Axial ConHtraints," ASCE J. Struct. Eng., Vol. 109, No. 3, 1983, pp . 773- 784.
Ocrc, J. M., and Weaver, W., Jr. , Matrix Algebra for Engineers, 2nd ed. ,

!\rooks/Cole, Monterey , Calif., 1983.

II Weaver, W., Jr. , Nelson, M. F., and Manning, T. A. , "Dynamics of Tier Buildings,"
ASCE J. Eng. Mech. Div. , Vol. 94, No. EM6, 1968, pp. 1455-1474.
1 Eisenberger, M., "Static and Dynamic Analysis of Plane and Space Frames with
1
Axial Constraints," Technical Report 48, J. A. Blume Earthquake Engineering
Center, Stanford, Calif. , Oct. 1980.

PROBLEMS
r,,:.t l.

Suppose that a uniformly distributed force by>(t) acts on half the length (from
x' = 0 to x ' = L/ 2) of the plane frame member in Fig. 6.2(a). Find the equivalent nodal loads at joints j and k for both member and structural directions ,
assuming that

Xjk

= 4 and

Yik

= -3.

td.2, Let the concentrated force Py(t) on the plane frame member in Fig. 6.2(a) act at
the point where x ' = 3L/4. For both local and global coordinates, determine the
II l,

equivalent nodal loads at points j and k, with Xjk = Yi k

t A moment M, (t) is applied at the midpoint of the plane frame member in Fig.
6 .2(a). If XJk = I and YJk = - 2, calculate the equivalent nodal loads atj and k for
both member and structural directions.

f1,L4, A lri angul ar distribution of force ~b 2(1) acts in they' direction on the plane frame
111c1nhcr in Fig . 6 .2(a). rind the equivalent nodal loads in local and global

o.~

d \rlCtillllH rm 111
1 und Y/4
4'
r, ,,l,,!I. 'l'w11 cum111ti11l1cl l11H'l'S P ~(I) 1111 11pplk d ut the third points of the plane frame

304
I t 11111ml Str 1101111111

r11crnhc1 In Fig h '(II) c ,111 1

directions usii1u I
11
q111 1 ,tit Ill 1111d11l louds in memhcr and ~,1111f
JI
11111 I ,1
I
6.2-6. Assume that two 1111 ,, 11 1 111
1I) ' " I ut tlw q1111rter points of the pl11111 Ir
member in Fi
l 11 1
g . 6.2(u). lktt1111111r iq11l v111l-11t nodal loads in Joe I
I I
.
coord mates for x1k = YI*
u u11t 11
6. 27. A parabolic distribution of force /J (t)
. .
frame member in Fig 6 2(a) F
.. 21s applied m the x ' direction on 1hr pl,11
I d .
. .
. or XJA
and YJk = - I find th
. I
oa s m member and structural directions.
'
e equ1vu t11t 111 I
'

ti '',"

'

"

6.3-1. Let a uniformly d'st 'b t d f;

the grid member /n ;i u e orce b, (t) act on ~alf the length (L/2 s ,\ '
I I
and global directions, ti!:.~:)~ ~al:~a:e e~u~valent nodal loads for both I,
6.3-2. Assume that a concentrated f;
P. ( ) . 1k
the grid member in Fig 6 3(o~ceF z t I~apphed at the point wherex '
IIII

a1 .
or x 1k - 3 and y I d

alent nodal loads in both me b

d
j~ -: , etermme tht ,qui
6 3.
.
.
m er an structural direct10ns.
,. 3. The gnd member m Fig. 6.3(a) has a
.
x ' = L/4. Find the equivalent nodal 10:~~entbMb(:) applied at the point wh
using x jk = YJk
s or ot ocal and global coordi11111

6.J -4. Repeat Prob. 6.3-3 for a moment My(t) in they direction
6 ..lS. Rep~at Prob. 6.3-3 for a moment Mx(t) in the x direction.
<,.J .6. A tnangular distribution of force (I - 1:)b ()
.
.
1 1
member in Fig 6 3( ) F
_
!>
acts m the z direction on tlw " Id
a or XJk - -4 and y - 3 b
"
loads in both member and struct 1
d' I* - ' o tam the equivalent 111111,II
ura coor mates.
<d-7. Suppose that two concentrated forces P. (t) ar
.
g~id member in Fig. 6.3(a). Calculate; uiv ~ applted at the q~arter points ol th
directions for X = 1 a d
- 3
q a ent nodal loads m local and luh1tl
1*
n YJk -
"
6.J -8. The grid member in Fig 6 3( ) b'
a moment - M (t) at x'. =. 2Lj;s ~u ~ecte~ to a moment My-(t) at x ' = l / J 111111
nodal loads in Y
member and struc.t olr xd~k - _I and YJk = -2, find the equiv11lr111
ura 1rect1ons.
c,.4- 1. Suppose that the space truss member in F
.
force bv"(t) applied over the segment where 46:(a), ~s a uniformly ~istrih11tc ii
ulent nodal loads in local and global di
. - x. - 3l/4. Determme equtv
CJA = {3, 2, 1} and CJ = {- 1 3 -2} rect1ons, usmg the relative coordi1111trN

lk peat Prnh. 6.I ~ wllh 1111 11111111,111 111tl11g in the y' direction at the point
I '
/, /5.
Ass ume that a unifo1111ly diNtilhuted force b (t) acts on a space truss member in
the di rection shown by Fig. 6.4(b). Using the relative coordinates Cj k = {2, 2, 3}
nrtd c1,, = {O, 5, O}, delermine equivalent nodal loads in member and structural
di rections.
fl ~ H. Repeat Prob. 6.4-7 for a triangular distribution of force [bi(t).
fl

t. Let the space frame member in Fig. 6.5(a) be subjected to a uniformly distributed

fl ~ L.
fl

I.

fl ~ 4.
fl 8 ~.

fl ~ <,.
fl II 7.

fl , H.

fi

(1.4-2. Repeat Prob. 6.4-1 for a un;fo~ly ~istributed ~

segment where L/ 3 s x ' s 2L / 3 .
orce b,(t) applied ove1 flu

fl

for the three unrestrained displacements. Reduce these matrices by eliminating

lhe rotations and retaining the translation. Then find the angular frequency of
vibration for the remaining system, which has only one degree of freedom.

ly
1~ 1

~/4i

11

force by(t) over its whole length. Find the equivalent nodal loads for both local
and global directions, assuming the relative coordinates c1k = {3 , 2, 4} and Cjp =
{- 2, 4, - I}.
Repeat Prob. 6.5-1 for a uniformly distributed force b,(t) over half the length
(O s x ' s L/ 2).
Suppose that a concentrated force Py,(t) is applied atx ' = L / 4 on the space frame
member in Fig. 6.5(a). Using the relative coordinates c1k = {2, 3, - I} and
c1,, = {-1 , 2, 2}, calculate the equivalent nodal loads in member and structural
directions.
Repeat Prob. 6.5-3 for a force in the z' direction at the point where x' = 3L/ 4.
A moment My(t) acts on the space frame member in Fig. 6.5(a) at the point where
t' = L/3 . Determine the equivalent nodal loads for both local and global direclions, assuming the relative coordinates Cjk = {O, 3, 4} and Cjp = {O, 0, -2}.
Repeat Prob. 6.5-5 for a moment in the z ' direction applied at the point where
r' = 2L/ 3.
Assume that a uniformly distributed force bx(t) acts in the x direction on the space
frame member in Fig. 6.5(a). With the relative coordinates Cjk = {I, -2, 2} and
c1,, = {- 3, 4, O}, obtain the equivalent nodal loads for member and structural
directions.
Repeat Prob. 6.5-7 for a triangular distribution of force (I - t )b1(t ) in the z
direction.

I I. The two-element prismatic beam shown in Fig. P6.7-1 is fixed at point 3 but free
to rotate at point I. Construct the stiffness and consistent mass matrices S and M

(1,4-J. Assume thut a concentrated force P,,(t) act

' member in Fig 6 4(a) For both
b s at x - 2L/ 5 on the space 11 11~~
equivulent nod~! ioads. with th me~ _er and st~uctural directions, find th,
c//,
{ I , I , 3}.
'
e reat1ve coordmates c;k = {I , I , I} 111111

(1.4-4. Hlpeut Prob. 6.4-3 if the force acts in the '

I'
1/,/5 .
Y direction al lhe point wh,11
,,.4-!i. I it II moment M (t ) be upplied at . t '
/
111 Fig . (>.'1 ( 11) . C'11lc111t111 the l'q11iv111~:~:11,m:l11I lt,4/1 on !he SJ)lll'C truss .nll111h11
/l
101 th, 1l'i11fl w 1,11111 11 1111111.~ 1
' ) 111 N 11 lornl 11ncl 1,1 loh11l d111ctlt11iN
11
,
1,
1111d 1 ,
{ I , l, 2),

306

Prnhl111111t

CD
1-

'

3L

4/1-

1:
11'1111111 l't,,7, I

5+
6/~ 3

307
3015

Frnmnd Struoturna

f'rohlnm1

6.7-2. Figure P6.7 2 shows u two 1lrml'lll pliN11111tl1 lw11111 thnt is free to trunNhlll
not rotate) at point I nnd is llx1d 111 polut I Ass1111hle the stiffness anti rnnI
mass matrices Sand M for the th111 11n11s1111incd displacements, and rctlu~o
by eliminating the rotation and keepl111-1 the trunslations. Solve the l't"t'IIY
problem to obtain angular frequenc ies and vibrational mode shapes tor th
duced system.

Figure P6. 7-4

. . P6 7-5 has the same properties E' P, A, and I, for

The plane frame sho.wn I~ Fig . .
t
the stiffness and consistent mass
each of its three pnsmatlc me~~e;:~;;e; ~ shown, neglecting axial strains .in
matrices S and M for the f;ur d
b liminating the rotations D1and D2, while
the members. Then reduce an
e
. the angular frequencies and mode
ietaining the translations D3 and D4. etermme
shapes of the remaining system .

J6

z
Figure P6.7-2

r.:;--,

03 \

!..--,_.....,;\V::::::3:__--1~ 4
6.7-3. The simply supported beam shown in Fig. P6.7-3 is divided into two flc,iu
elements with equal properties. Set up the stiffness and consistent mass mutrl
S and M for the four displacements that are unrestrained. Reduce these mutrl~
by eliminating the rotations and retaining the translation. Then find the angul
frequency of the remaining system, which has only one degree of freedom .

/3
y

D,

,+,
~J;;

I:

CD
e

3L

Y.,.
L

5+ 3
X

6/~
:1

F igure P6.7-S

. F

Figure P6.7-3

6.7-4. A two-element prismatic beam has no restraints whatsoever , as implied in liill

P6. 7-4. Construct the stiffness and consistent mass matrices S and M for the Hix
unrestrained displacements, and eliminate the three rotations. For the thrc1 '"
maining translations, calculate the angular frequencies and mode shntK'S.

1i P6 7~6

has 2El and 2pA for elements

and . A for their properties . Construct
I und 2; whereas, elements 3 and 4 hav~ ' S
M and eliminate the rotations
the stiffness and consistent mass matnc~\. anD a~d D Calculate the angular
U, , 0 2, and D,, while keeping thehtra~s a .o~~g ;ystcm s~eglecting axial strains
li cquencies and mode shapes for t e ,cmaim
,
111 the members.

fl 1 ,,. Assume that the plane frame shown m

JOB
I 11111111<1 H1111011111111

308

hnp

1rnhl1111111
1

n I n111H in !lig. P6.7-8. l)ctcnnine the

A 1,1id with tlll l'l' iq1111l 111 t, 11111th ,111 ,11~1\:1~11 !'or the three displacements at joint
HtlffncHti lllld 1.'0IIHIHh ' lll lllll-11 111111 111111,
I t'on and calculate the angular
,
II
ki'llll 1hr trunsf a h1 , embers in this

' Then clln11111111 1111' 1111II 111111,

structure,
"
p ,r each o t e m
1
lrcqucncy of lhL i11h111d ~YH
~O and E = 5G/2.
ussume that 21, 2 /1, ,/ , ,.N
,/ '

(1)

"

1,

l'II\

G)

o ..

- - -- -- - - - -

03/

o, / - - L ~

1- -

0
4

-----x

031

CD
y

5
'/

/--'

.,

Figure P6.7-6

(1.7-7. Figure P6.7-7 shows a grid consisting of two prismatic members with 1q1111I
properties. Assemble the stiffness and consistent mass matrices S and M frn lh
three di splacements indicated at joint 2. Reduce these matrices by eliminali111-11h
rotations and retaining the translation. Find the angular frequency of the ro11111l11
ing SDOF problem, assuming that Ix = ly = J, Tg = L/10, and E = sen
z

/
Figure P6.7-7

Figure P6.7 -8

D2

,__/

I'

tlll llNIIOII

wl

IIIL

111ul

displacement

311

t.111111 111 Irr I 1111111111n

Vl'l'lo1 IH

(3)

u - {u, v, w}
I 11t lltl1more , the 6

3 Stra in-displacement operator d has the form

Two- and
Three-Dimensional
Continua

7 .1 INTRODUCTION
1

7 ,2 STRESSES AND STRAINS IN CONTINUA

a
ay

a
az

(4)

a
ay
a
ax

. .
ak from Eqs. (3.2-2) and (3.2-3).
which the derivatives are t den
t n relationships for an isotropic maaddition
we
develope
stress-s rm
.
In
' such relations
h"ips can be wntten-.
>I<
As
before
,, , /' 1/ .
'
(5)
CT=EE.
111111

rc the stress vector is

CT

= {CTx ,

E was given by Eq. (3.2-lO).

our discussion to problems in two
For the moment, Let us specia ize.
ly those terms depicted in Fig.
,l1 111onsions, where the stress vector contams on
I l(u). Thus, in the x-y plane we have
(7)

=d

{<Tx, <Ty , 1"xy}

p1g. 7 1(b) ' which

re the strains indicated m
to these stressesda. fi .t simal eiement in the x-y plane. Hence'
hows a displaced and deforme m m e
11 11 i,train vector is

' ' 111 t'csponding

111 Sec . 3.2 we introduced the topic of stresses and strains in three dimension
strain-displacement relationships with respect to Cartesian corndl
these relationships will be expressed in matrix form as
E

(6)

<Ty, <Tz, 1"xy, Tyz, Tzx }

,1 11d the 6 x 6 stress-strain matrix . l'

1111<1 dclined
1111 ll' s. Now

E.

111

( 11
either static or dynamic analysis. The strain vector in Eq. ( 11
'

.,

= {Ex ,

(8)

Ey, ,'xy}

which

au

E
1

IH

110

a a
ay ax
a
0
az
a 0
az

CT

lo

111

Tht framed structures discussed in the preceding chapter consist of only sh.-1111
(or one-dimensional) members. However, we also wish to analyze solids 1111d
Nl1 uctures composed of two- and three-dimensional finite elements. In this ch111,
fl'I' we shall deal with two-dimensional continua in states of plane stress 11ml
plnnc strain LI] as well as general and axisymmetric solids. Our emphasis will
h(.) upon the isoparametric formulations leading to the most commonly l1Nl1d
tlc:mcnls 121. Computer programs cover dynamic analyses of plane-stress 1111J
ph11w Strain problems, general solids, and axisymmetric solids . In all cases lh
di screti zed continua are assumed to be linearly elastic with small strains 1111'1
displacements.

which applies

ax

=ax
-

av

eY =iJy

au

av

--+
ay -ax

(9)

,'xy -

Rcr. 2' und for generally an-

~ l'nr o, thol(OHlllly u11lM

ot,opic (m1/1111r1111/1) i1111ll' t 1n1N, St e

11,,111111/r 11111lllrit1I N, Nl'l'

l {(I I , "

J12

Two- and Three-Dimensional Continua

oc. 7.2

313

Stresses and Strains i n Conti nua

I'hese derivatives constitute strain-displacement relationships in two dimen~lons. As before, they may be written in matrix form as shown in Eq. (1), where
lhc linear differential operator d becomes

a
ax

yx

a
ay
a a
ay ax

(10)

d=

-J

dy
dx

1111d
y

L,

the vector u is

'

= {u,

(11)

v}

When a thin plate is loaded with forces in its own plane , the resulting
two-dimensional problem is called plane stress . Assuming that the plate lies in
lhc x-y plane, we can impose the following conditions of stresses and strains:
<J',

(a)

'Ty,

1'zx

=0

, =I=-

Yyz

= 'Yzx = 0

(12)

111 particular, note that the strain normal to the plate is nonzero. Using these
nmditions in Eqs. (3. 2-5), we find that
i)V

dy
If-- ~
,w

V I ily dy

E = -----=E
1 - V2

l~ : ~ 1
0

(13)

1- V
-2-

'l'his is the stress-strain matrix for plane stress in an isotropic material. In

uddition, the normal strain in the z direction is
(14)

The case of plane strain arises when a long prismatic solid has a constant
l't>ndition of loading normal to its axis. In this instance, the solid can be analyzed
11~ an infinity of two-dimensional slices of unit thickness, as illustrated in Fig.
I , l (c). However, the conditions given in Eqs . (12) must be changed to

V,V

i!u
U

i!x

<J', =I=-

dx

(ll)

(11) Stresses und (b) struins In two dimensions.

1'yz

T,x

=0

'=z

')'yz

Yzx

=0

(15)

The first expression indicates that the stress <J', in the axial direction is nonzero,
l'Vl'l1 though the corresponding strain '=z is zero. Substituting these conditions into
nqs. (3.2-5), we obtain
(16)
1111d

314

Two- and Three-Dimensional Continua

E - (I

+ v)~I

2v{ : v

>rvl

y
y

y'

x'

Th0 latter equation gives the stress-strain matrix for the case of plane strain in
1111 isotropic material.

i'

Now we shall consider rotation of axes for stresses and strains, starting in
three dimensions. For this purpose, the stress vector in Eq. (6) may be recast into
the form of a symmetric 3 x 3 matrix, as follows:

(J'

315

Stresses and Strains in Continua

Sec. 7.2

= [~:

(J 8)

z
z'

Tzx

(b)

(a)

In which complementary shearing stresses are included. Then the rotation-of

uxes lransformation for stresses can be stated as
( I 'JJ

1n th is equation the 3 x 3 matrix O'' is similar to O' in Eq. (18), but it cbntait1H
Ntl't'SNllS in the directions of inclined (primed) axes shown in Fig. 7.2(a). Thl'
I X J rotation matrix R in Eq. (19) has the form

y
y'

A1 1

Yp

A2 1

(20)

x'

,\31

111 the lultcr matrix the terms / 1, m 1 , and so on are slightly more efficient symbolN
Im the direction cosines A11 , A12 , and so on, that were explained in Sec. 3.fi
Nolu lhul lhe rows of R contain the x, y, and z components of the unit veclor11
I' , .I', und k' shown in Fig. 7.2(a).
Similarly, the strain vector in Eq. (2) may be restated as the 3 x 3 sy,11
111t1l1 k matrix

x
'Yxy
')'yx
y

'Yxz]
'Yyz

[
'Yzx

')'zy

(21 l

which in<.:ludcs dependent shearing strains . Again, the rotation operation is

i'
j~

Id)

(c)

Figure 7.2 0 rthogona1 aXes-three

. . dimensions: (a) inclined; (b) principal-two
dimensions: (c) inclined; (d) pnnc1pal.
.

111 this equation the symbol

nl principfil normal

O'p

.
1matrix (or spectral matrix)
represents a diagona

stressesu:lr ~'

l]

(24)

(2.11

whld1 the 3 X 3 matrix e' is similar to E in Eq . (21), but for inclined 11xt11
Wt cun obtain principal normal streS.\'<'.I' und I heir clircclions us the solut i1111
111 1111 ,ll{1n1 ul11, prohl<111 141. Using thut nmthod )'iws
111

er,,

1{1, er H /,

(.! I)

.1 fl (23) denotes the rotation matrix for the

111 nddilion, the symb~>l ~,, ; ~~) This mntrix is the transpose of the nor1,,/mliwl rms shown 111 hg. ,2 , ,
,111111/t d ,IR1'111,,1or 111,11ril c1,N I hnl IK,

.1111
Two- 011d 1111011 Dlrnonslonal Continua

Chap, 1

Rp = <t>X
(2')
lnwhichthc rowsof R,(orc l .
f<I>)
Simi larly
. .' l o umns o. N are normalized to have unit lengths
' prznczpa . normal strains may be calculated as
.
Ep = RpE R]
In tlu s expression the symbol Ep stands for a di

(2(11

Sec. 7.2

Stresses and Strains in Continua

Therefore, the stress transformation matrix '(,. is proven to be the transposed

inverse of the strain transformation matrix T,. The inversion implied by Eq. (33)
is not actually necessary, because expansion ofEq. (19) shows that the parts of
T" bear the following relationships to the parts of T,:

.
agonal matnx of normal strain-

l'1ll'rcsponcfing
to the stresses m
<Tp.
.
In subsequent work it will become nece
IIIHtrix E from one set of coordinates to ano~~d to trans~orm the stress-stnriu
purpose, we rewrite the expanded results of Eq. (~~/~~atton of axes. For th,~

e' = T,e
(27)
l lcrc the strains e and e' are in the form ofE ( 2).
d
.l'lm/11 tronsformation matrix T, in Eq. (27)qis
mstea ofEq. (21). The 6 X ,,

TO' = ['t.11
T0'21

TO'l2]
To22

[IT,,,
2 T,21

2T,12J
Tm

u' = E' e'

(35)

Next, we substitute Eqs. (27) and (32) into Eq. (35) , producing

TO'<T = E 'T,E
Then premultiply Eq. (36) with

/ .I2

(34)

As before, the subscripts 1 and 2 on the submatrices of TO' refer to normal and
shearing stresses, respectively.
Now the transformation of E' to E can be accomplished by first writing
stress-strain relationships in the primed coordinates as

/7
Ii

r:

317

2/1 /2

T;;', and use Eq.

CT = TJ E 'T.:E

(36)
(33) to find that
(37)

Thus, we see that

2/2/3

E = T;E'T,

(38)

which represents the transformation of E' to E. The reverse transformation is

E' ='(,.ET~

(39)

For the purpose of specializing axis rotations to two dimensions, we form

n symmetric 2 x 2 matrix of stresses, as follows:

(40)

This matrix can be used in Eq. (19) to determine similar stresses u' for the
Inclined directions shown in Fig. 7.2(c). For this case the 2 x 2 rotation matrix

ls
(oe')TO'' =
' l'h1111

ohf11l11

OET<T

( 10)

sin (}
cos (}

substitute the transposed incremental form f Eq

(27) into Eq. (30) to

(41)

111 which the rows contain the x and y components of vectors i' and j'. FurtherI l11111't'.

( 11,

we conclude that

er'

= 1;,o

wh111

'(;, 'J' ;- I

1110rc, the symmetric 2 x 2 matrix of strains becomes

E= [Ex Yxy]
'Yyx

(42)

Ey

whklr <.:ur1 h1; applkd in Eq. (22) to find strains E' for the inclined axes.
By solving II Nl'l'Olld Ol'(kf' t't/lnv11llll' prnhlcm , we cun also obtain prin1 lp11l 11011111d 1-1111:,i,t1, 1111d Nl111i111, in two di11111111,lo111,. Thus, tht l'igtnv11h1t prob-

318

Two- and Three-Dimensional Continua

Sec. 7.3

319

Natural Coordinates

lem associated with the 2 x 2 stress matrix in Eq. ( 40) is

,\;l)<I>;

(O' -

(i

= 1, 2)
Tl = 1

In this equation the symbol A; denotes the ith eigenvalue of O' , and <I>; represents
the corresponding eigenvector. To find A; we set the determinant of the
coefficient matrix equal to zero , as follows:

IO'

A; I I = 0

(44)

or

o-y T~ A;

I=

(45)

Expanding this determinant yields

AT - (o-x

0-y)A;

0-xO'y - T;y

=0

ll

11

i;

L.

(46)

From this quadratic equation, we obtain the roots

i; = 2

- 1

ll = - 1
(a)

(47)
3

which constitute the principal normal stresses o-p 1 and o-P2, They act in the
directi ons of the axes Xp and YP shown in Fig. 7.2(d). By substituting them into
the homogeneo us equations [see Eq. (43)] , we can find the eigenvectors, which
(when no rmalized) contain the components of the unit vectors ip and jp.
The strain transformation matrix T, in Eq. (28) can be specialized to two
din1ensio ns by removing the third, fifth , and sixth rows and columns to obtain

~~=~+

T. = [---:: - - - - ::
I - - - - --- - --42l,l2 2m 1m2 / l,m2

l2m,

Also, the stress transformation matrix T" becomes

T"

=[

If2

l2

mf2

m2

I
I
I

2/, m,
2/2m2

------- -4- - - --- - - -

i; = 1

(48)
2

s= - 1
y

L.

(49)

11/2 m,m2 I l,m2 + l2m,

which is related to T, in accordance with Eq. (34).

s = --12

ll = - 1
(b)

Figure 7.3 Natural Coordinates for quadrilaterals: (a) straight; (b) curved.

7.3 NATURAL COORDINATES

111 lllis Sl't'ti on we de fine natural coordinates for q1111drilatcrals and hcxahedrn in
prnp11rnt io11 for dl'Yl'lopment of isopurnnwt, k 1lt111111t 11 having these shapes . Wt
li11il11 with 11 <J11<ulr llt1t1r,il that has st11d ~ht 1dg111, 11~ ill11stratl1d in !lig. 7..l(i',)
'I'll!' poi nt l11lwl1d ,-: is tilt' Mr'o1111trl1 1,11111. !111 \\ hi, II lh1 1rnudi1111t1s in lhl' 1 1

pl une are
xH

= Hx1 + X2 + X3 + X4)

Ys

= HY1 + Y2 + y3 +

y4)

. (1)

(
) arc the x and y coordinates of po mt 1,
.
111 these averaging expressions, Xi, Yi
.. , .. 1 the same as the
N t ti I tl w gln1111.llrk t'l'llll' I' is no t ncccss,m Y
.
1111d 1-10 on.
o l' Ill
Al
i I th . llHllll' un till' di11w11.1im1f<,1.1
I
l l' llll Old

111 th1 q1111dtil11h11il

t\

NII S HlW II

321

1:10

lwn "'"' ll11nn 111111,1111111111111

11,111m1/ 111<1rdl,u111.1

t and T/

Ill poi111 /.:, it is i111portant l<~

Alli
ru

Cu 11 t1 111111

i','.H'.~ 1' f111 >j1 '1 ls1w rn.onlinutcs huvc

1("~

'1111:. / ,'J

<:hni,

!hell' rn 1)!111

pol111 o11 lhc clement. We see t~'

I ml Ihr Ii dlt l'l'fions arc different ut l'Vl'I V
1011
' I, T/ . I ulong edge 3-4, an~
~ 7dgc 1-2, ~ I _along l'd1,tr
pol11!1011 111 both the 1: and d"
.
,dong cdge 4-1 With linear i111r 1
!>
r, 1rect1ons form la . f 1 .
unywhcrc on the quadrilateral become '
u s or ocatmg a generic po1111

J_ / "

x=

f;x;

y = 2, f;y;

1=1

111

,,

. I

N11t11111t Cow dl1111lot1

11or this arrangcmcnl, terms in the coefficient matrix are easily obtained by
differentiating Eqs. (2). This array is called the Jacobian matrix J, which
l'Ontains derivatives of the global coordinates with respect to the local coordinates. Thus, we have

J=

g)(l - r,)
g)(l

r,)

f2 = !(I + [)(1 _
f4 = HI - [)(1 +

111

= x,g

Y,ri

X,')

= L f;,gx;

112 = Y,g =

L f;,gY;
i= I

i=l

r,)
r,)

(8)

[x,g Y,g]

Terms in this 2 x 2 matrix are

f 21 = X ,T/

( I)

= 'v

= Y.') =

122

+.

~Jl,T/X I

L AriYi
i= l

i=I

Casting these expressions into matrix form yields

(10)

= DLCN

The array DL in this equation contains derivatives with respect to local coordinates, as follows:
DL

= [ !1,g fi.g /3,g

!4,{

!1.ri Ari Ari Ari

(ti)

(9)

af a71
ar, ax
af ar,
ar, ay

122

I Ill Sl' I 1111ct1ons give the global coord t

f
.
l!~1111s of lhc natural coordinat
H ma es o a pomt on the quadrilaternl 111
hypl11holic), the local coordina~es~ gao~ever, because Eqs. (2) are bilinear (rn
ol Iill' global coordinates x and
n 71 cannot be expressed directly in te111111
W
11
y.
c w1 need derivatives of the function "
dl11plun111cnt relationships The h .
s J ! through f4 to use in stra in
1 for differentiation off([, r,) with
uspl'l'I lo x and y gives .
c am rue

af = af ag +
ax ag ax
af = af ag +
ay ag ay

112]

121

which the geometric interpolation functions '~e

!1 = Hl / j = !(1 +

[111

[-(1 =4
1

TJ)
-(1 - g)

(1 - r,)
-(1 + g)

(1 + r,)
(1 + g)

r,)J

-(1 +
(1 - g)

(11)

and the matrix CN consists of nodal coordinates in the arrangement

01'

[J.x] = [S.x 'Y/,x] [J, g]

!,y.
S,y 'Y/.y !.ri

wh1n the s m
.
( )
,,
. Y bol f. x means differentiation off th
I 11111s 111 lhc coefficient matrix of E 5
w1
_respect to x , and so on
1111, 11111,hlc lo solve explicitly "
t: q.d( ) ~e not readily available, because Wl'
1
ior !> an r, m terms f
d
'""'' I1ll' opposite approach and differenti
. o x an y. However, if w,
111lt produces
ate/with respect to sand r,, the chain

af = af ax +
ag ax ag
af = af ax +
a,, ax ar,

af ay
ay ag
af ay
ay ar,

f,1

f:: ::1
X3

X4

((l)

By comparing Eqs. (5) and (7), we can see that the coefficient matrix in
the former expression is the inverse of the Jacobian matrix. Using the formal
1
definition of the inverse, we can obtain
from J as

where Ja denotes the adjoint matrix of J and

quantity is calculated by

\JI = l,1122
('/)

(12)

Y3
Y4

-112]
f1 1 =

1)1

f = [.x., Y.cl [J., l

IJ.,,
x,,, Y.,, /1,.I

CN

Ji1l12

m
1 [

Y.ri

-x, 71

-y,g]

(13)

X,g

IJI is its determinant. The latter

= X,gY.ri - X,riY,g

(14)

To determine the derivatives of all of the functions with respect to x and y, we

cun apply nq. ('l) ll'Pl'lllctlly . Hence,

322

Two- and Three-Dimensional Continua

J-1[fifi.7]s]

fi.x] =
[ fi.y

(i

= 1, 2, 3, 4)

Sec. 7.3

(15)

323

Natural Coordinates

numerical integration for both straight-sided and curved quadrilaterals, as explained

in Sec
. 7.4.
.
Now
let us
consider
a h exa hed ron with straight. edges, as depicted in Fig.
7.4(a). At the geometric center (point g) the coordmates are

Altogether, we have
(16)
The matrix DG given by this expression consists of derivatives of Ji with respect
to the global coordinates. That is,

DG =

[fi.x fz.x .h.x f4.x]

ii .y fz.y h.y Ay

Xg

8L

X;

Yg

L Yi
8

=-

Zg

i=I

1=)

8 L Z;
i= l

T)

(1 )
7

Evaluating terms in DG, we find

]

Dc11 = 4 / J / [ -(1 - YJ)l22 + (1 - [)1 12]

1

Dc12

= 4 / J / [(1

I
Dc13 = 4 / J / [(1

- YJ)l22

+ (1 +

[)1 12]

{)112]

+ YJ)l22 -

(I

)-.
z

Dc14

= 4 /J /[ -(1 +

YJ)l22 - (1 - [)112]

1
DG21 = 4 /JI [(1 - YJ)l21 - (1 - [)l1i]

(18)

la)

1
Dc22 = 4 / J / [ - (1 - YJ)l21 - (I + [)111]
1

Dc23

= 41 JI [ -(1 +

Dc24

= 4 / J / [(1 + YJ)l21 + (1

YJ)l21

+ (I + [)Ji i]

- [)111]

lly this approach we are able to solve for De numerically.

Figure 7.3(b) shows a quadrilateral with curved edges that may follow
q1111drutic functions, cubic functions, and so on. Regardless of the complexity ol
llwsu functions, the natural coordinates g and YJ play roles similar to those for
11 quudriluteral with straight edges. The form of the 2 X 2 Jacobian matrix
1l111111i11s the same, even though the functions to be differentiated are of higl1t11
order. Because of the appearance of the determinant of J in denominator posi
11011s, Wl' 11suully cannot integrate terms explicitly to obtain stiffnesses, consis
lli11t ou,sscs , und equivalent nodal loaclli. lnstcnd , ii hlcomes necessury lo 11s1

=1

lb)

Fl111111 7.4

Nuturul coordinurcs for hcxnhcdrn : (11) straight; (b) curved .

(19)

I w11 nrn I 1111 n.- I>l111111111lur111I Cu11tlr11u1

Chnp

wht'll' ,,, I''.' "."d i, 1~1c the 0 1rt<:siu11 l'11111dl1111h11 of poi nts J thrnugh 8. 'l'lw fi ,,,.,
11lN11 Hhows t/1111,11.,101,le.1.1 natural coor,lln,,r,,, (.. '>'J und r Nol<: lhal 1:
11
fmt l l 7 (
' linear
b
b
11
>: 77 . J on face 3-4-8-7, nnd so 0 11 .' With
interpolation i 11 th
I.ti

~~ 'I und ( d1rcct1 ons, the location of any point in the hexahedron may be wrlflrn
s

= 2_ fix;

= 2, f;y;

= 2, f;z;

i=l

( .'ll I

i= I

111 whk h the geometric interpolation fu,nctions are

/,
./'i

ff{ I - {)(1 - T))(l - t)

!2 =

.f'..

HI + g)( I + 77)(1 HI - g)( I - 77)(1 +

.Ii

ff{ I I s)(I

Ii = k(I - {)(1 + 77)(1 - C)

16 = ko + t)(1 - TJ)(1 + c)
f s = k(I - f)(l + 17)(1 + ()

{)

?)

+ TJ)( l + {)

ko +

t)(1 - 11)(1 - c)

Ill

il-(

,.:1

325

Nntw 111 (.;oor dltrntuu

(1 + 77)(1 - ()
(I - 77)(1 - ()
(I
~)(I - I,')
(1 + {)(1 - ()
-(1
+
g)(l
()
I - {)(1 - ()
- ( I - {)( l - 77) - (1 + g)(l - 77) -(1 + g)(l + TJ)

(1 - 77)(1 + ()
- (1 + 17)(1 - () -(1 - T))(l + t)
(1 - g)(l - () -(1 - g)(l + {) -(1 + g)(l + ()
(1 + g)(l - TJ)
(1 - g)(l - 17)
- (1 - g)(l + 17)

I,')]

+ 17)(1 + t) -(1 + ~)(l +

(1 - g)(l + ()
(1 + {)(1 + ()
(1 - g)(l + 17)
(1 + g)(l + 17)

(1
('II

\ lso, the matrix CN becomes an 8 x 3 array of nodal coordinates. Thus,

llt'l'IIIIN.c th<:sc interpolation. functions are trilinear, the local coordinates i

1
d
' 1
I ' II . ' d.
.
x' y' an z.
' ucc. 11ncns1ons the chain rule for differentiation with res ect to lhr
11111\rntl rnordrnatcs for a hexahedron leads to the foll

3
P
owmg
X 3 l aco/,/, 111
111 ,,r,'II :
1111d , 1111rnot be expressed rn terms of the global coordinates

= [;~: ;~: ;~:] = [::

h,

1 32

x,,

] 33

r: :::]
YJ

z.,

(24)

() l,

CN

X1

Yi

Z1

-~2.

Y2

.~2.

Ys

zs

Xs

(25)

The inverse of the Jacobian matrix may be expressed as

(26)

'f't 1111s in thi s matrix are found by the differentiations indicated. Thus,
S

1 12

= 2, f;,Yi

I I

i= l

.!2, - ~ fi.r1X1
.I 11

==

2, f;,z;

= 2, f;,.,,y;

= 2, J;,.,,z;

i~ I

i= I

1 32

==

2, f; ,,Y;
i I

Wl' use Eq. (16) . In this case, the matrix

123

I I

L !t.,x,

l'o find the derivatives of all of the functions with respect to global coordinates,
DG consists of the following terms:

i= I

122

where the symbol Jarepresents the adjoint matrix of J, and IJI is its determinant.

11 3

] 33

(2 1)

= 2, J;,,z;

/4,x fs ,x Ax / 1,x fs.x]

!1.y !i,y /3,y !4.y fs ,y f6 s !1.y fs.y
f i,z f2.z

Az /4,z

(27)

fs ,z /6.z h,z f s.z

i l

AN hlIOi l',. lh.csc calculations may be arranged in the matrix format given b p 1
( I !I) In lh1s 111slance, the n:iatrix Dl is the following 3 x 8 array of derivt~i~1
will, Hs1wc1 to local coordmates:
"

u,

_ [/1,x A x Ax

DG -

t.,]

.Ii., .Ii., .li., A, 1~., /6,, !1.,

.Ii ,1/ .Ii.,, /1.1, 14.,, 1~.r, A,, Ari ls.ri
.Ii., .Ii., /i,, ./~., 1.~., Ji1,( !1., .IH.t

A hexahedron with curved edges appears in Fig. 7.4(b). Geometric interpolation functions for this solid may be quadratic, cubic, and so on. The natural
1oordinates g, 17, and ( will still be used for such a hexahedron in spite of its
111c11tcr complexity. The 3 x 3 Jacobian matrix in Eq. (22) remains the same,
11llllough the functions to be differentiated are of higher order. As for a quadril1111rt1l with straight or curved edges, it becomes necessary to use numerical
l11tl~ration to evaluate stiffnesses, consistent masses, and equivalent nodal
l1111ds.

'I

328

lw11

111111 11111111

l>l1111m11lo1111I Cc111tl111111

111

/JI

327

N11111n1 lo11l l111u11111tlo11

7.4 NUMERICAL INTEGRATION

The process of computing the valuc ol

Ix =

'l

x,

II

dtflnitc integral I see Fig . 7 .5(11ll

/(x) d x

t IJ

from a set of numerical values of the integrand is called numerical integ, 111 h 111
15]. The problem is solved by representing the integrand by an interpol11l l1111
formula and then integrating this formula between specified limits. Wht11 1111
plied to the integration of a function of a single variable, the method is rcl11111I
to as mechanical quadrature.
If interpolation formulas for numerical integration are polynomial~ 111
sufficiently high order relative to those assumed for displacement (or 0111111
functions , the integrations will be exact. Otherwise, the process of nunw1 h 111
integration introduces an additional source of error into finite-element analyi.1
The most accurate quadrature formula in common usage is that of 01111
which involves unequally spaced points that are symmetrically placed. To upplv
Gauss's method, we usually change the variable from x to the dimensiouh ~
coordinate g with its origin at the center of the range of integration, as sh11w11
in Fig. 7.5(b). The expression for x in terms of g is

= H(l

t)x, + (1 + t)x2]

( 'I

Substitution of Eq. (2) into the function in Eq. (1) gives

J(x)

= cp(g)

(a)

,,
(b)

Also,

( II
Then substituting Eqs. (3) and (4) into Eq. (1) and changing the limits 111
integration yields

3
2

Gr111.1.\" sformula for determining the integral in Eq. (5) consists of summing lh,
wcighted values of <f>(g) at n specified points, as follows:

(c)

(hi

In this cx.pression ~ is (he location of the integration point} relative to the tt111!1,
11 1wi~liti111< fa ctor f'or point j, and n is the number of point.1 ul whic:h c/,( ( I
is to lw tnku luted. Yult1es of these pururnt(l\rs nn: listed in Table 7. I.

N1 is

Figure 7.5 Gaussian quadrature.

1111

l\,\111 111111 tl111111 Ul11111111 lo1111I cw111!11111

<

11111

Nu111L1h;ul intcgrntinn hy, ,, 111111

,
I 'fl' t '.
,111 ,111,1rlm1t111 IS CXIICI for poly11011111tl I
'
.
,H IS , 011f y 11111' i11(r)tl111i
.
.
111tcgration of a linear functi , I
oo point rs required f'or tlw

Ml

,II

,,

dl'gl'l'C 2 11

,,11

und so on. For example, con~;~ ~.. ~:1. p11111t ''.' \nccd_ed for a cubi c poly1111111111I
ing in Fig. 7.5(c). First let n
I ~t111111,111r11 , 11,11,~ uncT
tron <!>W
3 - g2upp1111

'

lit

g, = 0

/? 1

mm

able 7 .1

=
that

=2

~ 1?1c/,(~1)
j

(;

0.5773502692

1.0

0.7745966692
0.0

0.5555555556
0. 8888888889

0.8611363116
0.33998 10436

0.3478548451
0.6521451549

0. 9061798459
0.538469310 I
0.0

0.236926885 1
0.4786286705
0.5688888889

0.9324695142
0.6612093865
0.2386191861

0.1713244924
0.3607615730
0.4679139346

0.9491079123
0.7415311856
0.4058451514
0.0

0.1294849662
0. 2797053915
0.3818300505
0.4179591837

Y2 l X2 f(x , y) dx dy

X\

=x+

(8)

xi + yj

g is

lhc rate of change of r with respect to

ar ax. ay .
- = - 1 +-J
ag ag
ag

(9)

iJr

ax
-

dT)

- dTj
ilTJ

iiy

d-ri

0.1012285363
0.2223810345
0.3137066459
0.3626837834

iJr

+ - dTJ
dT)

'l'hl'll from Eq. (6) we have

1E = R, </>((;,)
which is approximate. Next ' let n

(;, =

- (;2

=- ~=

I knee, we lind from Eq. (6)

= (2)(3) = 6

= 2 and

. f rom Table 7. J:
o btam
X

-0.5773 .

R,

= R2 =

dA

dT)

dT)

0.9602898565
0. 7966664774
0.5255324099
0.1834346425

(7)

llnwcver, this integral is more easily evaluated if it is first transformed to the

1111t11ral coordinates for a quadrilateral. We accomplish this by expressing the
llull'I ion f in terms of g and 1/ and using the limits - 1 to 1 for each of the
l11tlgrals. In addition, the infinitesimal area dA = dx dy must be replaced by an
11ppropriate expression in terms of d(; and dr,. For this purpose, Fig. 7.6 shows
1111 infinitesimal area dA in the natural coordinates . Vector r locates a generic
pnlnt in the Cartesian coordinates x and y, as follows:

R;
2.0

gD

. .) = 5.333 .. .
, hlrh is exact.
Now let us apply Gaussian quadrature to quadrilaterals in Cartesian coor1ll1111tcs. The type of integration to be performed is

YI

0.0

t ( 1)(3 -

I=

~O

(1)(3

= (2)(2.666

TABLE 7.1 Coefficients for

Gaussian Quadrature
n

329

N1111111tl1111t 1i1111u1111lt111

z
li'IKllr(! 7.6

Infinitesimal area in natural coordinates.

i!r

i)~

d~

:no
AINo ,

I Wt1 1111il I 1111111 l>lt11,1111l111111I <.1111tl1111n

tltl'

rntl' ol chungc ol

I'

With

1111,1111 I

I'

lo t/ IN

1111

d , I I iJ y I
ll tJ
ii 'Y/'

/Ir

o'YJ

t ,tll'S

ot' drnnge of,. wit h respect to , ,

== ( :~ d g x

:~ d 11) k

==

(ax~_ ax ay) dg d 11
ag a11 a11 ag

I;,,,,,s

Y.s I dg d11

Y.,,,

==

IJI dg d11

(18)

t t l the symbols a, b , and c denote

(I

The expression in the parentheses of Eq. (12) may be written as a 2

dctcrminant. That is,

dA

uml 'arc

ax
ay
az
or _ - i + - j + - k
ag - ag
ag
ag
a
ay
az
ar = ~i
+ -j + -k
a'Y/ a11
a11
a11
a +ayj + -azk
ar _ 2-i
~ - ai
ai
ai

I I 11

Substitution of Eqs. (9) and (10) into Eq. (11) produces

dA

T/ ,

When multiplied by dg and d'f/, the derivatives in Eqs. (9) and (10) fo1111 tW
11dj11ci;nt sides of the infinitesimal paraHelogram of area dA in the figun: I hi
111 cu may be detennined from the following vector triple product

dA

( I'/)

d I v,I I ~k

ar
(19)
ar
a dg
C = -d~
- _!
b = - d11
ai
a - ag
a11
.
.
as the edges of the infinitesimal pai::al78
lfo.isc vectors are shown m_ Fig. .
be determined from the following
!11lcpiped of volume dV. This volume may
Vl'Clor triple product:

(I I)

111 which J is the Jacobian matrix given in Eq. (7 .3-8), and IJI is its determina11t
ThuH, the new form of the integral in Eq. (7) becomes

( I II
Two successive applications of Gaussian quadrature result in
n

I ==

L L RjRd(g, TJd IJ(g, 1/k) I

( ISl

where Rj and Rk are weighting factors for the function evaluated at the point
((;J, 'Y/k). Integration points for n = 1, 2, 3, and 4 each way on a quadrilateral arl'
illustrated in Fig. 7.7.
Next, we tum to hexahedra in Cartesian coordinates, where the type ol
Integral to be evaluated has the form

==

iz2 JY2Lx2 f(x, y,

' I

YI

z) dx dy dz

(b)
(a)

(16)

Xi

lkfore integrating, we rewrite the function in terms of the natural coordinates

,, 'Y/, and Cand use the limits -1 to 1 for each of the integrals. Furthermore, we
must replace the infinitesimal volume dV == dx dy dz by an expression involving
tlg, d'f/, and d{ Toward that end, Fig. 7.8 shows an infinitesimal volume dV in
the natural coordinates. Also depicted is a vector r, which locates a generic point
ln the space. Thus,

(d)
(c)

.
-l(b) n = 2;(c)n =3;
Figure 7 .7 Integration points for quadnlateral: (a) n - '
(d) n = 4 (each way).

lw11 n111f

1111 1111

Olt1111111h1111,1 Co11t l111111

111 I II

ltmp11111111111tln Ou11drll11t11111I Im 1'1 11 1111 Strose end Pl ane Strain

333

1, ISOPARAMETRIC QUADRILATERALS FOR PLANE STRESS

II )(

AND PLANE STRAIN

<IV

lloilc c lement is isoparametric if the same interpolation formulas define both

tlt1 ).lt:Ometric and displacement shape functions. Such elements satisfy geo-

as well as displacement compatibility conditions. If the geometric interp11l11tion functions are of lower order than the displacement shape functions , the
, h 1111.mt is called subparametric. On the other hand, if the reverse were true, the
1 l11111t:nt would be referred to as superparametric [6]. However, most commonly
lf~l'd finite elements are either isoparametric or subparametric. Because isop11111mctric elements are usually curved, they tend to be more suitable than
1thpurametric elements for modeling curved geometric boundary conditions.
Figure 7. 9(a) shows the rectangular parent R4 of the isoparametric quad1 //,11,r al Q4 , which appears in Fig. 7.9(b). Conversely, the rectangle may be
11 111sidered as a special case of the quadrilateral, for which the natural coordi111tll'S gand 7J are orthogonal throughout the element. For either figure the generic
dl~placements at a typical point are
111111Ik

I /z
_ _ _ _ _ _ _J.,/

= {u, v}

(1)

N,,dal displacements indicated in Fig. 7.9 consist of x and y translations at each

11rn lc, as follows :

z
Figure 7.8

Infinitesimal volume in natural coordinates.

(2)

dV = (a x b) . c = / J / d( d7J d(

ax ay az
ag ag ag
ax ay az
aTJ aTJ aTJ d( dTJ dt
ax ay az

11,11 both elements we assume the displacement shape functions

4

u
<1111

in which J is the 3 x 3 Jacobian matrix

d /J / . .
.
revised form of the integral in Eq. (16) 'b:~omes IS Its determmant. Hence, lhl1

fl fl

/JI d(dTJd(

I, 11111

= f;q;

.Z: .Z: .Z: RjRkRtf(Sj, T/k, ?,) 1J(Q, T/k , ?,) 1

(4)

111 which
f.r

l = J k= I j= I

(i = l,2,3,4)

(.! I)

Three successive applications of Gaussian quadrature yield

n

(3)

i=I

111 these expressions the functionsJ;,!2 , Jj , andf4 are the same as those in the
v.10111etric interpolation formulas given by Eqs. (7.3-3). Therefore, the Q4
1 ltment is isoparametric, and previous statements regarding differentiation and
l11hration of functions hold true. Equations (3) may also be written in the matrix
U;

f1!((,TJ,t)

V = Lj;V;

i= l

ac ac ac
I=

= ""ifiu;

(.' 'I

Integration points for n = 1, 2, 3, and 4 each way number I 8 27

IcI
respectively.
, ,
, unc ,. ,

= [O
1 OJ
1 Jri

(5)

I h1 !'Cncric displacements u1 in Eq. (4) represent translations at any point due

q, al node i. As n furth t:r effi ciency of notation, we can write
1111 l'unction .fi us
h1lht displacements

./i

1(I

I 11H I I r111)

(6)

t.14

lwr, n111I

11111111

l>lt11n11alrn 111 1c0111r, 11111

ni

/ ,h

I opn11111u1ttlc 011udill11h11 nl11 for 1'1111111 btro1111 1111d Plono Str11l11

336

whl I l'
(7)
q6

T~ --+---:_-u~J-,.

Vnlucs of

-----

----Lli_____:J2tq~ q3

i-+--~~~~~2a~~~~~-/

g,

-I

-1

1/i

-1

-1

91

L,1~ ~tq-,

'Y/, for this element are listed in Table 7.2

TABLE 7 .2 Nodal Coordinates for Element Q4

Tdt
4

2b

g1 and

Similarly , the strain-displacement relationships for element Q4 may be

11111cisely expressed as

~; = B;q1

(i

a
ax
0

B1 = d f1 =
TJ

a
ay

0
a
f1
ay

[f;, /,]

0
f;,y fi .x

(9a)

ax

lfrfcrring to Eqs. (7.3-17) and (7 .3-18) , we see that the submatrix B1 can also
h written

q,_

B1

-+

G2

Dc11

Dc21

DG21

Dc11

(9b)

Next, we express the stiffness matrix for the Q4 element (with constant
lhlckness h) in Cartesian coordinates as

G1

l'lgurc 7.9

(8)

whore

(a)

L.

= 1, 2, 3, 4)

BT(x,y) E B(x,y) dx dy

(10a)

Ilowcver, in natural coordinates this formula becomes

(10b)
(b )

1111d

with two applications of Gaussian quadrature we have

Element Q4: (a) rectangular parent; (b) isoparamctric countc1p111t.

'=

II

L L R1RkBJ.kE Bj.k IJj.k I

k I J I

(10c)

338
loi, 11111111111,t, 1111nd1llnl111nl 1111

In this expression for nun11:r 1rnl 11111pr11111111, llw 1u1ttrix n1,4 und the <ktl1111111
J J1.d are evaluated al each inh.:gml 1011 point , whe re !he coordinates arc (/.,, t/
Similarly , the consistenl muss rnut1 ix lor l!lemcnt Q4 is
M = ph

LfT

1'1111111 t.11 111111 nrul l'ln1111 h1111l11

337

lhrntw e, liqs . ( I Oh) 1111d (III,) 111111plll y to

"""

(x, y) f(.x, y) d.x dy

rr

u,(~. TJ)E B(g. 'Y]) dg d'Y]

(14)

11111

or
M = abph

f, f fT(g, TJ)f(g, TJ) dg d'Y]

(15)

I h, resulting 8 x 8 matrices are given in Ref. 2 and need not b~ repeated here.
11,u,lly, the equivalent nodal loads in Eq. (12b) also take the simpler form

or
II

= ph L

IJ

L R1RdJ.k

Pb(t)

~ .k J Jj,k

f,f,

= abh

fT(g, 'Y])b(g, 'YJ, t) dg d'Y]

(16)

k E I j=I

I ,111nple 7.1

in which p and h are assumed to be constant.

In addition, equivalent nodal loads due to body forces may be statld 11
follows:
Pb(t)

=h

fT(x ,y)b(x,y, t) dxdy

or

111' 11 vc

numerically the consistent mass term M24 for the

Fig
/ 'l(h), using Gaussian integration with n = 2 each way. Assume that p an 6 aa;~
, 1t1INlunts and that the coordinates of nodes 1, 2, 3 , and 4 are (3, I), (8, 2), (6, ),
1 ',~). respectively .
.
.
( G
T
I Eq ( 11 c) we must set up the network of four mtegrat1on pomts or auss
o appG2,
Y G3
, and' G4 indicated in Fig. 7.9(b). In part1cu
Iar, ~or n -- 2 we have
l'illlls) GI,
11, Rk = 1 (from Table 7 .1), so that
isoparametric Q4 elemen~1~

M24

= ph

L L

(fit.k(/it.k IJj.k I

(a

k=l / I

I lw l'unctions/1 andfi are needed for evaluating the term M2 4 becau_se of the arrangement
111ntrix f shown in Eq. (5). Substituting/1 andfi from Eq. (6) mto Eq. (a) produces

or

111
n

Pb(t)

= h L L R1RkfJ.kb(t)j,k IJJ,k

h
J

M24 = f!._

k=I jE I

}6

where both Pb and b are functions of time.

Except in special cases, the integrals in Eqs. (IO), (11) , and (12) must h
performed by numerical integration to obtain approximate results. However, 1r
the element is rectangular, direct explicit integration may be used, for cithl'r
Cartesian or natural coordinates. Also, line loadings with g or 'YJ constanl 11111y
be handled by explicit line integrations. Of course, if the body forces consist of
point loads, no integration is required at all.
By specializing the quadrilateral Q4 to become the rectangle R4, we nm
derive explicit terms in its stiffness and consistent-mass matrices. For thi s cu~"
the Jacobian matrix and its determinant become

J= [

~]

JJJ = ab

in which a and bare half the width and height of the rectangle in Fig.

( 11)
7. 9(11)

L L [(1

- g2)(1 - 17)2] j,k IJj,k I

(b)

k - 1 j=I

111 Implement this formula, we first calculate the Jacobian matrix using Eq, (7 .3-1 0) , as
l11llows:

(1 - 11)
-(l +t)

l[

=4

(1 + 11)
(1

t)

-(1 + 17)]
(1 - ~)

38 21]
6 6
2

9 - 17
- 3-~

~]

5
(c)

1111'11 the determinant of J is

IJ I

H39 I I;

(d)

417)

1<1 11111 'l'nhlc 7 . 1 tlw vn lues of' { und 11 nt point Cll nr'l

I/V3 and

I/V3, and so on

336

lw11 1!1111

1111 1111

IJl11101111lon11I Continua

l oc. 7.5

lsoparametric Quadrilaterals for Plane Stress and Plane Strain

for !hi.) otl11.)r i11tl.)gralion poi11tN l vul11ull11M 1111 11,1111M in Eq. (b) at each of lht 11111
integration points and summing lht r1Nt1II N yl11ldN
I . I .I<) pit

Mi,,

which can be finalized using numerical values for p and h.

4

Now we shall consider a higher-order quadrilateral element that is b111.i 1l

on quadratic geometric and displacement shape functions. Figure 7. IO(a) doph I
the rectangular parent R8 of the isoparametric quadrilateral Q8 illustrat(d In
Fig. 7. IO(b) . We may consider the rectangle to be a special case of the qu11d1 I
lateral, for which the natural coordinates gand r, are orthogonal throughoul lh
element. In addition, edges of the rectangle are straight, and nodes 5 throu~h H
are located at midedges of sides 1-2 through 4-1. For both elements R8 and <)H
the nodal displacement vector is
q = {qi, qz, ... , q16} = {ui ,

V1, . . . ,

vs}

i=J

.ft

= i (1 + fo)(l +

.ft

= f (I - ( )(1 + 'Y/o)
= HI + (o)(l - r, 2)

f;

'Y/o)( - I

-u

---

LX

= 2, f;v;

(o

+ 'Y/o)

gl

I
15

q1-

2a

(a)

( I H)

i= l

----

I
I

= 2, f;u;

where

2b

q2

(1 7)

which contains x and y translations at each node. We also assume the fo!Jowi11 j1
quadratic displacement shape functions:
u

(i

= 1, 2, 3, 4)

(i

= 5, 7)

(i

= 6, 8)

(,

,,,

'l'hll
values
of 9 and 'Y/; required in these formulas [see also Eqs. (7)) are glvrn
i11 Table
7.3.

for element QB we take the geometric interpolation functions to be lht1

sunie as the displacement shape functions in Eqs. (19). Physically, this

11w111111

Ihut the natural coordinates gand r, are curvilinear, and all edges of the elc111c111
hcco,nc quadratic curves [7]. Thus, we locate any point on the quadrilatrn,I
(including point g) by the formulas
8

= 2, .ftx;
i= l

= 2,f;y;

(.!II)

l'lllment Q8 is seen to be isoparametric . Formulations of stiffnesses, co1111i~

lc1111 111usscs , and equivalent nodal loads for this element are very similar to thoNt
1111 t.1lt.1 111ent Q4 given earlier. Table 7.4 contains the necessary shape fu11<:lior11,
111111 Iheir derivatives with respect to gand r,. Numerical integration also follow~
lh1 s111m pattern as before, even though the lornl coordim1tcs arc curwd .
1111el

q 1 - . __ _ _ _ _ _ __

q2
Fl11111r 7, IO

Elunu.inl

(Ii)

Qll: (n) i 1:ct11111111l111 p111 \1111 : (h) iN01111111111tltric cou111crp11rt.

339

J40

tw
'l'Alll ,1!. 7, I

Nmlitl t

111111

11111111 I lh1111111 l1111nl < 111111111111

111111ll11nl1

tor

(,

- I

,,,

- I

lr.11111N1I
~

()

- I

- I

/ ,II

111,

1'11111111111 l>YNAl'li 1111 1'1101111,trnH nnd 1'111110 S1111ln

QH

TAIILE 7.5

Slrudur11I Dutu for Program DYNAPS

Type of Datu

341

hlem identification
1ctural parameters

Items on Data Lines

l
1

Descriptive title
NN , NE,NRN,IPS,NEN,E,PR,RHO,H

l'ln ne

TABLE 7.4

-3

Shape Functions and Derivatives for Element Q8

Ji
(I

stress (strain) data

n) Nodal coordinates
b) Element information
c) Nodal restraints

No. of
Lines

~( I - 71)(-{ - 71 - 1)/ 4

Ji.,,

fi.t
(2{ + 71)(1 - 71)/ 4

( I - m211

( I I- ~( I - 71)({ - 71 - 1)/4

(2{ - 71)(1 - 71)/4

(I+ m2r1

(I + m 1 + 71)({ + 71 - 1)/ 4

(2{ + 71)( 1 + 71)/ 4

o + m211

( I - ~( I + 71)( - { + 71 - 1)/4

(2{ - 71)(! + 71)/4

o - m271

(I - 2)(1 - TJ)/2

- g(I - 71)

- (I -

(l + ~( I - 71 2)/2

(I - 712)/2

-( I I ' )11

(I - f)(J + 71)/2

-[(I + 71)

(I -

( I - ~( I - 71 2)/2

-(I - 71 2)/2

-5

"

Dynamic Load Data for Program DYNAPS

Type of Data

2} /1.

- (I -

Jlor sequences of node numbers, see Figs. 7.9(b) and 7. IO(b).

TABLE 7.6

C1

~ )11

7 .6 PROGRAM DYNAPS FOR PLANE STRESS AND PLANE STRAIN

Now we shall discuss Program DYNAPS for dynamic analysis of thin plnll~ Ill
plunc stress or prismatic solids in plane strain. For this purpose it is assumed tha
H ttivcn continuum has been discretized using either Q4 or Q8 elemc111,, a
dtstribcd in Sec. 7.5. All such quadrilaterals in the analytical model huw th
1111111c thickness h, and the material is taken to be homogeneous and iso1111pl
Table 7.5 shows preparation of structural data for Program DYNAPS . I hi
lnhlc is simi lar lo Table 3.2 for plane trusses, described previously in Ser. I N
1lowcvcr, the second line of Table 7.5 contains four additional structural p11111111
1k1s . The first of these is an indicator for plane stress (IPS = 0) or plarw ,111,ln
(IPS
I). Next is the number of element nodes NEN for the quadril11lt111I
Whtrc NEN = 4 for Q4 elements and NEN = 8 for Q8 elements. The othtr 1w1
sl rurturul parameters included are Poisson's ratio PR and the thicknt~, II
hrr lhcrmore, clement information now consists of the element number I 111111
ll(l(k numbers one through four for element Q4 and one through cip.hl fur
tlt111tnt Q8.
Prcpuration of dynamic load data appenn, in Tublc 7.6, which is si 1111l111 111
Tuhk 4. 1 f'or plane trusses (sec Sec. 4. 10). l lowtYC.1r, the loud p11111111tll1s NI I

J, X(J), Y(J)
I, IN(I, 1), IN(I, 2), ... , lN(I, NEN)
J, NRL(2J-I), NRL(2J)

NN
NE
NRN

Dynamic parameters

No. of
Lines

Items on Data Lines

!SOLVE, NTS, DT, DAMPR

Initial conditions
(a) Condition parameters
(b) Displacements
(c) Velocities

NNIV

NNID, NNIV
J, D0(2J-1), D0(2J)
J , V0(2J-l), V0(2J)

Applied actions
(a) Load parameters
(b) Nodal loads
(c) Line loads
(d) Volume loads

I
NLN
NEL
NEV

NLN,NEL,NEV
J, AS(2J-l), AS(2J)
J, K, BL!, BL2, BL3 , BL4
I, BVI, BV2

( lround accelerations
(a) Acceleration parameter
(b) Acceleration factors
Forcing function
(a) Function parameter
(b) Function ordinates

NNID

l
l

l
NFO

IGA
GAX,GAY

NFO
K, T(K), FO(K)

"Pertains only to element Q4 (NEN = 4). For element Q8 (NEN = 8), we need three
node numbers (J, K, L) and six force intensities (BL!, BL2, ... , BL6).

,111d NEV in Table 7.6 must be explained. If NEL * 0, at least one line load
length) exists on the edgejk of an element. The types of line loads
1111 quadrilateral elements are illustrated in Fig. 7 . 11. For the Q4 element a
ll11111rly varying line load is specified by the force intensities BLl through BL4
1111,t arc listed in the table and shown in Fig. 7 . 11 (a). On the other hand , the Q8
, ll'111cnt in Fig . 7. 11 (b) hus two comprnwnl~ of quadratically varying line loads,
,h llrll'd by the force intensities BLI 1h1011gh BL6 (sec the footnote below the
111hll
t lmtc per unit

).

341

I W11 1111d lh rnn lll111111111lc11111I (;01111111111

111

I II

111, 11 111111 l>YNAI' 1111 l'ln11n

1,111111

nnd f'luno Strain

343

Ihm v11ry1ng
st,essts o., 0 1, 111., 1111t I <r,, tl1e, last
,
of which is nonzero forthe case
e1f plunc strain. I lowevlll', wt foun<.I il necessary to use~ = 3 eac? way :c~:r:}
ht lhc consistent-mass matrix of element Q8, which contam prod
c11111dmtic functions.

I 111n1>le 7.2
11 lllll 7 . 12 illustrates the cross section of a machine part that is s~bjecte? t~ a state of
Ill
Th
tis divided into Q8 elements and has a parabohcally d1str1buted step
I' nm Stram. e par
h'
J the physical
hmd on edge 1-5 , with maximum intensity bx at node 3. I n t 1s examp e
p,1111111ctcrs are:
y

L.

b,,

la)

b,s

/2
4

b,,
/1

L.

b,,

lb)

Figure 7.11

Line loads for quadrilaterals: (a) element Q4; (b) element QB.

If NEV =I= 0, at least one element is subjected to volume loads (fol"l'l' prr
unil volume). Here the notation implies that element I may have a unifiH11il y
distributed force BVl in the x direction and a second uniformly distributed fH l
13 V2 in the y direction.

In Program DYNAPS we take n = 2 each way Lo locale point s 1111

numerical integration. At each of the four points lhe computer evul1111t ts 1111

n,
IC'lic1u1 7.12

l'lnm Hlcnlc1 with ()H ,,1,111tc.llllN.

11

344

Two- and Three-Dimensional Co ntinua

E = 207 X 106 kN/m 2

R2

2Ri

= 0.02 m

v = 0.30

p = 7.85 Mg/ m3

h=Im

bx= I kN/m

Chap. 7

for which the material is steel and the units are SI.
We ran this data with Program DYNAPS, using IPS = 1, NEN = 8, DAMPR _
0:5 ,. and NUMI~T for the solution method. Translations of node 5 in the x and 1,
directions appear m the computer plots of Fig. 7 .13(a). Also, the normal stresses SX
SY, an? SZ near node 29 are plotted in Fig. 7.13(b). Maximum values of the nodai
translations are 0.06505 mm and 0.04775 mm and those 1"or th
I
56 20 MP
'
e norma stresses an
'

a, 11 6 8 MPa, and 51 . 89 MPa. (The maximum shearing stress accompan in

the normal stresses near node 29 is less than a twentieth of SX.)
y ~

Sec. 7.7

7.7 ISOPARAMETRIC HEXAHEDRA FOR GENERAL SOLIDS

Figure 7 .14(a) illustrates the parent rectangular solid RS8 of the isoparametric
hexahedron H8, shown in Fig. 7.14(b). In the former element, the natural
coordinates g, 1/, and ( are orthogonal everywhere. For either element the
generic displacements at a typical point are

= {u, v, w}

(1)

Nodal displacements consist of x , y, and z translations at each comer node,

which fill the vector:
q

DJ l
E

345

lsoparametric Hexahedra for General Solids

= {qi , qz, q3, ... ' q24} = {u1, Vi , W1,

. . . ' Wg}

(2)

Trilinear displacement shape functions may be expressed as

,._I:; 0, 06

Cl

v = Ifiv;

J;u;

w = Ifiw;

i=I

i= I

~ 0.04

lf)

(3)

i=l

where

I-

<

f;

(/)

k(1 + to)(l +

T/o)(l

+ (o)

(4)

;'. 0. 02

nnd

I-

<

T/o

-'
(/)

z
< 0. 00
o::
I-

o.os

o. l 0

0 . 15

0.20
(ms)

(a)

l 20

a.."'
:i::

SY

l 00

(})

w an
Cl

60

U; =

<

f;Q;

(i

f;

= [~

= 1, 2, ... , 8)

(6)

111 which

~ 40
(/)

~
(/'/

(5)

TJ;TJ

The formulas represented by Eq. (4) are the same as those given previously in
llqs. (7 .3-21). Values oft;, 1J;, and (; required for Eqs. (5) appear in Table 7 .7.
For the parent rectangular solid RS8, explicit integrations are feasible, and
Hliffnesses for an orthotropic material were presented by Melosh [8]. Furlhcrmore, consistent masses for this simplified element are easy to derive and
were given in Ref. 2.
Turning now to the more general isoparametric H8 element [9] , we take
lhc geometric interpolation functions to be those that were given earlier as Eqs.
( I .3-20). Because the geometric and displacement shape functions are the same,
lhc H8 element proves to be isoparametric. Equations (3) can also be stated as
lhc matrix expression:

er

20

Ct".

CJ) 0

( lil
1"1111111 7, 11

lfr~llllllNl N Il l pJ11111 Nflllill I ~11111pl!1 fll Nll1p 1111111

(7)

~]/;
1

Ai, lwl'ore, the generic displacements u1 in Eq. (6) denote translations at any
p111nt due lo the displ11cc mcnts q, at node i.
l1111ddilio11 , Hlnlin displut.'l'llll'lll nl11tionships 11111y be written efficiently as:
t I

ll1f1

(/

I , 2.

'H)

(8)

I w11 n11d 1111 nu l>lmonsionai Continua

Sec. 7.7

lsoparametric Hexahedra for General So lids

TABLE 7.7

,,

Nodal Coordinates for Element HS

g;

7/i

C,

-1
1

-1
-I

-1
-1
- I
-1
1

2
3
y

)-,
v,

.f

I -

347

-1

5
6
7
8

-I
1
1

I
-I
-I
I
I

-1

1
1
1

where
0
fi .x 0
0 f;,y 0
0 0 f;,z
f; ,y Ji,x 0
0 J;,z f;,y
f; ,z 0 f; ,x

ti

B;

(a)

= d f; =

0
D G2i

0
0

D G3;

DG2;

D Gli

D G3;

D G3;

DG!i

0
0

(9)

DG2i
DGli .

Terms appearing in the submatrix B; were discussed in Sec. 7. 3.

Next, we express the stiffness matrix for element H8 in Cartesian coordiIU\tes to be

fv BT(x,y,z) E B(x,y,z) dx dy dz

(10a)

In natural coordinates, this equation becomes

v,

und three applications of Gaussian quadrature give

5
i

/
w,

r
-

(b)

1'1,curl! 7.14
111111 .

II

rt

R j RkR1B[k,1E Bj,k,l

IJj,k,/ 1

(10c)

l=l k= l j= l

' l'his formula for numerical integration implies that the matrix Bj.kJ and the
determinant IJ j,k,1 I are evaluated at each integration point, where the coordinates
lll'l'

Element 118: (11) parent rectangular sol id; (b) isoparamctric countllf

11

=II I

(~, 'Y/k, (1).

Similarly, the consistent mass matrix for element H8 is

,,f1F, (r'u.

fT(x,y,z) f(x, y,z) dx dy dz

,,.l ll'(/ , 1/,tl l ,IU , r1,[)l,ll,/,,t1l

(1 la)

(1111)

340

lw11 111111 I l11u11,l)lrrn11111fw111I Co111l1w11

mi,

7.1

lsopuromotrlc Hexahedra for General Solids

349

or
Considering first the term M, 1 , we have from Eq. (15)
1l

M =PI

"

I I

ft

Rj RkR,ffk.l lj,(., j,JJ,k,/ 1

(a)

/= I k= I j = I

in which p is assumed to be constant.

Furthermore, equivalent nodal loads due to body forces have the f'cu 11

Ph(t)

fT (x , y , z) b(x, y,z,t) dx dy dz

wlll'rc
1

J; = kO -

( I 111)

g)(l -

TJ)O - ()

(b)

,11hslilution of Eq. (b) into Eq. (a) produces

or
M,, =

a:~p LL L

(1 - g)2 (1 - T/)2(1 - ()2 dg drJ d(

= Sabcp

(c)

27

or
11

Pb(t)

=L

1'1 occcding in a similar manner, we can find all of the terms in the first column of matrix
M, us follows:

L L RjRkR1ffk,1b(t)j,k,I IJj,k,d

l = I k= I j = I

(I

1
1

pV
(M)col.1 =

where both Ph and bare functions of time .

216

{8, 0, 0, 4, 0 , 0, 2, 0, 0, 4, 0, 0 , 4, 0, 0, 2, 0, 0 , 1, 0 , 0, 2, 0, O}

. F?r the rectangular solid RS8 , the Jacoblan matrix and its deterniirumt
specialize to

J =

r~ ~ iJ
0

(d)

111 which the volume V = 8abc.

JJj = abc

( Ill

I(ere _the constants a, b, and c are half the dimensions in the g TJ , 1 ~

1111
d_1rcct_
10ns, as shown in Fig. 7. 14(a). Thus, Eqs. (10b), (lib), and (l,2h)
111
simplified, as follows:

Now let us examine a higher order hexahedral element that is formulated

11Ning quadratic geometric and displacement shape functions. The parent rectan11lur solid (element RS20) is illustrated in Fig. 7.15(a), and its isoparametric
1 ounterpart (element H20) appears in Fig. 7 .15(b). For the rectangular solid the
1111tural coordinates g, TJ , and ( are orthogonal, and nodes 9 through 20 are
lol'nted at midlengths of the straight edges. Both elements have the nodal
1ll11placement vector

Q = {q,, q2, q3, , , . , q60} = {u,, V1, Wt , . . . , W20}

( II)
lllld

(17)

1h111 contains x, y, and

111tnl

z translations at each of the 20 nodes. Quadratic displaceshape functions for these elements are
20

u=

Ullcl

P,,(t)

= abc

r, r. f

20

J;u;

i= l

20

v = Ifi v;

w=IJ;w;

i=l

i= I

(18)

wh~rc
I fT(g, TJ, ()b(t, TJ, ( , t) dt dTJ d(

( lr1)

11:x111u11fo 7.3
Ass1 u11i11g .1hu1 P _is constant, let us derive the terms in the first column of the l'OIININII
r1111Ns mutn x M for the rectangular solid clement RS8 shown in Fig 7 J '1(u) JI
II 111
111
J>111poN1\ w1 tWL'd functi ons /, through};i given hy liq . (4) wi th vu lu~s ; ,. ,:
111k<111 lwn , '1'11hl c 7.7,
'
1 ,.,, l/1, _11111 L

i"

/,

H I + so)O + 7Jo)(l + (o)(to + T/O + (o - 2)

I,

,t( I

I,

Ht

/1 .t ( I

+ TJo)(l + (o)
TJ 2)( 1 + (o)( I + so)

- g2)( I

(i

= 1, 2,

... , 8)

= 9, 11, 17, 19)

(i = 10, 12, 18, 20)
(i

(i

= 13,

14, 1.5, 16)

( 19)

II o

IWtt 1t11ol 1111 1111 lll1111111nlo1111I Co11ll111111

"' / .U

Pro111 u111 l>YN/\1,ll

1111 (11111111111

351

Solids

1'AU1.11; 7 .H Nodnl Coordinates for Element H20

10

)-,

20

7
8
9
10

v,

. er

5--------e----..JI
6

II

-1

- 1

11

-1

1
1
-1
-1
1
1

- 1
1
1
-1

- 1
-1

12
13
14

- 1
-1

-1

-1

0
0
0
0

- 1

1
-1
0

0
1

-1
1

1
1
1

- 1

15
16

- 1

1
1

17

- 1

18

-1

19

1
1
1

- 1

20

-1

!1.{ = kO + 2g +
I >I'
X

T/

+ ?)(1 - 71)(1 - ?)

course, numerical integration is required for this element.

v,
5

t
-11

(b)

ll'll(Ut't' 7.15

1
1

111 this instance, the faces and edges of the element are quadratic surfaces and
111 ves, as indicated in the figure.
Terms in the Jacobian matrix for element H20 are the same as those given
111 Eqs. (7 .3-23), but with the upper index 8 changed to 20. Furthermore, the
1qu utio ns for element H8 will pertain to element H20 if the number 8 is changed
In 20 in appropriate locations. Derivatives fr.,, and so on, required for the
dvclopment of element H20 are easily obtained and need not be tabulated. For
1K11mple,

l'Ollll(l\lJ)Url .

- I

TJ;

t,

{,

- 1

t,

T/1

w,

(al

)-

2
4
5
6

2a

t,

-----,
17

Element H20: (a) parent rec tangular solid ; (b) isoparamclrk

Y11 '.111s ,0 1' ,, 'Y/1 , un~I (, for these formulas arc listed in Table 7.8. Ex 1,111 11
111111,1111,i1011s a.re poss ible for the subparametric parent clement.
I 120 clement 1101 in p,g 7 I5(b) , we use gco111111 I1
I I hir
I lhc 1sopurumctric

.
111 ipo 1tl1<'.1~ lunc11ons th al arc the same as the displacement shape f'unctionN 111
1<c111 ( 11) ). llrns,
.1(1

\'" ,/'I 11
I I

20

~ .l,v,
I

20

}: .fi z1
I

( 1111

1,8 PROGRAM DYNASO FOR GENERAL SOLIDS

111 this section we describe Program DYNASO for dynamic analysis of general
H111ids. Before using this program, we must discretize a given solid using either
I IH or H20 hexahedral elements (see Sec. 7. 7). The material of the solid is
.tHN umed to be homogeneous and isotropic.
The manner of preparing structural data for Program DYNASO is shown
111 Table 7. 9. Included among the structural parameters is NEN, which is the
1111mber of element nodes for each type of hexahedron. If NEN = 8, the hexaludru are H8 elements; and if NEN = 20, they are H20 elements . Therefore,
r ll'lllent information consists of the element number I and node numbers 1
011ough 8 for element H8 and 1 through 20 for element H20.
Table 7. IO gives preparation of dynamic load data for Program DYNASO.
l11 sl, we sec that there arc three possible initial displacements, initial velocities ,
1111d applied forces for each node, Also , lhc clement loads implied by the
p111m11ch.:rs NEI., NEA , und NHV ll' q11111 Home explanation . If NEL :/= 0, at
h111111u1t li111. loud (lon'l' pl' I' uni I h-n~tlt) 1111 ~tH011 lhc l'dgl' jk of' an c lement. For
, 11111111 I IH II lt11l'1til y v11 1yi ,11 11111 Ii 111d ii, d, f1111d hy tlw 1'0111 i111rnsitks Ill. I

Jb2

Iwo 11111 I '111111 l>l11111111lcm11I c0111'111111

TABLE 7. 9
Type of Data
Problem identification
Structural parameters
General solid data
(a) Nodal coordinates
(b) Element information
(c) Nodal restraints
a

'111c. / .8

Structurul D11tt1 for Program DYNASO

No. of
Lines

Items on Data Lines

1
I

Descriptive title
NN,NE,NRN, NEN, E, PR, RHO

NN
NE
NRN

J, X(J), Y(J), Z(J)

I, IN(I, 1), IN(I, 2), ... , IN(I , NEN)
J, NRL(3J-2), NRL(3J-l ), NRL(3J)
.._

For sequences of node numbers, see Figs. 7. l4(b) and 7. I 5(b).

TABLE 7.10 Dynamic Load Data for Program DYNASO

Type of Data

No. of
Lines

Dynamic parameters

!SOLVE, NTS, DT, DAMPR

Initial conditions
(a) Condition parameters
(b) Displacements
(c) Velocities

l
NNID
NNIV

NNID, NNIV
J , D0(3J-2), D0(3J-l), D0(3J)
J, V0(3J-2), V0(3J-l ), V0(3J)

l\ppl icd actions

(11) Load parameters
(b) Nodal loads
(c) Line loads"
(d) Arca loadsb
(c) Volume loads

l
NLN
NEL
NEA
NEV

NLN,NEL,NEA,NEV
J, AS(3J-2), AS(3J- l), AS(3J)
J, K, BL!, BL2, ... , BL6
J, K, L, M, BAJ , BA2, . . . , BAl 2
I, BVJ, BV2, BV3

( lround accelerations
(11) Acceleration parameter
(h) Acceleration factors

Items on Data Lines

Fmt'i ng function
(11) Function parameter
(h) Function ordinates

lhr ough BL6 . The first four are as shown in Fig. 7.1 l(a) for element Q4, and
rhc last two pertain to the z direction. On the other hand, the H20 element may
huvc a quadratically varying line load, specified by BLl through BL9 (see the
llrst foo tnote) . The firs t six force intensities have the same meanings as in Fig.
I . I I (b) for element QS , and the last three are for the z direction.
IF NEA -::f:. 0, at least one element has an area load on one of its surfaces.
I'ig ure 7 .16 shows the types of area loads for hexahedra. For element HS such
loads on face jklm are defined by 12 numbers, of which the first four (BA l
through BA4) denote force (per unit area) in the x direction, as indicated in F ig.
l . I 6(a). The next four (BAS through BAS) pertain to they direction, and the last
lour (BA9 through BA12) apply to the z direction. It is assumed that each
component of area loading has a bilinear variation over the surface jklm. On the
other hand, we take a biquadratic variation of area loading on a surface of the
1120 element. In this case the loads are specified by 24 intensities (see the second
l'ootnote). The first eight represent force (per unit area) in the x direction , as
depicted in Fig. 7 .16(b). The next eight are for the y direction, and the last eight
11rc for the z direction.
Volume loads BVl, BV2, and BV3 on both types of hexahedra simply
l'Onsist of uniform intensities of force (per unit volume) in the x, y, and z
directions. Also note that GAZ is included in the list of acceleration factors , as
for any three-dimensional analysis.
In a manner similar to Program DYNAPS, we use n = 2 in each of three
ways to locate points for numerical integration. Thus, there are eight such
points, at which the time-varying stresses CTx, CTy , CT,, 'Txy , Ty,, and Tzx are determined in Program DYNASO. Again, we had to use n = 3 each way for terms
in the consistent mass matrix of element H20 to retain sufficient accuracy .
11:xample 7 .4
'l'hc tapered cantilever beam in Fig. 7. l 7(a) is doubly symmetric, and the parabolas with
11pcxcs at the support determine the rate of taper. This beam is made of reinforced
l'Oncrcte and has a rectangular impulse of magnitude Pi and duration t 1 applied in they
direction at its free end. Physical parameters are given as follows:
E = 3.6 x 103 k/ in. 2

J
l

I
NFO

"'I t11111nh only ro clc_mcnt HS (NEN = 8).

353

1,011111111 DYNASO for General Solids

IGA
GAX , GAY , GAZ

NFO
K , T(K), FO(K)
For clement H20 (NEN

nodr numbe1h ond nine force intensities .

"Fm L' k1111.ml 1120 use eight node numbers and 24 force intensities.

= 20)

wc need rhrl'l'
'

= 120 in.

11

= 0.15
Pi =

400 k

2.25 x
t1

10- 7 k-s 2 / in.4

5 ms

where the units are seen to be US.

Figure 7 . I 7(b) shows discretization of a quarter of the beam into two H20 ele111cnts, with nodal restraints imposed for symmetric and antisymmetric deformations. To
define the geometry of this simple network, we need only state that the x-coordinates of
nodes 9 through l 2 are 9 0 in. , those of nodes l 3 through 20 are 60 in. , and those of nodes
1 1 lh rough 24 arc 30 in.
We used the foregoing data in Program DYNASO with NEN = 20, DAMPR =
II. I0 , 1111d solu1 io11 hy Suhprogrnrn NORMOD (wilh NMODES
12). Figure 7. I 8(a)

354

Two- and Three-Dimensional Continua

Sec. 7.8

Program DYNASO for General Solids

355

k
y

0.1L

)-,
z
0.3L

(a)

(a)

j2

Sym.

25

21

j3

13

- i,

29

32

12

/20

24

Typ.

Sym.
(ll)
(h )

li'l1111r1 7. 16

t\rcu loud~ for hcxuhcdrn: {u) tlcnll'nt IIH; (h) clcnw111 1120

Figure 7. 17

(a) Tapered c11111ilcvcr beam; (b) 1-120 clements.

t wu 111d lli11m 1>11 1111 111 1011111 Co11t l111111

."'

Ill

~.Lt

lnop111 1111111t1 ii I hHIIMIII N 1111 A,<11y11111111tllo tlolld

367

,h pil'IH II qu111 l(:r ol l hl 111t1111Httl111 l111p11I N1' 111 nodc 6; and they-translation at that point
I plotll'd in Fig. 7. I8(h), Nh11wi11g II muximum value of 0.09514 in . at t = 3. 75 ms.
\ 1~11 . the normal stress SX ncur node 25 is plotted in Fig. 7. 18(c), for which the
111l11i111u111 value is - 1.040 ksi at the same time .

10 0

Q.

;! 50

'-

(.0

~ 0

'

'

2'

<-so

,0 ISOPARAMETRIC ELEMENTS FOR AXISYMMETRIC

SOLIDS

\ 11 nx isymmetric solid is defined as a three-dimensional body that may be

d1vdoped by rotation of a planar section about an axis. This type of body is
_,1111ctimes called a solid of revolution. Cylindrical coordinates r, z, and ()
p111vide a suitable reference frame, as illustrated in Fig. 7 .19. We assume that
th1 body is axisymmetric with respect to the z axis and that a typical finite
1 l11111cnt is a circular ring. This ring element may have various cross-sectional
h11pcs, but we will deal only with isoparametric quadrilateral sections. Allhough nodes are shown as dots on the cross section of a ring element, they are
m1ually nodal circles. If the loads on an axisymmetric solid are also axisym111c1ric, we may analyze a representative cross section as if it were a two-

l-

<.

3 - 100

(a )

C
~

0.10

(.0

~ o.os

'..t,

Z,V

0, 00
2

Cl

;- -o.os

s)

<

-J
U)

-o. 10

,0(b)

>-

1--- -. I , 00

IJ'l

---

o.so

(\J

o.oo

10

T Cms)

-o.so

Iii

:;,:

~ - 1 .oo

I
1z

(c)

!figure 7.18 Tapered beam: (a) loud; (b) displucemcnt ; (c) stress.

Sec. 7.9

lsoparametric Elements for Axisymmetric Solids

369

Two- and Three-Dimensional Continua

358

dimensional problem. At first, only axisymmetric patterns of loads will h~

considered, but nonaxisymmetric loads also will be covered later in the section
.
For any point on an axisymmetrically loaded ring element, the gencrk
displacements are
u={u,v}
(I)

Figure 7 .20(a) shows the cross section of element AXQ4, which derives its
characteristics from the quadrilateral element Q4 in Sec . 7. 5. Bilinear displacement shape functions in matrix f are the same as those for element Q4, and the
strain-displacement submatrix Bi becomes

Tr~nslat~ons u an? v occur in the rand z directions, as indicated in Fig. 7.1'>

With ax1symmetnc loads, the translation w in the () direction is zero and tlu
shearing strains 'Yr& and y,8 are also zero. However, the figure shows f;ur typl'~
of strains that are nonzero, as follows:
(" )

Relationships between these strains and the generic displacements in Eq. ( 1)

seen to be
Er

_ au
- ar

111 r

av

= az

'Yrz

au

av

dZ

+ ar

a
ar
0

d=

r
a

az

(i = 1, 2,3 , 4)

(7)

fi., .h,r
which is obtained by using the operator d in Eq. (4) on submatrix ( from
Eq. (7 .5-5). The radius r in Eq. (7) is found as
(8)

In addition, the derivatives/;,r = Deli , and so on, are given by Eqs. (7.3-18) ,
except that r and z replace the coordinates x and y .
The stiffness matrix for element AXQ4 may be formulated in natural
coordinates as

K =

a
az

ff
1

<I)

= 21T

Jo27T BTE B \ J \r d(J dg d17

1

L, rl

BTE B \ J \r dg d17

(9)

Similarly, the consistent-mass matrix is

a
ar

Sym .

1J

+ 11)(1 - 2v)

Bi=

i= l

For an isotropic material , the stress-strain matrix is

(I

fi.,

In this instance the nonzero term 1/r in the third row of matrix dis a multipl1r1
of u, not a derivative.
Corresponding to the strains in Eq. (2), the four types of nonzero strcslil'II
depicted in Fig. 7. 19 are
(J' = { CTr, er,, (T9, Tr, }
( S)

r = "'Zfir;

These relationships are embodied in the differential operator

E =

21r(r + u) - 21rr u
=
=21rr
r

( I)
E,

.h,r

,,

1J

1J
II

1J

= p

rl L,
rl rl

L27T fTf \ J \r d8 dg d17

= 21Tp

fTf \ J \r dg d17

(10)

/\lso, equivalent nodal loads due to body forces are

Pb(t) =

L, rl r1T

= 21T

rl rl

fT b(t) \ J \r d8 dg d17
fTb(t) \J\rdgd17

(11)

( )
1J

I -

2 11

'l'ht MI X I 111111 y i1; H01tll111 tu till' I X I 111utii x 1111 pl1111, Nlnii11 In l\q ('/ ,). 11)

Numerical integration is required to evaluate Eqs. (9), (10), and (11).

The cross section of element AXQ8 appears in Fig. 7.20(b) . Its properties
nrc sitnilar to thrnic of the quadrilateral clement Q8 in Sec. 7 .5. Biquadratic
di splnL'l'tlll'III ~l111p, t 1111l'lions in 111alt ix f H1c the sam e as for clement Q8.

360

Two- and Three-Dimensional Continua

Chap. 7

Sec. 7.9

361

lsoparametric Elements for Axisymmetric Solids

Expressions for B;, r , K, M, and ph(t) are similar to those for element AXQ4
given by Eqs. (7) through (11), except that i = I , 2, ... , 8.

z
\

Example 7 .5

Find the consistent mass term M 35 for the axisymmetric solid element AXQ4 in Fig.
7 .20(a), using Gaussian numerical integration with n = 2 each way. Let the coordinates
of nodes 1, 2, 3, and 4 be (10,2), (14, 1), (15,5) , and (11 , 4), respectively.
From Table 7 .1 for n = 2, we have R1 = Rk = 1, so that the numerical form of
Eq. (10) gives
2

M 35

21rp

L L (ht.k(J3)j,k IJ,.k lrJ,k

(a)

k=I j= I

---.... u .
i
'

ks

In this case the bilinear functions h and h are needed to evaluate the term M35 because
of the arrangement of matrix fin Eq. (7. 5-5). Substituting Ji and.fi from Eq. (7 .5-6) into
Eq. (a) yields
(b)

To apply this formula , we first calculate the Jacobian matrix from Eq. (7. 3-10), as
follows:

(a)

(1 - 'l'/)

(1

+ 'l'/)

+ g)

(1

+ g)

-(I
z

' '\

-(1

+ .,.,)] 14
10 42~1
g) [ 15

(1 -

11

(c)

Then the determinant of J is

(d)

llvaluating the terms in Eq. (b) at each of the four integration points and summing the
results produces
(e)
M35 = 21.061rp = 66. l5p

-u.'

which can be finalized using a numerical value for p.

k==ser----..1__0- - r
I
(b)

Turning now to nonaxisymmetric loads , we can divide them into two sets
/ I 11 . The first load set is symmetric with respect to a plane containing the axis
or revolution , and the second is antisymmetric with respect to that plane. For
l'onvenience, the r-z plane is taken to be the plane of symmetry. Fourier
tl/'composition [12] of the symmetric loads form harmonic terms produces
Ill

Figure 7.20 Quadrilateral sections: (a) element AXQ4; (b) element AXQ8 .

b,.

Ill

b,j cos JO

b,

=I

J O

J O
Ill

ho

L IJ111 sin JO
I

b,j cos JO
(12)

Two- and Th ree-Dimensional Cont inu a

362

Chap. 7

Sec. 7.9

lsoparametric Elements for Axisymmetric Solids

363

where b,j, bzj, and b8j are functions of rand z only. When} = 0, we have be = O;
and Eqs. (12) become the case of axisymmetric loads. Otherwise, j = 1, 2,
... , m represent cases of nonaxisymmetric loads that are symmetric with
respect to the r-zplane. Figure 7.2l(a), (b), and (c) show the first harmonic loadN
for the r, z, and directions, respectively. If the loads were antisymmetric with
respect to the plane of symmetry, the functions sin Je and cos Je would bl
interchanged.
Generic displacements for nonaxisymmetric loads must include the trans
lation w in the e direction . Thus,

u = {u, v, w}

(IJ)

(a)

and we must also have ,',8 and r',o in the strain vector, as follows:
E

(14)

= { Er, Ez, Ee , Yrz, Yzo, Y,o}

Strain-displacement relationships developed by Love [13] are

Er

=
"1

au
ar

E,

av
az

1 aw

= -; + -; a8

av
r ae
1 au
aw w
-va=-- + - - r ae
ar r

au+av
-az ar

rrz -

Ee

aw
az

'Yo= - - + z

( 15)

(bl

I lcrc we see that the radius r appears in the denominators of several expressions
From Eqs. (15) we can form the operator d as

d=

a
ar

a
az

r
a
az
0
1

a
r ae

a
ar
1 a
r ae

r ae

a
az

a
or

( I hi

Figure 7.21

,.

Tlw slrcss vector for nonaxisymmctric loads musl contain r, 11 and

lollows:

er

!11,, 11,, 11 11 ,

11

1 ,,, 1111 )

7', 11 , 11 '1

( 1/)

Nonaxisymmetric loads: (a) b,1 cos 0; (b) b,1 cos O; (c) bo1 sin 0.

Stress-strain relationships are easily extended to cover six types of stresses and
llw corresponding strains. For example, if the material is isotropic [see Eq. (6)] ,
Wt' odd / 1\, - H,16
H/2( I + v) to form a 6 x 6 matrix E.
Tlw response of Hll nxisymmclric solid to II series of symmetric , harmonic
lo11ds rnusists of 11 1w1 il'S of symnwt, it, h111111011k, pentric displ11ccmcnts that
11111y Ill' t'X PH'HIH'd 111'

364

lwo- 1111d l 1111111 l>l111nr11l111ml Lo11tl11u11

n1
II/

=L

Uj

/ ,

1'1011111111 l>YAXSO for Axleymmotrlc Solids

10

cos J(:)

j O

=L

VJ

cos JO

IINl11g llqs . (2 1) in Eq. (22) , we find that

j O
Ill

w =

w1 sin}(:)

j=O

Again, i~ the l?a~s were antisymmetric with respect to the plane of syn111111, v
~he funct10ns sm J() and cos JO would be interchanged. Applying the opcr111111 d
m Eq. (16) to Eqs. (18) expressed in terms off, we find a typical partilio11 uf
the strain-displacement matrix to be

(B;)1

Ji., cos}()

Ji.z COS J()

0
0

Ji- cos}'()

}.Ji
- cos}'()

Ji., cos J()

.Ji Slllj
. '()
-1-

Ji., cos}()
0

.Ji
-1r

. '()

i,r -

f, {, f"
f, L

where k = 2 for j = 0, and k = 1 for j

(k = I) appears as a consequence of
2,r

cos J() d() =

(j = 0, 1, 2, . .. , m)
2, . .. , m. The latter co1111111111

sin 2 J() d(J = 1r

( 'II

i11 which
CJ

()

()

,}J

fT f IJ I r dg dYJ

(24)

fTcJc1b(t)j

IJ I rdOdgdYJ

fTb(t)1 !JI r dg dYJ

(j

= 0,

1, 2, .. . , m)

(25)

= {b,j,

bzj, be1}

(26)

Finally, the stresses for each harmonic response are

(j

= 0, 1, 2, . .. ,

m)

(27)

t >I course, such stresses, as well as nodal displacements, must be added at the
, 11tl of the analysis.

/ ,10 PROGRAM DVAXSO FOR AXISVMMETRIC SOLIDS

( "l

0
cos JO

77

b(t)j

~Ill}

Similarly, the consistent mass matrix for each harmonic set of symn111111
di splacements becomes

rco;/

Lf

( 111)

= 1,

[21T

Jo

k1rp

whore

BJE B1 JJI rd() dg dYJ

BJE B1 IJ I r dgdYJ

= k1r

f, f, f
= f f
=

hr

(ii fi) . '()

Slllj

( I!))

Ji.z sin j()

1liiN l'ormula is the same for j = 1, 2, ... , m, and it does not change for
,1 111tsy111111etric displacements. Thus, to determine frequencies and mode shapes
1111 nny value of j, we use the (variable) stiffness matrix K1 from Eq. (20) and
1111 (constant) mass matrix M from Eq. (24). However, in the latter equation note
11!111 k = 2 for the case of axisymmetric vibrations, where j = 0.
Equivalent nodal loads for each harmonic set of symmetric body forces
111k1 the form

p,,(t)1

where i = 1, 2, ... , nen and j = 0, 1, 2, ... , m.

An element stiffness matrix for each harmonic set of symmetric displ111
ments may be written in natural coordinates as

~=

365

Ill

(j

0, I , 2, . . . , 111 )

11

I , t us now consider Program DYAXSO for dynamic analysis of axisymmetric

~ollds with axisymmetric loads. We assume that such a solid has been discretized
11110 ring elements AXQ4 or AXQ8, which were decribed in the preceding
Nt1tion. As before, the material of the solid. is taken to be homogeneous and
IHOll'Opic.
With very few modifications , Program DYNAPS in Sec. 7.6 can be convm tcd to program DYAXSO. For example, the subprogram in DYAXSO that
H1%:rates the element stiffness matrix is practically the same as that in Program
I >YNAPS. However, when calculating stiffness terms there is multiplication by
1
11 r instead of h. Within the logic of that subprogram , the computer must
, vnluatc not only BJ.k and IJ1. k I, but also ,i. k at each numerical integration point.
'll 111ilur comments also apply to the generation of consistent masses and equiv11lt111t nodal lontls.
St rncturnl dnt n lor l'rngrum DYNAPS (s11 'l'uhlc 7.5) must be altered to
1H'l 1111111 for llw l11l I 11 11 11 !ht rnnt i,11111111 to ht 111 11 d y:1.1tl is 1111 nxisynrnwtric solid .

366

Two- and Three-Dimensional Continua

Chap. 1

Sec. 7.10

367

Program DYAXSO for Axisymmetric Solids

The structural parameters JPS and H must be deleted, and the nodal coordinat~N
x and y are replaced by r and z. The only significant changes in the dynamic

loud

data for Program DYNAPS (see Table 7 .6) are that line loads become area loatlN
and the acceleration factors GAX and GAY must be replaced by GAZ.
As in Program DYNAPS , we take n = 2 each way to locate points for
numerical integration on the quadrilateral sections. At each of the four points th~
computer evaluates the time-varying stresses a,, a,, a 8, and Trz. Once more, w~
needed to use n = 3 each way for terms in the consistent mass matrix of elemtnl
AXQ8.
Example 7.6
An axisymmetric titanium valve head is discretized using seven AXQ8 elements, 11shown in Fig. 7 .22. Acting on the lower surfaces of elements 2, 4, and 6 is an explosivt1
internal pressure p,, which is resisted by the valve seat (or restraint) at node 34. Enrh
edge of the finite-element network is divided into equal lengths between the nodes on thnt
edge. For this problem the physical parameters are
E

= 1.7 x

104 k/ in. 2

= 0.33

0.25 in.

(p, )max

= 4.20 x

z
I

~
~Stem
L

1 I 2'

l---~~~~ 6L ~~~~~~1

, I 3 _ _ _ _ _ _ _ _ _ _ _ _ -,

I
I
I

10- 1 k-s2 / in.4

I
I
I

1.53 k/ in. 2

b, = p,e~

(Ill

llcrc the symbol e ~ denotes the transpose of a unit row vector in the direction of 11 ',
which is normal to the tangential direction g' at the surface. The vector eT), may be fo11111I
by first calculating a unit row vector in the direction of g' as
1

- iI

and the units are US.

For this example we need to calculate equivalent nodal loads due to pressur, I'
on a surface of an axisymmetric solid element AXQ4 or AXQ8. Considering any of tl11
four surfaces of such an element, we can find the components of p, in the direction~ ol
r and z, as follows:

e,, = -[r,f

"

(!})

"\ I

Restra int

29

ti
2L

~~2~
7

k::::'.'.~ -/~ /

Z,f ]

I
I
I L

I ///

~30~
\ __,____ r

where
c = Y(r.d

+ (z.,,)2

(d

Second, from known orthogonality relationships, we can write the unit normal vcctrn In
the form
eT), =

- [ -z.,. r .f]

4L

l~//

~r::;i;es~:face

Then the equivalent nodal loads on surface nodes become

Figure 7.22
(ii

In this lormuln the dctrnnlnnnt of .J' is

Valve head with AXQ8 elements.

,11111

IW11 ,11111 I limn Dl1tw1111lo1111I Cot1ll111111

369

1111111111111 ""

Pi, = 2np,.

"
V)
V)

0::
a_

0, 50

0::

w
....

o.oo

so

I 00

200

sr

::. so
40

I")

w
C1 30
0

0::

i:'_j 20

ri .

z
V)

w I0

V)
V)

0:

'

V)

H.

200
r

(m s\

( b)

1r1gurc 7.23

r df

(h)

1', f

REFERENCES

(m s l

'

"'
-"
..,

11111N on these equivalent nodal loads are automatically determined by Eq. (h) for
111, ~sure in the positive sense of the normal direction "f/ '.
We processed the foregoing data with Program DYAXSO using NEN = 8,
11/\ MPR = 0.05 , and Subprogram NUMINT for responses. The computer plot in Fig.
I 'l(a) gives the variation of internal pressure acting on the valve head, and Fig. 7 .23(b)
i111ws time histories of the normal stresses SR, SZ, and ST at the integration point near
1111dl 34 . Maximum values of these stresses are 40.04, 34.54, and 50.67 ksi at time
80 ms .

w
~ I ,00

<!
z

f r1' [-z.f]
- I

11,

Valve head with internal pressure: (a) load; (b) stresses.

IJ'I =

I(),

r.e
If/

- z.e
C

11

'I'hls. ''.oh.ll'lninant tr~nsforms arc length instead of area because its second mw iN 11111
1111111:1.u l lo have unrt length. Expanding the determinant produces

IJ' I = :.[(r.d + (z.(') 2'1 = c

S11hNfil111i11g liqN. (a) , (d) , 1111d (g) into Eq. M , wt find 111 111

111

I'
II

Timoshenko, S. P. , and Goodier, J. N. , Theory of Elasticity, 3rd ed., McGraw-Hill,

New York, 1970.
Weaver, W., Jr., and Johnston, P. R., Finite Elements for Structural Analysis,
Prentice-Hall, Englewood Cliffs, N.J., 1984.
Lckhnitskii, S. G., Theory of Elasticity of an Anisotropic Body, translation from
Russian by P. Fern, Holden-Day, San Francisco, 1963.
Gere, J. M., and Weaver, W., Jr., Matrix Algebra for Engineers, 2nd ed. , BrooksCole, Monterey, Calif. , 1983.
Scarborough, J. B., Numerical Mathematical Analysis, 6th ed., Johns Hopkins
Press, Baltimore, Md. , 1966.
Zienkiewicz, 0. C., The Finite Element Method, 4th ed. , McGraw-Hill, Maidenhead, Berkshire, England, 1987.
Ergatoudis , B., Irons , B. M. , and Zienkiewicz, 0. C., "Curved Isoparametric
'Quadrilateral' Elements for Finite Element Analysis," Int. J. Solids Struct., Vol.
4, No. 1, 1968, pp. 31-42.
Melosh, R. J., "Structural Analysis of Solids," ASCE J. Struct. Div., Vol. 89, No.
ST4, 1963, pp. 205-223.
Irans, B. M., "Engineering Applications of Numerical Integration in Stiffness Methods, " A/AA J., Vol. 4, No. 11 , 1966, pp. 2035-2037.
lkgatoudis , J., Irons, B. M., and Zienkiewicz, 0 . C. , "Three-Dimensional Stress
Analysis of Arch Dams and Their Foundations,"-Proc. Symp. Arch Dams (Inst. Civ .
ling., London) , 1968, pp. 37-50.
Wi lson, E. L. , " Structural Analysis of Axisymmetric Solids," AIAA J., Vol. 3, No.
12, 1965 , pp. 2269-2274.
Snkoln ikoff, I. S. , and Redheffer, R. M., Mathematics of Physics and Modern
il'118i11eering, McGraw-Hill , New York, 1966.
Love, A. E. H., The Mathematical Theory of Elasticity, 4th ed., Cambridge UniverNily Press , Cambridge, 1927.

8
Plates
and Shells

371

I h11111111t IOI l'lnt111 111 ll11mll1111

1111, 11.~

1,2 ELEMENT FOR PLATES IN BENDING

11 IN possible to specialize an isoparametric hexahedron (see Sec. 7. 7) to become

,, plulc or a shell element by making one dimension small compared to the other
two. This type of modeling was introduced by Ahmad et al. [2] and applies to
,11111lyses of both thick and thin plates and shells. For analyses of flat plates, it
I~ 11lso necessary to restrict the other two dimensions of the modified element to
II, in a single plane. This section is devoted to the specialization of the isop111111netric hexahedron H20 to become a plate-bending quadrilateral called ele111,111 PBQ8. While an H8 hexahedron could also be specialized, the resulting
1111ight-sided quadrilateral would not be suitable for conversion to a shell ele11111nt later in the chapter.
Figure 8. l(a) shows the original H20 element, which has quadratic interpolation formulas defining its geometry . In order to understand the constraints
lll'l'ded to convert it to a plate-bending element, we first form a flat rectangular
~i,li<.I by making the natural coordinates
1/, and ( orthogonal and the (
dh11ension small. The resulting element appears in Fig. 8. l(b) as the rectangular
1111rent PQR8 of element PBQ8 before constraints. Note that groups of three
111dcs occur at the comers, while pairs of nodes are at midedge locations of
1 l'ment PQR8. By invoking certain constraints, we can convert each group and
p11'r of nodes to a single node on the middle surface, as shown in Fig. 8. l(c).
lht.: nodal displacements indicated at point i in that figure are

t,

8 .1 INTRODUCTION
When a plate is subjected to forces applied in the direction normal to itN own
plunc, it bends and is said to be in a state of flexure. For this type of prohh 111
we deal with flexural and shearing stresses and strains that are somewhat 1111111
ogous to those in a beam. However, the analysis of a plate is more compl11111 ~
because it is two-dimensional; whereas a beam is only one-dimensional.
On the other hand, a shell is three-dimensional, and its analysis is ,vitn
more difficult than that of a plate. In shells we must consider not only fltx11111I
1111d shearing stresses and strains, but also those associated with membrn11, 111r
i11 plunc) deformations.
Finite clements for dynamic analyses of plates and shells will be h1111r1I
upon those for general and axisymmetric solids from the preceding chuph 1
These specializations will automatically include the effects of shearing dl'lut
111111ions and rotary inertias , as in Mindlin's theory of plates [1] .
. Computer programs in this chapter perform dynamic analyses of pl111,11 111
bending , general shells, and axisymmetric shells. All structures that we nn11ly1
urc nssumed to be composed of linearly elastic materials with small struim, 111111
lllspluccmcnts . Guyan reduction (see Sec. 6. 7) is used in the plate and Nhrll
programs to eliminalc the nodal rotations and rl.ltain the translations.

ll/0

)-,
X

16

la)

Figure 8.1 Spcd11H1111l1111 111 ht K11hldron: (a) element H20; (b) rectangular parent
PQR8 of clllll1l'III l'll<)H 11111111 l 1t11Hl11il11tH: (c) constrained nodal displacements.

111 ,

u.7

373

I 11111111111 1111 1'1111011 111 llondlno

3

(l

.t

.t

) 'l

------ 2

h;
2

;t

~.u
lb)

21
)-,
_l
z

le)

t1

-3

h;

Figure 8.1 (cont.)

qi =

{qil, qiz,

qi3} = {wi , ()xi, Oyi}

(i = 1, 2, ... , 8)

where ()xi and Oyi are small positive rotations about the x and y axes. Relationship
between nodal displacements at a comer of element PQR8, a midedge of P()l{ H
and a node of element PBQ8 can be seen more clearly in Fig. 8.2. The two lyp
of constraints to be introduced are:

7/t

------ 5

4/j
6

the [ direction.
2. Normals to the middle surface remain straight (but no longer non11111)
during deformation.
Using these criteria, we can relate the nine nodal translations in Fig. 8.2(11) 111
the three nodal displacements in Fig. 8.2(c) by the following 9 x 3 consm1111t

Figure 8.2 Nodal displacements: (a) comer of PQR8; (b) midedge of PQR8; (c)
node of PBQ8.

Similarly, the six nodal translations in Fig. ~.2(b) ~e related to the three nodal
displacements in Fig. 8.2(c) by the constramt matnx
0

matrix:

0
0
1
0

0
0
0
0

hi
2

Cai= 0

J/'1

0
0

h,
2

()

()

hi
2

hi
2

hi
2

Cb,=

hi
2

(c)

(b)

la)

I . Nodes on the same normal to the middle surface have equal translation Ill

hi
2
0

0
0

(3)

hi
2

0
0

of size
6 x 3 If we were to apply each of these constraint matrices
in
h 1s
wIuc
1
t
lour locations , we wou ld be ublc to reduce the number of n~dal d_1sp acemen s
60 lo (8)(1)
24. Instead of following th1~ path , how(4)(9) + (4)(6)
I
iom we will pursue 11 11101t diil'l'I lorn1ulat1
,
on o f e 1ement PBQ8 m a manner
,vcr,
"imilur to that in Rd. I,
.

1
Figure 8 . I show Nthuu 111 l'IIIJH, ol t 011Nl11nt thickness h , with its neutra

l'l11t111 1111d Hh11ll1

1111. 11.2

376

I h11111111t Im l'lnl1111 In lh111cll11u

1111k pcndcntly of w and arc not related lo it by differentiation , as in a beam. In

lhlNcusc the displacement shape functions may be displayed in the matrix form

-,-

)I, V

f;

lJ

'X, U

i
/

h
2

3
6

fAi

y,

0.,

= 2., f;y;

i= I

0 lJ f;

[1

-1

(7)

0 0

f.I

= fA. + ~Y-2 fo
I

(8)

nnd

(6)

O
O OJ
0 0 0 f;

N111 fuce Iyin~ in the x-y plane. Its geometry is defined to be the same as thut I1,1
t'lll111cnt Q8 m Sec. 7.5. Thus,

= 2_, /;X;

1, 2, . . . , 8)

l'hcn

Figure 8.3 Element PBQ8.

(i

f;

l'o isolate terms in submatrix f; that multiply (h/2, we let

'-

"

t
-e
.

/!.2

(z

= 0)

(7 .5- 20)

(9)

i=I

where

fi = W +

fo)(l

= f(l -

{2)(1

f;

Clcncric

+ 7Jo)(-1 +
+ 110)

{o

+ 110)

(i = 1, 2, 3, 4)

= 5, 7)
(i = 6, 8)
(i

(7 .5- 1'))

'l'he formula in Eq. (9) will prove useful for obtaining the consistent mass matrix
nnd equivalent nodal loads for element PBQ8.
The 3 x 3 Jacobian matrix required for this element is
Xg

J =

displacements at any point off the neutral surface are

u

We assume that w,

0.,, and

= {u, v, w}

! II

Oy vary quadratically over the element, so thul

where z,,

Y,

Y.7)
0

X ,7)

z{)y

-z()x

z,,

and so on.

L f;Oyi
i= I

i= I

'l'he inverse of J becomes

-z"'
I'()xi
L, Ji

{,x

71,x

{;

~y

i= I

w = 2.,fiw;
/ I

whti:t z. (~,(~ I1~ ll~ese expressions the displacement shape functions.ti IUl' 1h,
N11111t us I hose 111 f!qs, (7 .5- 19) . Note thnl tht rotations o, und ()1, urt l'lmNrii

(10)

= h/2 and

OJ
0

J I = J* =

(11)

where (,z = 2/h.

We need cc1tnl11 dl 1ivutlv1N with respect to local coordinates, which are
mllccted into the l'ollnwlup I
I 1111111 ix:

J/6

I 1l111un ond Sholle

U,(:

v,,

h
'2,fi,l)y;

w,,
8

u,,,,

h
( 2fi.110y;

=I

v ,11 w,r,

- (1/i,/Jxl

u.,

v.,

h
2f;Oy;

w.,

377

Elomo11t l or Plates In Bonding

As with the submatrix f 1, we can isolate terms in B1 that multiply (h/2, as

f i.,w,

h
-( f; . 11 0x1
2
h
- J;ex,
2

i=l

'luc. 8.2

lollows:
f;,r,W/

0
0

0
0

BAi

Bs;

0
0
0

-f; 0
0 f;

b;

Tra~sformation of these derivatives to global coordinates is accomplished usin~

the 10verse of the Jacobian matrix, as follows:

0
0

a;

0
-b;
-a;

a;

0
b;

(17)

'l'hcn
[

: ::

U,z

:::

: ::]

J-1[:.: :.: :::]

V,z W,z

U,(

=I

i= I

V,(

BI

= BA +
l

W,C

h
h
( a10y1 -( a10x1
2
2
( 2h b10y; -( 2h b10x1
f; Oy1
-J; ()xi

h
(-BB'
2 I

(18)

h
'2BB

(19)

und

a;W;

B = BA

b;W;

( I 11

E4uation (19) will be convenient when finding the stiffness matrix for element

l'BQ8.

Stresses corresponding to the strains in Eq. (15) are

i11 which

(20)

a;= Jf,f;., + lf2fi,11

b;

= lf1fi., + lf2J;, 11

(lll

The five types of nonzero strains to be considered for element PBQ8

Ex

U,x
V,y

Ey

'Yyz

U,y + V,x
V,z + W,y

'Yzx

W,x

'Yxy

Then the stress-strain matrix for an isotropic material becomes

11111

1JI

E=
( I ~l

(1

v)(l - 211)

+ U,z

lly inspection ~f the second version of this strain vector, we can assemble 11 11
/lh purl of matnx B from terms in Eq. (13) as
0

h
(2a;

h
- (-b
2 I

h
- (2a,

h
(2b;

B,=
b,

- J,

(/1

()

,r,

(i = 1, 2, ... , 8)

Sym.

JI

1 -

JI

I - 2v

(21)

I - 2v
2(1.2)

1 - 2v
2(1.2)

l'his matrix is similar to that in Eq. (3.2-10) , but the third row and column
(1orresponding to u, and Ez) are omitted . Also, the last two diagonal terms are
111 vided by the form factor 1. 2 to account for the fact that the transverse shearing
Nlrcsses produce too little strain energy [4].
We may write the stiffness matrix for element PBQ8 as

= J' 1 J' 1

- J',J',

/',(11, 1 c''.11,,)

H 11:

HI.JI di; d11 d(

1

1:(n,,

,~ nu)I Jld{dTJdt

(22)

3/8

l1lnl111

1111d

ShollN

ChP

In this expression the malrices B11 and n11 .

array contains onJ t
h
urc hoth 01 .~lzc 5 X 24, but tht lull
Y erms t at are multiplied t rJ /2 1
.
.
through the thickness of the element gives
)Y .,, , . ntegrat10n of Ilq (

K = f , f/2BJE B11

BJE Be)JJJ dg dTJ

which must be evaluated nume . II I

.
multiplied by h/2 from the thir~~;:~f n this process the factors~ and hi(< II
hand h 3 /12. Thus th fi t
f IJ_J [see.Eq . (10)], producmg the lm1o
, e rs part o matnx K m Eq (23) d
shearing deformations, whereas the second
..
1~ ue t~ transvrr
deformations.
part is associated with Jlc1111,1I
The consistent mass matrix for element PBQ8 becomes

111

11.~I

.1 hns 1111ll y 111 pllll'l' of 1, /2; and b, in Eq. (27) now has the
111111ning of forcc pcr unit 11n11 . Note that this body force causes no equivalent
1111dnl moments.
After finding the lime-varying nodal displacements in the vector q(t), we
1 1111 evaluate stresses at any point in each element, as follows:
(29)
(J(t) = EB q(t)
111 this cxprcssion

I or best accuracy, these stresses should be calculated at the numerical integra111111 points (5].

1,1rnmple 8.1
1'111' the rectangular parent of element PBQ8 (after constraints are imposed), find the
111111;istent mass terms M11 and M22 , In the first case, Eq. (25) specializes to

=pf, f, f, fTfJJ I dg dTJ d(

=Pf

Mll = pabh

r(

i(r + Ci r + cirs) I
11

fs

In this equation the matrices f and f are b th

.
h11s only terms to be multiplied b B 'h/2 oI of s1~e 3 X 24, but the scrond
thickness produces
y
ntegratmg Eq. (24) ,through lh

2fif11

(a)

d{ dr,

M1 , = pabh
16

f1 f
-I

(1 - t)2(1 - r,)2(-1 -

g-

r,)2 d{ dr,

(b)

- 1

1'111forming the integrations indicated in Eq, (b) results in

M11 = ipabh

_ fl JI (

M - P _

LL /T

111hstituting the shape function/1 from Eqs. (7.5-19) into Eq. (a) yields

JI dg d,,, d(

11

379

1'11111111111l>YN/\1'111111 l'l11h1 111 llm1cll11u

h2 fJfe)
+6

JJ / dg dTJ

(c)

which is simply a fraction of the total mass. Similarly, in the second case we have

Again, the factors 2 and h2/6 are multi lied b / 2

part of matrix M consists of transl f p . Y,h from IJI. Hence, the hnl
1
rotational (or rotary) inertias.
a wna mertias, and the second part givt1
Equivalent nodal loads due to bod fo
using only matrix f as follow ,
y rces on element PBQ8 are calcu l11ll'd

M22

fl fl /T

pab h3 _
12 1

d{ dr,

1
= -pabh 3

96

(d)

S.

A,

which has units of mass moment of inertia.

Pb(t)

=f
=

2 f1

f l fJb(t) I JI dg dTJ d(

11,3 PROGRAM DYNAPB FOR PLATES IN BENDING

rJb(t)JJI dgd,,,

in which

b(t)

= {O, 0, b,}

( 171

und b: is force per unit volume in the z dir .

,
lhc factor h/2 from IJI and rewnt E (e2c6t)1~n . Alternatively' we may cxll m I
e q.
m the form
P&(t)

= {, {,

fJb(t)/JI

dgd,,,

( 'HI

Wc shall now describe a computer program named DYNAPB for the dynamic
111111lysis of plates in bending, which uses element PBQ8 from the preceding
~t'l'I ion. This program is constructed by modifying the part of Program D YNAPS
p111t11ining to element Q8 (see Sec. 7 .6), because the geometric and displacement
11h11pe functions are the same for both. However, in program DYNAPB the
11111trices K, M , and p,lt) must he handled according to the expressions devel11ptd in Sec. 8.2.
Table 8. I shows pH' p11111ti1111 of sln1ctural data for Program DYNAPB.
I '11111paring this d11t11 wllh 1l1111 l11 '1'11hll' I 'I, Wl' sec that the structural parameters

380

l'htht n11el Hhnll

TAUi.iC H. 1 NII 111'1 111111 1111111 for
Type of Data

1'1014111111

Problem identification
Structural parameters

Items on Datu Lines

Plate
(a)
(b)
(c)
a

bending data
Nodal coordinates
Element information
Nodal restraints

l
1

Descriptive title
NN,NE , NRN,E,PR,RHO, H

NN

J, X(J), Y(J)
I, IN(I, 1), IN(I, 2), .. . , IN(! , 8)
J, NRL(3J-2), NRL(3J-l), NRL(3J)

NE
NRN

!PS
andd NfEN are omitted. In addition, there are three possible nodal restnunt
mstea o two per node.

Dynamic load data for Program DYNAPB is given in Table 8 2 1-1, .

?ave three possible initial displacements, initial velocities and ~o-datt ~
mstedad of/wo per node (as in Table 7.6). Line loads act in th~ z direction u~:1,,: I
an e ge o an element and ma~ vary_ ~uadratically. Therefore, we require lh1 t1
;~de numbers and three for~e mtens1hes (per unit length) to describe them IN'1
g ..7. l !(b)J. Also, ~uadratically varying area loads (force per unit area rl
111
z direction) need eight force intensities for a complete de~criptio~ N~
1

TABLE 8.2 Dynamic Load Data for Program DYNAPB

Initial conditions
(a) Condition parameters
(b) Displacements
(c) Velocities
Applied actions
(a) Load parameters
(b) Nodal loads
(c) Line loads
(d) Area loads
Ground accelerations
(a) Acceleration parameter
(b) Acceleration factor

381

fllgu rc 8 .4 shows half of a square, symmetric, simply supported plate that is divided into
,,lght PBQ8 elements. Also indicated in the figure is a moving load P that travels in the
I' direction along the centerline, where nodes 5, 8, 13 , ... , 37 are located. We wish
lo determine the translational responses of the plate at node 21 due to the load moving
111 constant velocity and constant acceleration [6].
This problem is analogous to Example 6.5 for a moving load on a simplyNllpported beam divided into four flexural elements. However, the displacement shape
l11nctions for the PBQ8 plate element are quadratic instead of cubic. Therefore, the
Junctions given by Eqs. (7.5-19) must be used in Eq. (a) of Example 6.5, and y,(t)
11Jplaces x,(t) in both Eqs. (a) and (b). Of course , we need to extend Program DYNAPB
to handle one or more moving loads, as described previously for Program DYNACB.
Physical parameters in this example are
E = 69

106 kN/ m 2

a = 0.1 m

p = 2.62 Mg/ m 3

= 0.33

h = 0.025 m

P = 20 kN

lor which the material is aluminum and the units are SI. We ran this data on the extended
j3

No. of
Lines

Dynamic parameters

1110111 11111 DYNAl'II 1111 l'lnt n1 111 lln111ll1111

11.xumple 8.2

For sequence of node numbers, see Fig. 8.3.

Type of Data

11.:1

PIH, 7. I6(b) J. Finully, nolr th11l ll1l' llt'l'l' lrrution factor GAZ is for the z direction
1111ly,
As in Program DYNAPS, we lake n = 2 each way to locate points for
1111111crical integration. With a plate-bending element, the computer evaluates the
rl111c-varying stresses a:,, O'y , Txy , ry,, and T,x at each of the four integration points.
lllcause the displacement shape functions are quadratic, we found it necessary
111 use n = 3 each way for terms in the consistent-mass matrix of element PBQ8.

l>YNAPII

No. nl
LillCN

1111,

Items on Data Lines

I - j2

ISOLVE, NTS, DT, DAMPR

j1

NNID
NNIV

NNID, NNIV
J, D0(3J-2), D0(3J-l), D0(3J)
J, V0(3J-2), V0(3J-1) , V0(3J)

r------------7---v
/

Sym .

1
NLN

NEL
NEA

NLN,NEL,NEA
J, AS(3J-2) , AS(3J- l) , AS(3J)
J , K, L, BL!, BL2, BL3
I , BA! , BA2, ... , BAB

I, //

_L~ I/

y(t)--413

Typ.
21

j_

//

IGA
GAZ

Forcing fun ction

(a) Function parameter
I
NFO
(b) Function ordinates
NFO
_ _ _ __ _ _ ___J__ _ __J__ K, T(K), FO(K)

-------=-

Typ.

11'11111111 tl.4

Simply-supported plate with moving load .

382

Platos and Shells

Chap.

Sec. 8.4

Element for General Shells

version of Program DYNAPB , using Subprogram NORMOD to obtain responses. Coni

puter plots of Di, at node 21 arc given in Fig. 8.5. For the case of constant veloci ty
(VOP = 153.7 mis), the plot shows a maximum translation of 2.320 mm; and for
constant acceleration (AOP = 59.05 x 10 3 m/s2 , with zero initial velocity), we have 11
maximum of 2.034 mm. Their ratios to the static deflection of 1.472 mm (due to the loud
applied gradually at node 21) are 1.576 and l.382, respectively. As for the beam, lh1
values of VOP and AOP used in this example both give travel times equal to llw
fundamental period of the plate, which is 5.206 ms.

383

1 0 0
0 1 0
0

0
0
0

1 0

Ca;=

h;
2

YOP CON Sr.

0
0
0

hi
2
0

1 0

h;
2

0
AOP CONST.

h;
2

2,00

E
E

(1)

I .SO

When this 9 x 5 matrix is compared with Eq. (8.2-2), it is seen that columns
I and 2 have been added. Similarly, the constraint matrix Chi for a midedge node
of the rectangular parent becomes

( \J

Cl
0

.._ I . 00

1 0 0

<

-,
Cl

0
o.so

Cb;=
o.oo
3

5
<ms J

Figure 8.5 Translational responses at center of plate.

8.4 ELEMENT FOR GENERAL SHELLS

In this section we specialize the isoparametric hexahedron H20 to becorn1: 11

curved quadrilateral element for the analysis of general shells. Development of
lhc slw/1 element SHQ8 is similar to the technique used in obtaining c lcnw11l
Pl3Q8 (sec Sec. 8.2) for plate bending. However, the constraint conditions 1111
111odilied because two additional translations, u1 and v1, occur at each node of 1111
Nlwll eh:menl. Thus, the constraint matrix C111 for a corner node of 1h e rectn11g111i11
fllll'l'III clL111cn1 Isc-e Fig . R. I (b) l has two more 1:olumns thun before, us f'olfowi,

h;
2

0
0

1 0

h;
2

h;
2
0
0

h;

(2)

2
0
0

which is a 6 x 5 array that can be compared with Eq. (8.2-3). With five
displacements at each of eight nodes, element SHQ8 has (8)(5) = 40 nodal
displacements.
.
As with the plate element PBQ8, the general shell el~ment SHQ8 will be
formulated directly (2, 3]. Figure 8.6(a) shows the geometnc layout of element
SIIQ8, in which the coordinates of any point are

h [ l
L !, IX;l
Y1 + L fd-:j_
g

I I

%1

;]
m3;

I I

n 3;

(3)

'l'ht i11lll1pol11lio11 '1111Plt111111 / 1 11p111111i11g In Eq . (3) nrc given by Eqs. (7.5-19).

3114

' 110,

11.4

3811

I ltt1111rnt lor lln11n111I hnll

111111,, w1.; sec th ul thl th1r~t1l'HN I, 11111y vmy quadratically over the element. In
1uldltion , th1.; terms /11, 111 11 , und 1111 11n the direction cosines of a vector V3; that
IN normal to the middle surl'uct.l und spuns the thickness h; of the shell at node
I Figure 8.6(b) shows this vt:ctor, which is obtained as
(4)

z',{;

/,W

,.,,,

\i' ,,/

)- y,v

, ./

t1oints j and k in the figure are at the surfaces of the shell. In a computer program
1lther the coordinates of points j and k or the direction cosines for V3; must be
Hivcn as data.
Generic displacements at any point in the shell element are taken to be in
the directions of global axes. Thus,

/~v,
(a)

u,

u = {u, v, w}

C<,~

...

(5)

l )n the other hand , nodal displacements consist of these same translations (in
Hlobal directions) as well as two small rotations a ; and /3; about two local

tnngential axes x' and y', as indicated in Fig. 8.6(a). Hence,

q;

z',~

= {u;,

v;,

W;,

a;, f3;}

= 1, 2, . .. , 8)

(i

(6)

<lcneric displacements in terms of nodal displacements are

[
v'

UJ = I8f; [U;J + I8f; 'ih fJ.; [a]

V

V;

t=l

W;

(7)

/3,

, =I

In this formula the symbol ; denotes the following matrix:

fJ.;

-/2;

Iii

-m2;

mli

-n2;

nli

(8)

( 'olumn 1 in this array contains negative values of the direction cosines of the
Hccond tangential vector V 2;; and column 2 has the direction cosines for the first
tangential vector Vii, as shown in Fig. 8.6(b). These vectors are orthogonal to
the vector V3; and to each other, but the choice for the direction of one of them
Is arbitrary. To settle the choice, we let

(9)
Figure 8.6

(a) Element SHQ8; (b) nodal vectors.

'l'hcn
(10)

Ill' V II is parallel lot.\ , 111 l(q . (I)), thl' lntlCI' is replaced bye,.] Figure 8 .6(b) shows
locul generi c trunslnt11111H 11 1 111111 ,,, (i 11 llll' directions of V11 and V2 1) due to the
11od11I rotations /1 11111d ,~,. 11M1111 tl Vl' ly Tlwlr vulucs arc

386

Plates and Shells

v'

Sec. 8.4

1 0
f;

and so on.

/!!z
2 11

h;

h;

-(2 m2; ( m1;

2
h;
-(2 n2; (2h; nli

f;

(i

= 1,

2, ... , 8)

(12)

The inverse of J becomes

11nd
C,;

~ G! f f f}

~G

0 0
0 0
0 0

Then

-l2;
1, l h
-m2; mli
-;j_J;
-n2; nu

J* =

f y TJ.y

V,g

v,,,
v.,

(13a)

=I
i= l

fi.s

Ji.,,

0
0
0

fi.,

w.,
w,11

w.c

fi .,,
0

0
0
0
0
0
0

- C/;,gl2;
- (f;,,,l2;
-f;l2;
-(f;,gm2;
-(f;J,f1 m2I

0
0

Ji.,
Ji.1)
O O

f;

= fAi +

(fs;

(14)

O
0

U, y

= fA +

(fs

J=
We find the derivatives in matrix
8

x., = I I
I

x,s

~
I I

Y.c

( I<, l

Ji,,, 1,

... -4

I I

-f;m2;
-Cfi.,n2;
-(fi.,,n2;

(Ji.,n1;
(fi. ,,n1;

-}in2;

Jinli

~e:f

:' I

J*
0

...

ore.

W;

(18)

h;
-a
2 I
~{3;

0
J*

,'1

(19)

11

W ,(

U ,x

a;

U,y

b;

U,z

C;

V,x

0
0 0
a; 0
b; 0
c, 0
0 (11
()
/11
()
t'1

V,y

v .

~'

V;

Multiplying the terms in this equation, we obtain

J from Eq. (3), as follows:

8
h,
fi.,x, + I !i.d2'"

\,,,

z,,,
z. 1

U;

(f;,,mli
Cfi.11mu
fim1;

W ,z

z,s]

Y.t
x,,, Y.,,

[x,

. . .

( 15)

The last of these formulas will later be used to derive the consistent mass matrix
und equivalent nodal loads for element SHQ8.
The 3 x 3 Jacobian matrix required in this formulation is

Cfi.iu
CJ;,,,l1;
f;lu

Transformation of these derivatives to global coordinates requires that the inverse of the Jacobian mla:~x be aprp;~ed

und

(17)

t.y

[ f , 'T/,z (.,

We need certain derivatives of the generic displacements [see Eq. (7)] with
respect to local coordinates. These derivatives are listed in a column vector of
nine terms, as follows:

u.,
u,,,
u.,

(13b)

t,x 'T/,x {x]

In order to isolate terms in submatrix f; multiplying (, we let

f,,

387

h
-(~a;

Contributions of these terms to the generic displacements at any point are given
by the second summation in Eq. (7).
As for the plate element, the displacement shape functions in Eq. (7) may
be cast into the matrix form

h
-(-;j_l2;

Element for General Shells

,,,
I, ,, (. J I ,,
111

whldl

=I

l l

0
0

W,1

()

IV,1,

()

II', ,

()

. -d;l2;

d;lu

-e;l2;
e;l1;
-g;l2; g;lli
-d;m2; d;m1;
- e;m2; e;m1;
-g;m2; g;mt;
- d1n21

d1n1;

('1 1121

<11n11
{<11111

R1 ll i1

U;

V;
W;

a;

{3;

(20)

388

Plates and Shells

Sec. 8.4

(21)

Element for General Shells

As in a plate element, the normal stress a,, and the strain '=z' have been omitted.
Then the stress-strain matrix E for an isotropic material becomes the same as that
for element PBQ8 in Eq. (8.2-21).
To relate local strains in the vector E' to global strains in the vector E , we
can use the 6 x 6 strain transformation matrix T, in Eq. (7.2-28), as follows:
E'

U,x

V,y

'Yxy

'Yzx

(28)

+ V,x
V,z + W ,y
W ,x + U,z

(22)

Noti?g the second ver.sion of this strain vector, we may construct the ith part ol
matnx B from tenns m Eq. (20) as

B,=

a,
0
0
b,
0
C;

0 0
b, 0
0 c,
a, 0
c, b,
0 a,
(i

-d,121
-e1m 2;

d;lli
e1mli
-g,n21
g,nli
-e,L2, - d,m2; e,lli + d,m1,
- g1m2, - e1n21 g,m1; + e,nli
-d,n2, - g,12; d,n1, + g,Li,

l, 2, .

BA;

Matrix B' will contain only five rows, due to the deletion of the third row of T,.
Now we are ready to formulate the stiffness matrix for element SHQ8
using matrix B', as follows:

L, f f
= f f f,
=

(24)

Submatriccs B111 and Bn; are composed from Eqs. (21) and (23), but the actuul
dc111ils are omitted. Altogether, we have
(2~)

which will be useful when determining the stiffness matrix for the shell ch.lllll' III
The following nonzero stresses in the directions of primed axes will ho
l'ons idcrcd:

1111d

(29)

'8)

+ (Ba;

In these expressions the vector (J 1)norm. denotes the first row of the Jacobian
matrix normalized to unit length, and so on.
When calculating stresses in local directions, it is also useful to have

K
(2))

.Similar to the plate element, we can isolate terms in submatrix B, that multiply
( lo find

B;

(7.2-27)

w,,
U,y

'Yyz

= T.-E

However, the third row of matrix T, must be deleted, because E, is not to be

included in vector E'. For the purpose of evaluating T, at an integration point,
we need the direction cosines for vectors Vi, V2 , and V:i at the point. This may
be done with the following sequence of calculations:

For element SHQ8 we consider six types of nonzero strains, as follows:

Ex

389

(B')TE B' df dY] d(

(B~ + tBs?E(B~ + tBe)

1JI df dY] dt

I Iere the matrices B ~ and Be are both of size 5 x 40, but the latter array contains
only terms that are to be multiplied by (. Integration of Eq. (30) through the
rhickness* of the element leads to

K={,

[2(B~?E B~ + HBe?E BeJ

1

IJI df dY]

(31 )

The remaining integrals in Eq . (31) must be evaluated numerically, using two

integration points in each of the gand YJ directions [5] . In this process the factors
) ond } are multiplied by h;/2 from the third row of IJI, and Be also contains
rite same constant. Thus, we effectively obtain the factors h, and hl / 12 in the two
purls of matrix K.
The consistent mass matrix for element SHQ8 is

(32)

llw r01 1csponding slruins nrc

lt1,,

C30)

t\,

Y,,, Y, , Yl','I

111-11111ll l y 1111q1111ll11111l11111111h 1hr lllll ~111 "' tr1111~ 11111111111,

J 1111111111111111, 111

11111ln 1rd

JOO

I lt,11 a 1111d Sholl H

1t11i. U.b

Rcc:ull lhut the matrices f11 and fJJ arc bolh of r,1111 I
10, but the scc:on<.l hus 11111
lcnns to be multiplied by (. Integrating Eq . ( ll) th rnugh the thickness y1tld

=pfIfI[2flf11 + HEruJ /JI dt d71

TABLE 8.3 Strncturul l)utu for Program DYNASH

=
=

fff
1

ff
1

( I ~)

rib(t)

/JI dtd71

( II)

( I'\ )

As for the plate element, these body forces do not cause any equiyalent 111ul11I
111omcnts.
After the time-varying nodal displacements in the vector q(t) have lwrn
obtuincd, accompanying stresses in the element may be calculated for Im nl
(primed) directions . That is,
(t)

=E

General shell data

(a) Nodal coordinates; h1
(b) Nonna] vectors
(c) Element infonnation 3
(d) Nodal restraints

X(J), Y(J), Z(l) , H(J)

V3X(1) , V3Y(J), V3Z(J)
IN(I , I), IN(l, 2), ... , IN(J, 8)
NRL(SJ-4) , NRL(5J-3), ... , NRL(5J)

T,B q(t)

= E B'q(t)

TABLE 8.4 Dynamic Load Data for Program DYNASH

,,,,

No. of
Lines

Type of Data

Dynamic parameters

8.6 PROGRAM DVNASH FOR GENERAL SHELLS

111 Sec . 8. 4 we developed element SHQ8 for the dynamic analysis of ge1w111I
shells. Now we present a computer program called DYNASH, which is b11N1 ii
011 lhu~ clement. The easiest way to compose this program is to modify sta111
IIH.nts in Program DYNASO relating to element H20 (see Sec. 7.8) , from whh h
ulcment SHQ8 is derived. But in Program DYNASH, we must use the forn111l11given in Sec. 8.4 to construct the matrices K , M , and Pb(t).
Preparation of structural data for Program DYNASH appears in Tabfo H I
C'ompuring this data with that in Table 7. 9, we observe thatNEN is omiUcd 1111111
till' list of structural parameters. However, the symbol H is added to the list f111
11 r use where the thickness is constant over the whole shell. We also sec th11t 11( 1t
Is included with the nodal coordinates for 11 case where the thickness vuries ov, 1
llll l lt111l nl. The lint' of normal vectors in tlw f11hll t'onf nins u node 1111111lwr .I 1111d
1

J,
J,
I,
J,

NN
NN
NE
NRN

1hree orthogonal components of the vector Vj at the node [see Eq. (8.4-4)]. In
11ddition, five possible types of nodal restraints are given in line (d).
Table 8.4 lists the dynamic load data required for Program DYNASH,
which is similar to that in Table 7 .10 for Program DYNASO. However, now we
h11ve five possible initial displacements, initial velocities, and nodal loads inM
lcad of three per node. As for element H20, we may have a quadratically
vnrying line load on an edge of the shell element, with components in _the x, y,
11nd z directions. For this purpose, we need three node numbers and nme force
lnlensities (per unit length). Also, the area loads require 24 force intensities (per

~uch stresses should be determined at the sampling points for numerical intc~111
11011.

Descriptive title
NN,NE,NRN,E,PR, RHO,H

"For sequence of node numbers, see Fig. 8.6(a).

b(t) = {bx, by, b,}

0'

Items on Data Lines

I
I

1iroblcm identification
Structural parameters

rib (t) /J / dt d71 d(

In this expression, the load vector b(t) is assumed to contain components of 101,
(per unit volume) that are uniform through the thickness of the shell. ThuN,

No. of
Lines

T ype of Data

Because the factors 2 and j effectively become h; and hr /12, the first and sc1:111ul
parts of matrix M contain translational and rotational inertias, respectively
Equivalent nodal loads due to body forces on element SHQ8 may be fc,111111
using only matrix f 11 , as follows:

Pb(t)

391

1'1 011111111 DYNASH f<>t Clo111111tl llhollu

NNID
NNIV

1, D0(51-4) , D0(5J-3) , .. . , D0(51)

1, V0(5J-4) , VO(SJ-3) , ... , V0(5J)

Applied actions
(a) Load parameters
(b) Nodal loads
(c) Line loads
(d) Area loads
(e) Volume loads

I
NLN
NEL
NEA
NEV

NLN,NEL, NEA,NEV
J, AS(SJ-4), AS(51-3), ... , AS(5J)
J, K, L, BL1 , BL2, .. . , BL9
I , BAI , BA2, ... , BA24
I, BVI , BV2, BV3

Forcing function
(u) Jiu netIon p111111111t111
(b) F11mtl1111 rn.11111111

'=

!SOLVE, NTS, DT, DAMPR

Initial conditions
(a) Condition parameters
(b) Displacements
(c) Velocities

Ground accelerations
(a) Acceleration parameter
(b) Acceleration factors

Items on Data Lines

Nl10
~

NNID, NNIV

IGA
GAX, GAY ,GAZ
NFO
K , T(K), FO(K)

fllutoe end Shell s

392

unit area) to describe quadratically var ying components in the x, y, and

directions.
For element SHQ8 we use n = 2 in the gand 7J directions [see Fig. 8. 6(111 1
to locate points for numerical integration. At each of these four points tho
computer evaluates the time-varying stresses, a x', a y' , T.e y', Ty'z', and r ,'x' in lol'11I
directions. As for element PBQ8, we needed to use n = 3 each way for tc:1111
in the consistent-mass matrix of element SHQ8 to retain sufficient accuracy .

where the material is reinforced concrete and the units are US . We used this data. in
Program DYNASH with DAMPR = 0.05 and solution by ~ubprogram NUM~NT . Figure 8 .8(a) shows a computer plot of the ground acceleration, and the res_ultmg translntions in thex and z directions at node 5 are given in Fig. 8.8(b)_. The ~ax1mum values
of these displacements are DJ1 = 0. 82 l l in. and DJ3 = 1.430 m; at tlm~ t . = 7_2 0 ms
Also shown in Fig. 8.8(c) are plots of the flexural stresses in the x and y directions (at
~

100

Example 8.3
A quarter of a doubly symmetric cylindrical roof shell appears in Fig. 8. 7. This port11111
of the shell is divided into four SHQ8 elements of constant thickness. Note that the shlII
is symmetric with respect to the x-z and y-z planes. Consequently, nodal restraints 111111
be used to prevent translations across those planes and rotations in the planes, 11
indicated at nodes 4 and 14. On the other hand, simple supports at the ends of the ~hl'II
prevent translations in the x and z directions, as at node 20. A rigid-body grouml
acceleration D8 J(t) occurs in the z direction, and we wish to find the response ol tht
structure due to this influence.
For this problem the physical parameters are
E

393

Program DYNASH for General Shells

Sec. 8.5

= 3.6 x 103 k/in. 2

= 0.15

L = 100 in.

R = 3L

= 2.25 x

50

1000

1200

(.)

(ms)

<
-50
0::

c..,

,!_,

-100

(a)

10- 1 k-s 2/ in. 4

<I>=

h = 3 in.

'"

OJ3

40
C

(D8 J )max = 115.9 in./s

:::; 1

.oo

If)

Cl
0

Sym .

1.5L~::.L/

s/ 14u 17/
t----=

Sym.

G)

I
I

:/ 0 I

---71
// I

o.oo

,-

<

<fl
z
<

- 1 . 00

0::
I-

( b)

sx .

/I
// I

IVp

20

r.k"

... ._./_ _s____

l ____,s_ _ _2_ 1, /

_:; 0. 60
-

0-40

Cl O. 20

0::

o.oo

<

Typ.

W -0.20

~ -0.40
(/)

',

',

'-J

/ t ',
1

.~

R'--('''' I
'
I
~

o
....
(/)

0,60 -

(r l

1,111111,1, H.H ( 'ylli111t l111I lll'll 111111 (11) p11111111l 11n rk1111h111, (h) dlNpl11n1m111s, (,)
tk,111111 ~, 111~1 ~

394
Plates and Shells

Chap. 8

the upper surface) for the integration point near node 1. Their minimum values are SX' ::.
- 0.6350 ksi and SY' = -0.1456 ksi at time t = 720 ms. Of course, these stresses must
be added to the membrane stresses occurring at the same time.

Sec. 8 _6

395

Element for Axisymmetric Shells

z

8.6 ELEMENT FOR AXISVMMETRIC SHELLS

3

As shown by Ahmad et al. [7], it is possible to specialize a ring element with

an isoparametric cross section to become an axisymmetric shell element by
making one dimension small compared to the other. In this section we demon
Strate the procedure by specializing element AXQ8 from Sec. 7.9 to form a shell
clement called AXSH3.
Figure 8.9(a) shows the axisymmetric solid element AXQ8, for which thl
cross section is an isoparametric quadrilateral with eight nodes. As the first step
in the process, we make axes [ and 17 orthogonal and reduce the 1J dimension
to the thickness h. Thus, we form the rectangular parent AXSR3 of elem,111
AXSH3 (before constraints), as shown in Fig. 8.9(b). Next, we may introdut'l'
constraints to refer the displacements at each group and pair of nodes to thosl'
of u single node on the middle surface, as depicted in Fig. 8.9(c). The noclul
displacements indicated at point i in that figure are
'

(i

= 1,

---+- U ;

~l~______i___e

(a)

v',

'Tl

2, 3)

(I)
where a a small positive rotation about an axis normal to the t-11 plane. Figurl'
8. I0(a), (b), and (c) show relationships between nodal displacements at an end
of clement AXSR3, the middle of AXSR3, and a node of element AXSll.1,
respectively. The two types of constraints to be invoked are:
1 is

I. Nodes on the same normal to the middle surface have equal translations in
the 1J direction.
2. Normals to the middle surface remain straight (but no longer normul)
during deformation.

x',
(c)

(b)

Figure 8.9 Specialization of axisymmetric solid: (a) e!ement AXQ8; ~b) rect~n
gular parent AXSR3 of element AXSH3 before constramts; (c) constramed no a
displacements.

With these criteria we can relate the six nodal translations in Fig. 8. IO(a) to 1hr
lhrce nodal displacements in Fig. 8. lO(c) by the following 6 x 3 constraint

mutri x:

1 0

0
0

()

2
0

h1

C,,1

8 imilarly, the four nodal translations in Fig. 8: IO(b) a~e related to the three nodal
displacements in Fig. 8. IO(c) by the constramt matnx

( ))

"'

()

()

which is ol si:11 I

\, 11 wt w111

111

h;
2
0
h;
2

(3)

()

11pply Pq , ( >) 111 tht 111d~

1111<1

I\(( , (1) tit

.100

l'h,11111 n11d Uhnlln

1111,

3-r

- ,

------

,1~1

r,

397

---......

"",\

------

. 1 ..

/
3

h,

.,,~ '

l 1011111111 for Axl11y1tm1011 lo Shull

1, V

h,

y'

11.e

')

_J_

t
6

z'

-3

t
4
(a)

(b)

(C)

Figure 8 .10 Nodal displacements: (a) end of AXSR3; (b) middle of AXSR3; (c)
nocle of AXSH3.
(a)

the middle, we could reduce the number of nodal displacements from 16 to 1J

I lowever, we will take a more direct approach, which is similar to those in S11
8.2 and 8 .4 for plates and general shell elements.
Figure 8.11 shows element AXSH3, for which the coordinates of any po1111
may be stated as

[r]

[r;] +

f;

fiTJ~ [ l2;
2 m 2;
In this equation the direction cosines for the normal vector
z

zi

i= l

l2; =

cos 'Yi

y',ri,v'

t I)

i=I

m 2;

h;
2 ,

T/ -c, .

Vi are

h; .

2 sin Y;

= sin Yi

where y, is the angle between the r axis and the normal at node i. The geon 11111 h
interpolation functions in Eq. (4) have the formulas

!, = --(] 2

t)

= I

-g2

= {u , v}

(b )

(h I

11ssu111ing that the loads arc axisymmetric. These di1,placements cnn he exp11ss1cl
In tl1111s of the nodal displaccmcnls 111, i,1, 1111<1
us fo llowi.;:

,r,,

x',1;,u'

z'

'l'hercf'orc, the th ickness h may vary quadratically in the g direction.

Generic displacements at any point in the element are

11111111 R.11

(a) Element AXSH3; (b) nodal rotation.

308

Plates and Shells

[ U]

i=I

j; [Ui]

fiYJ!!!.
2

i=l

V;

r-

Sec. 8.6

,Y;J

Element for Axisymmetric Shells

399

These derivatives are transformed to global coordinates by the operation

sin
ai
COS 'Yi

(7)

Geometric details justifying these expressions appear in Fig. 8.1 l(b) . As for lh<'
~cneral shell ~Jement, we arrange the displacement shape functions from Eq. (7)
mlo the matnx format

fi

[1 0
0

h,

""i

- YJ sin 'Yi]
hf;
YJi cos 'Y;

= [oJ

01 00]/;

fsi

let us define

[00 00 -sin
'Yi]!!!./;
cos 'Yi 2

- r::

(H)

0
0

i=l

YJ,

To isolate terms in submatrix fi that multiply

fAi

(i = 1, 2, 3)

0
0

-d,s'
.ny,]
- ei

ai
bi

d; cos 'Yi
ei cos ,'i

Sill

'Yi

n
Vi
ai

(15)

where
(I})

(16)

Then
( I ti)

und

We consider four types of nonzero strains for element AXSH3. They are

u.,
v,,
u
r

( 11)

The _formula in_ Eq. (11) will be used later to determine the consistent nw11N
u1ulrix and equivalent nodal loads for element AXSH3.
The Jacobian matrix required for this element is

e=

=
9

u,, + v,,

'Yrz

= [;,:

:::]

( I lj

(17)

lJsing the second version of this strain vector, we form the ith part of matrix B
If'Om terms in Eq. (15) as

in which

ai 0
0 bi

and so on.
Bi=

The inverse of J becomes

(, YJ.,]

[ g_,

YJ. ,

We w_ill need derivatives of the generic displacements with respecl to 1h

ut

"'

11, ,,

L
I I

- drf; T/hi. sin y;

b;

ai

d; cos 'Yi - ei sin 'Yi

( I I)

lorn! coordmates, as follows:

u,,,

I!
r

1- 1 = J* =

!,,,
0
[ 0
()

-Ji.1;'YJ sin

-Ji sin y1
Ji.,'Y] cos y,
.fi

CON ,'1

(18)

(i = 1, 2, 3)

Sl111ilar to the general shell element, we isolate terms in submatrix Bi multiplying

1/, which gives

B;

y,

( I I)

-di sin 'Yi

ei cos 'Yi

= BAI +

'l7B0;

(19)

r,n,,

(20)

llor the whole matrix, we have

n nA
1lw 111st

10111111111

wlll lulp IIN to d111v1 tht htlll11nHM 11111td x

101 11l 111t11t

AXSI I I,

400

11111111, " '"'

m1011,.

,11 11.0

I 101110111 lur A><ly11111111tr lu 8hulle

401

Stresses in till d itl'l'flo11 111 ,,I p1 l1111d 11x1 i, (rnused by ux isy111111etrlc loud'9 1

Iti I.

er'
and the corresponding strains

r,

11

I' T, ')'}

fll'l'

e' = {e.,., e,, , 'Y.,y,}

II

II

(1

v)(l - 2v)

I - 2v
2(1 .2)

0
I

21, 12

2m1m2

(7. 2- J I l

e2 = {l2, m2} = {- m1, / }

1

( ' ~)

As for the general shell element, it is also convenient to have

= T.B

mar, I, n

Bn arc both of size 3

21rp

T]fa) /J /rd( d77

(28)

fyn

fA

HHa)/ J / rd(

(29)

We find equivalent nodal loads caused by body forces using only matrix
11 Thus,

Pb(t)

{,

frr fJb(t) / J / r dO d{ d71

fJb(t)/ J /rd{

= 41r

b (t)

(30)

9, buf flw 1 11

,
11 1 1

,,,,

fin ~

= {b,,

b,}

(31)

I hcsc body forces do not produce any equivalent nodal moments.

After solving for time-varying nodal displacements in the vector q(t} , we
1 101 determine stresses for local directions in the element using

u' (t)

r dO dg d71

111 this equation the matrices B,; a nd

T]fa?( fA

Nlllts in

(8.4 2HI

J he clement stiffness matrix may now be formulated using matrix D 1 'f'huI

1111d the components of force per unit volume (constant through the thickness) are

fo_r fhe purpose of calculating stresses in local directions . In this case,

wi ll h;~ve only three rows, due to the form of T. in Eq . (24).

fTf/J/rdOd{d77

.\N before, the first and second parts of M consist of translational and rotational
l1111rtias .

'fo cv~luate matrix T. at an integration point, we must find the direction coNI II
fo r unit vectors e, and e2 in the directions of x' and y' at the point. This
yI
11111
done as follows:

f f frr(B '?EB'/ J /

1,

E' = Te

B'

(27)

111 thi s case the matrices fA and f 8 are both of size 2 x 9 , but the second has only
h I ms that are to be multiplied by 71. Integrating Eq. (28) through the thickness

(4

which is of size 3 X 4 . Then the strain relationship becomes

e, = (J,)norm = {l1, m,}

_,

= 21rp {, { / fA

11

IT

pcB,:.?E B,:. + i(Bli?E Be] /J /r dg

r r frr

In or~e~ to relate local strains in the vector e' to global strains in the v1, 1t
we specialize the 6 x 6 strain transformation matrix T. in Eq. (7.2- 28) 1 t

M = p _
II

T =

= 21r

1lw remaining integral must be evaluated numerically, using two integration

p11l11ts in the gdirection [5].
The consiste nt mass matrix for element AXSH3 has the form

E=

K=

The stress <ry and th~ strain. Ey normal to the middle surface have been
1111 I
0111
Thus, the stress-strain matrix for an isotropic material has the form

E,

1111 ly te rms lo be mullipli(.)d by TJ. Integration of Eq. (26) through the thickness*
111 1hr clement g ives

E T. B q (t) = E B 'q (t)

(8.4-36)

l'h1sc stresses sho uld be calculated at the numerical integration points.

In Sec. 7. 9 we expressed nonaxisymmetric loads that are symmetric with
11 ~PL'Cl to the r-z pla ne as the Fourier componen ts:
~To Ni mpllfy lntl~f.1111lio11 thro1111h thl' lhilk,wss, 11,1111s In r 1111(1 rn11trl x .I contuining 1/ urc
11111hl'l1d ,

402

Plates and Shells

m

br =

brj

COS

Chap, I

j()

hz =

j=O

bo =

[cos J()
O
1 0
0
3

~.,t;

0
COS

o o o
(7.9-12)

b 8j sin J()

[u;]
V;

W;

1- l

0
0
sin }8

j()

L cl qj
j=O

(11,

L cj L f;Qu
j=O i= I

f, = 0
0

0
0

-J.,t;sj

f;,sCj
0

v ,s

=I
i= l

w,s

w,11

0
0

W,e

-f;,s17S;Cj
-.,t;s,.cj

0
0
0

jf;77s;Sj

-J.,t;sj 0
0
f; ' sSJ
0
0
0
jf;cj

s,.

= sin

y,.

sj
Cj

J=

772

0
0
0

f;, s17C;Cj
j;C;Cj
-jj;77C;Sj

f;, g17Sj

0
0

f;sj
jf;77cj

U;
V;

W;

(36)

h;

2a,.

%/3;

= sin JO
= COS j()

(37)

r ,1)

(38)

In this case the inverse of J is seen to be

.,t;

(i = 1, 2, 3)

( II)

h,

r' = J* = [ to,z

OJ

17,r
77,z 0
0 1

(39)

l )Hing this inverse matrix , we can transform the derivatives in Eq. (36) to global
,oordinates, as follows:

f/) ~ ; ~ ;:1./i
()

[ 0

fr

()

U,e

772 cos y,

()

0
0

rg

ANlwf'orc , lcl us is<>lale lcrms in submalrix f1 that arc multiplied by

1~11

f;,{Cj
0

(7.9- 31)

Also, we put the displacement shape functions into the matrix form
0

(35b)

'l'he required Jacobian matrix is

cos J()
0
0
cos J()
[
0
0

h;
-77 sin y,.
2
h;

u,s
u,11

C; =COSY;

in which

Cj

0 0

In the coefficient matrix of this expression, the following abbreviations are used

[0

Thus, Eqs. (10) and (11) apply equally well to any case of nonaxisymmetric
loads, but matrix f is now of size 3 x 15.
For this analysis we must determine the derivatives of the generic displacements in Eq . (32) with respect to local coordinates. They are

V,8

wh.crc t~e angle /3,. is a small rotation about the x' axis [see Fig. 8. ll(a) 1. Abcforc'. 1f_the loads were antisymmetric with respect to the r-z plane, the furn
lions sm 18 and cos JO would be interchanged.
Equation (32) may be stated more efficiently as

-sin y,.
cosy,.
0

= 0 0 0

fBi

v,11

Li77j
1= 1

(.l'I

h[-sin
Y;
cosy,. -

403

and

b,j cos J()

j=O
If the loa_ds were anti.symmetric with respect to the r-z plane, the functions sin jtl
un~ cos 18 woul~ be mterchanged. The response of an axisymmetric shell to lhi~
s~nes of hannomc loads consists of a series of hannonic generic displacemenlN
lor element AXSH3 these displacements are expressed as follows:

[u]

Element for Axisymmetric Shells

j=O

Sec. 8.6

r,. I k 1111

U ,r

u,,
II, 11

l''\11)

()

11', II /

fJ '
()

.I "'

~1

., ...

u, ,
u,'1
II ' II

11 ',11

(40)

404
l 1l11t1111 0 11d Shollo
Hue. 8.6

Element for A><lsymmetric Shells

405

Multiplying the terms in this equation produces

u,,

where i
Into

a;Cj
b;Cj

0
0
- d;S; Cj
0
U, z
0
0
-e;S;Cj
0
U, IJ
-JJ;sj
0
0
! (Jf;'r]s;sA)
0
V,r
0
a;Cj
3
0
d;C;Cj
0
V,z
0
=2
b;Cj
0
e;C;Cj
0
i= I
V,o
0
-JJ;si 0
-! (jJ;71c;sA) 0
W ,r
0
0
a;Sj
0
d;si
W ,z
0
0
b;Sj
0
e;s1
w , 11 i
0
0
jJ;ci
0
! }firJcih;
where the constants a;, b;, d;, and e; are given by Eqs. (16).

Ej

'Yrz

U ,z
W

'YzlJ

'YrlJ

-U,r IJ

,z

(H,)1

a:

w 11 )
.

r
w

I .
; rfic1

I
- 2rfi'rJS;cA

b,c1

I .
2r }/, 'f'JC/11

a1c1

(d;c1 - e1s1)c1

b,s1
("

'Yy' z',

(47)

'Yz' .,,.}

which is of size 5 x 15.

We write the stiffness matrix for each harmonic loading as

f f, f"
= f

(BJ)TEBJ JJlrdOdgd'rJ

br

+ HB~JTE B~J JJ Jr df

[2(B,11)TE B,;i

where k = 2 for j = 0 and k

muss matrix becomes

I for j

(49)

I, 2, ... , m. Next, the consistent

0
0

Ki=

- J,c;

Ez , 'Yx' y' ,

(48)

(I

- r J!,.11

= {Ex,

b; CJ

- ; jfis;

llonally, the 6 x 6 strain transformation matrix T. serves to relate E to E , as in

l'.q. (7.2-27), except that the second row is deleted. Another useful matrix is

uf

In this case the normal stress <Ty and the strain Ey are omitted. Also, the
HIrcss-strain matrix E for an isotropic material is similar to that for element
PBQ8 in Eq. (8.2-21), except that 33 instead of E 55 is divided by 1. 2. Addi-

(45)

lo which the following strains correspond:

11

+ 'r]Ba1

(46)

The strain-clisplac~ment relationships shown in the second form of this vccto

1 111
the same as those ~n Eqs. (7 .9-15). Using these relationships and Eq. (4 J ) , w
11111y construct the zth part of matrix B for the }th harmonic response, as follow

a19
0

BAJ

lll'C

+ V,r
+ V,11
w ,r

(44)

which will be used to find stiffnesses.

For the present analysis, nonzero stresses in the directions of primed axes

V,z

-r (u +

= (BA;)i + TJ(Bs;t

B1 =

E9

, m. As before, we can decompose (B; )j

11ncl state the total matrix in the form

U,r

E,

= 0, 1, 2, ...

1, 2, 3, and}

(B;)i

For no?axisymmetric loads on element AXSH3, we consider six typt ~

nonzero strains . Thus,
Er

- 2,)fi'rJC1s/11

-1).

I .

2r )/,rJ.1,.11h1

e1.11

(r1, Jiq!,')
2,. .,,

(50)

which is the same for j = 1, 2, . .. , m (see Sec. 7.9). Then equivalent nodal
londs for each harmonic set of body forces are

p,,(t)1 =

J', J', f" f'XcJc1b(t)1 IJlrdOdgd 'rJ

k TT

fi

2f'.\'l,(t)1 I .J 1, d

(/

0 , 1,2, , , , , 11/)

(5 1)

408

Plates and Shells

Sec. 8.7

Program DYAXSH for Axisymmetric Shells

in which only matrix fA is used. Components of force per unit volume (constant
lhrough the thickness) consist of
b(t)j

= {b,j , b,j, b0J

(52)

TABLE 8.6 Dynamic Load Data for Program DYAXSH

Type of Data

Finally, the stresses for each harmonic response may be written as

u'(t)j = E BJ q(t)j

No. of
Lines

Items on Data Lines

!SOLVE, NTS, DT, DAMPR

1
NNID
NNIV

NNID, NNIV
J, D0(3J-2), D0(3J-l), D0(3J)
J, V0(3J-2), V0(3J-l), VO(JJ)

NLN,NEA,NEV
J, AS(3J-2), AS(3J- l), AS(3J)
I, BAI, BA2, . . . , BA6
I, BVl, BV2

Dynamic parameters

(j = 0, 1, 2, . . . , m)

(5.l)

which are in the directions of local axes.

8. 7 PROGRAM DYAXSH FOR AXISYMMETRIC SHELLS

Let us now consider a computer program named DYAXSH for the dynanik
analysis of axisymmetric shells with axisymmetric loads. This program USl'~
clement AXSH3 from the preceding section. It is generated by modifying thl
part of Program DYAXSO dealing with element AXQ8 (see Sec. 7 .10), 011
which element AXSH3 is based. However, in Program DYAXSH we mw,t
formulate K, M, and Ph (t) in accordance with the equations developed 111
SL'C. 8.6.
In Table 8.5 we see how structural data is to be prepared for Progru111
l>Y I\ XSH. As for general shell data, the symbol H appears in the list ol
NII uctural parameters for a case where the thickness is constant over the whok
tihcll. In addition, H(J) is included with the nodal coordinates for a case whl'll'
thickness varies over an element. That line of data also contains the compone11t11
Y2R(J) and V2Z(J) of the normal vector Vi for node J [see Eq. (8.6-4)]. Finally,
three types of nodal restraints are indicated in line (c).
Dynamic load data for Program DYAXSH is displayed in Table 8.6. I ll'll'
we have three possible initial displacements, initial velocities, and nodal !0111111
instead of two per node. Area loads for element AXSH3 are similar to those Jor
AXQ8, but only the element number I (not the node numbers J, K, and L) lll'l'd
be given. Also, the acceleration factors GAX and GAY are replaced by GAZ I111
the uxisymmetric shell.

Initial conditions
(a) Condition parameters
(b) Displacements
(c) Velocities
Applied actions
(a) Load parameters
(b) Nodal loads
(c) Area loads
(d) Volume loads

NLN
NEA
NEV

Ground accelerations
(a) Acceleration parameter
(b) Acceleration factor
Forcing function
(a) Function parameter
(b) Function ordinates

1
1
I
NFO

No. of
Lines

Prnhlcm identification
SI I ucturul parameters
i\xlsy111mctric shell data
(u) Noclnl coordinates, etc .
(b) Element informatio n
(cl Nodnl rcslrnints
Fo1 Nl'(Jlltl ll l't'

ul

lllltlt 1111111htll N, Nl'l'

Fl11 t! I 1(11) ,

NFO
K, T(K), FO(K)

Example 8.4
The vaporous gas storage tank in Fig. 8.12 has a hemispherical top and a circular
z
I

Items on Data Lines

Descriptive title
NN, NE, NRN, E, PR, RHO, II

NN
NE
NRN

IGA
GAZ

With element AXSH3 we use n = 2 in the f direction [see Fig. 8.1 l(a)]
to locate points for numerical integration. At each of the~e two p.oint~ the
computer evaluates the time-varying stresses <rx, <r,, and Tx'y' m local direction~ .
Again, we had to use n = 3 when evaluating terms in the consistent mass matnx
for element AXSH3.

TABLE 8.5 Structural Data for Program DYAXSH

Type of Data

407

J, R(J) , Z(J), V2R(J), V2Z(J), ll(IJ

I, IN(T, I), IN(l ,2), IN(l ,3)
J, NRL(JJ-2), NRL(3J- I), NIU ,( \f J
1111

408

1'1111011 ontl Sholls

Sec. 8.7

Prog ram DYAXSH for Axisymmetric Shells

409

z
I

We ran the foregoing data with Program DYAXSH using DAMPR = 0.?2 a.nd
Subprogram NUMINT for calculating d7n~mic responses. The computer plot m ~1g.
8. 13(a) gives the variation of pressure ms1de the t~~' an.d Ftg. 8.~3(b) s~ows ti.me
histories of membrane and flexural stresses in the x direction at the mtegrat10~ fOmt
near node 37. The maximum value of the former stress is 205.2 MPa, and the mm1mum
value of the latter stress is -308.6 MPa (at the outer surface of the wall). Fu.rther
refinement of the finite-element layout near the base would produce a somewhat higher
llcxural stress at that location.

j2

j3

J-j1

16

..
a.

3 . 00

I:

0.4H

-+

~ 2.00

:::,
(/)

(/)

0::::

0.4H

a.

_l_

_J

I. 0 0

<
z
0::::
w
1z

- o. oo

(b)

(a)

cylindrical wall that is fixed at its base. This axisymmetric shell is divided into I H
AXSI 13 clements that become progressively shorter toward the base, where high be111ll1111
stresses arc expected. Surface pressure p, from an interior explosion causes dy1111111h
response of the shell , and we wish to examine stresses in the vicinity of node 37 d111 111
this inllucncc. Physical parameters in this example are
E = 207 X 106 k:N/m 2

v = 0.30

h = 0.025 m

,.:; rd(

a."' 200

"-

(p.,.)mnx = 3 MPa

f r-z J
fr

MEMBRANE
I:

p = 7.85 Mg/m 3

I")

when: the material is steel and the units are SI.

As in Example 7 .6 , we need to find equivalent nodal loads due to press1111 I'
which we assume acts in the direction normal to the middle surface of a shell ek11w11t
(lhl t / direction). These equivalent loads arc

p1, = 27Tp,,.

T Cms l

Figure 8.12 (cont.)

R = H = 2m

w
O
0

T Cm s l

0::::

<
w
z

-200
X

1111

(/)

'l'h1s 1'01111ult1 lcw 1>,, is the same as that in Eq . (7. I 0-h) , but in this case there is 110 111'1 d
1111 the pii111l' on f As before, signs for the equiva lent nodul londs urn 11uto11111t1111 lh
d11t1111il11ld hy liq. (n) l'or pressure t11ken to be positive in the positive sonsl of1h1 1111111111l
tliH'l'Iio11 r/.

( b)
lt'l1111111 II

II

VIIJ11UUII N )lllN Nllli ll)ll'

11111k:

(11)

Jond ; (h) Nll\'NNCS,

lnt11 nrut Hhnll1

REF

R NC

Ch

I. M1ndl11.1, R. 1)., "l111lucncc of Rotutwy Mo11111111111I Shl'llt on Flcxurul Mot11u11

Isotropic, Elastic Plates," J . Appl. Mtch ., Vol. 7 J, 195 1, pp. 3 1 38.
2. Ahmad, S. , lrons, B. M., and Z icnk.icwicz, 0. C., "Analysis of Thick 1111d lhl
Shell Structures by Curved Finite Elements," Int. J. Numer. Methods Elli( ,, Vul
No. 4, 1971 , pp. 575- 586.
3. Wea~er, W., Jr. , and Johnston, P R., Finite Elements for Structural A 11 ,,/1 ,/
Prent1ce-HaJ1, Englewood Cliffs, N. J., 1984.

Rigid Bodies
Within
Flexible Structures

4. Timoshenko, S. _P., and Woinowsky-Krieger, S., Theory of Plates and Shl'!/1 , , 11

ed., McGraw-H11l, New York, 1959.

5. Cook, R. D. , Concepts and Applications ofFinite Element Analysis, 2nd ed. WII, v
New York, 1981.
'
6. Yoshida,. D. M.'. and Weaver, W. , Jr. , "Finite-Element Analysis of BeaniH und
Plates with Movmg Loads," Int. Assoc. Bridge Struct. Eng. Vol 31 1971 l'I
179-195.
'
.
'
'
7. Ahmad, S., Irons, B ..M., a~d Zienk.iewiez, 0. C., "Curved Thick Shell nnd
Membrane Elements, with Particular Reference to Axisymmetric Problems," ,.,,,,
2nd Conj. Mat. Methods Struct. Mech., Wright-Patterson Air Force Base Ohio
1968, pp. 539-572.
'

9 .1 INTRODUCTION

Occasionally, the analyst encounters a structure containing one or more parts (or
bodies) that are very rigid in comparison to the other parts. Such bodies are
usually taken to be infinitely rigid [1], and the nodes connecting them to the rest
of the structure are constrained to displace in a pattern corresponding to the
rigid-body motions. These restrictions on nodal displacements serve to reduce
the number of degrees of freedom in a given problem. For example, the joints
labeled A, B , and C in the counterweighted plane truss in Fig. 9. l(a) would
ordinarily have a total of six degrees of freedom among them. However, the
rigid counterweight (shown hatched in the figure) , on which they are located, has
only three degrees of freedom. These three displacements are translations in the
x and y directions and a rotation in the z sense, as indicated by the arrows labeled
1, 2, and 3 adjacent to the rigid body. Thus, the number of degrees of freedom
in the problem is reduced by three due to the presence of the counterweight.
Figure 9 .1 (b) shows a second example of a building frame, for which the
analytical model is taken to be a rectangular space frame containing floor and
roof laminae . Each lamina is assumed to have infinite rigidity in its own plane,
but zero rigidity normal to the plane . The x and y translations and the z rotation
for any joint on a particular body are dictated by the corresponding rigid-body
motions of that lamina. These motions are indicated by three numbered arrows
at each framing level. Therefore, if ni is the number of joints at a given level,
the number of independent displacements at that level is 3111 I 3 instLud of 6111

411

412

lll11lcl tloclh11 within Flexible Structures

Chip

413

Rigid Bodies in Framed Structures

In this chapter we examine the effects of including rigid bodies within

framed structures and discretized continua. It is assumed that the bodies are
connected and supported by linearly elastic, flexible materials and that their
displacements are small. We also assume that the bodies are not rigidly connected to each other or to the supports, which would require additional constraints .

t_,

Sec. 9.2

)-,

9.2 RIGID BODIES IN FRAMED STRUCTURES

(a)

z
3

t_2
6

Figure 9.2 shows x, y, and z axes and six indexes p I, p2, ... , p6 for
components of actions or displacements at a reference point p on a threedimensional rigid body. The figure also depicts a typical joint (or node) j of a
structure connected to the body. If the structure is a space frame, there will be
y

Rig id body

/-5

~y
X

(b)

)-,

/
/
/Zp,
(c)

Figure 9.1 Rigid bodies within flexible continua: (a) counterweighlcd 11 uN,
(b) building with laminae; (c) shell with hub .

__________ ___ _y /
xP,

Appearing in Fig. 9. I (c) is a third example of a rigid hub at llll' l'l'lllrl 1

shell structure. If the geometry of the shell is general, displacements of no,l
111tn~lwd lo the hub urc construined lo displace according 10 the iil{11I t,1111
111011011s l11l>llcd I through 6 in the n~ure. Othcrwi1-K, ii' the shell Wl'11. uxl11v111
111l't11c. lhl 1111111hlr of iigid body motions would lw ll'Wl'l

11

1>'1111111 1>.2

Rigid hody consti uint~ .

Sec. 9.2

Rigid Bodies within Flexible Structures

414

~ix indexes j I, j2, ... , j6 for actions or displacements at point}, as indkul

111 thc figure. An offset vector rpj is directed from pointp to point} and has Sllll
l'0111~onents XpJ, YpJ, and zpJ We may calculate the statically equivalent al'flmt
~ ,~ point p due to actions at point j using the concept of translation of axes I >, I I
I hus,

II
In Ihis equation the symbol Aj denotes a column vector of six action comprn11 nl
111 point}, as follows:

A1 = {A11,

A12, ,

A16}

wh~rc the first three are forces, and the last three are moments. Also, the w11
(1) contains six statically equivalent action components at poi111 I'

A,, 111 Eq.

A = {A1 , A2 , ... , Ap6}

l<'inully, the transformation matrix

TpJ

in Eq. (1) has the form

where

Zpj -ypj]
0

Thus, transposition of CpJ merely changes its sign, which is an inherent property
of a skew-symmetric matrix. Note that the transformation in Eq. (6) is analogous
to that in Eq. (4.5-1) for small rigid-body ground displacements.
A similar transformation matrix TpJ can be derived for each of the other five
types of framed structures. Table 9.1 summarizes these matrices for beams,
plane trusses, plane frames, grids, space trusses, and space frames. Each of the
first five matrices in the table can be found by deleting appropriate rows and
columns from the sixth. For example, point p for a rigid lamina in a plane truss
would have three displacements, which are translations in the x and y directions

1. Beams (x-y plane of bending)

=[

Tpj

-zpJ

Pl

Thi s skew-symmetric submatrix contains positive and negative values ot th

n>mponcnts of the offset vector rJ These components are arranged in a 111111111
lhul pr,c~d~ccs the cros.s product of rpj and the force vector F1 at point}.
S11nilurly, the kmematically equivalent displacements at point j 11111y I
rnkulntcd from those at point p with the relationship

= T;o,,

1ft

11111 th~ ~ymbol DP represents a vector of six small displacement co1111>11t1l'III

ol lhl' 11gtd body at the reference point p. That is,

'1'111 11p1111l111 '1' 1'.1 i11 nq (h) is

Xpj

3. Plane frames (structure in x-y plane)

Tp;=[

b
-ypj

~]
I

Xpj

4. Grids (structure in x-y plane)

T p; =

I O
[~

b -t

YPi]

5. Space trusses (transposed to save space)

T;;

whl'H' thl.' first lhrcc arc translations , and the last three arc rotations. In 11dd111111
tlt1 Vl'l'tor 1>1 rnnsists of six kincmatically equivalent displacements at point

{D11 , l>p, . .. , D,c,)

b ~]

- ypj

[b ~ ~ -~; t -~]
2

D,, = {D,,1, D,,2, . .. , D,,6}

D,

[L ~]

T=[

-y~

D1

2. Plane trusses (structure in x-y plane)

(10)

-Xpj

i11 which l, is an identity matrix of order 3 and

CpJ

Xpj

TABLE 9.1 Transformation Matrices Tp;

for Framed Structures

:J

Tpj = [:;j

415

Rigid Bodies in Framed Structures

0 0

YPi

-Xpj

6. Space frames

(N
T,,,

0
0

()

:,,

z,,,
()

1,,1

0 0
0 0
0 0
Y,,1 I 0
0
0

,,,,

()

()

()

()

416

Rigid Bodies within Flexible Structures

Chap

Sec. 9.2

417

Rigid Bodies in Framed Structures

and a rotation in the z sense. But a joint} in the truss has only two displacemcnlN
which are translations in the x and y directions. Therefore, we keep only the flr "I
second, and sixth rows and the first and second columns of the 6 x 6 matrix 1;,1
Deletion of the other rows and columns from the sixth matrix results in the 3 x 1
array for plane trusses shown in the table. Note that the 3 x 3 transformal i1111
uiatrix pertaining to grids requires not only deletion of rows and columns , hul
also rearrangement as well. The reason for this is that the x- and y-rotation1, nl
Ille joint of a grid are taken before the z-translation [see Fig. 6. l(d)].
Now let us consider the task of incorporating rigid bodies into our ann
lyt ical models for framed structures. For this purpose, we use a member-orie111, ,I
<1pproach to transform actions, stiffnesses, and consistent masses at the end1, 11r
111cmbers to reference points (or working points) on the rigid bodies. Also, ii I
usually necessary to transform mass and mass-moment-of-inertia matrices l11r
lhe rigid bodies from their mass centers to their working points.
Figure 9.3 shows a space frame member i connected to three-dimensionnl
ri gid bodies at both ends. Points p and q on the bodies are taken as worhllljl
poinls to which information about joints j and k will be referred. Firsl, 1111v
lll'lions in vectors Aj and Ak at the ends of the member may be transformed 11110
sluti cally equivalent actions AP and Aq at the working points by the follow11111
/'l'llcrulizcd form of Eq. (1):

AP]
[ Aq

[Tp1 0 ] [ A1 ]
O Tqk Ak

(I I)

'l'hl' action vectors Ak and Aq are similar to Aj and AP in Eqs. (2) and (3). /\11111
lhl' transformation matrix Tqk is the same type as Tpj given in Eq. (4), b111 1111
points q and k. Equation (11) may be expressed more concisely as

(I )
in which
(II)
lllld

T; =

[;j

~J

( 11)

'l'ht trunsf'onnation matrix T1 is a combined operator that converts the m:ti1111" Ill
AM, (al lhc ends of the member) to the statically equivalent actions in A111 (111 lh
working points of the rigid bodies). These actions are all in the dircclirn, ~ of
Nl111cl11nd uxcs . Bolh actual and equivalent nodal loads can be trcall'd 111 lhl
I lllllllll'I'.

DiHplHl'Cllll'nls at joints j and k wi ll also be expressed in terms ol' lh11111 ti

I' 1111d ,1 hy 1111 txtcndcd fmm of Eq . (6), as follows:

I
I
T,~,
I
u,

l)A

o ,, n,. ,
'l',/1

l>,1

---

..__

--- ..__

I
I

--- --- -- -- --

I
I
--- --- ..J

Figure 9.3 Space frame member with rigid bodies at both ends.

The displacement vectors in this eq~ation corres~ond to the action vectors in

Eqs. ( 13). Equation (15) may be wntten more bnefly as
(16)
Whl'll'

( I 'I)

418

Rigid Bodies within Flexible Structures

Sec. 9.2

and

T;

= [;

point c, as indicated in Fig. 9.2. If the body is thr~e-di~ensio~al_ and has six
degrees of freedom, its mass and mass-moment-of-inertia matnx 1s

;JJ

.
In addition, we can transform the member stiffness matrix Ki from joinl
J and k to the reference points p and q. For this purpose, consider lh
action-displacement relationships

AM;= K;DM;
Substitution of Eq. (16) for DM; gives

AM;= K;T;D8 ;

(20)

Use of this expression in Eq. (12) produces

Asi

= T;K;T;D8 ;
= T;K;T;

(" )

The symbol_ Ka! denotes the member stiffness matrix for actions at points p urul
f/ due lo unit displacements at those points.
Similarly, the consistent-mass matrix M; for the member may be trn1111
lonncd from joints j and k to points p and q. Reasoning as above, but w11h
uccclerations instead of displacements, we can derive the fonnula

Mc=

0 0 m
f;u
0 0 0
-fyx
0 0 0
0 0 0 -Jzx

MaaDa

+ SaaDa

A 8 (t)

( '4)

To form the matrices in this equation, we assemble contributions from individ1111I

lllcmbers by the direct stiffness method, as follows:

,,,

S,111

= L KB/

Mao=

L" Mo;

Ao (t)

"e

= L As (t);

( 1)

This assembly process is similar to that described previously in Sec. 3.5, hut 1h
111otions arc at working points instead of joints. If a rigid body docs nol l'Xllll 111
11 particular joint, all of the offset vectors for members framing into that joinl 111
t11kl1t to hove zero lengths . Also, the rotational displacements arc 0111illtd lor
11011txislt111 rigid bodies i11 plane and space trusses.
As yet, lht muss 111111,ix M1111 in Eqs. (24) and (25) is tkvoid ol 11111
l11h11tio1'.s 110111 lht rigid hodits lht111stlws, whidt 11111y lw lht 1110s! 1111p,;111111t
111111~ li.11d1 111-t1<I hody 111 lltl 111111ly11nrl 111od1l lt11s Its 1111111 11/111t1.1,1 lornhd 141

0
0
0
-lxy

0
0
0
-fxz

fyy

-fyz

-lzy

fzz

(26)

In this symmetric 6 x 6 array, the symbol m denotes the mass of the body,
which is computed from
m

fvpdV

(27)

The mass moment of inertia lxx is obtained as

fxx

= fv p(y~

+ z~) dV

(28)

in which Ye and zc are they- and z-distances of a typical point in the body from
the center of mass. The mass product of inertia lxy is

fxy
I, Iere the sy~bol Mi_ re.presents the consistent-mass matrix for member i al jn1111
J nnd k, while MBi 1s its counterpart for the reference points p and q.
Equations of motion for all of the possible rigid bodies in a framed sll 111
lure may be written as

0
0
0

m 0 0
0 m 0

(l I)

I lence, the matrix relating ABi to D8 ; is

K8 ;

419

Rigid Bodies in Framed Structures

PXcYc dV

(29)

where x is the x -distance from point c. Other moments and products of inertia
in matri~ M have similar definitions. The negative signs on mass products of
inertia in Eq~ (26) result from Euler's equations for small motions of rigid bodies
14]. Table 9.2 contains various forms of matrix Mc required for all types of
framed structures.
The matrix M for a rigid body may be transfonned to a working point p
hy an operation si;ilar to that for consistent masses of membe_r i ~iven in Eq.
(23). However, only the points p and care involved, as shown m Fig. 9.2. The
required congruence multiplication is

Mp = TpcMc TJc

(30)

1n this equation the transformation matrix Tpc is of the sam_e _form as ?1a~ix TpJ
discussed earlier, except that c replaces j. Note that for a ng1d body_ m either a
plnne truss or a plane frame, the matrix Tpc must be the same (and of size 3 x 3).
Also, for a three-dimensional body in either a space truss or a space frame, Tpc
111 11gain the snml' (1111d of si7,l' 6 x 6) .
.
.
As the st~o11d 1111p 111 11HNl'11thling cquutrons of motion , we must add the
11111 ssis 1111 d 11111s1, 1111111,rrl! M111 111111111 lt11 tht ri~id bodies to the matrix M1111 in Eq.
()I) 'l'lti1, piwi. 1111 1111 1111v.1111111l ti 1111111 i,1 M 11 11, 111, lollowi,;

420

Rigid Bodies within Flexible Structures

TABLE 9.2 Mass and Mass-Moment-of-Inertia

Matrices Mc for Rigid Bodies
1. Beams (x-y plane of bending)

I~J

Mc= [~

2. Plane trusses and plane frames (structures in x-y plane)

Mc= [ ~
0

i..l,- - -; ; - - - ~ -- - - ~ - - -- -+---;;---

~z ~]
0 I,,

3. Grids (structure in x-y plane)

Mc

f xx
- f yx

-fxy
f yy

OJ
0

j~

....,
~

4. Space trusses and space frames

0
m
0
M, -[} 0
0
0

0
0
m
0
0
0

0
0
0

0
0
0

f,x

-fxy

- f yx

! yy

-fzx

-1,y

,,;

g"'

J]

..,c::
-a.
"'
.s..,"'

-ly,

I,,

....,
"b

M1s =Mes+

00

Mpk

( II)

.$
-0

'@i

62
...,.

k=l

wkrc 111, is the number of bodies.

After solving the eigenvalue problem for the augmented form of Eq. ( 1 11
wt cun add modal damping to our analytical model. Initial conditions 111111
g1ound motions may also be included, along with applied actions, if dcsii11I
SttIion 9. 3 describes a program named DYRBPF for dynamic analysis ol 111,1111
hodil'S in plane frames, using the member-oriented technique given aboVl'
In any case (especially when the mass of the structural framing is 111 h
1wgltclcd), we could use the mass center c of each rigid body as the wor~ln
p11l111 p. With grids, space trusses , and space frames, it may also be convl11lr11t
lo 11.~t f)l'incipal body axes, for which the mass moment of inertia suh11111111
wtlhin M, is diagonal. If principal mass moments of inertia and the diHt111111
ol pri11l'ip11I body axes are not known in advance, they may be found by solvl11
1111 ligl'Hvaluc problem of order 2 or 3 for each rigid body . This method is 1,i111i1111
111111111 lor principal stresses described previously in Sec. 7.2. lf one or 1111111 ut
fill' i111plitd rigid bodies do not exist, we could also eliminate the displ11n111111t
111 11111ssk ss 11odts by stuti, r<'d11ctio11 , as discussed in Sec. 6.7 .

Q\

...
QI

So

K1rn111111t, >.1
111111 11 I Hli11wH II p1111i11u 11111 pl111111 1111NN wilh lwo dgid 1l'l'l11111;11l111 l11111ii1111 111111111 l
liy II pilN11111t11 1111111lw1 r 111 1111111~ / 1111d A 11,1 11~ dl'Vl'l11p 11111l11h111l1111N111 lhl' 1111111h
1111d 1hr 111111111111 l111111d11111p1tl 1q11111i1111N 111111111111111111 p111t1l ~ /I 11111l 1/ llt1 ll11, lipid 11111(1

"'c::

'

.....

422

Rigid Bodies within Flexible Structures

1UC.

9.2

423

Rigid Bodies in Framed Structures

We shall include actions, stiffnesses, and consistent masses from member i, ns w II

masses and mass moments of inertia from the bodies.
First, we consider actions at points j and k in the column vector

AM; = {Ah Ak} = {A,1, A12, : Aki, Ad

Statically equivalent actions at the working points p and q may be obtai111d II II
transformation matrices for plane trusses from Table 9. I , as follows:

TP,

= [

~ ~J

-4L

3L

ABi = T;AM; = {Ap, Aq}

= {A11, A,2, -L(4A,1 - 3A12),: Aki, Ak2 , 4L(Ak1 - Ak2)}

K.ii

K; =

[ Ki\i

K,k]
Kkk

x 4 stiffness matrix for member i, using Eq. ( I ,

0.64
Sym.1
0.48
0.36
i - IOL -0.64 -0.48 10.64
EA

3l

4L

- 4L

2
-8L

64L2

12 0
0 ]
mi
Mei= 0 12
0
12 [
0 0 25l 2

(h)

1'1nnsformation operators required here are the 3 x 3 arrays

-0.36 ! 0.48 0.36

,:, n l-:L ; :1

0.64
0.48

0.36

- I.l2L

- 0.84L

Sym .

0
Xqcz

1.96L 2

------------------1---- 0.64
- 0.48
l . l2L l 0.64
I 0.48

- 0.48

-0.36

0.84L

- 0.64L

- 0.48l

l.l2l 2 J 0.64L 0.48l 0.641.'

5pAI,
1

r~

~] = [

~]

-2l

I hey are applied to convert matrices Mc I and

111"rnrding to Eq. (30). Thus,

6
T

Mp= Tpc1 Mc, Tpc, -

m1

[ -12L

Sym.

2
0

()

(i)

Mc 2 to the reference points p and q,

0.36

Third , we write the cons istent-mass matrix for member i from Eq. ( I ~ \
Thul is,

M,

-282

l0
I 8L

which is found using Eq. (23).

.
Turning now to the rigid laminae, we take the followmg mass and mass-moment111 -inertia matrices from Table 9. 2 for plane trusses:

Then we transform matrix K; to the working points p and q with Eq. (22), p1111h11 I

EA

(g)

- 0.48

= lOL

5pAL -8L
6L
502 J
-------- - ----i-1
0
-4L I 2
3
I

Substituting these arrays into Eq. (14), we form the operator T; and use it in h1 1I
Thus,

Second, we set up the 4

Sym.

M,,

m2

1,,, i M,. 2T ,,,2 - 3

61,

61,

~~L]

(j)

321}

')

'1'11111sl11111111111111 nl 11111ld x M, lo llll' wm~11111 points 11 111ul ,, yi1lcls

1lw 11 11, 1 o r thl'Nl' 111 rnys 1111n111111ts s 11h1111111 Ix M,,1,1 in Eq . (Hl, 1111d the Nl'l' ond is nddcd 10
M,,.,, Thr N Htr11 t111till H th1 11h111 II Vl'N Ht11hd Ill lh1 '1111 11111111)\ 111 th1 t'Xlltllpk

424

Rigid Bodies w ithin Flexible Structures

ftoc. 9.4

9.3 PROGRAM DYRBPF FOR RIGID BODIES IN PLANE FRAMES

As an example of programming for the member-oriented technique, wr

briefly describe Program DYRBPF for dynamic analysis of rigid bodies 111 pl I
frames. To simplify the procedure, we take the mass center of each rigul 111
as one of the nodes in the structure (as well as the working point for tht h111I
Within the logic of the program, every member is assumed to have a rigid h,
at each end, unless proven otherwise by the input data. This idea providt I
key to easily extending the programs for framed structures to incl ud1 11
bodies.
Building upon Program DYNAPF from Sec. 6.6, we can add rigid hrnll
to plane frames and create Program DYRBPF. For this purpose, the line c1111t
ing nodal coordinates in the structural data (see Table 6.1) must be aug111111I
by adding BM(J) and BI(J). These terms represent the body mass m and 1111 111
moment of inertia /, 2 of the body with respect to its center of mass. Al so , 1t1 II
element information we add XCJ(I), YCJ(I) , XCK(I), and YCK(I) , wh11 h
note the x and y components of the offset vectors r cJ and r ck at ends j a11d A C
course, if there is no body at one end or the other, such offsets are set tq111I
zero.
In Program DYRBPF we assemble the stiffness matrix S8 8 , lh1 111
matrix Mon , and the action vector A 8 (t) by assessing one member al a ti1111 I
Eqs. (9.2-25)]. To the second of these matrices we add the mass-inerl111 t 11
for the rigid bodies to form the augmented mass matrix
0 , as given hv I
(9.2-31). After the equations for rigid-body motions have been solved , w,
find time histories for displacements at the ends of members using Eq. (') ' I
Then time-varying member end-actions are calculated from these mc111hr 1 111
placements by premultiplying them with the member stiffness matri x.

425

Rigid Laminae in Multistory Buildings

L-

---.1.--

j2

-i1

4L

j3/ i

M;

- - -- ---x

/ ;:;
/

Example 9.2

Figure 9.5

Figure 9. 5 shows a steel plane frame supporting a single rigid body, which is 11 1111111
cube of size Lon each si.1e. We shall analyze the dynamic response of this conllp11111tl
to ground acceleration D 8 1(t), caused by an underground blast.
If all members of the frame have WlO x 45 cross sections, we can Ninh
physical properties as
E, = 3 .0

104 k/in. 2

p, = 7 .35 X 10- 7 k-s 2 /in. 4

A = 13 .2 in . 2

L = 36 in .

I, = 249. 0 in . 4

Rigid body in a plane frame.

111,1. 9.6, the first of which depicts the impulsive x-g~oun.d ac~eleration due to the
underground blast. The plot in Fig. 9.6(b) gives the resultmg_t1me history of x_-translat10n
111 node 3 (the center of mass of the rigid body). Also, Fig. 9.6(~) contains y~ots of
lli1tcvarying member end-moments AM6 for element.s 2 and 4. Max~mum (or rrumm~m)
vnlues for these three types of responses are 0.03943 m., -157 .3 k-m., and 97 .06 k-m. ,
1111 of which occur at time t = 52 ms.

where the subscript s implies steel. Al so, the relevant properties of the d gid l1111l y

t>,

2.25 x 10

k-s 2/ in .'1
I

Afl, I.'

Ill

{), [ , I

1.050

10

ks

lt,4 RIGID LAMINAE IN MULTISTORY BUILDINGS

/ i11

2.268 k s 2-in.

for whk h lltt' s11hst1lpl ,. dt11otcs rnut1tl1.

,
Wt 11111 1111' loi,1101111 <111111 011 1'10H111111 l)Yltlll'I ; 11Nl111,1 S11hprnp111111 NI IHMC
Wllh NM< II >1 1S
/ 1111d I >AMI'!<
() 0 1 l'l11M111 111111111 p1111h111~ 1111 111111p111l I plnl

M1tllistory bui ldings arc usually constructed in tiers of one, two, or three .stories
limi;. I lcm.'l' , llw 1111111l' ti<'t huildil1Rs 15, 61 may be u_sed to descr~be_the
1h yNl'rnrwrs th111 ah111111d 111 l111w cilks. Tlw stlc l skl'l\.\lons of these lull buil~mgs
1 l!IININI nl ll jllll'l' 1111111111 111111 IIIIIHI l11q111nll y 1111w thtir llll' itllll'l'I, lllrttl\At'<l 111 llll
111 11

Rigid Bodies within Flexible Structures

426

Chlll

Sec. 9.4

427

Rigid Laminae in Multistory Buildings

Slab or deck

400

.,

y-beam

'- 200

x-beam

60

20

(.)

<

80
(ms )

a:: - 2 00
{!)

- 400

P,(t)

T---J--.----+----ll'---.1
Y FP

(a)
C

.:; 0 , 0 4
I")

~ 0 . 02
0

i - - - - --

< o. oo

20

80
(m a )

X , c - - -- - + I

k---

;: -0. 0 2

- - -- -X,p-- - --,.... P,(t)

- --

--X,; - --

- -- - --1

<

..J
(/)

a::

Figure 9.7

-o. 0 4

I-

( b)

I
)(

ELEM,2

100
C
I

~o
~
lO

< - I 00

( c)

li'IKurc 9 .6

Rigid body in a plane frame: (a) ground acceleratio n; (b) displut'l'1111111 .

{c) 1'1HI lllOIIIClltS.

111 Ihogonnl ( or l'L'<.: langu lar) pnllcrn . Figure 9.7 shows a lypical floor (111 11
pl1111 of II litr building, (.'()Jlsisling o f x hcnms, Y beams, und a sluh 111 ,It, k ti
wr shull l'llllN1dtr lo tw I igid i11 its own plnnL'. This l11111i1111 is fk x1hh 111 I
dtrt'\'111111 11<1111 1111 to ii ~ plt11a 1111d tntdd Ill' 1li s111ti:,,td wilh plult' lw111lt11ai I
1111111 , A 1111,lt-1 l11rl 111111r 1111111111111l y !Hird upp1011d1 IN t11111cl11d1 11 tr 111111111 v
ol lhl' Ninh 111 tlw 1,11,.'I .,r, 111111 111 1 111 Ii lw11111 Thl''1' hr11111 ~ 11111111 111111 v rt

Typical floor plan of multistory (tier) building .

rnlumns that are parallel to the z axis (not shown in Fig. 9. 7). An origin of
1cference (or structural) coordinates may be located at any convenient point and
llrobably would be taken at the base of the structure. The point labeled Fin Fig.
').7 lies at the intersection of the z axis with the plane of the floor. Also, point
,. represents the center of mass that is tributary to the floor, and point j is a typical
joint at this framing level.
Because each horizontal lamina is taken to be rigid in its own plane, all
points at that level are constrained to displace in a rigid-body pattern. Such
,notions consist of translation in the horizontal plane and rotation about a vertical
11xis. Thus, the rigid-body motions of the lamina in Fig. 9.7 may be described
I1y the x and y components of the translation of point F (considered to be attached
to the body) and its rotation about the z axis. Moreover, the corresponding
displacements at any other points on the lamina, such as joint j or the center of
111uss c, are related to the motions of point F by the concepts of Sec. 9. 2. On the
11t her hand, the x and y rotations and the z translation at a typical joint j remain
11s unconstrained displacements.
For simplicity in the analysis , we assume that the geometric layout of the
li 11ming is lhe same at ull levels, that there is only one distinct lamina at each
llvcl, and thut lhL:rt nn oo shear walls or lateral bracing in the building.
Mc.111bc1s of llw spnl'<' t 1111m nrc 111kcn to he prisnmt ic and of a lincorly elastic
11111111inl Wt ulNo IIHNIIIIII" 111111 joi11tNIn llw r, 111111 IIH' I i1tiil 1111d thllt displ11L'l'll1Cl1I S
1r l11t i VI' lo ~'111111111 1111 111111111

428

Rigid Bod ies w ithi n Flexible Structures

Because every member in the analytical model is parallel to one ot II

structural reference axes, there is no need for rotation-of-axes transformali1111
Figure 9.8(a) shows a beam with its member axis Xm parallel to the x ux1~ I
11ddition, its principal bending axes Ym and Zm are assumed to be parallel lo 11~
y and z. Action and displacement indexes j I, j2, andj3 at the j end of llll' lllh t
i denote rotations in the x and y senses and translation in the z direction . 'I h I
counterparts kl, k2 , and k3 also appear at the k end of the member. The (1
stiffness matrix Ki for such a beam is the same as that for the grid mcn1h1 1 II
Sec. 6.3. Thus,

"'
,.,- - - - -

I"-

'.zr
-

"~.

'~.s

------=--------'< -.

:s
VJ

,;:
oj

S:l

-~

~~

0.

GlxL 2
4ElyL 2

0
0

-6EJ L

lI
I

0
0

l O
2
2ElyL i -6ElyL
6ElyL l - I2Ely
0

E
::,
0u
~

12Ely

- ------- - --L- ~- ------ -----2

-GfxL2

C:

Sym .

-in'~

(I

GlxL

;,...

'<

0
0

2
a:l

=;;i

12Efy

0.

As hcf'orc, the symbol Ix represents the torsion constant, and ly is the 1w11111
1110111ent of the cross-sectional area with respect to the Ym axis.
The second type of beam to be considered has its axis parallel to thl' I' 11 I
as indicated in Fig. 9.8(b). In this case Ym is chosen to be the member uxlN, 1111
the principal axes x 111 and Zm are parallel to axes x and z. Action and disph1cr111 H
Indexes j I through k3 correspond to those for the x-beam. Therefore, tlw h
111ember stiffness matrix for the y-beam is

K, = [~j
Kki

Kik
Kkk

E
oj

..8

e
in'

,;:
oj

><

.;

'<

0.

Q)

!:Q
~

4EI.,, L 2

I..'

Syni.
GlyL 2

..

Q\

I
I

,., - - - -

I
I

0
12Elx
. ------------1
-----2ii;i}
0
6ElxL I 4ElxL 2
- GI y L 2
0
I
0
0
- 12/, 1-61,L
6El,L
0
6ElxL

QC>

::,

.!:/1

....
,,l

GlyL 2

12W,

Noh thul for lhis type of me mber the cross-sectiona l properties h11 v1 11
Ny 111hols lwtnusc of lhc orientation of the memhcr uxis . T hut is, / 1, is 1111w th
1t11 slo11 rn11s1u11l , whlk / 1 is the strnnd 111011wnt ol' the cmss Sl'<:lio,wl 111111 wit
11'Nlll'l'I lo till' 1111 uxis

430

Rigid Bodies within Flexible Structures

oc. 9 .4

Figure 9.8(c) depicts a typical column with its member axis z,,, p11111II
the z reference axis. Its principal bending axes x,,, and y111 are assu,mil I
parallel to axes x and y. Action and displacement indexes at joints j u11il A
numbered in a sequence that expedites the process of transferring co11l11h111
terms from the member stiffness matrix to joint and floor stiffness matrkr~ 11
is , indexes j I, }2, and }3 at the top of the column correspond to joinl dr~pl
ments that are not directly associated with the rigid-body motions of tlu
above. These displacements are the same types as those discussed previou~I
the beams. Similarly, indexes kl, k2, and k3 at the bottom of the colun111 p ti
lo joint displacements not directly associated with the rigid-body motio1111111t
floor below. However, the indexes j 4, j 5, and j 6 at the top and k 4, k 5, 111111
ul the bottom correspond to displacements that are directly dependt111 11
rigid-body motions of the floors above and below. They consist of Lru11~l11II
in the x and y directions and a rotation in the z sense at both levels j 111111 A
We can represent the 12 x 12 stiffness matrix for a column as a p111 1l1 hll
urray composed of 3 x 3 submatrices, as follows:

K;

J_;;_{~K_~L-[Ki, I I
K31

K11 = K 22 =

Kt3

_E

K,,

!'his equation is similar to Eq. (9. 2-1)' but point F replaces point p. In this case
1he vector Ai contains

Ai

= {Aj4,

(6)

Ajs, Aj6}

4~

2/.,

2;;
- ly

_ 6E
K42 - /} I,

ro

K11

I
f

(a)

()

KJ,

lhe x,,,;::t~~:ex;:~sence of the rigid laminae, certain geometric tr~nsforfmationhs

.
. t . and c to those at pomt F or eac
ure required to relate the motions at p~m s J
. t . and their
lraming level Figure 9.9(a) shows actionsAj4, Ajs, and Ai6 at pom J .
.
. 1ents AFI, An, and A F3 at point F. Values of the latter actions are
Nlat1cal
eqmva
n1lculated from the former' as follows:
(5)
AF = TFjAj

4fx

K2, - I [

K,11

For this member the cross-sectional properties are defined as follow.s:hA, = art:
.
t t and I x and I y = second moments of area wit respec o
/ = torsion cons an ,
1

Sym.

111 !his matrix subscript 1 denotes action and displacement indexes of lyp,
./2, und }3. Subscript 2 represents indexes of types k I, k2, and k3. S11hN1 111
Nlnnds for indexesj4,j5, andj6. And subscript4 is for indexes k4, k5, 111111
Note that the stiffness matrix in Eq. (3) is also partitioned to scp11111h I
1111to11strained displacements at level j from those at level k, as well as I111111 I
tonstrnined displacements at both levels. The submatrices in Eq. (3) 1111

K,i

431

K32 I K 33

! K42

K11

Rigid Laminae in Multistory Buildings

()

~l

I JW,,

()

()

I JFI,

()

(I

()

<;/ I '

(IJ)

li'IMIII ,, 11 ,11 I h 111111 II II ll llll " l11111111IIIIII N (II) 111111111 11l11l l1111MhlpN, (t,) d lNpl1111111111nl
1rl11lh111ijhlp~

Sec. 9.4

Rigid Bodies within Flexible Structures

432

which has the same meaning as Eq. (9.2-22), except that the subscript F replaces
8. The results of Eq. (14) in expanded form become

Vector AF is
AF= {AFI, AF2, An}

_!_(~~-~---- - -

and the transformation operator TFJ has the form

T,1 - [

K Fi

Dj

_L ,!, ~]
= TJ:jDF

Again , this equation is similar to Eq. (9.2-6), but with point F replacing poh
p . Here the vector DF contains the terms

TJ:, = [~

TFk K42

TFk ~

TT

Fk

(16)

Mee= me O 1
(

0 0

where me is the mass at that level. Furthermore, the mass-inertia matrix with

respect to point Fis obtained as

tI

0
1

0 0
which is the transpose of the matrix in Eq. (8).
Consideri?g no:"' the stiffness matrix for the column in Eq. (3), w1 11111
[runs.form certa1? of its submatrices to rigid-body coordinates. This opir ntlt
n.:qu'.rcs a combmed action-displacement transformation that makes usl' 111 ti
11111tnx
Sym.l
tI

T,.'k
'l'h1 tr:u11,1'0 1111ation matri x 'I; conlains f'our 3 X 3 suh111atric1:s on till' d111~1111 I
of whrch thl' first 1111d Nl'tond 11r1: id1:ntit y mntriccs. The third nnd fo11r11l ul
111atri11s lltl' ol till' typt giVl'n by Eq . (8), 11nd thtir suhs<:tipts dt noh' j11lt11 I
th1 11001 11h11Vl' 1111d joint ~ in tilt flow lwlow INt't' Fii:. <) H(c)I , 'l'lll' ,h ~11
nl 1hr

(15)

1 0

t 11

Also, the transformation operator becomes

- - - - - -

TFJK33 TJ:1
I
! TFk ~ 3 T TFj
I

und lhe vector Di is

= {D,4, DJ 5 , DJ6}

l TFJ K 32

Sym.

Note that the transformed stiffness matrix KFi is still partitioned in the manner
indicated for K, in Eq. (3) . When the submatrices in Eq. (15) are evaluated, we
find that the equalities among them are the same as those given in Eqs. (4),
provided that reference points at all levels lie on the z axis.
In lieu of more detailed information, we assume that the mass of the
building tributary to a given framing level is uniformly distributed over the area
of the lamina. In that case the center of mass c will coincide with the centroid
of the area, and every floor will have the same radius of gyration re with respect
lo point c. Thus, the mass and mass-moment-of-inertia matrix with respect to
point c for the lamina at level may be written as

t Ill

DJ

= _ K 21 -+-- K22 - ~-[ TFJ K31

TFk K41

~u~ing next to .displacements, we see in Fig . 9.9(b) the effect at polnl

of a n ~1d-body rotation DF 3 at point F This rotation causes a negativl'
lran.slation ~nd a positive y-translation at point j. Hence, the kinemuti1 1111
equivalent displacements at j due to rigid-body motions of point F are

l11111 11 lll111111f11111

433

Rigid Laminae in Multistory Buildings

n1l111111111frll11110111111f11x tH'

(17)

This transformation is of the same type as that in Eq. (9.2-30). The symbol rF
In Eq. (17) denotes the radius of gyration of the floor with respect to point F. It
Is related to that with respect to point c by
(18)
which is a familiar expression for translation of axes from elementary dynamics [4).

Horizontal forces P.,(t) and Py(t), applied at a typical framing level, appear
In Fig. 9.7. ln general, these forces are eccentric with respect to the reference
point F, and the stalically equivalent actions at that point are
A,:(t)r

T,..,,A,, (t) r

(19)

In whkh
A,,(/l1 (/',(I) , /\(()\,

(10)

434

Rigid Bodies within Flexible Structures

Sec. 9.5

435

Rigid Bodies in Finite-Element Networks

and

When the forcing influences are horizontal ground accelerations, the resul11111
inertial forces at level are

A 8 (t)e = -me{D8 1(t), D82(t)}

where D81 (t) and D82(t) are accelerations of ground in the x and y directions . I h
forces in Eq. (22) act through the mass center at each story. Therefore, llwy n
eccentric with respect to point F by the distances XFc and YFc (see Fig. 9. 7), 1111
these terms replace xFP and YFP in Eq. (21).
The transformation relationships discussed above prove useful frn lh
purpose of writing equations of motion for a tier building . Our method h
handling such equations will be described in Chapter 10 when we study lh
modified tridiagonal method for dynamic analysis by substructures.

)-,

9.5 RIGID BODIES IN FINITE-ELEMENT NETWORKS

I

In this section we shall consider rigid bodies that are embedded wllhll
linite-element networks of the types described in Chapters 7 and 8. ln NIH h
situation, we cannot avoid having more than one connection of an elenw111 111
particular rigid body. Therefore, the member-oriented approach for 1111111
structures (see Sec. 9.2) will be abandoned in favor of a body-oriented 111,11,,,
Figure 9.10 illustrates an analytical model consisting of rigid 111111111
connected by Q8 elements for plane stress or plane strain. By omitting till' 11111
bodies temporarily and excluding support displacements, we can write equ11111111
of undamped motion for the nodal displacements in the finite-element 1w1w111~
as fo llows:

[DA]

MM MAF]
[SM
[ Mfoi\ MFF Op + SFA

[DA] = [AA
(t)]
Ap(t)

SAF]
SFF DF

II

In this equation the subscript A refers to nodes attached to rigid bocli111, 111
subscript F denotes free nodes (see Fig. 9.10), at which the displact1111111 ~ 11
indcpcndcnl. For an arbitrary sequence of numbering nodes , the lcr 11111 11111
l1ihuti11g lo Eq. (I) must be rearranged to put type A first and typcF st.Korul N11
ll't us hri ng in the rigid bodies and recogni ze that nodes of type A arc <:0111111,1111
lo lllOVl' with the m. Thus, we need lo trunsform actions, stiffnesses, 1111d 11111 I
knt 111aSSl'S frrnn nodl'S A to wol'king points on lhc rigid bodies. For lhiHp111p11
Wl' dHll)Nl' lh, IIIUNH l' l'llll'IS IIS lilt WOl'kinp poinls.
llip1111 1J Io 111tl1c1111H rh, 11111N.~ ,1111,1 , 1 of hody /.. 1111d 11 1yp1111 I 111111
wh111 l'111111111, 111 1111111l11d l1uk11111 /..I ,/.. ' , 1111d /.. I al p111111 , 1 d11111h I

Figure 9.10

Rigid laminae connected by QB elements.

l'igid-body motions, which consist of ~ransl~tions in ~hex an~ y directions and

rotation in the z sense. At node J the mdexes J 1 and J2 repr~sent ~nly
lranslations in the x and y directions (for a Q8 element). Correspondmg actl~ns
nt node j are forces in the x and y directions, and their statical equivalents at pomt
,.k become
11

(2)

which is similar to Eq. (9.2-1), but point ck replaces pointp. Actions in vectors
Aj and Ak are

(3)
In the latter vector we have forces in the x and y directions and a moment in the
sense. The transformation operator in Eq. (2) is

(4)
which hus tlw
I'

sa111t

Hq 1111111 11 1 (

1)

ll

frn 111
ll

11N

that for plane trusses in Table 9. I , with k replacing

I ic 1111 I y

Io ti w / 111 11odt 011 I ht /.. 111 11 p1d hod y,

11111 I

II II II 11odt~

Sec. 9.5

Rig id Bodies w ithin Flexib le Structures

436

437

Rigid Bodies in Finite-Element Networks

attached to the body are considered, the equation becomes

DA1l
DA2 = [TI1l
Th D
[~:.:' ~i, "

Ank = TkAAk
The expanded form of Eq. (5) is

(12)

When all rigid bodies are taken into account, we have

(t\)

Here the vector D8 consists of displacements at mass centers for all of the rigid
contains the dependent displacements at all of the attached nodes.
bodies, and
Of course, T'};A is the transpose of matrix ToA in Eq. (8) .
Now let us return to Eq. (1), consisting of the equations of motion for
attached and free nodes, devoid of contributions from rigid bodies. To account
for the presence of the rigid bodies, we form a transformation operator T , as

DA

AAn1 k
where nj is the number of nodes on the body. Each subvector in AAk contains tw
~crms, as in the first of Eqs. (3); and each submatrix in Tk is a 3 x 2 arniy , 11
111 Eq. (4).
Considering now all of the rigid bodies in the system, we express the nrt h1n
trunsformation relationships by

follows:

An= TnAAA

T =

l II th is equation the vector AA contains actions for all of the attached nodes, 1111 I
A,, consists of their statical equivalents at mass centers for all of the r igid hod I"
'l'ht transformation matrix T8 A in Eq. (7) is a large, sparse array contuinln
1nibmatrices of type Tk in diagonal positions, as follows:

TBA=

(13)

DA= TlADB

[T~A

:J

(14)

in which IF is an identity matrix of order equal to the number of free nodal

displacements . Th~n we have
(15a)

Sym.

T1
0

T2

Tk

Also,

'"

(15b)

Therefore , premultiplication of Eq. (1) by T and use of the relationships in Eqs.

Tnb

( 15) produces

where 11,, is the number of bodies in the system.

Similarly, the kinematically equivalent displacements at node j in 11,
'). 10 may be calculated from those at point ck by the formula

DJ = TiDk

whh.: h is of the same form as Eq. (9.2-6). In this case the displacement

Vl'l

M 8 n MnF J
[ MFB MFF

[Dn]
[Snn SnFJ [Do] _ [Ao(t)J
DF + SFB SFi DF - AF(t)

(16)

(II

Terms with the subscript B in this equation refer to motions of the rigid bodies
in the system. Block stiffness and mass submatrices in Eq. (16) are

1111

(17a)

llll'

t 111

11 1111 nodes on lhc kth ri gid body arc included , the tru11sf'orrnatio11 in
ht'l'Ol lll'S

h(

and
(17b)

1U

llccnusc the nrntri x 1'1111 has the form shown in Eq . (8), we can express Eqs. (17)
IIIIH l'

1111 whi1 h 1111' 1,i p11rnh-d 1111111 i11

('Xplil'itly

llH

S
11111

'I'PS IU , 1'l' ft

S1111

S/m

' liS~, 1

( I 811)

438

Sec. 9.6

Rigid Bodies within Flexible Structures

and
(18h)

fork = 1, 2, .. . , nb and C = 1, 2, .. , nb, Matrices Snnk. e and SMk, e in Eq.,

( J 8a) are submatrices of S88 and SM that contain terms coupling body k wi lh
body C, and so on.
To complete the equations of rigid-body motion , we need only add 1hr
mass-inertia matrices for the bodies to matrix M 88 , as follows:
nb

M;n

= Mnn + I

Mck

( 111)

Program DYRBPB for Rigid Bodies in Plate-Bending Continua

Sec. 8.3 (see Tables 8.1 and 8.2) are needed to form action, stiffness, and
consistent-mass matrices for all nodes not connected to supports. We also
introduce information about the rigid bodies that enables the computer to distinguish attached nodes of type A from free nodes of type F . Then the desired
matrices may be generated in the partitioned form shown by Eq. (9.5-1).
As mentioned previously, the sequence for numbering displacements at a
node of the PBQ8 element is z-translation, x-rotation, and y-rotation. We also
use the same sequence of displacements at the mass center ck of the kth rigid
body in the analytical model. Therefore, the transformation matrix for corresponding actions at point ck due to unit actions at node j becomes

k= I

where Mck is drawn from Table 9.2. For a typical two-dimensional body k in Fill
9. 10, the matrix Mq is the same as that for plane trusses and plane frames giwn
in the table.
We shall now discuss briefly other types of rigid bodies in other types 111
discretized continua. If we have three-dimensional rigid bodies embedded in 11
network of H8 or H20 elements, the transformation matrix Tkj becomes the s1111111
ns that for space trusses in Table 9.1, except that p is replaced by k . Also , lh
11u1ss-inertia matrix Mck for a typical rigid body is the same as that for spm
lrusscs and space frames in Table 9.2 . Otherwise, the process of settin~ up
tquations of undamped motion remains the same as that described abow 1111
two-dimensional continua, but numbers of actions and displacements art 111
ncased.
Rigid bodies supported by discretized plates in bending are analyzed 111 11
11111nncr analogous to that for grids, except that the theory is body-oril111111
instead of member-oriented. Also, the sequence of nodal displacements for th
plate-bending element is not the same as for the grid element. As shown i11 hj
(8.2 I), the z-translation W; is taken before the x- and y-rotations Ox; and IJ,,
Three-dimensional rigid bodies connected by networks of general 1,lu II
ckmcnts have characteristics similar to their counterparts in space f'r111111 11
1lowcvcr, the rotations in the x, y, and z senses at an attached node are ac11111II
dlpcndcnt components of the independent tangential rotations a; and {31 (s1.t I 11i1
8 .() ).

Finally, rigid rings in axisymmetric solids or shells with axisy1111111'lll

lo11ds would be restrained to translate only in the z direction. If loud'I 111
11111111xisymmctric, such rings can also translate and rotate in the r and () dlll'l
lions und rotate in the z sense .

9.8 PROGRAM DYRBPB FOR RIGID BODIES IN PLATE-BENDING

CONTINUA

'1'11 illusll'lllt tlw 1111lun or prog rumming for Ila hody rn itnll'd Hppmud1 , w,.,1111
dl Nl'IISN Prn~ tlllll l>YRI\PII 1111' dynun1il' 11111il ys 1~ 1)1 li 1ii d hodi,N s11pp1111id I
pl111111 Ill lll'111ll11p 1"01 lhl N p11tpUHI', till' Nltlldll111l 1111d d y 1111111ic loud d11111 1111

439

T kj

= [ :~
-xkj

~]

(1)

O 1

This array is taken from the third, fourth, and fifth rows and columns of the last
matrix in Table 9 .1, with p replaced by k.
Similarly, the mass-inertia matrix for rigid body kin a discretized plate is

M,, -

[I ~;~ ~!~l

(2)

which is drawn from the third, fourth , and fifth rows and columns of the last
matrix in Table 9.2.
Supplementary rigid-body data required for Program DYRBPB appear in
Table 9.3, which conveys the number of bodies NB in part (a). Then in part (b)
we have NB lines of body-node data containing the body number K, the number
of nodes NJ(K) attached to the body, and the body-node numbers JB(K, 1)
through JB(K, NJ(K)) . Properties of the rigid bodies are given in part (c), wh~ch
indicates NB lines with the body number K, followed by values of the followmg
six terms:
XC(K) = x-coordinate of point ck
YC(K)

= y-coordinate of point ck

BM(K) = mass of body k

XXI(K) = mass moment of inertia l:o:k
XYI(K)

= mass

YYI(K)

= mass moment of inertia l yyk

product of inertia

lxyk

We also need supplementary dynamic-load data for Program DYRBPB , as

shown in Table 9.4. Information contained in this table consists of initial condilions and applied uctions for the rigid bodies. In part (a) of initial conditions we
huvc the 1111111lll'r of' bodies with initial displacements NBID and the number of
hodics with i1111111I w lul'llitH NBIV. In part (b) arc NA()) lines with the body
1111111h11 K 1111d lh11, po10,1hll' 11111i11l displ11l1111,11ls 1)1\0( 'K 2), DB0(1K I ). und

Rigid Bodies within Flexible Structures

440

Chap

TABLE 9.3 Rigid-Body Data for Program DYRBPB

Type of Data
Rigid-body data
(u) Number of bodies
(b) Body nodes
(c) Body properties

No. of
Lines

Items on Data Lines

I
NB
NB

NB
K, NJ(K), JB(K, I), JB(K, 2), ... , JB(K, NJ(K))
K, XC(K), YC(K), BM(K), XXl(K) , XYl(K), YY((KI

"Supplements (and follows) structural data in Table 8.1.

TABLE 9.4 Rigid-Body Dynamic-Load Data for Program DYRBPB"

No. of
Lines

Items on Data Lines

lnitiul conditions
(u) Condition parameters
(b) Displacements
(c) Velocities

I
NBID
NBIV

NBID, NBIV
K, DB0(3K-2), DB0(3K-1), DBO(.IK I
K, VB0(3K-2), VB0(3K-l), VBO( IKJ

Applied actions
(u) Load parameter
(b) Rigid-body loads

I
NLB

NLB
K, AB(3K-2), AB(3K-1), AB(3K)

Type of Data

11

Supple ments (and precedes) dynamic-load data in Table 8.2.

DII0(3 K). Also, in part (c) we see NBIV lines containing Kand three poN~lhl

initial velocities VB0(3K-2), VB0(3K-1), and VB0(3K). Because thesl' i11ltlAI

rnnditions are given for rigid bodies, their effects on attached nodes nuo,t h
to111putcd within the program and are not included as dynamic-load data 1111 th
Ntrncturc .
Actions applied directly to the rigid bodies are listed as the second lypr or
d11tu in Table 9.4. The only parameter required in part (a) is the number of 1111111 d
hod ii.ls NLB . In part (b) the data for rigid-body loads consist of the body 1H11nh
K untl applied actions AB(3K- 2), AB(3K- l), and AB(3K). This forcl.l und th
lwo nwmcnt components are assumed to act at the mass center ck, as drn " th
111L11tiul l'orcc in the z direction caused by ground acceleration. Of course, 11\'I 11111
11t 1111y other points on a body could also be handled, but they would requi11 dnl
Im lot'11tio11s as well as magnitudes.
As in Program DYNAPB, the rotational displacements at free nod1111 11
tli11il11atud by Guyan reduction . After solving the equations of motion 1111 rt.aid
hodils und f'rcl.l nodes, we can find time histories of displacements for 1111111 h I
11odts with Eq . (9.5 13). Then the displacements in vectors 0 11 and D, 11111y I
11Nld to obtain ti111u vmying stresses in the finite clements.
1,:,ui111pfo

'> ..,

ll1t11r 'I 11 d1ph IN hull ol II " y11111111tdr Hlit1rl1111, \'IIII NiH(lllU ol II q11111h11 111 11 -1111
, 111111111 dlH~ H11pp,111rd hy II ph111 111 lw11dt111l 1111 pl111t ih d1vid1d 111111 two I'll

442

Rigid Bodies with in Flexible Structures

elements that are fixed along edge 3-5-8 and symmetric with respect to a plane co111111n
ing edge 8- 10-13. A step force of magnitude P, = 0.1 kN acts in the z direction 111 th
mass center of the first rigid body, which is labeled point c 1 in the figure. Phy~1111I
properties of the plate are

E = 103 x 106 kN/m 2

= 0 .34

= 0.05 m

= l = 0.05 m

0.1s

w
a o.
0
z

= 8 .66 Mg/m 3

<

h = 0.002 m

h1

=5 =

[ 2(1

I xy =

16)
m,-Rf(!
1r 2 - 91r

0.01 m

.,

:c

hf] =3112X
10 -8 Mgm 2
.

"' -10

w
a
0
z

= -8.913 X 10- 9 Mgm 2

0 .2 0

o. 15
fl
Q

0 , 10

0,05
0 00

Q..

+12

We used this data in Program DYRBPB with DAMPR = 0.05 and direct 1111111 r
icul integration by Subprogram NUMINT. For this purpose, half the load was applll,I In
l'nch of two cases to find symmetric and antisymmetric responses. For the sy1111111 t1I
l'Use, restraints against x -rotations are required at nodes 10 and 13. However, 1111 th
unlisymmctric case, we need restraints against z-translations and y-rotations 111 1h11
nodes. Figure 9. l2(a) shows computer plots of the resulting z-translations at poi111, 11n
1
lhl' rig id body for the symmetric and antisymmetric cases. In part (b) of the fig1111,
111.~o sec the z-translations of node I for the two cases. Finally, in part (c) 1111 11111
hi stories o f the flexural stress SX at the integration point near node 3. Of course, lhl IW
1111 vcs in each of the plots must be added to obtain total values for the responsck 111v II

0 . 05

-, 0. 00

a::
<

-2 0

w
z

whe re the symbols R, , h 1 , and m, denote the radius, thickness, and mass of the qu111 t
disk .

ID

10

1rRf
m, = ph,- = 1.700 X 10- 4 Mg
4
16)
I.u =IYY =m, R, -2
4 -97T

0.20

and both the plate and the solid are made of brass. The rigid body is attached to 110,I
I , 4 , 6, 9, and 11, and its properties are as follows:

R,

443

References

Chap. 9.

X -30
U)

( C)

Figure 9.12

REFERENCES
W
W Jr "Dynamics of Elastically-Connected Rigid Bodies," in Develeaver,
. , .,
.
d w A Shaw Pergamon
opments in Theoretical and Applied Mechanics, Vo1. 3 , e
,
Press, New York, 1967, PP 543-56Z.
of Framed Structures, 2nd. ed. ,
.
2. Weaver, W ., Jr., and Gere , J . M ., Matrix Analysis

!.

Van Nostrand Reinhold, New York, 1980.

.
3. Weaver , W . , Jr. , and Johnston , p R , Finite Elements for Structural Analysis,
Prentice-Hall, Englewood Cliffs, N.J. , 1984.
.
.
4 . Beer, F. P., and Johnston, E . R. , Jr. , Vector Mechanics for Engineers: Dynamics,
4th ed., McGraw-Hill, New York , 1984.
.
.
.
Analysis
5. Weaver, W. , Jr., and Ne1son , M . F . , "Three-Dimensional
6
385-404of Tier Bu!ldings " ASCE J . Struct. Div., Vol. 92, No. ST6 , 196 ' pp.
.
. .
.
6. Wea,ver, W . , Jr., Ne lson , M . ~-, and Manning , ~ - A. , "Dynam1ct4~~~:~7:_u1ld.1ngs,
. " A<'(
Ml'C"h. Dtv. Vol. 94, No. EM6, 1968, PP
o} ' /
" , t , F,"'
,
("i

"''""l'I' '),

ll lt1111d hmly 111 11 pl1111 (11) hrnly fl1111Nllllh1IIN, (h) 11nd11l 111111Nl111 11111N, 11)
111~111111 ~l11on

(cont.)

'

10

Sec. 10.2

445

Guyan Reduction Methods

y

Substructure
Methods

10.1 INTRODUCTION

When the number of degrees of freedom in a structure becomes very lar!,w, w

1u;cd to divide the analytical model into substructures. Figure 10.1 ill ust1111
such a case, which is a computer plot of a large radio telescope antenna desi~n I
as u space truss. This type of structure consists of a reflector and a s11pp11rl
structure. We can take advantage of the facts that the reflector has two pl1111
or symmetry and the support has one such plane. Therefore, we need 111111ly
only a quarter of the former and half of the latter, using appropriate restrni111- 1111
tht planes of symmetry. Also, the more complicated reflector can be dividld 1111
substructures and analyzed by one of the methods in this chapter.
For analysis by substructures, we must distinguish techniques thul 111
1rn ituhlc for statics from those more conducive to dynamics. In static 111111ly~I
1111 nodal displacements for a substructure can be eliminated from equl lth1111111
tq1111tions during a frontal reduction procedure [1, 21. However, in dy1111111I
111111lysis, we need to retain a certain number of scattered degrees of frccdrn11 with
low stiffnesses and high inertial actions, for which approximate cq11nl11111~ 1
111011011 cnn be written. Three approaches that work well for dynamks 1111 th
1i1di11gonul 11wlhod (with modificutions), tht parnllcl elimination lllllhcul , 1111 I
lht to1111HH1t111 111mk ll't'hnique . Euch ol' these suhslnic1111c 111c1hods will I
d('Sl'iilwd in till' l'IISlllll~ Sl'l'liOIIS or thi s d111pl11.

Figure 10.1

Radio telescope antenna structure (pointing to zenith).

10.2 GUYAN REDUCTION METHODS

Tridiagonal Method for Substructures in Series

Pigurc 10.2 shows n two-dimensional discretized continuum divi?ed into

s1thstruclurts 1h111111l l'lllllll't'lcd in series. Suhstructurc numbers appear in ~oxes
lwlnw till' llg111t 111,11 1111 My111hol (1 dl'IHlll'S II lypil'ul s11hstruct11re, 1111d 11, 1s lh(:

446

Substructure Methods

B,

Chap, 11)

Sec. 10.2

Guyan Reduction Methods

447

B,

(4)

Now we will apply Guyan reduction (see Sec. 6. 7) to decrease the number of
degrees of freedom in the t'th substructure. For this purpose, let the displacements of type Ae be dependent upon those of types Be and Be+i Thus,
(5a)
Figure 10.2 Substructures in series.

in which

number of substructures. For each substructure we have interior nodes of lyp,

A (or Ae) an~ i~sulating boundary nodes of types Be and Be+ 1 on each side. Tiu
sy~bol Be md1cates boundary nodes common to substructures
1 and /
while Be+1 represents those joining substructures and + 1. Let us form 11 11
column vector of nodal displacements as

e-

(5b)
Equations (5b) are of the same form as Eq. (6. 7-7b). The acceleration relationship similar to Eq. (5a) is

..

DAe

{on

I'

DA p Da2, DA2, .

, DAn,' Dsn,+}

( 11

..
= TAeBeDBe +

..

TAeB e+1 Dne+,

(5c)

To reduce the equations of motion in Eq. (3) to a smaller set, we construct the
transformation matrix

With this sequence, the stiffness matrix for all nodal degrees
of frccd11111
h1..comes

(6)
Sa,s 1 Sn,A,
SA1B1

SA1A1

SA,82

S82A1

SB2B2

S82A2

SA2B2

SA2A2

S=

Then we have
( 'I

SAnsA"s

SA"sBns+I

Sn11s+ IA11s

Sn,,s+1B,1sf

(7a)
and

m~t1:ix Sis a tridiagonal array of submatrices, this approach is rdr 11111

lo ns. lhc tridiagonal method. Without writing it, we can also observe tlllll 11 11
t1111s1stcnt mass matrix has the same tridiagonal form.
Suhstructurc .e contributes terms to the equations of motion as follow

MeDe + SeDe
111 l'Xpaudcd form, thi s equation is

Mu,n,
M, ,11,
0

M11 1A1
M,1,1,
M11 111 11

= Ae(t)

(7b)

De=

lll1..11usc

Substitution of Eqs. (7) into Eq. (3) and premultiplication of the result by TJ
produces

( 11
(8)
In this equation the reduced matrices are

St

'I'~ Sr'I}

S
' "'
11 rll r

Ist,,

111,

(9a)

448

Substructure Methods

Chap. 10
Sec. 10.2

A i (t) =

TI A e(t) = [

~;e
ABe+1

Guyan Reduction Methods

449

(9h)

Example 10.1

(9tl

The unrestrained beam in Fig. 10.3 is divided into two substructures, each consist!ng of
two equal prismatic flexural elements. Let us determine the coefficient matrices s;B and
M!8 by the tridiagonal method, as required in Eq. (11).

Modified submatrices in Eqs . (9) have the definitions

s;eBe = SBeBe

+ SB eAe 'I:i eBe

MrieBe = M BeBe + T Jen eMA eBe

( 1011)

+ MB eAeTAeBe + T JeneMA eAeTAeBe (!Oh)

A;e = Ane + T JeBe A At

(!Ot)

11nd so on.
Finally, the reduced equations of motion for all of the boundary nodes tu~,
the form
(I l l

Substructures contribute to the matrices in Eq. (11) by the usual direct assenihl v
proccss. The reduced stiffness matrix is still a tridiagonal array of submatrit,,11
ns follows:

Figure 10.3 Beam divided into substructures.

For substructure 1, the stiffness matrix in Eqs. (3) and (4) becomes

6
3L 2L2 i
-------t--- 6 - 3L i 12

0
( I I)

L2

0
Subrnatrices of S!n having single asterisks receive contributions from onl y 11111
.~ubstructu re, but those with double asterisks have two contributing 1rnh
Nltuctures. The assembled mass matrix M%
8 in Eq. (11) has the same tridia111111,1l
u1111 11gcmcnt of submatrices as the assembled stiffness matrix in Eq. (12) . S1111
ll 11rl y, the assembled load vector is

A*()lJ t - {A*
BI' A**
8 2 , A**
83 ,

A***
Dn,, An,,,+, }

I
I

- 6

-3L

l 3L

I
I

! - 3L

2L2

Terms in this array are drawn from Eq. (3 .4-24) . Similarly , the consistent-mass matrix
for substructure I is
156

22L

4L2 i

--- - - - -- - -l- - -- -

l 312 .
l O 8L i
- - - - - - - - - ~ - - - - - - - - -~-- - 0
0 l 54
13L i 156
0
0
- 13L -3L l - 22L

( l II

wli il'h aga in has single and double contributions. For convenience in co111p11h 1
pr ogrnmming, we assemble the reduced matrices while proceeding fro111 11111
1Hil11,t1uctun..: lo the next, thereby increasing the number of type B di splrn:t111111t
lk cuusc the assembled stiffness and mass matrices both have It idl11v11111il
form~. vibn11 ionnl and dynamic analyses can be more efficient 131than 101 111111 1
111t thods wlwrt the rn111t iccs arc filled. The l'o llowing example dl'111<11111l11111
Hppl lt-a1lo11 ol lhis npproul' h to II lwum 1ypt ol st111tt111t.

(a)

l O 4L
- --- - --T------- 1--3L

54

13L

- 13L

- 3L2

(b)

4L2

in which the terms arc obtained from Eq. (3 .4-26).

Next, we use Eqs. (Sb) to fo rm TA tne and T AcD e+ i for substructure 1, as follows:
'I\

1/11

1
S ,1 1fl I S ,1 1/1 1

"'

21

2/,

/,

(c)

450

Substructure Methods

TA 1B2 = -S;iA I SA 1B 2 =

IO

-L

Sec. 10.2

.------~------------/ 78

-L

With these submatrices, the transformation operator in Eq. (6) may be written as

T I --

M* _ pAL
BB 105

lo O
I
B,
_9___~-L~___Q_.
O

J _

In 1

TA1B1

TA1B2

In 2

= 4L

L2

2L

l 2L
-L 1 3

,:-3

- L2
-L

- -- -- -i---- - - - - -

'

0 / 0

4L

(i)

EI 3L 4L2 I
---------f---- Sym.]:'
2L3 -3 -3L I 3

SI - TI S1 Tl = -

2L2

3L

-3L

4L2

0
0

\ 27

13L

i- 13L _ -6L 2 _ -

(j)

78
\
22L __ 8L 2 \

B,

Modified Tridiagonal Method

( fl

Sym.

1,

8L

which may now be used in Eq . ( 11).

3
T

Then the reduced stiffness matrix becomes

*-

22L

0
0

4L

[I]

r------ --,----------1
I 27
13L \ 156
I
\
2
2
II______
-13L -6L
16L
II] II
____ TI _ _O____
__ _ I
I
I

4L

451

Guyan Reduction Methods

and

__!_[ 2 L
4L

Chap.

Now we consider a modification of the tridiagonal method for substructures in series that abandons the idea of insulating boundary nodes. Instead,
there are three definitions for types of displacements, as follows: (1) Subscript
Ae implies dependent displacements in substructure C; (2) subscript Be represents
dependent displacements at boundary nodes joining substructures C and C+ 1;
and (3) subscript Ce denotes independent displacements associated with substructures 1 through C, including boundary nodes common to C and C+ 1. The
dependent displacements of types A and Bare to be eliminated, but the independent displacements of type C will be retained. Note that the number of type C
displacements increases for each consecutive substructure. Also, they may be
scattered within the substructures as well as at boundary nodes.
At each stage in the reduction process, the partially-formed equations of
motion are symbolically the same as in Eq. (3). Thus,

MeDe + SeDe = Ae(t)

(14)

However, the expanded form is now

f."or substructure

2, we can proceed in a similar manner and find that

s;

st

Ml= Mi'

< \,11slquc11!1y, assembling the reduced stiffness and mass matrices f O

b
111111 J yields
'
r m su st ruct1111 I

------------

,- 3

I/ 3L
s,~"

m
2/,'

OJ

4L2

1-1/,3

I
2//
3L

()

()

()

()

fl,

8/~-i---

I/

,I '

Sym.

fJ I

[II

IJ 2

,I/ '

II I

fl

/11

(15)

Men

Mee

De e

ScA

Let the displacements of types Ae be dependent on those of types Be and Ce, as

follows:
(16a)
where

I/
I/

:::J[;:J [t
+

(hi

11)

TA1llr
Also,

Wl'

= s;;,1 ,St1 r1J1

Ti1,cr

=-

s;~A eSAtCt

(16b)

( I Cw)

huvl' llw lll'l'l'il'l'Hlion rclnlionship

If

ll1,

II

'l\1111>11,

'I\,, 11),

452

Substructure Methods

Chap. 10

Then the transformation operator becomes

Tc=[~: T~cJ
0

le

( If I
c

With this matrix , the displacement and acceleration transformations are

( l Hnl
und
De=

[2:J =
..
De e

Te

[~BJ
De e

C, the mass matrix is diagonal, as in the next example. For such a cas~,
transformation of the associated eigenvalue problem to standard, symmetnc
form should involve factoring the mass matrix instead of the stiffness matrix.
Also, an efficient overlay technique for stiffness terms is demonstrated in the
following example (see also Sec. 10.3).
Example 10.2
Figure 10.4 shows a rectangular plane frame divided into _two substructures, each of
which consists of a beam and the two columns below. For this structure we shall find the
reduced stiffness matrix in Eq. (21) by the modified tridiagonal method , using the
following assumptions . Tributary masses having values of m and 2m are lumped at the
two framing levels, as indicated in the figure. Flexural rigidities of beams 1 and 4 are
equal to 2E/ and 4E/, while those for columns 2, 3, 5, and 6 are equal to El. Members
are all prismatic , and axial strains are to be neglected.
y

( IHhl

Substituting Eqs. (18) into Eq. (14) and premultiplication of the equation hy Tl
yidds

Me*

[Dn]
..
+ Sc* [Dn] = Ae*(t)
c
c
De

De

st = TISe'fe = [s!B s:c]

Ste e

Sen

TIMe'fe = [M! M%c]

McB Mtc e
At(t) = TIAe(t) = [A!]
Mi

T7

( IV)

QJ

'l'hc l'educed matrices in this equation have the definitions

CD

-- -Da,

L
z

* Dc
.. + Sfc Dc = A7,(t)J,,,
[M,c

/4

- oc,

Da,

[I]

Modified submatrices in Eqs. (20) are the same as those given by Eqs . ( IOI
t'Xt'llpt that subscript Ce replaces Be+ 1
Wl~en proceeding from one substructure to the next in a series, displ111 1
111111(s of type Be become a subset within those of type Ae+1. Also, for lht In 1
suhslrncture, displacements of type B,,, are usually omitted. Thus, tht 1111111
1q1111tions of' motion become
l 'II
which invol ves displacement s of' type C for the whole structure .
111 the nwdilkd tridiu1,onal muthod , cli111i11ution (or release) of du,pli111
1111111111 ol hotli 1y1ws I\ 1111d II l'llllHl'S s7, 1111tl M1, in Hq . (21) 10 ht IIJlc d
I lt1Wl'Vl'I , ii l1111qwtl lllllNSt'H 1111' II Mttl ill ('lllljlllll'llo11 wllh lll'l't' ltmliot1N 111 l\' 11

-oci

---

(.ltlhl

DA,

(.!OnI

Ac e

453

Guyan Reduction Methods

Sec. 10.2

/ 5

/ I

/ 6

2L

.1

Figure 10.4 Plane frame divided into substructures .

Fm suhNhul'1 1111 I , w1 Sl'I up the 6 x 6 stil'f1wss 111ntrix in Eqs. {14) und ( 15), as

lollows:

4114

li111J1 1r1u.h1111 M11thrnh1

1

,110.

10.2

(11 1y1111

llnd111.. tlot1 Molhoda

455

,I/

S,

WI

-u

Syn1. A1

I. '

It '

J;'

()

,I '

I;'

()

21.1

3L

3L
-3L

3/~

3L

- 3L

- 3L
A,

In !his instance, (hem arc no displacements of type B2 because the bases of columns are

llxcd. Also, displacements of type C2 are the same as those of type C1 in the first
ijilbstructurc. Terms in matrix Si from Eq. (o) may be added to those in matrix S2 from
I !q. (p) to produce

II,

II

12

-3L ! - 12

B1

116L2

(' ,

12

S2

C1

* + S2 =

21
153

--

SI

3IL2

116L2 i

(q)

-36L--36il --126-

wl11l'h 1s J)tlt l_i(ioncd according to displacements of types A 1, B 1, and C 1 'fhl'll I 'q~ 1II
lo form T A e11e and TA ece for substructure I. Hence,

9L !' - 126

9L

306

IIH' IINl'd

TA

5 A1A1
-1

5 A- 1A
11

101 -

T A 1C:1

S 1[-4 -~J
A1 8 1

A 1C 1

'J'his superposition of terms represents an overlay technique that will prove useful for
1omputer programming in Sec. 10.3.
In preparation for elimination of type A2 displacements, we determine TA 2c 2as

15

3[-1 11]

SL - 1

II

TAC 2 -_ _ 8, A_21A 2 s'A 2 C 2

I JNin,-: lhcs1.: suhmatrices, we compose the transformation operator in Eq. (I '/) 11111 111

-4L
L

/ -9

I
I

-9

O / 0
ISL
O

L
- 4L

15L
O

I
--------~------0
0

0
0

/ ISL
/ 0

0.2449

Ti

B1

36L / 126

36L

- 36L

-36L

!-

126

0
ISL C ,

I,

I;;::
()

6L

Sym.l :'
126

"

A,

L[

~}~~- ~:S)~.!_~-L

-L

(s)

S*2

rs T = L3
2

El [

14.45
- 17.39

-17.39]
40.65

(t)

Parallel Elimination Method

A more general manner of substructuring does not produce coefficient

matrices composed of submatrices in a tridiagonal pattern. For example, the
two-dimensional discretized continuum in Fig. 10.5 is divided into four substructures with interior nodes of type Ae and insulating boundary nodes of type
Be having no special arrangement. In this approach the meaning of subscript Be
is that it includes independent nodal displacements for substructures 1 through
As in the tridiagonal method, the technique is to eliminate dependent displacements of type A 1 from each substructure and to retain a growing number of
independent displarcmcnts of type Be.
For each st1p in the reduction process , substructure
contributes the
following tc11111, In 1111' l'q1111tions of mo tion:

e.

11 I

0
(' ,

l/,

= _ _!_

()

[TA2C2]

0.061221

This matrix contains actions of type C due to unit displacements of type C, with those
of types A and B eliminated (or released).

hy Hq . (20H) .
1'1ol'l'l'di11g l'rnm su~structurc I to substructure 2 (sec Fig. 10.4), Wl' 11 ch 1111
iltNpl11c1m1111s_ ol lype lJ in the former to become type A in the htllur. Adcl111 1111 I
11111111h1111011Nfrn111 suhslruc1url' 2 to the stiffness matrix arc:

WI

(r)

Now we can determine the final reduced stiffness matrix to be

IIN HIVl' II

Si

l c2

S*
[ cc

1-~l~--~~~l----

EI
'f'/' S I T1 = l 5L3

1 [0.2449 0.06122]
L 0.2449 0.06122

Then the abbreviated transformation matrix T2 becomes

'l'h111tfon: , the reduced stiffness matrix Si is

St

A1

---------j-------

1
ISL

()

17

<n )

Sec. 10.3

456

Substructure Methods

457

Modified Tridiagonal Method for Multistory Buildings

Chnp

and

..

De=

B,

=
[1?.AJ
Dn e

TeDnc

(26b)

Substitution of Eqs. (26) into Eq. (22) and premultiplication of the result by TJ
gives
(27)
A,

A,

The reduced matrices in this equation are symbolically the same as in Eqs. (10),
but displacements of type B are different.
As for the tridiagonal and modified tridiagonal methods, we assemble
stiffness, mass , and load matrices in a direct fashion while proceeding from one
substructure to the next. We could also devise a modified parallel elimination
approach that abandons the notion of insulating boundary nodes and introduces
retained displacements of type C. In that case, the displacements of types A and
B would both become "slaves" to the "master" displacements of type C.

B,

B,

A3

A.

10.3 MODIFIED TRIDIAGONAL METHOD FOR MULTISTORY

BUILDINGS

Figure 10.S Substructures in parallel.

This. cx~ressio n again contains the same symbols as Eq. (3) but its ex 1
Vlll'H10 0 IS

'

I 11111

MAB] [~A] + [SM SAnJ [DA] _ [AA]

[MM
MnA Mnn Dn e
SnA S88 D8 e - A8 e

( \I

which is diffcrentfrom Eq . (4). ThedependenceoftypeA d' l

type JJ,. is written as
e isp acement" upon

DAc = TA elJ eD11r

I'

I ll'rl' .the n~atri ~ TA.i!J,e has the s~me formula as the first of Eqs. (Sh), hut
111t11111ng of IJ1 1s different. T he required transformation operator is

T1 =

II

lh

~:'l.

Thii. hu.~ lh.t uppt'lll'!lllt'l' of Eq . (6.7 9), but it pl)rlltin s to only onl) s uhs 111111111
Now thl' d11q1l11tt111t11t 111111 lll't'tll'rn1io11 tn111sl'o1111Hlions urt

n,

1) I
1

u,,

I
t

'I', I >11 I

( 1t111)

We now apply the modified tridiagonal method described in Sec. 10.2 to twoand three-dimensional multistory building frames. Most planar building frames
can be handled in the high-speed core storage of a large-capacity digital computer without dividing them into substructures. However, we wish to use the
same technique for both plane and space frames in order to take advantage of
inherent similarities while explaining these two applications.
For a multistory rectangular plane frame, we take_the analytical model
illustrated in Fig. 10.6. The frame is assumed to have linearly elastic prismatic
members, rigid joints , and fixed bases. We also assume that there are no shear
walls, diagonal braces, or setbacks in the building. Each substructure in the
figure consists of the beams at the framing level A (above) and the columns
between levels A and B (below). Displacements of type A appear at the upper
level , and those of type B are at the lower level. From the joint displacements
labeled in Fig. 10.6, we see that axial strains in the beams are to be omitted;
whereas, those in columns are to be retained. The reason for keeping the latter
strains is that their influences are known to be significant in analyses of tall
buildings r4, 5) . Therefore, at each framing level there is only one translation
l),,.r in the x direction and its corresponding lateral force AFt Thus, the subscripts
in the fi gure match those for the modified tridiagonal method in Sec. 10.2,
l)X<.:Cpl that F rnplaces C. Nole that the numerical subscripts o n displacements of
types A n11d I/ (lcf'I to liphl ) nrc I, 2, . . . , 211,, whcrn 11,. is the number of
l'OIIIIIIIIS , AINO, till' Nlli>M'ripl ~ llll lhmll' 111 typl' F (lllp lo botto m) Hl'l' I , 2, .. .

458

Substructure Methods

Sec. 10.3

Modified Tridiagonal Method for Multistory Buildings

Ym

459

Level A

-/),

n
1~
(j)
t
-~- '-------=------"7 ----

k2/ k

i2/ i
y

Xm

/
(a)

B __ loA, _

Leve l A

t0A3

DA2nc-

~~~....C....~~~~~-4~~~--;

_,,

,
Ym

DA4

DA2nc

Level 8

83

DB2nc- 1

Level A

_,,

.
"-.

- ~ -

--13

j2/

/
084

11

082n c

t
/

(j)

_,,

/
Level 8

---x

~!_

}--,

Figure 10.6 Analytical model for Program DYMSPF.

n,, where ns is the number of stories. Finally, this analytical model has tr ih11t111

masses m 1, m2 , , mn, lumped at the framing levels, which includt 11111

tributions from beams and columns.
As for a tier building, every member in Fig. I0.6 is parallel to om 111 th
structural axes; so no rotation-of-axes transformations arc rcquirld . 111:1 111
10.7(a) depicts a beam with its member uxis I',,, pnrallel to tht 1 11xiN 111111 It
principal bending axis z,,, purnlkl lo lht ., uxis . Aclio1111nd displ11c11111;111 1111h 11

- k 3 - -- - Xm

/ k2/t

-~

Zm

(b)

Figure 10.7

(a) Beam parallel to x axis; (b) column parallel toy axis.

1 2 k L and k2 imply that the 4 x 4 member stiffness matrix is the same as

.I 'J '
'
that in Eq . (3.4-24) for a flexural element.
. .
.

On tht othtr h1111d , Fig . I0.7(b) shows a column with its member axis 1m
purnllcl to 1lw v 11xt'l 1111tl ti s prinripnl bending axis z,,, ugain par~llel to the z axis.
Action urnl dt ~plm i 11 w 11 1111d1x1s 1111111ls.J 1111d k ml' 1111111lwnd 111 a s1q11c11ct thut

480

i,11h11t1111 turn Mnll10d1

line. 10.3

IN l'<'.IHhll'.i;l' fo~ l1'1111Sll' lll11p 11111111111 ~llll111~Nl'S lo joint 1111d Horn 11tl
111ut11cts. I hus, indcxcs1 I 1111111 'Ill lhl' lop ol lhc column arc lhl' .~1111 11 11 t
Im II lw11111 ul lcvcl A , whllL A I 1111d A.> 111 lht bottom coincidc with tho~ h
hl11n1 at level B. However, lhl' indlxcs j I al the top and k3 ul lfll' hull
crn IL'spo'.'~. to the x-~ranslations or tloors al levels A and B. Wu muy w, 11 I
'1 "< <> s~1flness matnx for a column as a partitioned array consistinp ol ,
s11h111utnces, as follows:

r,

4L

-r,
K,-L3 0
_ El ,

Ir
I O 4L2 I
--------~--------t----0
6L I O
6L I 12
0
-6L I O -6L I -12
0
2L2

.i I

}2
A

JI

Sym.

2 _J __ _
________

}2

kl

kI

k2

j3

k2

II

II

j3

12
k3

k3

whid1 .is a rearranged version of Eq. (6.2-1). Note that subscripts in Eq , ( t)

l11l..111 111 the sequence A , B, and F
Wh~n stiffn~ss~s for beams and columns in substructure arc m.~r 111 hl
lhl resultrng matrix 1s

Sn- e
Bl11111s cont~ibute ter~s only to submatrix SM, but columns contribull' 11, nll
till' su~n~utnces. Matnx Se augments stiffnesses from previous subst 111r111,
11 prcl~n1111~
1ry lo reduction . From the modified tridiagonal method , Ntrlln
ll'lh1L't1011 formulas for each substructure are

= Sm1 + s,IA TAB = Sou

- SIJA S,J SAIi

s;,. = s,JF + s,1ATAF = s,JF -

s:,.

Sn:

I1

s,1As;;-J sA,..

I \I

+ s,.i\ TM' = Sn: - Sri\ SA) SA/o'

Whl'll'

T,1/1

- SA) SAIJ

TA/o'

- SA) SA/o'

(~

A~ llw .~Lrics elimination proceeds from one substructure to the ntxt, thi 1111111 1
nl 111ws 111.1d <.'ol1111111s of _types A and B remains equal to 211 ThLHlorr ,
11d1'.111~. d1spl11L'l''.1,'L'nls of .type 11 i.11 subs1r'.1tturc (' to hL'L'mnc tyrw A 111 ~111
Hlllll lt11l ( I I . I his 1l'qtlll'l'S pl11c1ng 11111(11L'l'Nof' typl' JI 11110 posillonic nl IVI
, \ 111111 hJs, ( \) llll' itpplll'd IO l'lll'h SllhStllll'lllll' f'lll'II lll'W l 'Olllllhllllllll ll frt1I
,

461

the next substrucllltl' 11re added to the residual arrays from the preceding substructure. On the other hand, the number of rows and columns of type F
Increases from 2 for the first substructure tons for the last substructure. For each
11cw story (except the last), we pick up one more lateral displacement, so that
matrices s;F and SJF keep expanding in size. This overlay technique requires
1omputer core storage for only one substructure and n, framing levels.
The forward elimination procedure is completed at the lowest story, where
l'Olumn bases are assumed to be fixed. At this stage, we have thens equations
of undamped motion
(5)
In which MFF is a diagonal matrix of lumped masses and AF(t) is the vector of
!literal forces. Damped or undamped story displacements may then be found
using either the normal-mode method or direct numerical integration from Chapter 4 or 5.
After time histories of story displacements DFhave been determined, other
Items of interest may be calculated in a backward-substitution procedure. Starting at the lowest level and working upward , we compute joint displacements DA
from
(6)

For this purpose, it is necessary to have saved back-substitution matrices TAB and
'l~F for each story during the forward elimination process. When applying Eq.
(6) recursively, we redefine displacements DA for any particular level f to
l above. Of course, the vector D0
become displacements D 8 for the level
Is null for the lowest substructure.
Time histories of member end-actions for beams may be obtained by
placing appropriate terms from DA into a 4 x I member displacement vector
l)Mi Premultiplication of this vector by the 4 x 4 beam stiffness matrix K;
yields:
(7)

e-

Sym .J

S1t1

Mudl!lnd I I lrlln11111111I Molhod lor Mulllstory Buildings

in which the end-action vector AM; contains a shearing force and a bending
moment at each end. For columns, time histories of end-actions are found in a
similar manner, but vector DM; contains two terms each from DA, Do, and DF.
Also, the 6 x 6 column stiffness matrix K; in Eq. (I) is used in Eq. (7) to
produce six member end-actions . Of course, these time-varying end-actions for
hcams and columns could be added to any static actions existing at time t = 0.
Turning now to three-dimensional multistory tier buildings, we must acrnunt for the rigid lamina existing at each framing level. Figure 10.8 shows
~ubstructure
consisting of a rigid lamina, the beams at level A , and the
l'Olumns below . Recall that stiffness matrices for x-beams, y-beams, and zl olumns were giwn in Eqs. (9.4-1), (9.4-2), and (9.4-15) and that the last is
trnnsl'ormed lo I lit id hndy rnordinates . In addition, formulas for the transformed
11111ss inert in 1111d ~1111 y 11111d 11111trices appear in Eqs. (9 .4 17) and (9 .4- 19) .

e,

462

Substructure Methods
/

Sec. 10.4

I
I
I/ /

/
/
/

1-------t
I

I
I
Level A

x-beam
/

--------- z-column

/
/

/
/

Level 8

1?.

10.4 PROGRAMS DYMSPF AND DYMSTB

/
/
/

- ----,

In this expression , TT is the transpose of the matrix in Eq. (~.4-13~,. and DF,.
contains story displacements from levels A and B i~ the last s~x P?s1t1ons.
In retrospect, it is interesting to note that the axial const~a~nts in .the ~eai:ris
of a multistory plane frame are analogous to those due to th~ ng~d laminae in tier
buildings. Recall that each framing level of the pl~ne f~ame 1.n F1? .. 6 has only
one translation in the x direction. This implies infinite axial ng1d1t1es for the
beams and we can visualize a one-dimensional rigid body at each framing level.
As for' the laminae in tier buildings, these unseen constraints in plane frames
serve to reduce the number of degrees of freedom in the analytical model.

463

Programs DYMSPF and DYMSTB

/
1----

In this section we discuss Programs DYMSPF and DYMSTB for dynamic

analysis of multistory plane frames and tier buildings. These pro.gram~ use
auxiliary storage as well as core storage to analyze two- and three-d1mens1onal
.
building frames by the modified tridiagonal method:
Starting with the multistory plane frame, we give the outline for Program
DYMSPF, as follows:

Outline of Program DYMSPF

z

Figure 10.8 Typical story framing in tier building.

For tier buildings , assemblage of stiffnesses and the forward elimio11111111

proceed the same as with plane frames. However, the number of rows 11ml
columns of types A and B is 3nc because each joint has three unconstrn111r1J
displacements. Also, the number of rows and columns of type Fis 3n., dur 1t1
the fact that each lamina has three rigid-body displacements. The overlay 111 h
nique also works the same as for plane frames, and computer core storugl' 111 I
only contain information about one substructure and 3ns rows and colu11111~ Ill
matrices of type F
Calculation of story displacements in tier buildings and the buck WIii I
substitution process follows the sequence given for plane frames. But to oh111111
column end-actions, we must transform story displacements from the rel111m
point F to joint j at the top and joint k at the bottom. Therefore, Wl' haw

l. Read and write structural data

a. Structural parameters
b. Bay widths and story heights
2. Generate story mass and stiffness matrices (by substructures)
a. Clear stiffness and mass matrices
b. Read and write story mass and member information
c. Augment residual stiffness matrices with S1
d. Calculate reduced stiffness matrices by Eqs. (10.3-3)
e. Place story information into auxiliary storage
f. Shift matrices of type B into locations of type A
g. Repeat steps b through f for each story (top-to-bottom)
.
3 . Determine frequencies and mode shapes
a. Convert eigenvalue problem to standard, symmetnc form
b. Calculate eigenvalues and eigenvectors
c. Write natural frequencies (cycles per second)
d. Transform, normalize, and write modal vectors
e . Normalize modal vectors with respect to mass matrix
4 . Read and write dynamic load data
n. Dynn111ic.: parnmctcrs
..
b. l11itinl conditio11s (stmy displnl'l' lllcnts nnd vcloc1t1cs)
c , Appli,d 1111111m1 ( 1 loil'l'~ 111 lrn11ii11l' llvi-l M
)

464

Substructure Methods

d. Ground acceleration (in x direction)

e. Forcing function (piecewise-linear)
5. Calculate story displacements
a. Set up modal damping matrix
b. Cal_culate story displacements for each time step
c. Wnte and/or plot story displacement-time histories
d. Find and write maximum/minimum story displacements
6. Determine. member end-actions (by substructures)
a. Retrieve story information from auxiliary storage
b. Calculate joint displacements DA
c. Cal_culate member end-actions for each time step
d. Wnte and/or plot member end-action time histories
e. Fi~d and write maximum/minimum member end-actions
f. Shift elements of DA into D8
g. Repeat steps a through f for each story (bottom-to-top)
1:ublc JO . I shows preparation of structural data for multistory plane frnr11
Str uctu~al parameters i~ the table include the number of bays NB and the minih
of sto:';~ NS. For _d1.men~ions of. bay widths and story heights, Prn~IIIIII
DYMSJ .1 requ1r~s mm1mal mformation. Each line of that data gives the nurnh
N~JM of sequential _occurrences, followed by the repeated dimension of tlw hll
w1d1h BW (left-to-right) or the story height SH (top-to-bottom). Each of thr N
blocks of substructure data contains the superimposed story mass SMA
(t,o p lo-bottom) and member information (left-to-right). For the latter wt 11r J
i-:1.vc only the number NUM of repetitions, followed by the cross-sectionnl 111
of th~ member AX or A Y and its second moment of area ZI with respctt to th

, ,,, axis.

TABLE IO.I Structural Data for Multistory Plane Frames

No. of
Lines

Type of Data
l'mb 1cm identification
St,ucturnl purumctcrs
D11ncn,,ionN
(11) Buy widths
(b ) Story heights

llt'IIII I~

('nh111111N

"N'I hl111~~

NUM, BW
NUM, SIi

ft

SMASS

--

A~ 11q11i11d

Descriptive title
NB, NS,E, RII ()

Suh, tI Ul' llll C d11tu11

(11) Stoiy lllllSN
(h) Ml111lw1 info,mution
)

Items on Data Li

,~

NlJM, AX, /I
NlJM, AY, t'I

Sec. 10.4

465

Programs DYMSPF and DYMSTB

Preparation of dynamic load data for multistory plane frames appears in

Table 10.2. The dynamic parameters are the same as before, and the initial
conditions involve NS story translations and velocities at time t = 0. Under
applied actions we see the load parameter IAF, indicating whether story loads
exist or not. As shown in Fig. 10.6, each story load is a force An in the x
direction at framing level The ground-acceleration and forcing-function data
in the table have been discussed before in conjunction with other types of
structures.

e.

TABLE 10.2 Dynamic Load Data for Multistory Plane Frames

Type of Data
Dynamic parameters
Initial conditions
(a) Condition parameters
(b) Story displacements
(c) Story velocities
Applied actions
(a) Load parameter
(b) Story loads
Ground accelerations
(a) Acceleration parameter
(b) Acceleration factor
Forcing function
(a) Function parameter
(b) Function ordinates

No. of
Lines

Items on Data Lines

!SOLVE, NTS, DT, DAMPR

IND, INV
00(1), 00(2), ... , DO(NS)
VO(l), V0(2), .. . , VO(NS)

a
a

IAF
AF(!), AF(2), ... , AF(NS)

1
I

!GA
GAX

1
NFO

NFO
K, T(K), FO(K)

As required .

Example 10.3
Figure 10.9 shows a two-bay, ten-story plane building frame having a rectangular layout.
Beams in this structure are all steel rolled sections of size W 21 x 55, but the steel
columns vary, as follows: (Cl) W 8 x 31, (C2) W 10 x qO, (C3) W 12 x 85, and (C4)
W 12 X 106. The length Lis 144 in., and the mass superimposed at each framing level
is 0.06 k-s2 /in.
For this frame we have two dynamic loading conditions. The first is an atmospheric
blast that causes the triangular force An in Fig. 10. lO(a) at each level , except t~~ top
(where it is half as much). The second consists of rigid-body ground acceleration D81 in
the x direction that has the sawtooth shape in Fig. 10. IO(b). We analyzed the frame for
these two load ing conditions using Program DYMSPF, with a damping ratio of 0.10.
Resulting time histories of the translation D1,, at the top level are given in Fig. 10. lO(c) .
I .oud case (11) l' llll\t'\ 11 mnximum response of 4 .090 in . at time t = 0.480 s, while the
11111xi n111111 1x1111~H111 lrn lontl 1nsc (h) is 4 \'i7 111. nt tinw t
2 . 16 s. After those
IIIIIXIIIIII, 1hr 1np1111\1, il1111llll\h h1111111w ot d11111p111p

i;111>atr uutur u Muthocl11

41111

467

Soc. 10.4

1'10111111111 l>YMSPf' and DYMSTB

AF, (k)

o~I\
5

- - - - - ' - - - t (s)
0.04

0.2

0.4

0.6

0.8

(a)

0.1

Du, (g l o -----+-----1-----~------l--2.0

t (sl

-0.1

C2
10@ l

(b)

4
3
2

t (s)
011 (in.)

- 1

ly

- - ~I

-2

Dy,

7,7,:--

~~

2L

'}l

Fl1t11n I0. > Ten

1

~~

Nl ory cxumplc

1hr Program l)YMS1'J1

-3
4
(c)

Figure 1().10

Results for ten-story example: (a) atmospheric blast; (b) ground

m;cclcrntion ; (c) 1r11nslutinn 1),. 1 due to (a) und (h).

488

Sec. 10.5

8ul111tr ucturn M11thncl1

469

Component-Mode Method

The outline of Progrn111 l>YMS I B lrn tin buildmgs 1s s11111l111 111 1h11
Program DYMSPF for plane frames. Ilowcvcr, structural <.lulu in p111 t I 1111 11 I
building is given in the sequence \ direction, y direction, and negat1v, ,II
tion. Also, for a typical floor plan, we must provide Boolean dutu ti,, 11111111
joints within a rectangular x-y pattern. That is, an existing joint is i11d1111t ,I h
a one in an integer matrix, while a nonexisting joint is indicated hy 11 1, 111 It
part 2 of the outline, the program calculates the location of the ccn111 111 111
(point c) for each story. Also , the mass-inertia matrix is generated with ll ~I
to point F, as shown in Eq. (9 .4-17). In part 4, the dynamic load d11t 11 1, 1111
extensive than for a plane frame. Three types of initial displaL1111r111~ 1111
velocities are possible for each lamina. In addition, we handle both , nn
y-forces applied at each level, as well as x- and y-components of gm1111d 111 I
erations. Therefore, the data must also define two independent pictcw1,, 1111
forcing functions (for the x and y directions).
If shear cores, bracing, and setbacks are included in the analy11rnl 11111 I
of a tier building [6], we need more data than that described ahll\T WII
substructures, it is also possible to analyze soil-structure interaction I I I 1111d I
calculate inelastic responses [8] of tier buildings.

10.5 COMPONENT-MODE METHOD

The original ideas for the component-mode method are attributed to I 11111 y I I
1lowever, Craig and Bampton [10] also made useful improvements. In thr htlt
work, generalized displacements in a substructure consist of a limitLd 1111111
of vibrational mode shapes and a finite number of nodal displacement, 111 Ill
luting boundaries. The component-mode theory produces mass (or dy1111111I
coupling between the modal and nodal accelerations in a typical suhsl111d\l
Figure 10.11 shows a substructure C that is arbitrarily located within
discretized continuum. Nodes of type A are indicated at interior position, . whll
those of type Bare at boundary locations. Undamped equations of mot11111 l111 lh
substructure may be written as

Me De + Se De

= Ae(t)

(I

whil'h is the same as Eq. (10.2-14). However, the expanded form is

1111w

] [DA] [SAA SABJ [DA] [AA I

[MM MAB
MBB Ds ( + SsA s,JB
= A,,
M11A

D11

With boundary nodes restrained and zero loads, we can set up

11~lnv1llue problem

e for component-mode method.

2, ... , m" where m, is the number of retained modes. By normalizing the

modal vectors with respect to the mass submatrix MAAl, we can state the
relationship
(4)

In this expression, the symbol DNe represents a vector of m, normal coordinates,

and l>Ne is the normalized modal matrix.
For nonzero boundary displacements (occurring statically), we have

(5)

111111

Figure 10.11 Substructure

where
"'Iv~ lh

TAB e = -S;leSABt

(6)

Now let us define a transformation matrix Te as

111 wl11l'11 (1)11 is the ungulur frtqut'ncy of motk I und <1>11 is tlw n111,,p1111,ll11
111111k sl1111w. Frn this ll11111ld vihrnt101111l 111111lys1s, tht 111od11I indl'Xl'' 1111 1

T( =

[<l>N T,111]
0

I,, ,

(7)

470
Substructure Methods

Sec. 10.6

This operator relates the displacements in vector De to generalized displacemtnl

111 a vector De, as follows:

De =

[~:l T{~:l
=

= TeDe

(Hu)
8

De= Te.De

(Hh)

.Substitution of Eqs (8) int0 Eq (1) d

yields

an premult1phcat10n of the result by 'I J

= Ae(t)

1\1NB] [~NJ +
Moo Do e

[ro;,,O

_o

S88

J[DN]
= [AN]
D e
A e
8

(10)

= ct>iMM cf>N = I m,
MNB = ct>iMAA TAB + ct>XMAB = MIN

MMN

T1 MAB + MBA TAB + T1 MAA TAB

( I I ft I
( 11 h)

(lld

Also , the stiffness submatrices become

SNN

= 1>isM cf>N = ro m,2

sNo = ct>XsAA TAo + 1>isAB = o = siN

Son = Sno + T Xn SAB
und the applied-action subvectors are

( I '11)
( I h)

( I l)

AN= 1>rAA
Ao

(I In)

= Ao + TXoAA

= L Ae(t)

(=I

(J 5)

(=I

where n, is the number of substructures . Of course, the substructure matrices in

Eqs. (15) must be expanded with zeros to become conformable for addition.
Now the equations of motion in Eq. (14) can be solved for the displacements in vector D by either the normal-mode method or direct numerical integration. Then displacements in vector DA e for each substructure are found using Eq.
(5) as a back-substitution expression. In the next section we shall apply the
component-mode method to trusses, for which some additional complications
arise.

M I) I S
llw ll~~l111hlld 11111ldn~

In earlier chapters of this book, we considered only axial deformations for

dynamic analyses of trusses. However, inertial and body forces also occur along
the members in transverse directions, which cause flexural deformations as well.
These influences are most significant in trusses composed of only a few members. We have found that the best approach for analyzing such structures is the
component-mode method [ 11, 12]. By this technique, a member with both axial
and flexural deformations constitutes a substructure. Treating the member as a
simply supported beam, we include a limited number of its exact vibrational
mode shapes as flexural displacement shape functions.
Figure 10.12 shows a prismatic plane truss member i with local (primed)
and global (unprimed) axes. We shall handle such a member as a specialized
finite element acting as a substructure within a plane truss . Flexural displacement
shape functions in member i are taken to be the vibrational mode shapes of a
simply supported beam [13]. These mode shapes will be superimposed on linear
displacement shape functions due to translations at joints (or nodes) j and k.
Thus, the generic displacements u' and v' in the x' and y' directions may be
written in terms of modal and nodal displacements as

u: = f Df

(1)

or

ur

.M. '.

tq1111111111

n, -

A(t)

Me

( I lh)

:; El(l, ( l )I"!, the symbol Im, den~tes an identity matrix of order m,. Also, .., in
<I 12a 1s the spectral matnx for the substruct r

"'
1111g11l11r frequencies w2 2
2 .
.
u. ~ contammg sq11111(11
Nllh11111trix c.lefincd ab~~:r~
Wm, m diagonal P?S1t1ons. The most i111prn l11111
11fi11R terms between accel~rati;~toEqi;n~ DJJ,), which represents dy11a1111, '""
T
N
JJ ,
o HSSl'lll~le e~1uutions of motion for all substructures we simpl udd th
11111 1
1 ins 11 PPl'm111g 111 Eq . (9), u~.ing the direct stiffness 111~ll10d. Thl'~l~ull I
In lhi~

ns -

10.6 COMPONENT-MODE METHOD FOR TRUSSES

The mass submatrices in this equation (without the subscript e) are

Moo = Moo +

471

(1))

In expanded form, this equation is

[ ~,,
MoN

t J

~:;~~ !:t !~:on~;,;~~:o~e~;ti~:::i~s DNe in the first part and D e in the scrnnd

MeDe + SeDe

Component-Mode Method for Trusses

I)
Jill

A(t}

( Ill

(2)
In this equation the symbols DA; and D8; represent the displacement vectors

1>At -=

DA,
0A2

D,11

D111

Ds,I
Ds2

(3)

DJo
1)/,,1

472

Substructure Methods

Sec. 10.6

473

Component-Mode Method for Trusses

y'

I
D~,
D~, __...
j

v'

tr-----+.::...._....:.....;_
t- _ _0_ _ _ _ t
u'

-<J -

- -- -

~ -- - -- - L~-----~
(a)

1~

(e)

(b)

(f)
(c)

Figure 10.12 Plane truss member with local and global axes.

'l'l-1111s in vector DA; are the amplitudes of a limited number m, of retained 11111111111
111odc shapes for a simple beam, as depicted in Figs. 10.13(b)-(d). In add1 111 111
the terms in vector D 8; are the four translations D 81 through D 84 at joints 1111111
k, shown in Figs. 10.12 and 10.13(a).
Matrix fin Eq. (1) contains displacement shape functions, and its 111 I
s11h111ntrix is

(d)

Figure 10.13

= <I> = [

f
A

0
0
0
sin 1r{ sin 21r{ sin 31r{

.. ]
...

Displacement shape functions for plane truss member.

in which the linear differential operator d; is

111 which the di mcnsionless coordinate is g = x ' IL. Appearing in the scco11d 111w
111 this s11hmulrix are the natural mode shapes for vibrations in a simply .,Ill'
po, tLd prismatic beam. To keep them dimensionless, these mode shapcs 1111 11111
1111111111lii'.l'd with respect lo the mass of the member. On the other h1111d, th
Nl'l 'OIHI sul>lllatrix in f has the form

0
I

g
0

OJ
g

II,

cl, I'

111 ,

II,, I ,

(7)

L 2 ag2 ;

The first submatrix in B; has the resulting form

t ~,

'lhtsl' li11l'11 r shupl !'unctions result froin unit displuccments of' 1>;11 th ro11ph /J 111

IINl11dic111td i11 Figs. 10. Uk) nnd (!') .

S1111i11 diNplllL'L'llll' llt nl11tio11ships f'tu lhi s L'lt111t11t 11111y lw st11tld nil

.o

y' a2

0
0
0
I 7T2 [
8 Ai = d;f" = Y L2 sin 1rf 4 sin 21rf
9 sin 31r(

.. .l

(8)

und the second is

I'll/

cl, r,,

llllllll,

('))

474

Substructure Methods

Chap. 10

Sec. 10.6

475

Component-Mode Method for Trusses

The stiffness matrix Sf in local directions for member i has the known frn 11 1

s: =

BrE B; dV

[soM snB
oJ

2
,

(tm

Iv ,,r,r, dY 1

M,m -

pAL 0

6
[

In this equation the stress-strain matrix E is

(I f l

where 12 is an identity matrix of order 2. The first submatrix in

-sMi --

[~

f BT

7T4EJ,
Ai E BA; dV = 2L3

_ (pAL
-2-w"',

2 )

Sf

becontl'H

Sy1

16

SBBi -

r BB;E Ba; dV = -L
)v
1

(II)

Mf = ( pf f dV =
Jv

Mf

[M~ M~J
MBA

(17)
1 0

2 ;

[A~J

(18)

Aa ;

A,; -

[by'bx']

(19)

has the form

~Tb'(t) dL -

f];b'(t) dL =

1[:~f::]~,

dL

and the second is

An;

l [g ~b;i::1

( Ill

Mos ;

is

thy'

dL

For local (or member) axes , the undamped equations of motion for small
displacements of the plane truss member are

As before, the3mbol I,,,, represents an identity matrix of order

submatrices in Mf are

v p<l>'I' f 11 r/V - pAL [

7T'

()
'

1111d

(21)

Mf Df + Sf Df = Af (t)

T\it'ill/I

(20)

0 0 ;

which is the same as matrix K' in Eq. (3.5-25) .

. Next, the consistent mass matrix Mf (in local directions) for memhl't 1 111
obtamed from
1

Af (t)

Sym.]

1
0 0
-1 0

r r1 b'Ct) dL

Ji

b'(t)
The first subvector in

The first submatrix in

(I 'I

EA

Here the vector b' (t) contains body forces (per unit length) in the x' and y'
directions, as follows:

and the second submatrix is found to be

Sym.l

which is the same as matrix M' in Eq. (3.5-32).

Also, equivalent modal and nodal loads in local directions are calculated
as

A; Ct)

81

1/2

()

1/2

I/ l

()

1/J

111,. < >11 111

By rotation of axes for the parts of type B, Eq. (22) may be transformed to global
(or structural) axes to become

M;D;
(If,)

(22)

+ S;D; = A;(t)

(23)

The displacemcnls D/ and.accelerations fi; in Eq. (22)_are related to the corresponding vectors 1)1 ond 1)1 in Eq . (23) t,y the expressions

n;

ft,1>1

f>;

ft,n,

l111h11111 11l11ti1111 -.hlpH tlu 1011111111111111,mM 111111Hl11111111lHlll 11111111 ,i 111

(24)

476

Substructure Methods

Chap, 10

( 1~)

Note from the form of matrix R; that only the nodal displacements D
transformed fr~m structural to member directions. (Displacements D 81 th;.:11;":,
'f?n4
are shown m the x and y directions in Fig. 10.12.) Submatnx R'811n
. E q. ( ~,)
lS

Rn;=

[R

OJ

0 R;

( '" I

where

= [ _ Cx

Cy]
Cx i

Cy

(.! II

Dire~tio~ cosines in the rotation matrix R; were defined in E s (3 5-24) 11

substitut~ng Eqs. (24) into Eq. (22) and premultiplying the J~t~er by R} w~
convert it to Eq. (23). In that equation we have the following matrix procl:H 1~
-

S;

A;(t)

( '')I

( 1111

c;
=

EA
L

c2y

CxCy

[ -c;

c2X

-cxCy

s:]

( 111

-::12]
Cx/3
.

- R" TA'

Ill -

/Ji

(I - f)(c,.bx - cyby,) I
+ c,by)

'
dl
1.
g(c,bx - Cy by,)
[
{(cyb, + C\ bv,)
.
I

-1

01 -

(I - g)(cyb,,

( I\)

or

Noll tlu.11 submutrix. M11111

M f1111 is invariant with rotution
axes
I
Afl~r tlw rotat,011
nxcs lntnsf'ornrntion f'or cuch nw111h<.1r is ~0111plrt I
I I(' l'ljlllll IOIIS of Ill()( ion for ( ht whok H( 1111.'ltll'l' 11u1y lw USSl'lllhlt:, (O t11il11l1'.'

or

Mi>

Sh

i=I

A(t)

L A;(t)

(35)

i=I

where m is the number of members. The equations of motion for the structure
can now be solved by the normal-mode method or by direct numerical integration. Either approach may be preceded by a vibrational analysis, from which a
damping matrix can be established for the whole structure, as before.
Extension of the theory for plane trusses to the analysis of space trusses is
straightforward and appears in Ref. 12. The primary change to recognize in three
dimensions is that each member has two principal planes of bending, defined as
the x '-y' plane and the x '-z' plane. But the flexural mode shapes in each plane
are still the sine functions sin 7Tg, sin 27Tg, sin 37T{, and so on. Definition of a
principal plane in space may be aided by using a third point p in addition to
points j and k, as explained in Sec. 6.4.

( 'HI

= RTMiR'.
= RT A; (t)
l

s = LS;

10.7 PROGRAMS COMOPT AND COMOST

'T- '
= R;
Sf R;

The results of these operations are as follows:

Son,= R.];,Snn,R.n,

477

Programs COMOPT and COMOST

in which

R; =

R;

Sec. 10.7

A(()

( 11)

Now we briefly discuss Programs COMOPT and COMOST for dynamic analyses of plane and space trusses by the component-mode method of the preceding
section. These programs calculate responses to initial conditions and piecewiselinear forcing functions that may be either applied actions or translational ground
accelerations. Using the normal-mode method, we determine axial forces and
bending moments in the members, as well as time histories of modal and nodal
responses.
Example 10.4
Figure 10.14 shows a plane truss having only two members, with a step force P applied
in the x direction at the quarter point of member 1. Both members are prismatic and have
the same values of p, E, A, and /,. Although realistic sizes were used, we take the
dimensionless parameter AL2 / I, to be 250 for this problem.
Dimensionless frequencies for the truss solution without flexure are w i = 1. 000
and wi' = 3.023, which have been normalized by dividing them by the first. Table 10.3
contains twelve such dimensionless frequencies for the structure when five vibrational
modes are included for each member. Also, Fig. 10.15 shows the corresponding mode
shapes of the truss for the first four modes. If members 2 and 1 are taken separately as
simply supported beams, their fundamental frequencies become only slightly more than
those for modes 1 and 2 in Table 1.
Figure l O.16 gives responses of the dimensionless displacement Di, plotted
against the dimensionless time r *. Herc the dimensionless displacement is obtained by
dividing the dynamic value by the static value for the same load . On the other hand, the
dimensionless time , ,.,. is time r divided by the period of the first mode for the truss
solution without lkx11n. The curves in H11, 10. l<i 111l' lnhckd with encircled numbers 0 ,
I, ... , 'i t11 lntlil'11ll' how 1111111y lkx11rnl 111odl'S 1m1 1111111h11 llll' i11d11tl1d i11 thl nnnlytic-nl
111od1I Wr 11111 Hr1 1h111 1111 Hol11tu111 with 1111 lhK111r I~ q111h 1lllh11111 1111111 lhos1 wi th

478

Substructure Methods

Chap. 10

Sec. 10.7

479

Programs COMOPT and COMOST

\
\
\
\
\
\

\
\
\
3L

(a)

(b)

(c)

(d)

----x

~I.-3L-.I
Figure 10.14

Plane truss with two members.

TABLE 10.3 Dimensionless Frequencies

for Two-Member Truss
Mode

I
2

w lf<

Mode

w lf<

0 .204

0.117
0.7%

3.034
3. 163

11

I , I 8()

1.628

I .MH
J II I

10
11
l :l

~. 146
t ~, ,

H )ti/

Figure IO.IS Mode shapes for two-member plane truss: (a) mode 1; (b) mode 2;
(c) mode 3; (d) mode 4.

400

;11lin11111ill1111 MulhocfN

Sue. 10, /

1110111111111 C:O MOI' I 1111d COMOS1'

2.5

D;t-~-t

2.0

T
l

1.5

6--------....:o----...,__,..;u-----s

'

1.0

1.0

Figure 10.16 Displacement time histories for D

Figure 10.17

Plane truss with np panels.

;,.o

1.5

'/.

0.5

t.
1.6

flexure included in the members. The first three flexural modes in membor I ll'lld 111
dominate the response because the load is applied directly to that member. Addi118 1111,,h
4 and 5 to each member does not improve the accuracy of the response very 111111 Ii
Example 10.5

1.5

1.4

1.3

Figure I0. I7 illustrates a truss with an indefinite number of members, having u Nhp 11111 1
P applied in they direction at the center of member 1. By increasing the numbLr ol 1111111 I
frn111 I to 11,,, we can study the effect of this parameter upon the dynamic rcsp1111~1 / I
of joint I in they direction. All members of this truss have the same values for 11, I . I
nnd /,; and we take the dimensionless parameter AL2 / I, to be 2250.
Por this example the truss solutions without flexure will be compmld uv.11111-1
solul ions with one flexural mode included for each member. In Fig. I0. I 8 w1 pl111 1111
rulio R of the maximum response of D 2 for the model with flexure to th at wilho111 lln 1111
versus the number of panels n,,. The ratio R approaches unity as the nurnlwr ol p11111 1incrcascs. Thus, the flexural deformations in the members have little cffcl'I 011 1hr 1111111
rusponscs when the number of mem bers becomes large.

1.2

0.9

l'l uun J0. 18

Displacement ratio versus number of panels.

481

482

!h1h11t1uut11111 M11tlu1d1

R NC

I. 11011s, 13 . M. , "A Fronlnl Solul11111 1'1111{111111," /111 . J . Numl!t. Method., ll'IIH V11I
No. I, 1970, pp. 5-32.

i. llinton, E., and Owen, D. R. J., F/11/tl' Weme11t Programming, Act1dl111h 1111
London, 1977 .
\ , Wcuvcr, W. , Jr., and Yoshida, D. M., "The Eigenvalue Problem for B1111d1d 1\11111 I
ccs," J . Comp. Struct., Vol. 1, No. 4, 1971, pp. 651-664.

Notation

'1 . Weaver, W., Jr. , and Nelson, M. F., "Three-DimensionaJ Analysis of Ti11 ll11tltl
lngs," ASCE J. Struct. Div., Vol. 92, No. ST6, 1966, pp. 385-404.
~. Weuver, W., Jr. , Nelson, M. F., and Manning, T. A., "Dynamics of Tier ll11lldi111-'
ASCJ~ J . Eng. Mech. Div., Vol. 94, No. EM6, 1968, pp. 1455- 1474.
(1 . Wcuver, W. , Jr., Brandow, G. E., and Manning, T. A., "Tier Buildings with ~lu 11
Cores, Bracing, and Setbacks," J. Comp. Struct., Vol. I , Nos. 1/2, 1'1/1. I'll
57 84.
7. Wcnvcr, W. , Jr., Brandow , G. E ., and Hoeg, K., "Three-Dimensioruil ~1111
Slructure Response to Earthquakes," Bull. Seismal. Soc . Am., Vol. 63, No. I, l 1J/ I
pp. 1041- 1056.
H. Wcuver, W. , Jr., and Bockholt, J. L., "Inelastic Dynamic Analysis of'Tit11 ll111hl
lugs," .I. Comp. Struct., Vol. 4 , No. 3, 1974, pp. 627-645.
1

1 ll11rly , W., C. , "Dynamic Analysis of Structural Systems Using Co111p11111111

Modes," A/AA J., Vol. 3, No. 4, 1965, pp. 678-685.

10, ( 'ndg, R. R., Jr. , and Bampton, M. C. C., "Coupling of Substructures for Dy1111111h
Annlysis," A/AA J., Vol. 6, No. 7, 1968, pp. 1313-1319.
11 , Wluvcr, W. , Jr. , and Loh, C. L., "Dynamics of Trusses by Componc111 Mn1h
Mclhod," ASCE J. Struct. Eng., Vol. 111, No. 12, 1985, pp. 2526- 2575.
12. Loh , C . L., "Dynamics of Trusses by Component-Mode Method," Ph .Jl ,/11
.11rratio11, Department of Civil Engineering, Stanford University, May I 98'1
11 , Timoshenko, S. P., Young, D. H., and Weaver, W., Jr., Vibration Proli/11111 111
l:11t,1i11ttrl11g, 4th cd., Wiley, New York, 1974.

1. MATRICES AND VECTORS

Symbol
0
A
B
C

D
E
F
G
H
I

J
K
L

Definition
Null matrix
Action vector (also coefficient matrix)
Strain-displacement matrix
Strain-stress matrix (also damping
matrix and constraint matrix)
Displacement vector
Stress-strain matrix
Flexibility matrix
Constraint matrix
Characteristic matrix (also Householder
matrix)
Identity matrix
Jacobian matrix
Element stiffness matrix
Lower triangular matrix
Mass matrix (also concentrated
moments)
Concentrated force vector

483

484

Notation

Symbol
Q

s
T

u
V

X
y

z
A

<I>
b
d

e
f

j
k
p
q
u

Definition
Factor in QR algorithm
Rotation matrix (also factor in QR
algorithm)
Stiffness matrix
Transformation matrix (also tridiagonal
matrix)
Upper triangular matrix
Eigenvector matrix
Vector of unknowns
Vector of unknowns
Vector of unknowns
Spectral matrix
Eigenvector matrix
Body force vector for element
Linear differential operator for strain
displacement relationships
Unit vector
Interpolation function matrix
Unit vector
Unit vector
Unit vector
Nodal load vector for element
Nodal displacement vector for elc11w111
Generic displacement vector for elc111rnt

2. SUBSCRIPTS FOR MATRICES AND VECTORS

Symbol
A
8
F
/,
M

N
I'
N
II

Definition
Nodal displacements eliminated (<1/.w1
attached)
Nodal displacements retained (al.10
body and boundary)
Free (also floor or fram ing leve l)
Lumped
Member
Nornrn l rnordinntes
J>ri11l'ip11I rnordin111ts

485

Notation

Symbol
b

d
e

f
g
j
k

C
m
n
p
q
I'

s
X

y
z

Definition
Body
Damped
Element
Forced
Ground
Index
Index
Index
Index
Number
Number
Working (or reference) point
Working (or reference) point
Radial direction
Structure
x direction
y direction
z direction

3. SIMPLE VARIABLES

Symbol
A

B
C
D

E
G
I
J
L
M

R
'f'

l{~Nll llllll'd

(}

l111tlul

\'

Definition
Area
Constant
Constant
Displacement
Young's modulus of elasticity
Shearing modulus of elasticity
Moment of inertia (second moment
of area)
Polar moment of inertia
Length
Moment
Force
Radius
Period
Strnin l'llorgy
l 1111r11t 111l l'lll'I HY

486

Notation

Symbol

X
a
b
C

f
h
j
k

e
m
n
p

q
r

s
t

u
V

w
X

y
z

Definition
Work
Generalized action
Constant (also acceleration)
Constant
Constant (also damping constant)
Constant (also displacement)
Base of natural logarithm
Interpolation function (also frequency ,
cycles/ sec)
Thickness
Index for ... (also v=T)
Index for . . .
Index for ... (also spring constanl)
Index for ... (also length)
Number of ... (also mass)
Number of degrees of freedom (also
damping parameter)
Action at element node
Displacement of element node
Radius (also cylindrical coordinate)
Segment length
Time
Translation in x direction
Translation in y direction
Translation in z direction
Cartesian coordinate
Cartesian coordinate
Cartesian coordinate (also cylindric11l
coordinate)

487

Notation

Symbol

Definition

Rotation or angle (also Hilber constant)

Magnification factor (also Newmark
constant)
Shearing strain (also damping ratio and
Newmark constant)
Increment
Normal strain
Dimensionless coordinate
Dimensionless coordinate
Rotation or angle (also Wilson constant)
Direction cosine (also eigenvalue)
Frequency coefficient
Poisson's ratio
Dimensionless coordinate
3.1416 ....
Mass density
Normal stress
Shearing stress
Curvature
Twist (dOx/ dx)
Angular frequency (also angular
velocity)

/3
)'

5
E

(
rJ
()

A
,
V

t
'TT

a
T

</>

!/;
w

5. PROGRAM NOTATION

Symbol
AO( )
AOP
AB( )
AE( )
AF( )

4 . GREEK LETTERS

Symbol
~

Definition

...,.

Increment
Summation

<I>

FU1Klio11 or ,nod<.
A11~11l111 I 11q111111 y

ALPHA
AM( )
AN( )

AR( )
AS(

AX(

Definition
Initial accelerations of nodes
Initial acceleration of moving load
Actions applied to bodies
Actions at element nodes
Actions at free nodes (also actions at
floors or framing levels)
Hilber parameter a
Actions at ends of members
Actions in normal coordinates
Support reactions
Actions at structural nodes
Cross Sl'l'lionnl nnas A,

488

Symbol
BETA
BI( )
BL1 , BL2,.
BM( )
BS1,BS2, .. .
BV1,BV2, . . .
BW
CME( , )
CMS ( , )
CV( )
CX,CY, CZ

DO( )
DAMPR

DBO( )
DE(
DF(

)
)

DM( )
DN( )
DR( )
DS( )
DT
E
E l ,E2, .
EL( )

F( )
FO( )
G
GAMMA
G AX, GAY, G AZ
II

l,J, K, L
IA('
IAF

489

Notutltm

Ocfi11ltlo11

Nlw111111k puramclcr f3
Body i11crtit1s
Jntcnsilics of line loads
Body masses
Intensities of surface loads
Intensities of volume loads
Bay width
Consistent mass matrix for clc111t'11I
Consistent mass matrix for strntl111,
Characteristic values (eigenvalul'N )
Direction cosines c.,, cy, and c,
Initial displacements of nodes
Damping ratio
Initial displacements of bodies
Displacements of element nodes
Displacements of free nodes (also
displacements at floors o r 1'11111111111
levels)
Displacements at ends of membtrN
Displacements in normal coorcli,1111111
Displacements of restraints
Displacements of structural nodes
Duration of time step /:l.t
Elasticity modulus
Elasticity constants
Element lengths
Frequencies (cps)
Function ordinates
Shearing modulus
Newmark parameter y
Ground acceleration faclors for 1, I' ,
and z directions
Thickness
Indexes
Indicalor for imposing axial co11s1111l111 ~
lndicalo r for actions al lloors

Symbol
ID( )
IBO( )
!G A
IML
IN( ), JN(
IPL
IPS
IR,IC
IRO
!SOLVE

IWR
JB( ' )
JNO( )
LN
NB
NBID
NBIV
NC
NDF
NE
NEL
NEN
NEO
NES
NEV
NFO
NJ
NLB
NLN
NLS
NMODES
NN
NNA

Definition
Displacement indexes
Element numbers for output of stresses
Indicator for ground accelerations
Indicator for moving load
Indexes for nodes of elements
Indicator for plotting
Indicator for plane stress or plane strain
Row and column indexes
Indicator for eliminating rotations
Indicator for method of solution
Indicator for writing
Body-node numbers
Node numbers for output of
displacements
Loading number
Number of bodies (also number of
bays)
Number of bodies with initial
displacements
Number of bodies with initial velocities
Number of columns
Number of degrees of freedom
Number of elements
Number of elements with line loads
Number of element nodes
Number of elements for output
Number of elements with surface loads
Number of elements with volume loads
Number of function ordinates
Number of joints (or nodes) on a body
Number of loaded bodies
Number of loaded nodes
Number of loading systems
Number of modes
Number of nodes
Number of nodes of type A

490

Symbol
NNI>
NNF
NNII >
NNIV
NNO
NNR
NRL(
NRN
NS
N'l'S
NUM
OM EGA(
p

Pl II (

PR
)

R(

I{( '(

RI 10
SH(
SIi
SS(
SX ,SY, .
T( )
TIME

VO( )
VOP
VHO( )
VN(

VS(

X( ),Y( ), Z( )
XC'( ),YC( ), ZC( )
XC'.I( ),YCJ( ), ZCJ( )
XC'K( ),Y CK(
X I( )
YI(

). ZI(

), ZCK(

l)()llnlllon
N11111lw1 of nodal displucc11K111s
N111uhcr of nodes of typu F
Number of nodes with initial
di spl acements
Number of nodes with initial wlrn llh
Number of nodes for output
Number of nodal restraints
Nodal restraint list
Number of restrained nodes
Number of stories
Number of time steps
Number of repetitions
Angular frequencies w
Moving load
Eigenvectors ct> (mode shapes)
Poisson's ratio
Rotation matrix
Radii of gyration of rigid bodies with
respect to centers of mass
Mass density p
Element stiffness matrix
Story height
Structural stiffness matrix
Stresses
Times
Time
Initial velocities of nodes
Initial velocity of moving load
Initial velocities of bodies
Velocities in normal coordinates
Velocities of structural nodes
Nodal coordinates
Coordinates of point c
Components of offset vectors
Components of offset vectors
Torsion c:onslanls /, of cross scc:lions
Sltwul 111011wnts of arcu /1, 1111d I of
l ' I OHM ~111io11s

General
References

TEXTBOOKS ON STRUCTURAL DYNAMICS

(CHRONOLOGICAL ORDER)

1. Rogers, G. L. , Dynamics of Framed Structures, Wiley, New York, 1959.

2. Norris, C. H. , et al. , Structural Design for Dynamic loads, McGraw-Hill, New
York, 1959.
3. Biggs, 1. M. ,Introduction to Structural Dynamics, McGraw-Hill, New York , 1964.
4 . Hurty, W. C., and Rubinstein , M. F., Dynamics of Structures, Prentice-Hall, Englewood Cliffs, N. J., 1964.

5. Lin, Y. K., Probabilistic Theory ofStructural Dynamics, McGraw-Hill, New York,

1967.

6. Przemieniecki, J. S., Theory of Matrix Structural Analysis, McGraw-Hill, New

York, 1968.

7. Rubinstein, M. F., Structural Systems-Statics, Dynamics, and Stability, Prentice

Hall, Englewood Cliffs, N. J., 1970.
8. Fryba, L., Vibration of Solids and Structures under Moving loads, Noordhoff,
Groningen , The Netherlands, 1972.
9. Fertis, D. G. , Dynamics and Vibrations of Structures, Wiley, New York, 1973 .
10. Clough, R. W. , and Penzien, J. , Dynamics ofStructures, McGraw-Hill, New York ,
1975.
11. Belytschko, T., Osias , J. R ., and Marca!, P. V., Finite Element Analysis of Transient
Nonlinear S1r1U'/11ml ll<ltGvior , ASME, AMD , Vol. 14, 1975.

12. Bathe, K . J,,

1111d WIINon, E. L ., Numerical Methods in Finite Element Analysis,

Prcnlicc-llnll , I111plnv1111cl C'lilTN, N. J., 1976 .

491

492

{1111111r11l llnftunm 111

13. Blevins, R. I)., Floll' l11rl11, ,rl I l/1111tlr111.,, V1111 Nostrnnd lfrlnliold , NI'\\ , 1
1977.
14. Simu, E., and Scanlan, R. II ., Wind HJ)ict.1 011 Strt1ct11r,s, Wiley, New Y111 ~, I J
15. Meirovitch , L., Computational Methods in Structural Dy11a111ic.1, Sljlliufl
Noordhoff, Alphen aan den Rijn, The Netherlands, 1980.
16. Craig, R. R ., Structural Dynamics, Wiley, New York, 198 1.
17. Paz, M. , Structural Dynamics, 2nd ed., Van Nostrand Reinhold , New Ylll ~. l'Jk

TEXTBOOKS ON VIBRATIONS
(CHRONOLOGICAL ORDER)

1. Rayleigh, J. W. S., The Theory of Sound, Dover, New York, 1945.

2. Den Hartog, J.P., Mechanical Vibrations, 4th ed. , McGraw-Hill, New YOI~, lJ~h
3. Myklestad , N. 0., Fundamentals of Vibration Analysis, McGraw-Hill, Nm , 111h
1956.
4 . Jacobsen, L. S. , and Ayre, R. S., Engineering Vibrations, McGraw-lltll , N
York, 1958.
5. Bishop, R . E. D. , and Johnson, D. C., The Mechanics of Vibration, C11111l11hl11
University Press, London, 1960.
6. Tong, K. N., Theory of Mechanical Vibration, Wiley, New York, 1960 .
7. Church, A.H. , Mechanical Vibrations, 2nd ed. , Wiley, New York, 19(,l
8. Crandall , S. H., and Mark, W. D., Random Vibration in Mechanical S11t, 111
Academic Press, New York, 1963.
9. Bishop, R. E. D., Gladwell, G. M. L. , and Michaelson, S., The Matri, A1111/\ 11
of Vibra,ion , Cambridge University Press, London, 1965.
10. Chen, Y., Vibrations: Theoretical Methods, Addison-Wesley, Reading, ~111 ~
1966.
11. Anderson, R. A., Fundamentals of Vibrations, Macmillan, New York, I% I
12. Vernon, J.B ., Linear Vibration Theory, Wiley, New York, 1967.
11. Vierck, R. K., Vibration Analysis, International Textbook, Scranton, Pu. , J 11(, I
14 . llaberman, C. M. , Vibration Analysis, Charles E . Merrill, Columbus, Ohru , l!JriN
15. Thomson, W. T., Theory of Vibration with Applications, Prentice-Hall , Englrn 11, ut
Cliffs, N. J., 1972.
l <i. Timoshenko, S . P., Young, D. H ., and Weaver, W. , Jr. , Vibration Prohl,1111 /11
H11,:i11eering, 4th ed., Wiley, New York, 1974.

17. Meirovich, L. , Elements of Vibration Analysis, McGraw-Hi ll , New York, Jilt,

IH. Newland, D. E., An Introduction to Random Vibrations and Spectral 1\11,1/\ 111
Longmans, London, 1975.
I9. Tse , F S., Morse, I. E., and Hinkle, R. T. , Mechanical Vibrations 'J'h,1111
Applications , 2nd ed., Allyn and Bacon, Boston, 1978.

,111

493

TEXTBOOKS ON FINITE ELEMENTS

(CHRONOLOGICAL ORDER)

1. Przemieniecki, J. S. , Theory of Matrix Structural Analysis, McGraw-Hill, New

York, 1968.
2. Desai, c. s., and Abel, J. F., Introduction to the Finite Element Method, Van
Nostrand Reinhold, New York, 1972.
3. Oden, J. T. , Finite Elements of Nonlinear Continua, McGraw-Hill, New York,
1972.
4. Martin, H. C. , and Carey, G . F., Introduction to Finite Element Analysis, McGrawHiU, New York, 1973.
5. Norrie, D. H. , and de Vries, G., The Finite Element Method, Academic Press, New
York, 1973.
6. Strang, G. , and Fix, G. J., An Analysis of the Finite Element Method , Prentice-Hall,
Englewood Cliffs, N . J., 1973.
.
7. Gallagher, R. H., Finite Element Analysis Fundamentals, Prentice-Hall, Englewood
Cliffs, N. J. , 1975.
8. Bathe, K. J., and Wilson, E. L., Numerical Methods in Finite Element Analysis,
Prentice-Hall, Englewood Cliffs, N. J. , 1976.
9. Hinton, E. , and Owen, D.R. J. , Finite Element Programming, Academic Press,
London, I 977.
10. Desai, C. S . , Elementary Finite Element Method, Prentice-Hall, Englewood Cliffs,
N. J., 1979.
11. Cheung, Y. K., and Yeo, M. F. , A Practicallntroduction to Finite Element Analysis,
Pitman, London, 1979.
12. Hinton, E., and Owen, D.R. J. , An Introduction to Finite Element Computations,
Pineridge Press, Swansea, Wales (United Kingdom), 1979.
13. Owen, D. R. J. , and Hinton, E., Finite Elements in Plasticity, Pineridge Press,
Swansea Wales (United Kingdom), 1980.
14. Cook, R.' D., Concepts and Applications ofFinite Element Analysis, 2nd ed. , Wiley,
New York, 1981.
15. Becker, E. B., et al., Finite Elements (five volumes) , Prentice-Hall, Englewood
Cliffs, N. J., 1981-1984.
16. Bathe, K. J., Finite Element Procedures in .Engineering Analysis, Prentice-Hall,
Englewood Cliffs, N. J. , 1982.
.
17. Huebner, K. H. , The Finite Element Method for Engineers, 2nd ed., Wiley, New
York, 1983.
18. Weaver, w., Jr., and Johnston, P. R. , Finite Elements for Structural Analysis,
Prentice-Hall, Englewood Cliffs, N. J., 1984.
19. Segerlind , L. J., Applied Finite Element Analysis, 2nd ed., Wiley, New ~ork, ~985.
20. Zienkiewicz, O. C . , The Finite Element Method, 4th ed., McGraw-Hill , Maidenhead, Berk shill', Fnj.\lnnd, 1987.

494

General Referenc11

TEXTBOOKS ON MATRIX ANALYSIS OF STRUCTURES

(CHRONOLOGICAL ORDER)

1. Laursen, H. I., Matrix Analysis of Structures, McGraw-Hill, New York, 1966.

2. Martin, H. C., Introduction to Matrix Methods of Structural Analysis, McGruw
Hill , New York, 1966.
3. Rubinstein, M. F, Matrix Computer Analysis of Structures, Prentice-Hall, 111
glewood Cliffs, N. J. , 1966.
4. Hall, A. S., and Woodhead, R. W., Frame Analysis, 2nd ed. , Wiley, New York
1967.
5. Willems, N. , and Lucas , W. M. , Jr., Matrix Analysis for Structural Engin1'1'11
Prentice-Hall, Englewood Cliffs , N. J. , 1968.
6. Beaufait, F. W. , Rowan, W. H. , Jr., Hoadley, P. G., and Hackett, R. M. , Co11111111,,
Methods of Structural Analysis, Prentice-Hall, Englewood Cliffs, N . J ., 1970
7. Rubinstein, M. F. , Structural Systems-Statics, Dynamics, and Stability, Prc11ll1 1
Hall, Englewood Cliffs, N. J., 1970.
8. Wang, C. K. , Matrix Methods of Structural Analysis, 2nd ed., International '1'1
book, Scranton, Pa. , 1970.
9. Meek, J. L., Matrix Structural Analysis, McGraw-Hill, New York, 1971.
10. Kardestuncer, H., Elementary Matrix Analysis of Structures, McGraw-Hill , Nr w
York, 1974.
11. Vanderbilt, M. D. , Matrix Structural Analysis, Quantum, New York, 1974 ,
12. McGuire, W., and Gallagher, R.H., Matrix Structural Analysis, Wiley, New Y111k
1979.

13 . Weaver, W., Jr., and Gere, J. M ., Matrix Analysis of Framed Structures, 2nd Id
Van Nostrand Reinhold, New York, 1980.
14. Meyers, V. J., Matrix Analysis of Structures, Harper and Row, New York , 1'1"1
15. Holzer, S. H., Computer Analysis of Structures, Elsevier, New York, 19H~

Appendix A
Systems of Units
and Material Properties

A.1 SYSTEMS OF UNITS

The two most commonly used systems of units are the International System (SJ
units) and the United States Customary (US units). The first of these is called
an absolute system because the fundamental quantity of mass is independent of
where it is measured . On the other hand, the US system has force as a fundamental quantity. It is referred to as a gravitational system because the unit of
force is defined as the weight of a certain mass, which varies with location on
Earth.
In the SI system, the three fundamental units required for structural dynamics are mass (kilogram), length (meter) , and time (second). Corresponding
lo mass is a derived force called a newton, which is defined as the force needed
lo accelerate one kilogram by the amount one meter per second squared. Thus ,
we have
1 N = l kg m/s 2
which is based on Newton's second law that force = mass x acceleration.
In the US system, we use force (pound), length (foot) , and time (second).
(Note that the unit of time is the same for both systems.) Corresponding to force
IN u derived mass, which carries the name slug. This quantity is defined as the
11111ss that will be n<.cclcrulcd one foot per second squared when subjected to a
lorcc of one po1111d I ll'11tt,

lhlll t'Olllt'N I 111111 lltt 111111111111 1111111N

lcH('l'/ncnkrnlion ,

496

Systems of Units and Material Properties

Table A. I presents conversion factors for calculating quantities in SI unit ~

from those in US units. The factors are given to four significant figures, whid1
usually exceeds the accuracy of the numbers to be converted. Note that stress i~
defined in SI units as the pascal. That is,
1 Pa= l N /m 2
TABLE A.I

Conversion of US Units to SI Units

Quantity

US Units

Length
Force
Moment
Stress
Mass

inch (in.)
kilopound (kip or k)
kip-inch (k-in.)
kip/inch2 (k/in. 2 or ksi)
kip-sec2/inch (k-s2/ in.)

x Factor
2.540
4.448
1.130
6.895
1.751

10- 2

10- 1
103
X 102
X
X

Sec. A.2

Material Properties

(GN/m 2 or GPa) , acceleration ii in meters per second squared (m/s2), and so on.
[Note that the force kilonewton corresponds to the mass megagram (Mg) .] Also,
in US units we give force Pin kilopounds (kips or k), length Lin inches (in.),
modulus E in kips per square inch (k/in. 2 or ksi), acceleration ii in inches per
second squared (in./s2 ) , and so on.

A.2 MATERIAL PROPERTIES

= SI Units

meter (m)
kilonewton (kN)
kilonewton meter (k.N ru)
kilopascal (k.Pa)
megagram (Mg)

To analyze solids and structures composed of various materials, we need to

know certain physical properties. For structural dynamics, the essential material
properties are modulus of elasticity E, Poisson' s ratio v, and mass density p.
Table A.3 gives these properties in both US and SI units for some commonly
used materials. Note that the shearing modulus G is not listed in the table
because it can be derived from E and v.

For any numerical problem in structural mechanics , we must use a co11.111

tent system of units . By this we mean that all structural and load parameters 11111~t
be expressed in the same units within each system. Some examples of consisll'III
units for force, length, and time appear in Table A.2. For instance, in SI( I ) wr
must express an applied force Pin newtons (N), a length Lin millimeters (111111 1,
the modulus of elasticity E in newtons per square millimeter (N/mm 2). 1111
acceleration ii in millimeters per second squared (mm/s2 ), and so on.
TABLE A.2
System

SI

us

49:

Consistent Systems of Units

Force

Length

Time

(I)
(2)
(3)

newton
kilonewton
meganewton

millimeter
meter
kilometer

second
second
second

(!)
(2)
(3)

pound
kilopound
megapound

foot
inch
yard

second
second
second

When programming structural dynamics for a digital computer , it ,~ ,

pecially important that the system of units for input data be consistent. <llhr1
wise, units would have to be converted within the logic of the program, thl'll h
restricting its usage. For example, if in US units the length L were given 111 I I
and the modulus E were expressed in pounds per square inch , the progr11111 w1111lll
need to convert either L to inches or E to pounds per square fool.
For all of the numerical examples and problems in this book wt 11~l' !'Ith
SI(2) or US{2) in Table A. I . Thus, in S r units we lake force Pin ~il1111rnt1111
(kN), length /, in lllL'tcrs (m), modulus H in gig111Hwtons 1wr squ1111 1111 I

TABLE A.3

Properties of Materials

Modulus of Elasticity E
Material
k/in. 2
Aluminum
Brass
Bronze
Cast iron
Concrete
Magnesium
Nickel
Steel
Titanium
Tungsten

1.0
1.5
1.5
1.4
3.6
6.5
3.0
3.0
1.7
5.5

X
X
X
X
X
X
X
X
X
X

104
104
104
104
103
103
104
104
104
104

GPa

69
103
103
97
25
45
207
207
117
379

Poisson's
Ratio 11
0.33
0.34
0.34
0.25
0. 15
0.35
0.31
0.30
0.33
0.20

Mass Density p
k-s2 /in. 4
2.45
8. 10
7.80
6.90
2.25
1.71
8.25
7.35
4.20
1.80

X
X
X
X
X
X
X
X
X
X

I0- 7
I0- 7
I0- 7
10- 7
10- 7
10- 7
10- 7
10- 7
10- 7
10- 6

Mg/m3
2.62
8.66
8.34
7.37
2.40
1.83
8.82
7.85
4.49
19.2

"Numbers in this table are taken from J. M. Gere and S. P. Timoshenko, Mechanics of
Materials, 2nd ed. , Brooks/Cole, Monterey,. Calif., 1984.

Sec. B.1

499

Inverse Iteration

on the left-hand side, we find that the reverse iteration converges to the fundamental mode. To simplify notation, let

X; = <I>;

(4)

Then Eq. (3) beco mes

(5)

Appendix B
Eigenvalues
and Eigenvectors

which will be used in the iteration procedure.

To begin inverse iteration , we first assume an approximate shape of the
fundamental mode. The usual arbitrary choice for this starting vector is a column
of ones. Thus,
(X 1) 1 = {l , I , I , ... , 1}

(6)

Substituting this vector into the last of Eqs. (4), we calculate (B 1) 1 as

(B 1) 1

= M(X 1) 1

(7)

Then solve the simultaneous algebraic equations in Eq. (5) to obtain

(Y1) 1
8 .1 INVERSE ITERATION

All eigenvalue solution routines are iterative because we seek the roots ot tho
<'11<1racteristic equation (see Sec . 3.6) , which is a polynomial of order n, 1111
mogencous action equations of motion provide the form of the eigenvul11
problem to be solved. Thus, we have
S <I>; =

WTM <I>;

(i

1, 2, . . . , n)

(I)

s = UTD u

()

In this lype o f factorizatio n, D is a diagonal matrix; and U is upper trinn itlllM

with vulues of unity in diagonal positions. Substituting Eq . (2) and
I/ A
11110 Eq . ( I ) produces

,vi

Ur I) U A1<l>1

(8)

This expression is only an approximation, unless the estimated mode shape

satisfies Eq. (5) exactly. Also, a first approximation to the eigenvalue A, may be
found by dividing any term in vector (Y1) 1 by the corresponding term in (X1), .
That is,

(A) =
I I

(}j),

(1 :s j :s n)

(XJ,

(9)

If (};) 1 is chosen to be the largest (positive or negative) term in vector (Yi) ,,

normalization with respect to that value gives us the second trial vector:

which is a slightly rearranged version of Eq. (3.6-4). Direct (or for11111tl)

ifl'f'otion of Eq . (1) would involve substitution of trial vectors for <I>, 011 lh
ll'ft hand side and evaluation of WT on the right-hand side. This techniqul' 11 I
converges to the dominant (largest) eigenvalue w?, and the corresponding l' ijll'll
wtlor <1>11 J lowever, to extract the smallest angular frequency first , we muNI 1111
" '"''I',\'' (or inverse ) iteration. For this second approach, let A; = I / "'i 1111~
fnclor the stiffness matrix S by the modified Cholesky method [2 , 3], as folloWM

= (A 1).(X 1) 1

(X,h = (A,), (Y1)1

(10)

in which the normalization constant is (11. 1) 1 itself. This procedure is repeated

until the eigenvalue A1 and its associated eigenvector X, are determined to some
specified accuracy.
In the kth iteration, the recurrence equations for the steps described above
are:

I. Calculate vector (B 1)k as

(11 )

M <1>1

ll y 11si11H 11 iul Vl'l'lors for <1>1 011 the ri ghl hand side of Eq

( 12)

600

Eigenvalues and Eigenvectors

App. B

Sec. B.1

501

Inverse Iteration

3. Find the new trial vector (X1)k+ 1 to be

2 .72
M = m
[

(I J)
where (A 1h is the largest term in vector (Y1)t. To check convergence of thl'
eigenvalue, we use the expression

which nd is the number of significant digits of accuracy desired.

Bathe [4] recommends calculation of (wr)k = 1/(A1h from the Raylcixh

D =

and

t//10tie11t
(

2) _
WI k -

(XThS(X1h
(XT)k M(X1h

<l>TM <l>i

= 0

(i = 2, 3, ... , n)

( I~ )

( It!)

Adding this modal constraint condition to the eigenvalue equations [see Eq . (III
Wl' cnn red uce their order from n ton - l . For the reduced equations , the Sl'l 11ml
1111llk becomes dominant and also may be calculated by inverse iteration I hi
dtl1111ion iteration sequence is repeated to extract as many modes as dl'"tt ~d
1lowl'Vl'I', to retain accuracy in each iteration, we must M -orthogonali tl' rn, h
tll'W v1.c tor with those found previously. This may be accomplished 11111111
<,'111111 Schmidt orthogonalization, as shown in Ref. 4.
E111111111lt H. I
N11w w1 sh11II apply inverse iteratio n to extract the fundamental mode from lh1 1l1t II
vu ltH pt 11hllm I'm the plane truss in Fig. 3. I I (a) (sec Sec. 3.5). For thi s truss 1h1 NI ti In
111111t Ix wns found to he

0.36

w ltr1r ,1

/ /\ / /

Al1.o,

0. 1(1

0.481
0.48

0.48

I .M

tlt1 1011Mis h111 11111 ~11 11111t1i x

IN

t ,1

(b)

0.2647

1.000

s[l.3

_[l
U =

whkh produces much faster convergence. Unfortunately, improving the rail' 111
l'11t1vergence of the eigenvalue has no effect on the rate of convergence of 1111
l'l~lt1vcctor. However, improvement of eigenvector convergence can be altat11ld
wtlh spectral shifting, as described later in this appendix.
After the fundamental mode has been determined, it is usually elimi1111hd
I111111 the eigenvalue equations by the process of deflation [l , 4]. For this ll'l'h
11iq11l'. we express the orthogonality of the eigenvectors <1> 1 and <l>i with n~pr, I
lo lhl' mass matrix M, as follows:

3.28

for which m = pAL/6.

In accordance with Eq. (2), we factor matrix S to obtain

( 14)
111

60

- 0.2647
1

-0.3529]
1.333

(c)

(d)

As the starting eigenvector, assume

(X 1) 1 = {I , 1, 1}

(e)

In the first iteration, we evaluate (B 1) 1 from the last of Eqs. (4), as follows:

(B1)1 = M(X1)1 = m{3 .72, 4.28, 3.28}

(f)

Then solve for (Y1) 1 from Eq. (5):

m
(Y, ), = - {8.000, 23.12, -2 .426}

(g)

Normalize this vector with respect to its largest term to find the new trial vector
1

(X,)z = - (
) (Y1 ) 1 = {0.3460, 1.000, -0.1049}
A1 I

(h)

In this iteration the first approximation to the eigenvalue is

m
(A 1) I = 23.12-

(1')

which is the largest term in vector (Y1)1.

Results of successive iterations are listed in Table B.l. We see that convergence
(to fo ur significant digits) for the reciprocal A1 of the eigenvalue wf occurs in six cycles
of iteration. However, the eigenvector X1 requires another cycle to attain the same
nccuracy. Their final values are
m
2
1
s
w, = - = 0.04501 A, = 22.22(j)
A,
m
s
1111d
X,
{0.2314, 1.000, 0.2472}
(k)

wl11l'11

1111 1111 ~111111 11~ 1lt11~1 111111111

hy Pro~rnm Vllll'T 111 Ex11111pll' l I

Sec. B.1

503

Inverse Iteration

Spectral Shifting

r-

~
~

8:

"<I"
N
- 0 r<">0"<1"
NON

c:;;....;c:;;
I

To improve the rate of convergence of eigenvectors, we use the technique

known as spectral shifting. For this purpose, let the constant ai be a number
close (but not equal) to WT, which is the eigenvalue to be calculated. Then
subtract aiM ct>i from both sides of Eq. (1), and divide by
ai to obtain

wr -

N
N

c--i

(17)

S* A; l>i = M l>i

where
"<I"

'D

8:

;;; 8~

NON

o....; o
I

'
~8~
NON
0 ....;c:;;

N
N

s* = S

c--i

V)

8:

A;=---

wT-

wr.

N
N

c--i

s* = u*To* u*

..

"""

8:

"""

<"l
<"l
N

Also, the eigenvalue

0
V)

"""
N

o....; o
I

<"l

!"i

2
W

8:

r' '
N

'D
"<I"
N

c:;; ....;c:;;
I

Q.)

8:
~

8:
A

C 8:
--'--

~s'

o....; o
I

-- ..
9
~

1
=-*+a
Ai
I

(20)

which need only be computed once (after convergence).

Spectral shifting may be used to aid the following tasks:

r-

!"i
N

1. Extraction of modes near an expected resonance (ai--" n), where !l is the

angular frequency of a periodic forcing function.
2. Extraction of modes corresponding to repeated roots (ai--" w;,) of multiplicity m, including rigid-body modes (ai--" O)
3. Extraction of sequential modes by using (ai--" wT) after deflation

0
N
~

(19)

wr becomes
I

<"l

(18)

ai

By this manipulation , the dominance of A; in Eq. ( 17) is greater than that of Ai

in Eq. (3), because in the second of Eqs. (18) we see that A;--" oo as ai--"
After the spectral shift, inverse iteration proceeds as before, except that the
factorization of s* is expressed as

'D

~
~

- aiM

"""

V)

!"i
N

If the stiffness matrix Sin Eq. (1) is semidefinite, one or more rigid-body modes
exist. In that case, a small (but finite) spectral shift away from zero makes s*
in Eq. (19) positive-definite. Then by inverse interation we can extract the first
rigid-body mode. Elimination of that mode from the equations by deflation
allows iteration for the second rigid-body mode, and so on. A similar procedure
is required when calculating nonzero repeated eigenvalues.

..,.;
N

~
"'IE:

Example B.2
Let us use a sp~ctrnl shif1 to make the second mode dominant for the plane truss in
Example n . 1. Thi tollowl ng shift

(f)

no4

Elgonvaluoe and Eigenvectors

App

Sec. B.2

w1ll ,.11ll11w, cxtrm:tion

) or the second eigenvalue and eigenvector w1'thout dcflat1011. I111111
I i11st o , 1,qs. ( 18,
111
the modified stiffness matrix becomes
-0.72

0.3808

S"' - S - a2M = s -0.72

[
- 0.48
1111ltmi1.11tion

or S"' as in Eq.
D*

-0.48
0.48

0.48

(1111

0. 3808

- 2.182

-oLJ

111111

TABLE B.2 Iteration of Second Mode

Cycle

(X2h

(X2)1

(X2h

(X2h

(X2)

(X2)s

Vector

1
1

0.8904
-0.1400
1.000

0.8677
-0. 1701
1.000

0.8673
-0.1715
1.000

0.8673
- 0. 1715
1.000

0.4592

(19) yields

- 0.8208

(11)

505

Transformation Methods

.!... (Ai)t
m

-92.68

-84.2 1

-83.27

-83.25

-83.25

B.2 TRANSFORMATION METHODS

_
[l
U* = 0

- 1.891
I

- 1. 261]
0.1~59

Io J

As lw lon:, we ussumc the starting eigenvector

(X2)1 = {I, I , J}

q,1

tlw first iteration, the vector (B2 ) 1 is computed to be

111

Jql

(82), = M(X2), = m{3.72, 4.28, 3.28}

AV=VA

wll1d1 rnnt uins the same values as Eq. (f). Solution of Eq. (17) for vector (Y2 ), p, rnhll
m
(Y2)1 = ;{- 82.52, 12.97, -92.68}

(I)

N111 11111 li z11lion or this vector with respect to its largest term gives
I

(X2h
Ill

= (A;), (Y2)1 = {0.8904,

When most or all of the eigenvalues and eigenvectors are desired, transformation methods [5] prove to be more efficient than inverse iteration. In this
section we describe the Jacobi, Givens, and Householder transformation procedures, as well as QR iteration. To confirm ideas, a numerical example follows
the discussion of each approach.
As a preliminary matter, we assume that the expanded eigenvalue problem
has been converted to the standard, symmetric form:

- 0.1400, 1.000}

as described in Secs . 3.6 and 4 .2. In Eq. (1) the symbol V denotes an orthogonal
modal matrix of n eigenvectors that are normalized to unit lengths and listed
column-wise. Also , the eigenvalues ,\ 1 , ,\2, . . , ,\11 appear in diagonal positions
of the spectral matrix A.
The basic process in all transformation methods is to diagonalize matrix A,
as follows:
VTA V = A

wh1l'11
(At), = - 92.68~

UI

.!'

l,u11

l'uhk 13.2 ~h.ows .the resu lts of successive iterations. In this case conver~Mlt (hi
At and the second eigenvector X 2 occurs in fi w I y 1I

Nignilk-111_11 d1g_11s) for both

l-111111 vnlms for llu s mode arc

At

- 83.25

111
,\'

W2 =

I
7*'

+ a2 = 0.3480

"2

1111d

X,
wh11

h 1111 11i 11111 11111111 lo thos1

1

j0.8<17.l, - 0. 17 15, 1,000}

111 Ex11111pk

l.4 ,

Ill

(11)

( l)

(2)

When this is accomplished, we have found not only the spectral matrix A but
also the normalized modal matrix V.
Jacobi Method

Sinee its development in the nineteenth century [6] , the Jacobi method has
enjoyed extensive usage by mathematicians, scientists, and engineers throughout the world. The essential idea in this approach is to zero a selected offdiagonal term A,,., = Aqp of the coefficient matrix A in Eq. (1) . This is accomplished by oplrnting upon that matrix with a generalized rotation matrix R.kin

step k, us follow~

1508

I lunrwnl11n1 1111d t l11n11vncto11

The form of the

II

II

Soc. B.:.>

rotutiou 11111111 11 Is

1rorrnformotlon Methods

and Eq. (8) becomes

(10)
which is an approximation of Eq. (2).
The Jacobi method may be applied selectively to annihilate the largest
off-diagonal term in matrix A. However, this approach requires searching for
that term before it can be put to zero. More commonly, we operate in sweeps,
systematically annihilating all of the off-diagonal terms by rows or columns.
Each sweep of this kind requires approximately 2n 3 multiplications. Equations
(6) give the formulas to be used for altering either the upper or lower triangular
part of matrix A.
During a particular sweep, terms that are zeroed do not necessarily remain
zeros. However, the square root of the mean of the squares (RMS) for offdiagonal terms is reduced after each annihilation (and even more dramatically
after each sweep).

sin fh-------- row I'

-sin ()k

cos 8c - - ----- row q

I
I
I
I
I
I

I
I
I
I
I
I

col. p

col. q

In this rather sparse array, we define the terms

(Rpp)k
(Rpq )k

= (Rqqh = cos fJk

= - (Rqp h = sin fJk
t= p

(Ruh = 1

(i

(Rij h

(i or j

=0

(;")

Example B.3

1,111

With one sweep of the Jacobi method, determine approximately the spectral matrix A
and the modal matrix V associated with the 3 X 3 array

or q)

(~d

t= p or q)

(111)

= (APP cos 2 fJ + 2Apq cos () sin fJ + Aqq sin 2 fJ)k

(Aqq)k+ , = (App sin2 () - 2Apq cos () sin () + Aqq cos2 fJ)k
(Aq)k+, = [(Aqq - App) cos ()sin()+ Apq(cos2 fJ - sin2 fJ)]k
= (Aqp h+ I
(App)k+r

(A,1,)H r = (A;p cos () + A;q sin fJ)k = (Ap;)k+,

+ Aiq cos

(Aiq)k+1 = (-AiP sin fJ

8)k

= (Aq)k+ ,

(fut)

tan 28k

Io~ I -s 'Tr/ 4.

After nr rotations, Eq. (3) yields

T
R,,, ... R2R,A 1 R"T""
1 R2
A

....

..

"'T
R,,
"" A
r

Tlwrcforc, the matrix of orthonormal eigenvectors is

V -= ""T
"T
Kr R
2

""'
K I'
11,

-0.48
1.44

0 ]
-0.48

- 0.48

1.92

(a)

In the first transformation of matrix A, we shall make the term A,2 = A21 equal to zero.
Therefore, p = 1 and q = 2, so Eq. (7) becomes

((1h)

tan

281

2A12
= (2)( - 0.48) = I
2
A11 - A22 0. 64 - 1.44

(b)

From this expression we determine the angle 8, and its sine and cosine as
( (1\:)

(<111)

81 = 0.4380

sin 81 = 0.4242

cos 81 = 0.9056

(c)

Then the first generalized rotation matrix in Eq. (4) takes the form

((10)

'l'lw unglc 8k is chosen so that the term (Aq)k+ , becomes zero. Thus, fro111 I 11
(C,l') we have

( 2Ae9 )
Ap - Aqq k

0.64
-0.48

A =

'l'ht! multiplication in Eq. (3) alters only terms in rows and columns p and , 111
1
mutri x A , which become

whlrc

507

It =

0.9056 0.4242
- 0.4242 0.9056
[

0~ ]
(d)

and the first transformation in Eq. (3) produces

0.4152
,

, T

A2 = R , A1R1 =

- 0.2036

0
1.665
-0.4347

- 0.2036]
- 0.4347

(e)

1.920

As a conNl'< t111,mt, lht fll'sl rotation has reduced the RMS of off-diagonal terms in matrix
A from OAH 111 II I 11lI
h11 1111 N1111ud 1111us twm11ti o n, WL' ,unl..1 A,, A II tq1111l 10 11111, so 1h11t

608

Eigenva lues and Eigenvectors

tan

" _
2 u2
-

2A13
A11 - A33

---

App,

Sec. B.2

0.8927

(2)(-0.2036)
0.4 152 - 1.92 = 0 2706

(1)

= RTRI RI =

Thus, we have
sin (Ji= 0.1317

82 = 0.1321

-0.0~727
[

- 0.05727
1.665

O ]
- 0.4309

-0.4309
1.947
for which the RMS of off-diagonal terms is 0.3074.
The third transformation involves zero1ng A 23 -- A 32, wh'ch

1 gives
2A23
tan 283 =
= (2)(-0.4309) _
A22 - A33
1.665 - 1. 947 - 3.o53
I knee,

(h

= 0.6272

sin 83 = 0.5868
Sil the third rotation matrix becomes

R,- G

-0.5868

tan Ok = (Ap-i ,q)

cos 83 = 0.8097

I I

( j)

0.5~68]
0.8097

( I)

-0.04637
1.353

0.03361

0.03361]
0

0 .30881

0
1.353

O
0

2.259

2 2 'ii)

nddllhlll, IIJlJll llXi lllillt' l'i)tl'II Vl'l't\ll ,~ Ill\' l'llk 11) 11(1d IIN

0
0
0

(12)

(Ill)

. le ent <~ 11e sweep, the diagonal terms in matri x A4 represent ,m,d 111110 11
the eigenvalues of the original matrix . That is,
g

111

(1::)2

An-2 =

l11111t1011s lllr

Rotation in the 2-4 plane yields

(k)

wlt111 tlw RMS of off-diagonal terms is O 0405 1 Note that each off d'
t I
.
.

- 1agona I te1111 IN 1111w

n 111st 1111 order ol magnitude smaller than at the beginning of th
At ti . . I r I .
.
e sweep .

(11)
k

and so on. After we clear the first row (and column), matrix A has the form

0.3881
- 0.04637
[

Ap- 1,p

where Ok :s 7r/2 . For the first row of matrix A, rotation in the 2-3 plane gives

Suhstituting It into Eq. (3) results in

A,1 = RJ A3 RT =

0 .8026

(ll

0
0.8097

0.5817

(o)

Instead of annihilating the p, q term (as in the Jacobi method), Givens [7]
proposed annihilating the p - 1, q term instead. Then a zeroed term will remain
zero during a forward sweep. For this purpose, we set Eq. (6e) equal to zero to
obtain

Using R2 in Eq. (3) yields

A3 = R2 A2 RI =

0.1317

0.1523]
-0.5767

(h)

- 0.1317

0.388 1

-0.4 135
0.7005

Givens Method

0.9913
0

R2 =

0.4205

Modal vectors appearing column-wise in this matrix are automatically normalized to

have unit lengths.

cos 82 = 0 .9913

and the second rotation matrix is

,

509

Transform ation Methods

1111

T his technique produces a tridiagonal matrix T in a finite number of steps,

consisting of n, = (n - 2)(n - 1)/2 rotations. The entire sweep requires approximately 4n 3 /3 multiplications.
While the Givens method does not lead directly to the spectral matrix, we
consider it a useful preliminary to the Householder method, which is described
next.

Example U.4
By tlu1 <JivrnN 1m,1h111I, 111111Hforn1 th1 followln1,14 X 4 11111trix to tridiogonnl l'o1n1:

~---------111111111111111111111111111111
610

I l11u11vnl110111111d l:lgo11vooto1H

A=

0.48

1,0()

- 0.48
U<,

1.64

- 1.00

1.36

0.48

Sec. B.2

oq

Substituting this matrix into Eq. (3) yields

(p I

1.64
3 with a rotation in the 2-3 plane. For this p11IJH1~1.
We
start
by
annihilating
the
term
A,
Eq. (11) results in
3
-1 = tan - I A1
- = tan -

01

1.00
-0.48

A12

Then we have

Transformation Methods

= 1.123

1.360
I.109

A3 = R2A2RI =
[

-0.4463

-l.109
1.412
-0.4
463

- 0.1953

1.538

-o.iL,]

which is now in tridiagonal form.

1.689

(q)

Householder Method

sin 01 = 0.9015
and the first rotation matrix R. 1 becomes

cos 01 = 0.4327

(1)

r
i
0

R, -

0
0.4327 0.9015
-0.9015 0.4327
0

Using this operator in Eq. (3) gives

A Householder transformation operator [8] causes all of the terms in a

vector to become zero except the first, which becomes the length of the vector
itself. The operator has the form

f]

AT

A,-R,A,R, -

[ -1.109
13~

-o. 1092

l.412
-0.1092

l.588

0.4327

04~27]

(II

0.2077

0.2077

1.640
In this case only one transformation is required to clear the first row (and column) 011f~hh
of
triangular
part. The term A 14 = A4 1 was zero initially and remained unaffctt,cl h\
thethe
operation
above.

82

= tan

= tan

-I

A23

0.4327
- 0.1092

Then fhc second rotation matrix

R.2

= pT = p - 1

fl

s2

= L aJ

(16a)

J=l

0.2447

0.9696

0.9696

0.2447

e=--

. Vc1c

(16c)

In Eq. (16b) the sign of s is taken to be the same as that of a 1 , so that there is
no possibility of getting zero for the first term in vector c. The unit vector e in
Eq . (13) is the result of normalizing vector c to unit length, as indicated in Eq.
(16c).

is

(16b)

(Ill

cos 82 = 0.2447

, O}, wheres

(15)
We form the matrix P by first creating e in Eq. (13) from the vector a , as follows:

= - I.324

(14)

Pa=b

c = {a1 s, a2, a3, . .. , a,.}

From thi s we get

sin 82 = - 0.9696

(13)

Hence, P is equal to its own inverse.

Consider a vector a to be converted to b = {s, 0, 0, .
is the length of a. That is,

The second step in this example is to annihilate element A by a rotation l11 tl11
3-4 plane. Thus, Eq. (11) produces
24
- I A24

- 2e eT

in which e is a col umn vector of unit length. The matrix Pis both symmetric and
orthogonal, so that
p

-1.109

=I

Now let us express the product Pa in Eq. (15) in terms of the vector c.
Using Eqs. (13) and (16c), we obtain

fi

(\\ I

CT)

Pa = \1 - 2cTc a = a Rcsull s ol flll'

l1111!'l'

2c(c Ta)
(cTc)

(17)

producls (in pi11e11Jlll'sts) from llw strnnd form of Hq . ( 17)

512

Eigenvalues and Eigenvectors

App. II

Sec. B.2

Transformation Methods

513

are
Recurrence equations for the kth step are the same as those in Eqs. (19) and (21),
except that k replaced 1 and the lower limit on j is k + 1.
After n - 2 Householder transformations, we evolve the tridiagonal
matrix

era= (a 1 s)a 1

+ 2 aJ = s 2 a 1s
j =2

c Tc

= (a1

s)

II

2 aJ = s 2 2a 1s + s 2 = 2(s

j =2

a 1s)

(22)
where the Householder matrix is

Substitution of these expressions into Eq. ( 17) produces

2

=a

p a

s)
2(s a 1s)

_ 2c(s

a1

= b

( 18)

as desired.

In a similar manner, the last n - 2 terms in the first column of th1

coefficient matrix A may be zeroed . With this objective, we let

2 AJ.1

= {O, A21

S1, A 31, A41, . . . , A,.1}

( 191,)

( 19\')

VcTCi

From the same matrix A used in Example B .4, we shall create a tridiagonal matrix T,
using Householder transformations. To begin, let us annihilate the first row and column
outside the tridiagonal region, as shown by Eqs. (19) and (21).

I - 2e 1 ef

L" AJ,

= (-0 .48)2 + (-1.00)2 = 1.230

SI

= 1.109

C1

= {0, A21 s1, A 31,

e,

(z)

C1

C1

\/c"fCi =

l.

878

A41}

0
0

P,A1

(2())

- 0.~327
- 0.9015

A2 = P1A1PT

s,
0

()

,\'

,\'

(a')

(b')

-0.9~ 15 0 ]
0.4327 0

1.109

1.412

0.1092

0.1092

1.588

(c')

. -0.4327 0.2077

s,

A2 = Pi A1 p I =

[1300
1.109

Also, the last n - 2 terms in the first row of matrix A may be zeroed as will
by
X

{0, -1.589, -1.000, 0}

P1 -_ I - 2e, e1T -

S1

= {O, -0. 8464, -0.5326, O}

-r~

( 19d )

Then

(y)

j =2

( 1911)

e, = -C1- -

P,

Example B.5

sf =

"

j=2

C1

(23)
3

Equations (22) and (23) require approximately 2n /3 multiplications.

= {s, 0, 0, ... , O}

ST=

= Pf PI, .. PI-2

ff

=a_ c

-04~27
0.2077

(d')

1.640

Matrix A2 is the same as that found before in Eq. (t) by the Givens method, except for
the signs on several off-diagonal terms.
Similarly, for the second row and column we have
II

s~ =

A].2

= (0.1092) 2 + (-0.4327)2

0.1992

(e')

j- 3

S2

= 0.4463

C2

= {O,

(.'I l
t ,

(f')

0, A 32

S2,

Ad = {O, 0, 0. 5555 , - 0.4327}

C2

CJ

Vt !t'J

0.704 fI

{O, 0 , 0 /HKI/ ,

0 .61 tl~f

(g')
(h')

514

Eigenvalues and Eigenvectors

P2 = I

0 I
2e2eI = [ 0
0

A,~ P, A,P, ~

1.109
1.412

0
-0.4463

-0.4463

1.538

0.1953

= Pf PI=

[i

- 0.9015

oiLl

(J')

1.689

0.9696

T k+I

-08~411

( (' ' )

0.4196

(28)

(29)

= QITk

Rk

(30)

This formula gives us the means for determining both of the desired matrices,
as will be explained next.
At the operational level for this method, we wish to annihilate a lowertriangular term (1j;h of the tridiagonal matrix Ti,, where j = i + l. For this
purpose, we set equal to zero the product of row j of a Jacobi rotation matrix and
column i of matrix Ti,, as follows:

-sin O;(T;;)k

cos 8;(1j;)k = 0

(31 )

From this expression, we have

Assume that Householder transformations have converted Eq. (I) to

tan

TW=WA

(.?I)

. (}
'

(1j;h
C;

sm = - -

or

(32)

-1!.

cos (} . =
I

(T;;)k

-C;-

(33)

where

V=HW

(34)

the form

Then the Jacobi rotation matrix (R;)k premultiplies matrix Tk (for i = I, 2, ... ,
n - l) to produce

T1 = Q1R1
which is known as Givens factorization. The symbol Q1 in Eq. (26) dcnoll'II ,111
o~thogonal matrix obtained by n - 1 Jacobi rotations, and R 1 is an upp11
triangular array. The QR algorithm l9] derives its name from the factors in h1

(26).

'

Q'/' and

(T;;T,)

Consequently,

(}~ul

.
Prcniultiplication of Eq. (26) by
gives

e; =

where

= T1 into

QITkQk = RkQk-A

where n1 is the number of factorizations.

The recurrence algorithm expressed by Eq. (28) indicates that we must
generate Rk and Qk and multiply them in the sequence shown. From that equation
we see that the definition for Rk is

0.2447

QR Algorithm

Now let us factor T

W = Q1Q2,, Q,!f

(k 'l

515

which iterates to the spectral matrix A. The modal matrix Vis calculated from
Eq. (25b), in which

where

0.2206
- 0.1059

Transformation Methods

(I')

Again, matrix A3 is the s~e as t~a.t in Eq. (x), except for the signs on off-diagonal tc 1 IIIN
Thus, we see that the desired tndiagonal matrix is

0
-0.4327

Sec. 8.2

than T1. Recursively, we have

oL]

0
- 0.2447

0.9696 0.2447

['1.109
360

App, II

postmultiplication by

Q 11~111
1

(35)
After the first operation in this sequence, matrix Tk is no longer tridiagonal; and
after n - 1 operations it becomes the upper triangular matrix Rk in Eq. (28) .
Comparing Eq. (35) with Eq. (30), we see that
(36a)

( I/)

l'lw 1ww 1111111 ix 'I'.! is unotlwr 11 idingo1111l 111 rny huving s11mllt1off di11g111111I

111111 111

or

( Hih)

516

Eigenvalues and Eigenvectors

App. B

App. B

References

Therefore, matrix Qk is defined as the product of the transposes of n - I Jacobi

rotation matrices.

Example B.6
Now we shall apply one cycle of QR iteration to the tridiagonal matrix T obtained by the
Householder method in Example B.5. For the first rotation (in the 2-1 plane), Eqs. (34)
and (33) give

C1

= YTf1 +

T:i.1

sm 81 =
cos 81

n1 = Y(l .360)2 + (1.109)2= 1.755

1.109
1.755

- = - - = 0.6320
C1

Ti.1

=- =
C1

1.360
-1.755

TABLE B.3 First Cycle of QR Iteration

(R2)1

(m')

[!

0.7749

0
0

T32

(J

sm 2 = C

T-i.2

COS 82 = -

C2

T52

= Y(0.3935)

0
0

0
I

00 01
0
1 0
0

(p')

2 + (-0.4463) 2 = 0.5950

- 0.4463

0.3935
0.5950

= 0.6613

Vr ~J +
43

srn 03 = -

C3

7;,

T~3 = Y(0.7578) 2 + (0.1953)2 = o.7826

0. 1953
0 .7826

= - - = 0 .2495

0.7590

cos 0, - Ci == 0. 7826

(r')

R1

= 0.9684
(I{ift,((,)1'1'1

1.109
0
0

0
0

[0.7749
0.6320
0
0

(t ' )

(11 1 )

0
-0.4463
1.538
0.1953

oL]
1.689

(R1)1T1
1.752
0.3935
-0.4463
0

[Lr

-0.2821
- 0.3459
1.538
0.1953

J,,]
1.689

(R2R1)1T1

ot]

1.752
0.5950
0
0

[17550
0 0

0.9684

= (RTRIRD1

R1

0ll83] [1755
0

- 0.4180 - 0.4591
0.5152
0.5629 - 0. 1450
-0.7501
0.6404 -0.1650
0
0.2495
0.9684

-00.1291
'~"]
1.689

= (R3R2R1)1T1

1.752
0.5950
0
0

0
0

-0.282 1
-1.383
0.7578
0.1953

-0.2821
-1.383
0.7826
0

Qi = (RfRillr)1

-01~65]
0.5466
1.604

(x')

is listed in the lower left-hand position.

Using Eq. (27), we now compute a second tridiagonal matrix, as follows:

(N')

T2

Ri Qi =

2.468
0 .3760
o

0. 3760
1.342
- 0.5870

.0

0 .0000 0.0000
-0.5870 0.0000
0.6375 0.4001
0.4001

(y')

1.553

Note that computed zeros must appear in the upper triangular part of matrix T2 to satisfy
the symmetry guaranteed by the congruence transformation in Eq. (27) . We also see that
the RMS of off-diagonal terms is reduced from 0 .6993 (for matrix T1) to 0.4645 (for
matrix T 2). Spectral shifting can be used in successive cycles to accelerate the rate of
convergence, as explained in Sec. B. l.

(v' l

REFERENCES

(w ' l

l . Timoshonko , S. P., Young, D. 11. , and Weaver, W., Jr., Vibra1ion Problems in
l~11J(l,111rl11}( , 4th L'd., Wiley, New York, J1)74,
l. W111v,,1, W , Ji , 1111d Ci111, .I . M , Motr/1 /\1111!1w/,1 1!/' l111111NIS1r111111r1,1, 2nd 1d.,
Vn11 Nrn1t111111l lfr l11h11ld , N1w Yrnk, l11HO

'l'lw third rot11tlo11 mulrix (R,) 1 ulso uppeurs in Tublc B.3, ulong with the prodlll'I

l1111ddil in11 1 th, 11111t11 ~

~]
~]

(q')

'l'uhle 8 .3 contains the second Jacobi rotation matrix (R.2)1 below the first, and the l'l'Ntrll
of multipl ying it with (R1)1T1 is shown below the latter matrix.
The third rotation (in the 4-3 plane) involves
c3

[!

0
0
0.9684
-0.2495

Qi

= 0. 5950 = -0.7501

=- -

0
-0.7501
0.6613
0

["ffi

1.109
1.412
- 0.4463
0

(R3),

Tnhle B.3 shows this matrix in the upper left-hand position. After multiplying it with
mu1ri x T = T1 (in the upper right-hand position) , we list the product (R. 1) 1 T1 in tla
positi on below T1.
For the second rotation (in the 3-2 plane) , we have

C2 = YTi2 +

0
0.6613
0.7501
0

(0 1)

0.7749 0.6320

-0 .6320

T1

0.6341 0
0.7776 0
0
1
0
0

-0.6341
0
0

Then the first 4 x 4 Jacobi rotation matrix becomes

(R1)1 =

(R1)1

[ o.m,

(n')

= 0.7749

517

618

I l111111vnh1111 nnd I l11111w1u:tor

At>t> I

3. Weaver, W. , Jr. , and Johnston, I'. R., Vi11it1 J:'l,1111111.v for S1m, 111m/ A111,/n11 1
Prentice-Hall , Englewood Cliffs, N.J ., 1984.
4. Bathe, K. 1. , Finite Element Procedures i11 Engineering Analysis, P1c11t1n 111111
Englewood Cliffs, N .J. , 1982.
5. Wilkinson, J. H., The Algebraic Eigenvalue Problem, Oxford Univcrntly l'll'HN,
London, 1965.
6. Jacobi, C. G. J. , "Uber ein leichtes Verfahren die in der Thcoiie d11 Sh,
ularstorungen vorkommenden Gleichungen numerisch aufzuloscn," Crl'l/1' 1 I ,
Vol. 30, 1846 , pp. 51-94.
7. Givens, W. , "Numerical Computations of the Characteristic Values of a Rl11I !i\ 111
metric Matrix," Report No. ORNL-1574 , Oak Ridge National Laborut01 y, I >11t.
Ridge, Tenn. , 1954.
8. Martin, R. S., Reinsch, C., and Wilkinson, J. H., "Householder's Ttid111pn1111I
ization of a Symmetric Matrix ," Numer. Math ., Vol. 11, 1968 , pp. 18 1 l<l'i
9. Parlett, B. N. , and Kahan, W., "On the Convergence of a Practical QR Algw 1111111
Proc. IFIP Cong., 1968.

Appendix C
Flowchart
for Program DYNAPT

1. SUBPROGRAM SDATPT FOR SUBPROGRAM VIBPT

a. Problem Identification
Descriptive title

Descriptive title

b. Structural Parameters

He;iding: STRUCTURAL PARAMETERS

Subhead: NN NE NRN E RHO

NN. NE. NRN, E. RIIO

NN . NE. NRN, E. RHO

Read and write struc1ural

parameters.

1120

I l11woh1111

1111

1110111 11111 IJYNAI' I

App. C

I lowol11111 Im Pt Ol)l'Ot'II OYNAPr

521

<. Nollul ('oordlr1111t,

Heading; NODAL COORDI NATES

Subhead: NODE X Y

I
I

J. X(J), Y(J)

J
I

Read and wri1c 11od11 I

coordinates.

J, X(JJ, Y(J)

e. Nodal Restraints

L_

CONTINUE
NND = 2*NN

d. Element Information

Calculate number of

nodal displacement,
possible.

r1

Heading: ELEMENT INFORMATION

Subhead: ELEM. J K AX EL CX CY

I
I

NRL(J) = 0
3

L_

r----1

I
I
I
I
I

I
I
I
I

I, JN(f), KN(!), AX{!)

XCL = X(KN(JJ) - X(JN(l))

YCL = Y(KN{I)) - Y(JN(J)J
EL(!) = DSQRT(XCL *XCL + YCL*YCL)
CX(J) = XCUEL(I); CY(!) = YCUEL(l)

I, JN(l), KN{!), AX(!). EL{!), CX(I). CY{!)

L ____ _

Read infonnmion ltu

Element I.
Compute x and y ~01111~,
ne111s of clement lc11ptl1
XCL and YCL. tlw Jt,1111111 I I
and lhc direction l'Cl\1111~
CX and CY.
Write infon11111io11 lrn
Elemc n1 I.

CONTINUE

Heading: NODAL RESTRA INTS

Subhead: NODE NR! NR2

Clem nodal rcs1111i111

list.

I l11w,1h 11 lnr 1 1 11111111111 I VNAl'l

22

,-1

I h1w1.h 11t lo1 1'1ot:J1 mi OYNAPT

App

t
I
I
I
I

4 J

I
K. NRL(2*K-1). NRL(2*K)
I
I
K. NRL(2*K-1), NRL(2*K)
I
I
4
I
L-CONTINUE

Rend 11ml W1lh 11111111

,cstruinl I M

L __ _

I
I
I

NNR = NNR+NRL(J)

,-

f---

I
I
I
I
L_

Cal ulah.: 11111111111 111

degrees ol l1,11h1111 NI I

J = I. NDF

K = J. NDF

SS(J.K) = 0.0
CONTINUE

b. Calculate Element Stiffness Matrix

6 J = I, NND

NI = NJ+ NRL(J)
ID(J) = J-NI

TOP

Write NDF 11nd NNII

Initialize N l 11111111

1
I
I

Otheiwise, put rD(J)

among supp rt re train ts.

2. SUDPROGRAM STIFPT FOR SUBPROGRAM VIBPT

a. Clear Structure Stiffness Matrix

Calculalc 111111111< 1 111

nodal rc11ul111, NNII

I'. Displacement Indexes

r--

UE

END

L--

NUMBER OF DEGREES OF FREEDOM (NDF)

NUMBER OF NODAL RESTRAINTS (NNR)

99

NDF = NND - NNR

ID(JJ = NDF+NI

RETUR

lni1ializc 1111111h 1 111 11101 I

rc,1rui111, hi 11111

,-1

NRL(JJ > 0

523

By clcf'uull, p111 lllt l1

lllllll11g 1'11:1 1h11li11 I 1111 111

r-----

I
I
I
I

I
I
I

SCE = AX(l)*E/EL(I)
SCEXX = SCE*CX(l)*CX([)
SCEXY = SCE*CX(l)*CY(I)
SCEYY = SCE*CY(l)*CY{I)
SE(l.1) = SCEXX; SE(l.2) = SCEXY
SE(l ,3) = - SCEXX: SE( 1.4) = -SCEXY
SE(2,2) = SCEYY; SE(2.3) = -SCEXY
SE(2,4) = - SCEYY; SE(3,3) = SCEXX
SE(3,4) = SCEXY; SE(4.4) = SCEYY

Compute stiffness
constants and ti II
upper triangular
part of SE, as shown
by Eq. (3.5-26).

f lowi:hnrt for l'rournrn l>YNAl'I

624

App C
App .C

525

f luwohurl for Progrom DYNAPT

c. Calculate Element Displucemcnt Indexes

e. Fill Lower Triangle of Structure Stiffness Matrix
IDE( l)
IDE(3)

= ID(2*JN(l) - l); IDE(2) = ID(2JN(I ))

= ID(2*KN(I) - I); IDE(4) = ID(2*KN(l))

Eq,. (J . 5 30)

d. Transfer to Structure Stiffness Matrix

JMI = J -1

r---

I
Check for row restrnint

r--

I
I

Check for column rc,11,11111

I
I
I

..
I

= IDE(J ):

IC

= IDE(K)

L __

SS(J,K) = SS(K,J)

ITEM

= IR; IR =

SS(IR,IC)

I
IL __

L ____

L _____

IC; IC

= ITEM

CONTINUE

7
CONTINUE

Detenn ine row and col

umn indexes for structur ~
stiffness matrix.

= SS(IR, IC)+SE(J, K )

If not. interchange row ,111d

column indexes .
Transfer a term in mat1 1x
SE to matrix SS, u, i111
plied by the fi rst of Eq~

Place an upper triangular tenn

into a lower triangular position.

RETURN

Check whether row intk~ i,

less than or equal to coh111111
index.

I I
I
I
I
I
I
I
I

IR

I
I
I

I
I
L

END

3. SUBPROGRAM CMASPT FOR SUBPROGRAM VIBPT

This subprogram is similar to STIFPT. but the consistent
mass matrix is generated instead of the stiffness matrix.
4. SUBPROGRAM STASYM FOR SUBPROGRAM VIB
a. Decompose Stiffness Matrix and Copy Mass Matrix to CMU

(3.5- 14).

CONTINUE

r-

I , NDF

CONTINUlo

CONTINUE

I
L

r
U(J,K) = SS(J,K)

n ansfer stiffness matrix to U.

626

Flowchart for Program DYNAPT

App

App.C

Flowchart for Program DYNAPT

527

r--DECOMP(U, NDF, & 3)

Dccompo~c Mil Im" 111,1111

inro the factorl'd lt11111 I 11
as given by Eq. (.I " 11

r-

r---

2 J

L .-----''----,
CMU(J.K) = SS(J,K)

STIFFNESS MATRIX DECOMPOSED

Copy stiffness matrix to CMU.

I. NDF

,-

STIFFNESS MATRIX /\ND MASS MATRIX NOT POSITIVE-DEFIN ITE

CMU(J.K)

= CMS(J,K)

Copy mas~

1111111 ix 111 t

Ml 1
STOP

c. Calculate

u-T
7

b. Decompose Mass Matrix and Copy Stiffness Matrix to CMU

INVERU(U, NDF)

,-

Call Subprogram INVERU to

obtain the inverse transpose of U.

I. NDF

,I

d. Transform to Standa rd, Symmetric Form

(I)

Prcmultiply by

u-T

U(J,K) = CMS(J,K)

lhtnsfcr mu,~

11111111\ 111

I'

r----1

I = NDF-11+1

Set up decreasing index I.

DECOMP(U, NDr, & 6)

Dcco111pm1 11111" 11111111,

11110

Slll+N l.SS M/\TRIX NOT 1os 1l"IVI,

1) 1'.l IN I J'I :, M/\SS M/\TRI X Dl :C'OMPOSED

the iill'lorl'd

1111111

I 11

,---I
I

'

I I
I I
I t
I

JI

I , NDF

NDF JI f I
Tl.MP
00

Set up decreasing index J.

1111<.l ini1iali1c TEMP to zero.

I lowt hnr I for 1'1ou1nm llYNAl 1 1

I
I
I

I
I
I
I
I
I

I ,--I

I
I I
TEMP =
I I
I IL ___
I

App C:

X K

T EMP

l ,J

U(J.K)CMU(I.K)

P1 c111u lt1ply C'MlJ hy II 1

indicmcd 111 tht. 111 , 1 ,,1 I 11

(3.6 11 ).

1lowr.hurt fur 1'109run1 DYNAPT

828

IL __ _ _

IL _____ _

RETURN

CMU(l,J) = TEMP

I IL ____
I
I ______
L

(2) Postmultiply by

u-

END

Put result back into ('~ ll I

5. SUBPROGRAM EIGEN2 FOR SUBPROGRAM VIB

Calculate eigenvalues and

eigenvectors to double precision.

10

6. SUBPROGRAM TRAVEC FOR SUBPROGRAM VIII

a. Calculate Angular Frequencies

r----I
I

13 II = l,NDF
TOL
I = NDF - 11

+I

= 0.000001

Set up decreasing i ndex I

1
, 'l'I

IOl11tl lll'I' JIii /l' III

~ l)ll.'IIVllilll'\

I
I
I

r-- --

12 J = l ,NDF

I
I

I I
I

lf mass matrix wa, de

composed, go to 2.

I
I

TEMP = 0.0

Initialize TEMP to zero

r--1

CV(I) = 1.0/ DSQRT(CV(I))

r---

I I I
I
I

I I

I II

I
I I
I I
I
I

I
I
I
I

TEMP = TEMP

L ___

CMU(K.J)*U(I .K)

Postmultiply CMU by U i, II\

in the first of Eqs. (3 .6 11 l

II

Calculate angular frequencies from the first

of Eqs. (3.6- 12).

r1

CMU(l ,J) = TEMP

Put result back into CM lJ

Check si ze of eigenval ue .

1130

I l11w1 hnr t fur l'rour nr11 l>YNAl' I

C V(!)

= DSQRT(CV(I))

App,

c:

I lowohurt lnr Proumm lJYNAPI

Calculait: 1111g11J11r lh
qucncics from rlw 111NI

c. Normalize Eigenvectors with Respect to Largest Values

ofEqs. (3.6 19),

I
I
I

,----I

CV(!)

L __

= 0. 0

Set small cigcnvullll'

equal to zero.

b. Back-Transform Eigenvectors

I
I

r----

I
I

II

r----

I
I
I
I
I
I

I
I
I

Initialize SUM to ~crn

= SUM

U(K,J)*PHl(K ,I)

Back-transform cigcnv1'l'l111 ,,
as in the second of Eq.~. ( I !I I r,

Search for largest value

in vector I.

I
I
I
I

CI = DABS(PHl(J,I))
C2 = DABS(BTG)

I
I

C I > C2

PHl(J,I) = SUM

B IG = C l

If C J exceeds C'2, n:placc BIG wit h (' I .

CONTINUE

r----

I
I
I

P Hl(J, I) = PHl(J,l)/BIG

I
I

Set up constants C I
and C2.

L __

+ L __

I
L ____ _

r----

L ___ _

I
I

Initialize BIG to zero.

I
I

I I
S UM
I I
I I
I L __ _

531

L ___ _

10

Put SUM into PIii.

7
RETURN
EN D

Nom1alize vector I with

respect to B IG.

..
I l11w1 hurt

101

I'I ou111111 IIYNl\l' 1

"''

/\pp, (

I lowoh111 t for f'to111 nrtt tJYNI\P I

533

7. SUlll'ROGRAM 1rnswr FOi{ su1wno(;IUM VIIWI'

a. Reorder Angulur l<'rcc111cnclcs 1111<1 Elgcravcl'lcu',~

I
lDCMP = I

If lite

I
I

lllllSS 11 11111 iX Wu,

decomposed. o lo I

NDP2 = NDF/ 2

Divide rhe n11111lwr 111

degrees of f recd om hy

..
1

I
I
I

r---

J = NDF - I

I ; TEMP = CV(!)
CV(!) = CV(J); CV(J) = TEMP

I
I

Interchange sy111111c111t11lly

placed angular frcquc11l'il''

I
I

I
I

I
I

I
I

I
l

I
I

L __ _

If K corresponds to a
free displacement, go to 4.

+
I

EE(J) = 0.0

I
I
I
I

For a restrained displacement, put a zero into the

expanded eigenvector.

EE(J) = PHI(K ,1)

L_

For a free displacement

transfer a nonzero term.

TEMP = PHI(K ,l)

PHI(K, IJ = PHJ(K,J)

I
I

Interchange sy111111ctric11lly
placed eigenvectors.

I
I

PHI (K,JJ = TEMP

MODE (!)
ANGULAR FREQUENCY (CV(!))
NODE DJI DJ2

Write headings for modal

outpu1.

+
I
L __ _

J,EE(2*J - I).EE(2*J)

Write cigc11vec1or for

mode I.
7
CONTIN UE

b. Write Angular Frequencies and Expanded Eigenvectors

c. Normalize Eigenvectors with Respect to the Mass Matrix

r--1

I
I

I
I
I

r------

,rI
I

Set index K equal 10 ll)(J ).

I
I
t

,----

I
I

Initialize SUM to zero.

1134

I low1 hull for l'1ou1n111 l>YNAl'I

A1111

App. C

1lowchort for Program DYNAPT

535

8. SUBPROGRAM DYLOPT FOR PROGRAM DYNAPT

a. Dynamic Parameters
I

I
I
I
I
I

I r--I
I
I
I

SUM = SUM

L ____

I
I

I
I

I
I

I
I

Prcmultiply ihl.! 1111M 11111

trix by the trnm,po,lcl
eigenvector:
ISOLVE,NTS,DT,DAMPR

EE(J) = SUM

Put SUM into Iii: (1111

temporary Monigc).

Read and write dynamic parameters.

ISOLVE,NTS.DT,DAMPR
SUM= 0.0

r--I

L_

SUM = SUM

r--L_

b. lnitial Conditions

10 J = I.NDF

+ EE(J)*PHl(J,I)

SUM = DSQRT (SUM )

Reinitiali1:e SUM to 1<111

10

I
I

+ PHl(K ,l)CMS(K ,J)

I
I
I
I

Heading: LOADING NO. (L N) OF (NLS)

Subheads: DYNAMI C PARAMETERS
!SOLVE NTS OT DAM PR

8 K = l,NDF

II

Postmultiply by thl.! l'lp

envector.
Calculate nonnalil11t1n11
constant.

J = 1,NDF

r- -

I J = I ,NND

I
I
DO(J)
I
L_

= 0.0; VO(J) = 0.0

Clear initial displacement

and velocity vectors.

II

L _ ______

PHl(J, I) = PHl(J. 1)/ SUM

Divide eigenvector hy 1h1

normalinuion con~t,1111

Heading: I NITIAL CONDITIONS

Subhead: NNID NNIV

12

NNID,NNIV
RETURN
END

Read and write initial

condition parameters.

536

Flowchart for Program DYNAPT

App. C

App.C

537

Flowchart for Program DYNAPT

(I) Initial Displacements

Heading: INITIAL VELOCITIES

Subhead: NODE VOi V02

If there are no initial

displacements, go to 4.

r-I

Heading: INITIAL DISPLACEMENTS

Subhead: NODE DO! DOZ

r- 1

I
I

Read and write initial

displacements.

J,D0(2*J - l),D0(2*J)

5 I = 1,NNlV

J, V0(2* J -

I
I
I
I
I
L __

Read and write initial

velocities.

1), V0(2*J)

J,V0(2*J - 1),V0(2*J)

5
CONTINUE

I
I

J,DO(Z*J - l),DO(Z*J)

r--

I
I
I
I
I

L __

CONTINUE

r--

Set index K equal to ID(J).

K = ID(J)

K .;; NDF

I
I
I
I
I
I
I
L __

6 J = l ,NND

VO(K) = VO(J)

If K corresponds to a
free displacement, put
VO(J) into VO(K).

I
Set index K equal to

11)( I)

l __

K .;; NDF

DO(K) = DO(J)

lf K corresponds to ;i
free displacement, p 111
DO(J) into DO(K).

CONTINUE

c. Applied Actions

r--

CONTINU E

L__

(2) l11i1ial Velocities

If' 1hcrc un.: no ini1i11I

vcloci1ic>. go 10 I

8
AS(J)

0.0

Heading: APPLIED ACTIONS

Subhead: NLN NEL

Clear action vector.

ftJff

Flowchart for Program DYNAPT

NLN,NEL

Read and write load

parameters.

App.C

r----

L___ _

I
I
I
I

PBS(I) = (2.0* BLI + BL3)*EL(l)l 6.0

PBS(2) = (2.0*BL2 + BL4) *EL(l)16.0
PBS(3) = (BLI + 2.0*BL3)* EL(l}l 6.0
PBS(4) = (BL2 + 2.0*BL4)*EL(l)l6.0
JI = 2*J N(I) - l;J2 = JI + I
Kl = 2* KN(l ) - I; K2 = Kl + I
AS(JI) = AS(JI) + PBS(!)
AS(J2) = AS(J2) + PBS(2)
AS(KI) = AS(KI) + PBS(3)
AS(K2) = AS(K2) + PBS(4)

I
I
I
I

9 l = l,NLN

J,AS(2* J - 1),AS(2* J)

Read and write scale

factors for line loads.

l,BLI ,BL2 .BL3 ,BL4

I
I
I
I
I

Heading: NODAL LOADS

Subhead: NODE AJ I AJ2

r--

l ,BLI ,BL2,BL3,BL4

I
I
I

If there are no loaded

nodes, go to 10.

J = l,NEL

II

( I) Nodal Loads

539

Flowchart for Program DYNAPT

App. C

Calculate equivalent
nodal loads, as in Eqs.
(4.10-4).

II
CONTI NUE

Read and write scale

factors for nodal loads.

d. Ground Accelerations
J.AS(2*J -

I
I

L __

l), AS(2*J)

12
Headi ng: GROUND ACCELERATIONS
Subhead: !GA

Read and write acceleration parameter.

(2) Elc,ncnt Louds

If' there arc no clements

with line loads, go to 12.

If there are no ground

accelerations, go to 16.
I llndin)l: U NE LOADS
S11hlw11d : El.EM . IJLI ... BI A
Heading: ACCELERATION FACTORS
Subhead: GAX GAY

1140

Flowchart for Program DYNAPT

App.C

App. C

I
I

Read and write acccll1ti

tion factors.

GAX,GAY

GAX,G AY

( J) Fill Vecto r GA

Flowchart for Program DYNAPT

r---

+ L1
I

14
SUM = SUM

I
I
I
I

with Ground Accelerations

13 I = 1,NN

rJI

= !0(2*1 -

I
I
I
I
I

L__

Augment AS(I) with the

negative of SUM.

15
CONTINUE

= ID(2*1)

Calculate displacernl'III
indexes JI and J2 .

17

Set the displacement

index K equal to ID(]).

I
JI ,a; NDF

J2 ,a; NDF

I
I

I); J2

GA(J I ) = GAX

GA(J2)

= GAY

If JI is free, put GAX

into GA(J I).

If 12 is free , pul GAy

mto GA(J2).

I
I
I
I
I

L_

K ,a; NDF

AS(K)

I
I
I
I

AS(J)

lf K is free, move the

corresponding action up
the list.

17
CONTINUE

e. Read Forcing Function

CONTINUE

N = NLN

+ NEL + IGA;NFO

r ----

'

13

(l) Multiply /\ccclcrntions by Mass Matrix

I
I
I

Multiply mass and acceleration terms.

CMS(J,K)*GA(K)

16

AS(l) "'. AS(!) - SUM

L_____ _

r--

541

Add the parameters

requiring a forcing function.
If N = 0, skip reading a
forcing function .

Sci lhc displucc11wn1

index J cquul 10 11 )(I)
If' J is l'CSlrai11cd, N~lp
the 111uliiplit111it111.

l11i1i11lin SlJM 11 , itiu

Heading: FORCING FUNCTION

Subhead: NFO

Read and write number

of function ordinates.

llnwil11111 tor l'1u111iu11 l>YNAl'I

,-

18 J

Rcud and write sulN11pl ,

lime, a11d f'w1c1iou 01<11111111

K,T(K),FO(K)

I
L-

643

DYNAfJI

21 I

l ,NTSS

'

1,NFO

l11011rt1111

I
I

lk11d 11111, l l lN( IION tll<lll NA ,-~

S11hhc11d; K 'l 'IMI lA( 1()1(
~

r---..-

I lowoh111t lot

App.C

I I
I I
I I
I I
t I

STEP(J) = FO(K)

L __

1s

Increase J by J.

(I -

l)* DST EP

Detennine step and ramp

for lime increment wilhin
a piecewise-linear segment.

21

I
I
L____ _

r - - -K-.'_r(_
K..L)
,F 0-(K-)--~

RAMP(J)

= DSTEP

22

RETURN
END

I'. ( 'nknlntc Step und Ramp for Each Time Increment

r--1

20 J

1.NTS

9. SUBPROGRAM TRANOR FOR SUUl'ROGAM NOUMOD

a. Read and Write Number of Modes

STl iP(J )

I
I

L __

0.0; RAMP(J)

Clear slcp and r,1111p

vectors.

20
NMODES
CONTINUE

N l'OMI

,---

= 0.0

NFO

22 K

I: J

NORMAL-MODE SOLUTION
NMODES = (NMODES)

lniti,liizc NFOM I 1111t1 .I

l ,NFOM I

b. Calculate Transformation Operator <I?;;'

I
I

s_s__ <
_.,_.<K
_ 1_1J___rc_K_lJ_11_)1_'__,

r----

2 l = l ,NMODES

If' NTSS

'~----

2 J = l ,NDF

0, skip 1111 ~1111

anti nunp cakul111ion,.

C.1icula1c number ol 1i111r

slcp~ for a picccwisl'

I incur M
.lgnicnl.

llST/\JI

( IU(K

I I)

-i-

l'O(K)J/ NTSS

Culcullllc 1hc nuc ol

chnngc ol nnli11u1l, wilh
1c~pcc1 lo linlL' ,

lnilialize SUM to zero.

1144

Flowchart for Program DYNAPT

App.C

r-- -

I L'
I
I

SUM = SUM

+ PHl(K ,l)*CMS(K,J)

Calculate <I>;

<I>IM.

-,L__

L - - -

App.C

TR_o_r_u_.J..,..
l _= _s_u_M_-..J

I
I
I
I
I

Flowchart for Program DYNAPT

r--1

SUM = SUM

L _____ _

I
I
I

AN([) = SUM

I
I
I
I

RETURN
END

COMMON.

Initialize DSUM and VS UM

to zero.

4 I = I.NMODES

+ TROP(l ,J)* DO(J)

+ TROP(l,J)* VO(J)

Premultiply DO and VO
with TROP, as in Eqs. (4. 3 ~).

I
I

OM EGA = CV(I)

( I ) Response Co nstant s

l)NO(I)

DS UM; YNO(l) = VSUM

Put DSUM and VSUM i1110

ONO an VNO.

(1

I
I

I
I

1,NMODES

For a rigid-body mode ,

go to I.

I
I

cl. 'lh 111Nfor111 A1>pllccl Aclfons

Put CV(!) into OMEGA.

I
I

L __ _

l __ __ _

OM EGA2 = OMEGA*OM EGA

OMEG AD = OMEGA* DSQRT(l.0 - DAMPR* DAMPR)
CN = DAMPR* OMEGA: EX = DEXP( - CN* DT)
C = DCOS(OMEGAD*DT); S = DSIN(OMEGAD*DT)
C l = EX*(C + CN *Si OMEGAD)
C2 = EX* S/OMEGAD; C3 = ( 1.0 - C l) /OMEGA2
C4 = (DT - C2 - 2.0* CN* C3) /(0MEGA2* DT)
CS = - OMEGA2 *C2; C6 = Cl - 2.0*CN*C2; C7 = C2: CS = C3/DT

For a vibrational mode,

determine the constant's
CI through C8 given by
Eqs. (4 . 10-3).

-r AF equivalent to AS in

IO. SUBPROGRAM TIHIST FOR SUBPROGRAM NORMOD

a. Calculate Response for Each Mode

r------

DSUM = DSUM
VSUM = VSUM

r- ---

Put SUM into vector AN.

Put SUM into the transformat ion operator TROP.

r---

I I'

Premultiply vector At
with <Pi. as in Eq. (4.4-4).

4 I = l ,NMODES

DSUM = 0.0; VSUM = 0.0

+ PHI(J,l)*AF(J)

c. 'lhmsfonn Initial Conditions

,r----

545

l11 i1i11litl SlJM 10

111, 1,

I
I

I
I

Cl

C5

1.0; C2
0.0; ('6

DT; C3
1.0: C7

DT DT/2.0: C4 = DT * DT/ 6.0

C2; C'8
C3 !l)T

For a rigid-body mode,

use contants from Eqs.
(4 .9 9) and (4.9- 10).

546

Flowchart for Program DYNAPT

App, C

Flowchart fo r Program DYNAPT

App.C

54'

(2) S1cp-hy-S1cp Rc,pon,c

I
I

DNJ = DNO(I); YNJ = VNO(l)

lni1ializc DNJ and VN.1

12. SUBPROGRAM NUMINT FOR PROGRAM DYNA

a. Read and Write Integration Parameters

r----

I
I
I

ALPHA.BETA,GAMMA

I
I

L___ _

I
I
I

DN(l,J) = Cl *DNJ + C2*VNJ + C3AN(I)

*STEP(J) + C4*AN{l)*RAMP(J)
VNJ = C5*DNJ + C6*VNJ + C7*AN(I)
*STEP(J) + C8*AN(J)*RAMP(J)
DNJ = DN(l,J)

I
I

Calculate modal n:sponscs using Eqs. (4. IO I l

and (4. 10-2).

DIRECT NUMERICAL INTEGRATION

ALPHA = (ALPHA) BETA = (BETA)
GAMMA = (GAMMA)

CONTINUE

b. Calculate Structure Damping Matrix

L _____ _

(I) Delcrminc Factor SA

r-----

RETURN
END

I
II. SUBPROGRAM TRABAC FOR SUBPROGRAM NORMOD

r-----

2 J = 1.NTS

I
I
I

r-----

1
Initialize SUM to , cro .

r--1

SUM = SUM + PH l(K,l )*DN(l,J)

Back-1ransfonn cfo,phtn
mcnt, with Eq . (4 I ~)

L_ _ _ _ _
DF(K,J) - SUM

IWl'l JRN
lN I>

SUM = SUM + CMS(J,K)*PHl(K.I )

I = I.NMODES

L
I
I

I
I

Initialize SUM to zero.

I L

r---

Compute Cl = C)iJ .

t - - - - -

I
I
I
I
I

Cl = DSQRT(2.0*DAMPR*CV(I))

Place SUM intu thl' ll\'t'

di~phtCCllll' lll lllllll IX I)(

I
I

L____ _

SA(J.1) = SUM*CI

Multiply M by <t>N.
Complete factor SA =
M<I>New needed for Eq .
(4.6- 17).

548

f l11w11l110 I 101 1'11111111111 l>YNAI' I

App. C

Flowchart for Program DYNAPT

549
(2) Mul11pJy SA 111111 II l\ 1II INfHl1<1

r-------

r----

l
I
I

r----

'

I
I
I
I
I

I
I

r-----

I
I L
1

r---

I L'

I
l

I
I
I
I

SUM

= SUM

+ SA(K, J)*SA(J,J)

CS(K ,J)

6
SUM = SUM - SS(J ,J )*DO(J) - CS(l,J)*VO(J)

+
Apply Eq. (4.6- 17).

AA(!) = SUM

l ______ _

AF(l)*STEP(J )

Set up right-hand vector in Eq. (5 . 1- 4).

= SUM

L___ _

SOLYER(SA,AA.AO,NDF)

Solve Eq. (5.1 - 4) for inirial accelerarions.

c. Calculate Initial Accelerations

r-

r-

I
I

'

I- I
I
I
I

L_

9 I = 1,NDF

1
AO(]) = 0.0

If rhe mass matrix is

singular, use zero inilial accelerations.

l_

SA(l,J ) "' CMS(I.J)

Copy mass 111a1rix lo SA.

5

d. Calculate and Factor Equivalent Stiffness Matrix

FAC.Vf'OR(SA, Nl)/~&8)

F11c1or 11111~s 111atrix. If'

ir is si,1guhu; go 10 8.
CI
C3

= 1.0

ALPHA; C2

= 1.0/( BETA*DT*DT)

= GAMMA/( BETA*DT)

Detennine consrants
for Eqs. (5.5- 15) and (5.5-33).

680

I l11woh1111 ror l'r our IHn llYNI\I' I

I\JJp, C

I. NI> ~

t
II J

1,NDF

AA([)

= SUM

SA(],J) = Cl*SS(l,J)
+ C3*CS(!,J)

+
I
L __ _

C2*CMS(l,J)

&1s.(5.5 15)111111 (\, 111

SOLVER(SA,AA ,DD,NDF)

I
I

DF(l, I) = DO(!) + DD( !)

AO(!) = AO(!) + C5*DD(I) - Q(I)
VO(]) = VO() ) + C6*DD(I) - R(I)

CI = 1.0/ (BETA*DT) ; C 2 = 1.0 / (2.0*BETA)

C3 = GAMMA / BETA; C4 = (GAMMA*C2-l.O)*DT
C5 = Cl/OT; C6 = GAMMA*CI

(5.5- 13).
f. Calculate Responses for Subsequent Time Steps

Q(J) = C I* VO(J)
R( I) = C3*VO(I )

+
+

C2*AO( l)
C4*A0(1)

r-------

Obtain Qand 1 Vl'l'lui.

from Eqs. (5 .5 10)
and (5 . 5- 1 I ).

12

r-------

,---

'

I
I
I
I
I
I
I

I
I

I
I

I r----IJ K

I. ND!'

L{

SlJM

.~lJM I C'MS(l ,K)' Q(K) I C'S(I,K) R(K)

Add
tc 1111,

lllllSS

lllld d11111pl11p

111 llq . (~ , 111)

I
I
I

+
I
I

Q(l)
C l *VO( I)
R(I ) = C3*VO(I)

L __ _

r --- --

15

L __ _

Dete rmi ne cons1u111s 1111

Eqs. (5.5- 10) 1h101111h

Apply Eqs. (5.4- 16), (5.4- 17),

(5.4- 18), (5.5- 12), 1111d (5 .5 13) 10
find lhc 10111! rcsponsl!s.

12 I = l,NDF

L__ _

Solve Eq. (5.5- 14) for the

incremental displacements.

r---

Factor cquiv11lc111 still

ncss matrix . If 11 I,
s ing ula r, go to IOI ,

c. Calculate Response for First Time Step

I
I
I

Complete the calculation

of terms in Eq. (5 .5- 16).

CONTINUE

Calcul111c l!qulv11ll1 111

stiffness n1u11 ix hy

II

FACTOR(SA ,NDF,& 101)

r---

+ AF(!)*RAMP(I)

14

L______ _

I
I

551

,---~
L ____

I lowol11111 for Program DYNAPT

+
+

C2* AO(!)
C4*A0(1)

Use Eqs. (5.5-)0) an9 (5.5-J I) to

obtain vectors Q and R.

552

Flowchart for Program DYNAPT

App.C

App. C

553

Flowchart for Prog ram DYNAPT

13. SUBPROGRAM RES2PT* FOR PROGRAM DYNAPT

a. Read and Write Output Selections

tI tI
I
I
I
I
I
I
I
I

I
I

r----~
17

r~~~~~~~~-'--~~~~~--,
SUM = SUM + CMS([ ,K)*Q(K) + CS(l ,K)
*R(K) + ALPHA*SS(l ,K)*DD(K)

I
I
I
I

AA(l) = SUM + AF(l)*(STEP(J)

+ RAM P(J) - STEP(J - I) - RAMP(J - I))

L _____ _

Heading: OUTPUT SELECTIONS

Subhead: !WR IPL NNO NEO

Apply Eqs. (5.5- 16) and (5.5-34) to

de termine the incremental
load vector.

IWR. IPL, NNO,NEO

I
I
I

(JNO(J), J

I
I
I

SOLVER(SA,AA,DD,NDF)

Solve Eq . (5.5- 14) for the

incremental displacements.

(IEO( l), I

DF(l,J) = DF(l,J - I) + DD(l)

AO(I ) = AO(!) + C5*DD(1) - Q(l)
VO([) = VO(l) + C6*DD(I) - R(I)

IL __ _

Read and write nodes

selected.

l ,NNO)

NODES: (JNO(J), J = l ,NNO

__
) __ 6

,r--I
I
I

CONTINUE

I
I

18

I
I
I

Read and write output

parameters.

IWR ,IPL,NNO,NEO

Read and write elements

selected.

l,NEO)

Use Eqs. (5.4- 16), (5.4- 17),

(5.4- 18), (5. 5- 12), and (5.5- 13)
to o btain the total responses.

ELEMENTS: (IEO(l), I = 1.NEO)

19

CONTINUE

l ______ _

20
I , write parameters

to plotter.

b. Transfer Displacements from DF to DS

RETURN

IOI

If equivalent sti ffness

FACTORIZATION OF EQUIVALENT
STIPrESS MATRIX f'AILED

matrix is singular, write

this message; and stop cal
cu lat ions.

,---1

t
STOP
END

4 I = l ,NND

Set K equal to ID(l).

I
For convcnicmc, i mplied itcrntivc control stntcmcnts arc
u,cd tn tlm ,11hp1u11n1111 l.011ic11I output unit, urc 6 f'or printer
(ltltl / IOI (lhlllt'I

064

Flowchart for Program DYNAPT

I
I
I
I

Fl owchart for Program DYNAPT

Otherwise, put zero into

DSO (for a restrained displacement).

DSO(l) = 0.0

r!WR = I

L_

DS(I.J) = 0.0

If IWR = I , write initial

displacements.

For a free displacement,

put DO into DSO.

DSO(I) = DO(K)

,---

,-

I
I

L_

DS(l,J) = DF(K,J)

Also, put OF into OS.

L _ __ _

I
I
I
I

I
I
l".

Write Displacements

,---1

The node number K is fo1111d ,

parameter KO is comp111cd,
and the number of r'Cst111i111.,
is obtained.

JNO(KK); KO = 2*(K - I)
NRL(KO + I) + NRL(KO + 2)

I
I
I
I

i
I

TIME,(DSO(KO + K), K = 1,2)

J,TIME,(DSO(KO + K). K = 1.2)

Also, put zeros into OS.

IPL = I

Initialize J and TIME to zero.

J = O; TIME = 0.0

If IPL = I, write K to plotte[

IPL =

I
I

555

free displacement, go to 2.

App. C

If K corresponds to a

I
I

App.C

If J = 2, there arc 110

displacements al node
K; so 11011c arc writhlll.

llc11di111:1: l)ISPLAC'EMENT TIME

I II S'J ORY FOR NODE (K)
Suhh,ud : S'l'lll' 'l'IME DJ I D.12

I
I
I
I
I
I

I
IL ___

,--I
I
I

I
I

I
I
I

Initialize MAX and MIN

terms.

8 J = 1,NTS

I
I
I

5 K = 1,2

DMAX(K) = DSO(KO + K)
DMIN(K) = DSO(KO + K)
TMAX( K) = 0.0; TMIN(K) = 0.0

I
I
I
I

Similarly for IPL = I.

I
I

I
I

TIME = DT*J

Compute TIME al step J.

r-I
I
I
I
I
I

DS(KO + K. J) < DMAX(K )

DMAX(K) = DS(KO + K. J)
TMAX(K) = TIME

L __

If displacement DS
go to 6.

< DMAX ,

Otherwise, replace DMAX

and TMAX with current
values.

111111

1 lowuhut I lo t l't or,1111t1 l>YNAl' I

I
I
I

I
I

I
I

I
I
I
I

7 K

I
I
I

DMI N(K)
TM JN(K)

Otherwise, 1q il11l'1 111\ II N

11nd 'l'MI N w ill1 111111111
value,.

DS(KO+K, J)
'J'JM E

IWR

J,TIME.(DS(KO I K, J), K

J- .---------.J
11'1.

I
L
__ _

1,2)

If IWR

[WR = I

TIME OF MINIMUM
(TMIN(K), K = 1,2)

Write times of minima.

d. Calculate and Wr ite Axial Forces

I, w,IJ1 d lN pl111111 11111

TI ME,(DS(KO + K , J), K

1,2)

Similarly for 11'1 ,

r-- - - I
I
I
I

CONTINUE

14 II = l,NEO

l = IEO( II)

!WR = I

MAXIMUM
(DMAX(K). K

1,2)

(J

Wri1c 111aximu111 d1 Np h1, 1

rncnt,.

11'1.

(DMAX(K), K

1,2)

Sim ilarly l<ir plo1111

TIME OF MAXIM UM
(TMAX(K), K
1.2)

Write 11n1c~ uf 11111, 111111

J.

Heading: AXIAL FORCE TIM E

IWR

[ I WI{

Find dc111c111 1w111hc1 I .

HISTORY FOR ELEMENT (I)

Subhead: ST EP TIME AM I

'
iI

CONTINUE

I
I
I
I

1,2)

Similarly for plotter.

L____ _

I I

I ll\llN

I
I

I
I
I
I

(DMIN(K), K

I
If' displ11t'l'IIK'1II I >S
go 10 7.

DS(KO I K, J) -> l)MIN(K)

I I

IPL= l

L__
I

557

I
I

Flowchart for Program DYNAPT

1,2

App. C

,r--

Ar,11 (

~ I NIMl lM
")M IN(K), K

l.2)J'1
W11111111111r1111111cllspl1111
11111111N,

I
I
I
I
I
I
I

+
I
I
i
I
I
I
I

t
I
I

IPL = I

JI= 2*JN(IJ - l ; J2 ~ JI + I
Kl = 2*KN(l) + I; K2 = Kl + I
SCM = E*AX(])/ EL(I)
J = 0: TIME = 0.0
AMO = SCM*((DSO(K I) - DSO(Jl))*CX(I)
+ (DSO(K2) - DSO(J2))*CY(I))

IWR

Write I to plotter.

Compute displacement indexes , and so on.

Calculate initial axial

force.

6
If !WR = I , write initial
axial force.

1
IPL =_._ _ _~
-T-IM_E_,,AMD
1

: imilarl y for IPL = I.

668

Flowchart for Program DYNAPT

App. C

Flowchart for Program DYNAPT

559

AMDMAX
TMDMAX

t
I
I
I
I
I
I

App<

= AMO; AMDMIN = AMO

0.0; TMDMIN = 0.0

Initialize MAX and MIN

terms.

r---1

I
I

I
+ t

TIME = DT *J; AMO = SCM*((DS(K 1,1)

- DS(J I ,J))*CX(l) + (DS(K2,J)
- OS(J2,J))*CY([))

Calculate TIME and AMI>

!WR = I

I
I
I
I

If axial force AMO < i\MllM \

go to 11.

! WR = I

AMOMAX = AMO; TMOMAX - T IME

II

If axial force AMO /\M l >t-.ll N

go to 12.

AMO > AMOM IN

AMOM IN = AMO: TMOM IN = T IME

M INIMUM(AMDMIN)

I
I

Write minimum axial force.

IPL = I

Similarly for plotter.

I
I

Otherwise , replace AM l)M/\ \

and TMDMAX with cu11 ,111
values.

Write time of maximum.

AMO < AMOMAX

TIME OF MAXIM UM
(TMOMAX)

!WR = I

I
I
I

TIME OF M INIMUM
(TMOMIN)

Write time of minimum.

L _____ _

Othe1wise. replace /\MDM IN

and TMDMIN with Clllll' ltl
values .

14

CONTINUE

RETURN
END

12

!WR=

14. SUBPROGRAM DECOMP* FOR SUBPROGRAM STASYM

I, write axi11I 111111
IPL = I

I
I

L __ _ _

Similarly for IPL

I,

13

IWR

L ____ _
I
I

CONTINUE

t
M/\X IMUM( /\MDM/\ X)

I
I

I
I

3 l = l ,NDF

t
I

r -----

Write 111uxi111u111 11x111I lrn, 1

IPL

(/\Ml)M /\X~

--r

SUM = A(l,J); Kl = I - .t

r---1

L_

I
I

SUM = SUM - A(K,l)*A(K ,J)

Si111il11rly !or pli11111

Sec Co1111i1111r

l'ro11111111.1'.f111 S1r111111n,I /\1/(//1w/,1, hy W, Wcuvcr.

Ii , ll V1111 Nost1 aml , l'1111t1to11 , N J , 1%7

Arguments: (A ,NDF, *)

11110

I lowoh1111 for flroornm IJYNAPr

App. C

Flowchart for Program DYNAPT

t
I I

561

SUM = 0.0: JMI = J - I

I
I
I
I

A(l,l) = DSQRT(SUM)

I
I
I
I

r--1

2
SUM = SUM - A (KJ)*A (K,J)

L__ _

A(J,I) = SUM*A(J.J)

A(l,J) = SUM /A(l,I)

RETURN

I
L ___ _

END
CONTINUE

16. SUBPROGRAM FACTOR* FOR SUBPROGRAM NUMINT

RET URN
END

A(I . J) ,s; 0.0

15. SUBPROGRAM INVERU* FOR SUBPROGRAM STASYM

r-

I I = 1,NDF

ArgumcnJs: (A,NDI )

I
A(I .I) = 1.0/ A(l .1)

NI

r----I

lI

L ___
+
I

11,111

,-----

3 I

NDF -

l,N I

I
I

r----

+ I
I
I
I

IPJ

.I J

I I I

11'1 ,NI)!'

I
I
I

SUM = A(l ,J); II = I - I

I
I
I

SUM

I
L ___ _

= SUM

- A (K ,l)* A (K ,J)

A(l,J)

= SUM

Sl'\' M111r/1 ll1111/1w/.1 11{ F ri11111r/ .\'11111 111111, 211d ,d by W

W1i1Vl' I, Jo 1111,i J. M (llH, V1111 Nos111111d 1{11 11h11Jd , N1w Y1 11 i-, l'IKIJ

A rguments: (A ,N , *)

562

Howchart for Program DYNAPT

App

App. C

Flowchart for Program DYNAPT

I
I

SUM = A(J,J)

+
I

! r--'II

I
I

TEMP = A(K,J)/A(K ,K)

SUM = SUM - TEMP*A(K,J)

I
I IL__ _

I
I
I

1 -.- 3 I = I .N
IL _
3
__.____ _

A(K,J) = TEMP

X(I ) = X(]) / U(I.I)

I
tL ___ _

,----

4
A(J,J) = SUM

'II

RETURN
END

l = N - 11 +1
SUM = X(I)
K2 = I + I

I
17. SUBPROGRAM SOLVER* FOR SUBPROGRAM NUMINT

I
I

r----

A rguments: (U, B,X ,N)

2 I = l,N

I
I
I
I
I
I
I

SUM

I -

r--L

SUM

L ___ _

= SUM

- U(I. K)* X(K)

5
X(I) = SUM

RETURN

END
SUM = SU M - U(K, l)*X(K)

L ___ _

' Ihid

= 8(1); KI =

I
I
I

,r---

2
X(I)

SUM

56:i

Ar1ewo1 11 tu l'r uhlome

565

6
2.3-7. D = \ {j()8 sin flt

Lf, 88g
G cos
1271T4
2
2.3-10. D = d cos flt
35
2.3-8. D

2.3-9. D = ~; d sin flt

2.4-1. u =

Answers to Problems

13f sin (flt -

2.4-3. u,. = e

_
_ ( 2nfl )
8 - tan 1 w 2 _ ff

- 8);

-n,(

M fl

\N cos wdt +

+ Nn

wd

Pl 3
2.4-4. D = 0.01445 sin(flt - 8)
El

mal3

2.4-6. D = -3. 185

2.5-1. P (t) =

ut

8 = tan- C)~fln2)

8);

ma .
- f3T
sm (flt

2.4-2. u* =

*=

.
\
sm wdt)
Ml
2.4-5. 88 = 9.086 ~G cos (flt - 0)
7Tr

cos(flt - 8)

2.4-7.

/3rcs = 50, 25, . . . , 2.5

;:Pi(sin flt + i sin 3flt + . . )

P (cos flt - j cos 3flt + .. )

2.5-2. P (t) = ;

CHAPTER 2

2.2-1.
2.2-2.
2.2-3.

/3ii
1
/3ii T-- 2 L -Vm
/mL
w = L 't;;:;i;; ! = z1rL -v-;;i>
4 /3ii
2 /3ii
7Tl /mL
w = L -V-;;;i; f = 7Tl -V-;;;i ; T = 2 V3Ei
- l /3EI . - l
/3EI. T -- 2 L V(2;i,
w - L Vi;;:;i, . t - z1rL Vi;;:;i, ,
w
1

1r.

M.

V-;;;E ; f
2,-2

2.2-5. w

2.2-6. w =

{3;G

r2

u V2pA ; f = 7Tl
/3ii

L -V-;;;i ; f

2.3-1. u = [

11114

d sin

!lt

/mL

{3;G

7Tl2

ma
I
2.3-2. u * = - [ _ (fllw
)2 k
1

2ML
cos
2.3-4. 80 = - - 31rr 4 G

ma1.'
. cos
45111

nt

~2 - ~1r (sin flt

a0
2 5-5 u = .
k

/2pA

V2pA ; T = 7 V~
4 /3ii
7Tl /mL
7Tl -V-;;;i ; T = 4 V3Ei

2.3-6. 0 "'

2.5-4. P (t) =

1r

L.J

1 1

flt - ! sin 2flt + ! sin 3flt - ...)

+ !sin 2flt + ! sin 3flt + ...)

2
3

a1 cos (i flt - 8;) + b, sin(iflt - 8,)

k \!(1 - i 2 fl 2 / w2 ) 2 + (2yifl/ w) 2

Pi

2.5-7. u = 13.06 k cos flt

. flt
2.5-8. u = -6.2 1 Pi
k sm

T = 21rVAE.

_ (~ /w) 2 ] d cos flt

PL 3
2.3-3. D '""'
W sin flt
144
2.3-5. /)

ME.

= Z7r V-;;;E ;

3_ Pi (sin

2.5-6. u = 6.70t sin flt

1T.

2.2-4. w =

2.5-3. P (t) =

n,

t'OH

Ht

2.6-1. u =

kPi(
sm flt

2.6-2. u =

2.6-3. u =

~
[w
t, kw

2.6-4. u =

u=

P,

[1 - e- n,(cos wdt + : d sin wdt)

k (1 -

fl sm
. wt)\[ _ (D,/
1 w)2
- -;
1

J
2

t - 2n + e- n,(2n cos wdt - wl - n sin wdt\ ]

cos wt)

[cos w(t - t,) - cos wrl

Wd

(0 st s ti )

n1111

At111wor11 to Problems

I',/ 1

co.~ 1,1 (t

7Z lcos w(t

11 -

-i

I I

1,)

Ml

cos

11)]

t,

= -Pi [ - cos

wt

(0

wt,

sin wt - sin w(t - t ,)]

+-------

wt,

k
P,
2.6-6. u = -(1 - cos wt)
k

[
u >I< = - a1
- 1
w2

(t

= - :: [cos w(t -

!1

- -2w

wt 1

2.6-9. u* = - ~2 [ ~ - - 2 -( t - cos
u*

=-

= 0

11 10

I ii

(t 1

,,1

(t I

tr

2.7-5.

112() ..,

2,7-fl.

//1 0

2.546

Pi

0.3 186t

U1 0

= kPi

3.4-2. pb(t)

= {1, 3} b2L

3.4-3. K =EA[-; ~: -~]

3.4-4. M = p1sL[ :
3L
1 -8
7
-!
mx1L
3.4-5. pb(t) = {I , 4, 1} bxL
3.4-6. Pb(t) = {5, 3}

rl

Pb(th = {6x

3.4-11 . Given
I)

:x

(0

Wl1

wt +

/1)

COS W

I)

wt)]

(0

2.7-3. Given

"'!:. I

U1 0

2.7-JO.

11 10

I iJ

pAL
pAL
; etc.
M,11 = 3.28 - -; M,12 = O; M, 13 = 0.64
6
6
3EA
EA
EA
L ; S,12 = - L; S, 13 = - L; etc.
3.5-3. S, 11 =
2
2

I)

4pAL
_ .
_ pAL .
M,1 1 = - -; M,12 - 0, M,13 - - - , etc.
3
6
EA
EA
3.5-4. S,11 = L; S,12 = O; S.,13 = -L; etc.

/1 )]

2.7-4. Given

2.7-7.

O; M,13 = 0.06 pAL; etc.

EA
EA
EA
3.5-2. S,11 = I.64L; S,12 = 0.48T; S,13 = -L; etc.

(t - ( >]

Pi
0.9 135 k
I'1

-~]
2

60

EA

3.4-8. pb(t) = {9, 2L, 21, -3L} b2L

3.4-12. Given

= 2. 72 p~L; M,, 2

12

12

6xL, 3x 2L - 4xL2 + L3, - 6x 2

EA

M,11

P,
U20 = - 4 k
Pi
2.7-9. 1110 = 1.379 -(
2.7-6.

3.5-1. S, 11 = 0.64T; S,12 = - 0.48L ; S,13 = O; etc.

a ~{ w; ,,2[cos wt - cos w(t - t,) + wt, sin w(r w

2, 7-2,

I)

+ cos w(t - t,)}

l.7- 1.

( r1

t 1) - cos wt]

a 1 - ao [ sin w(t - t ,) - sin

/ 1)

(0

t 1)-- -sin
+sin-w(t
- -- - wt]
wt,

<.1.>2

+ 2b2} L

(0 s I I i)

Wt1

2.(,-8. u* = - ao
- (I - cos wt) - -a, -- 2 a- 0 - - -sin- wt)
u*

(I i

2.6-7. u* =_ a~(!._ _ sin wt)

V1

{2b 1 + b2, b,

3.4-7. pb(t) = {L - x, x}

- ti)]
u = -P, [ 1 - cos wt - -t - - t, + -sin -w(tk
t2-t1
w(t2- ti)
[
sin w(t - t1) - sin w(t - t2)J
u = P,
- - cos wt + - - - - - - - - - k
w(t2 - t1)
W

II

(IJ

sin wt)
2.6-5. u = -Pi ( l - cos wt - -t + k

CHAPTER 3

3.4-1. pb(t)

[cos w(t - t2) - cos w(t -

567

pAL
pAL
M,11 = - -; M, 12 = O; M,13 = 0.64--; etc.
3
6
EA
EA
3.5-5. S,1 1 = I.64L ; S,12 = 0.48L; S,13 = O; etc.

I. 102 ,..

M, 11 = Zp:L ; M, 12 = O; M,,3 = 0.06 pAL; etc.

ti68
An11wurn to f'tuhl1t1111

EA
3.5-6. S, 11 = U; S,.,2 = - l; Sm == O; etc.
2
4pAL
_ .
_ pAL.
M, , I = -3-; Ms12 - 0, M,13 - 6 ' etc.

Answers to Problems

3EA

3.5-7.

S,11

M,

'

4EJ
T;

S ,1 2

pAL

T2EJ ; Ssl 3 =

8EJ
T;

3.5-8. S, 11 =

s,,

M
sll

3.5-10.

S , 11

S , 12

6E!
u;

13pAL. M i2
35 , s

8EJ

T;

S , 12

W 2.4

3.7-4. (a) w 1

__ pAL 3 .
140 , etc.

-u;

Ssl3

2EJ

T;

13pAL2; etc.
420

_ 0.5412,
L 1.307

@.
Vp'

= 0.9393~

/E ; <f>1 = r-21
yp

W1,2 -

3.6-3.

W1 ,2

3.529

3.6-4. w,,2 = 0.6049,

l 1.815
3.6-5.

W1.2

5.546 , 21.04
=
L2

3.6-6.

W1 , 2

= 15.14,
L228.98

3.6-7.

W1,2

==

<1>1 -

<I>1 _ [
-

<1>2

<1>1 -

<1>
2

81 .98 /El
~
ypA.;

h
.
.
modes ape 1s unity.

==

,.2

v;x;

[OJ
1

W1 ,4

(b)

W2,J

l.271, 1.927
L

3.7-7. (a)

W3,4

==

1.352, 2.218
L

W1,2

0 , 1.225
l

0.6650, 2.909
L

==

/E.

Vp'

<I> _ [
I -

1 ] . <I> == [
l
0.8801 ' 4
-0.6091

/E. <I>
Vp'
/E.
Vp'

-[

<I>3 _ [
-

1 ]
0.08687 '
I

] .

-0.3242 '

l] . _[

/E ; <f>1 = [J
yp

_ [ (9 sin w1 t)/ Wt + (sin w2t)/ w2

43
. -1. D - (3 sin w,t)/ wt - (3 sin w2t)/w2

, <f>2 -

<1>

= [

<1>

== [

1 ]
-0.5

. <I> = [

'

r 0.1m,
, 1.1 ,,)

10

_ [ 1.207 COS W1 t - 0.2072 COS Wz( d

3
4.3- . D - 0.5000 cos Wt t + 0.5000 cos W2f

0.707 1

- r c-2 sin W1t)/w1 + (2 sin W2t)/w2]Dot

4
4.3- . D (sin w, t)/w, + (4 sin w2t)/w2
5

81.98

c,,

JDoi

_ [I.961L cos w, t + 0.03923L cos Wzt] PoL2

43 2
D 2. 701 cos W1 t + 0.2990 cos W2t
6EJ

[1]1

/El
w2 = ~ ypA.

{ij

==

CHAPTER 4

= r-31
1

L
] . <I> = [
L ]
-0.3947 2
- 26.6 1

(b)

53 .28, 139.2

_ r]J
2

/El
~ ypA.;
22.74

; <f>i -

1 -

3.7-2. (a) w 2 =

w, .J

/El.
vpA.,

-[l]
o ' <I>

<I>2

3.7-6. (a)

(b)

_ [1].
3 '

/E.
vp'

/E.
yp'

/El. -[ -II],
V04.
13.32, 34.79 /El . <I> _ [
l
]
L2
V04.
-0.707 1

3.7-1. (a) w1 =

(b)

'1

27 cos w, t - 3 cos W2t] Pol

4.3-5. D == - [ 81 cos w 1t + cos Wzt 18EA

0.1m ,

3
(l.015L sin w, r)/w, - (0.0J506L sin w2r)/w2 1 D0,
4. -6. D == [ ( 0.4006 sin <JJ,t)/w, I (0.4006 sin <u2r)/w2 f ,

, .

1 ]
0.10n

J
@.<I>== [ 1.344
l ]<I>= [
1 J
== 0.6581,
L 1.974 Vp'
'
-1.092

= r2.414J
1
'. <I> = r-0.4142
1

,I>

== [

/E mode shape 1s. uruty.

.
T1 134 yp;

3.6-1. Answer not provided.

3.6-2.

J
16. 86

W1 , 4

etc.

3
2pAL
13pAL2 .
__ pAL3 .
M
- , M sl 2 = - 420 ' Ms 13 sll -- 105
140 , etc.

<1>
4

== [

(b)

u' etc.
-

L2

3.7-5. (a) w2,J == 1.485,L 1.732

M , 13 -

llpAL2. Ms1J =
210 ,

16.64, 95.21

(b) w 2 =

_ 6EJ.
Ssl3 -

6EJ

(b)

6EJ
2EJ
u;
Ssl3 = T; etc.

Ss12 =

12/
u;

.
0, etc.

_
M,13 -

3
2pAL
13pAL2.
Ms11 = ~ ; M,12 = - ~ ,
3.5-9.

w1 , J

O; etc.

pAL3 .

=105' M,12 = ---,

140

{ii. <I>i = [ -0.5868

e J ; cf>
V04.
53.28, 139.2 /El. <I> = [
I
J.
=
L2
vpA,
-0.1011
0.9258 /E
.
== ~ yp; mode shape 1s unity.
==

3.7-3. (a)

4.).7,

I)

: / (} u

~os 10,1

-7.365
1
2.055

670
Answers to Problems

4.3-8. D = [ 2.707 cos

-1.914

COS

1.293 cos w 2 t
Wit+ 0.9142 COS Wzt

Wit+

14/

4.4-9. D t

_ [ (1.000 - 1.221c,
4.5-2. D - 1.681c,
_ [ - 1.000
453
-1.000
D-

0.1465(t - ~i sin w1~/wt

=[

+ 2(1 + 4(1 -

/3,/wt + 9{3z/ w~] Pi cos flt

3{3,/wt - 3{3z/w~
10m

.!:_

sin w 2 ~/w~ t,m

- 2b2
4b

I -

m =

t - wz sin w2t / w2

+ 0 .2205c2)L] ~
+ l.68lc2
L

4 47.

I)

rJ

I /31Pi L sin flt

- I
88wrm

- 0 .05776(1

4.4-8.

I)

r
w lu11 111

0.04084 r

where

+ 0.1816(1 -

m = pAL
420

wheres= EAIL ; mi = pAL/6; m

= 3.28m1;

_ [(-1.000 + 1.053c, - 0.0532c2)l] (J

456
-1.000 - 0.4156c, + l.415c2
'
D where

3.28 pAL
6

Wi

cos w 2 t)/w~

I
W2

1
(tJ1

/IA/

I '0

.
SIil

CHAPTER 5

5 .2-1. and

Pi

5.22,

U10

= 0.9819k

5.2-4.

Ut 6

= 0.3622k

5.2-6. u20

W1 f ) / (1/11

5.2-11. Given

/\I J

sin

c1 = cos wit
c2 = cos w2t

10,,)/1,d

111/I

5.2-8.
5.2- IO.

P,
- 4.004k

111 0 ""

P,
0.3 146k

II 111

I , 107

I'1

1.

Pi

5.2-3.

Uzo

= -3.063k

5.2-5.

U20

= -2.595k

P,

5.2-7. u,o = 0.9182k

5.2-9.

U1 0

bi

= /3i/wt;

b2 = {3z/wi

c1 = cos w1 t
c2 = cos W2t

b2 = /32/ Wi

o.05776b1 - o. t882b2J so, . fl

458
D = [ -0.04084b 1 - 0.133lb2 --;;;- sm t
b, = {3i/ WT; b2 = {3z/ Wi

w,,)!wr+ 0 . 1882(,
,,'., sin ,,,,,)/M
t I 0, I 1.l 1(,

_!_ sin

c, = cos w, t
C2 = cos W2t

4.4-6. D = r - 0 .001745L(l - cos w,t)/wt - 0.006828L(1 - cos w 2 t)/w?]Mi

0.0006885(1 - cos w,t)/wf
pAL3
where m = - 2 10

sin w,r)/wt
)

J~ cos .at

cos Wzt)/w'i]_!j_
cos w 2 t)/w~ 10m

where

~i

+ 1.207c, + 0.207lc2 JLO

+ 0.5000c1 + 0.5000c2
'

4.5-7. D* = - [ J b2 (J, sin flt

2(1 - cos w,t)/wt

(1 - cos w,t)/wt
h
2.72 pAL
were m =
6

4.4-5. 0

=r-

(r-

bi = 13d wt
b _ r.i. / 2
2 - JJ2 W2
(0.6s + 3m,n2)b1 - 3(1.8s + mifl2)b2J d . r>
455 D - [
- - sm ~u
.
- 3(0.6s + 3m,fl 2)b1 + (I.8s + miil2)b2 lOm

Wz

+ 0.8536(- ~

where m = pAL

4.4-4. D

* _ [ 2b,b

4.5-4. D -

J_ sin w1 t\ I wt - 0. 353/t - J_ sin w) / w~1

w,

si =
s2 =

(b) D = [(0.02675{3i/ wt + 0.2878{3z/ w~)l] M 2 sin flt

0.03683/31/wr + 2.193{32 /wi
mL2
pAL
where m =
; (Dz)A, = (D,)A 2
210

\'

4.5-1. D* - [ -9s, - Sz
-3si + 3s2 lOt,

4 .4 2 (a) D = [(0.01943/31 /wt + 0.03775{3z/w~)l] P1 sin flt

0.02675{3i/ w,2 + 0.2878{3z/ w 2
ml

O. 353/t -

J~

(b) D = [3(1 - cos w,t)/wt - 3(1 - cos w 2 t)/w'i]_!j_

(1 - cos wit)/wt + 9(1 - cos w 2 t)/w~ 10m
3.28 pAL
where m =
; (D2)A, = (D1)A 2
6

=_

~ 2 sin w2t)/w~1___!:_
1
)
z l08t1m
12 t - wz sin W 2 t /w2

pAL
where m = - 6

4 .4 1. (a) D = [ 9(1 - cos Wit)/wf + (1 - cos w 2 t)/w~]_!j_

3(1 - cos wit)/wt - 3(1 - cos w 2t)/w~ 10m

4.4-3. D

+ 24(r-

_ - [t

t - (sin w2t)/w2

571

]M, L

[t + (2 sin Wzt)/wz]D01

4.3-9. D =

Answers to Problems

= J .387 kPi

s = s + 3mfl2
s = 2El/l
m = pAL3 /420

572

Answers to Problems

5.2-12. (D1)20

=
=

5.2-13. (D1)max

- 19.94~; (D2ho

-0.9992~; (D2)max
s

-19.85~

S'l

Answers to Problems

5.4-4. (8) U15 = 0.3676t (b) U15 = 0.3659t

= 0.7950~
s

5.4-5. (a)

U 20

= -2.513t (b)

U20

= -2.541

5.4-6. (8)

U20

-3.942~
k (b)

U20

5.4-7. (8)

Uio

0.9087~
k (b) U10 = 0 .9119~
k

5.4-8. (a)

U10

0.3191~
k (b)

5.4-9. (a) U10

l.372t (b)

U10

l.377t

5.4-10. (a) U10

J.094~
k (b) U10

1 .098~
k

EA

wheres= L

For
problem sets 5 .3 and 5 .4 ' parts (a) and (b) denote average- and linear-acceleration
methods.

5.3-1. and 5.3-11. Given

5.3-2. (a) U10 = 1.031
5.3-3. (a) U20

(b)

U10

-3.023~ (b)
k

U20

5.3-4. (a) U1 6 = 0.3676t (b) U16

l.016t
-3 038~
.
k

- 2.513t" (b)

U20

-2.541

5.3-6. (a)

U zo

-3.942t" (b)

U20

-3.963t"

5.3-7. (a)

U10

0.9087t (b) U10 = 0.9119t

5.4-13. (a) (D1)max = - 0.9974~ (b) (D1)max

= 0.7917~

(D2)max

5.3-10. (a)

uw

l.094t" (b)

Uw

uio

l.37?t

- _

l.098t"

= 0.7860~
s

5.4-1 and 5.4-11 Given

5.4-2. (a) U1 0 =

5.43. (11)

Pi
- .1.02.\ k (h)

11 111

.03 ]

(b) U10

11 111

(A tj)1

+ -

-i>1 -

'

-s;

Aj+I = Aj+I +

1 0 lH /'i

c -Qj + s _Rj

AtJfr. R 1 = -D1 - MD
.. - (At1 )2D..1
1 1
2

At
5.4-17. C = C +-M + ___!S
At1
2 '

'

AA=
AA1 + MQ1 + SR1
1

where Q1 = 2DJ; RJ = -AtJDJ

5.4-18. M = M + Atic + (AiY s AA1 = AA1 +

2
4'

l.0J 6t
.

At1

where -Q =

(D2)max

5.4-16. M - M +

= 0.7917~

= 0.7861~

Ati

Atj

1
5.3-13. (a) (D 1)ma. = -0.9976: (b) (D 1)max = -0.9994~
(D2)max

5.4-15. C - C + -,.-M + -S; Ai+t = A1+ 1 + M Q1 +SR1

1.1tj
2
2 ,
.. flt ..
where Qi= - D1 + D1 R1 = -D1 - ___!D1

(D2ho = - 19.51 ~

- 0.9993~

4
4,
.. _
2
.
where Qi= (Ati)2Di + Mi Di+ Di; Ri = Ati Di + Di

5.3-12. (a) (D1ho = -20.45~ (b) (D1ho = - 20.32~

s
(D2)zo = -19.47~

5.4-14. S = S + (AtJ1M + tltj C; Aj+I = Aj+I + M Qj + C Rj

0.3 191 ~I (b) U10 = 0.3176t

5.3-9. (a) U10 = l.372t (b)

(D2)max

t"

Uzo

U10

0 .3176~
k

5.4-12. (a) (D1ho = -21.22~ (b) (D 1)20 = -21.40~

= 0.3659t

5.3-5. (a)

5.3-8. (a)

U10

-3 .963~
k

c -Q1 + s -R1-

1174

Annwur " to l'rublome

At,

5.4-19. S* '"' S I (A:)2 M I ~lt1 C;

A1,, I M

Q/

I CR/

"
6 D1
+ 2D1;
R/ == -;.3 Di -1 2D1
. I At D1
= (A 6 )2 D1 + -;.1
u.tj
u.lj
u.t1
2
5.4-20. C* = C + AtM
+3
2
At1S ; Aj+I
* = A1+1 + M QJ* + S RI *

where

Q/

2 1 + D-
1 R -* = - D. - 2At1
1
where Q*
= -D
D
1
1
'
A t1
3 1

5.4-21. M *

AfJC + ~
(AtJ)2S ;
=M +2

-(At1)2 D
6
1

= A~+l + CQ* + SRj *

A*
j+ I

CHAPTER 6

6.2-1. pi:i,)

Lby-{t) pi:k{t) =
Lby(t)
78
18 l92
OJ l92;
[OJ
[1IL
-SL

At R * = -D - At-D. - - --='---D'J
(At-)2 .
1
where Q.*
= -D 1 - -2D
1
1
2

;,

2
5.4-22. C* = C + A M + Atj S; AA/ = AA; + M Q/ + S R/
u.tj
3

Q/ = 2D1;

where

Pbi(t) =

. - T
(At)2
R/ = -AtiDi
D j

=M +

~jc + (At

-ArD
R*
J
1,
J

where Q*
J

s;

AA/= AA1 + C

Q/ + s R/

6.2-2. p;/t)

= S + /3(AtJ 2 M + /3 Ati C;
1

Ai+1 = A1+1 + M Qi + C R
1
1 - 2(3

[1~]

PPit)
5.5-1. S

=[

54
72 ]Lby-(t)
960
-25L

P~f); p;k(t) = [ 5~
-9L

3L

2
= -AtD - (AtJ
2 1)
J

=[

55L

5.4-23. M*

234]Lby-{t) ; Pbk(t)
312
960
[

-10

to

Py,(t)

64\/2 ; p,,k(r)

=[

3V2L

J~f)
- 54

54

/~,,(I)
64V2

- 9V2L

M,(t)
6.2-3. p~(t) = {O, - 6, -L, 0 , 6, - L}4L

00

where Qi= /3(A tJ Dj + /3 Ati Di+ ~ D i

2

and

R1 = /3 Ati D1 A

(1 - "jjy) Di o

y) At1D 1

1-

213
1
5.5-2. C = C + +
M + /3 yAt S; A1+1 = A1+1 + M
y u.tj
1
1-y
where Q1 = - - D1 + - D
1
y At1
y

lt + s Rj

and
5.5-3.

- D1 - (1 - ~) Mi)1 -

G-

~)(At1)2 D
1

M=

M + y dtjC + /3(Af_j) 2 S; Aj+I = A1+ 1 + C Qj

where Q1 = -D1 - (1 - y) At1Di

and
5.5-4.

R.1 =

R.1 =

-D1 - At1Di 1

C = C + -y Atj-M + /3 ydtj S;
where Q1 =

;D

1;

R.1 =

G-

+s~

/3)(AtYD1

AA.j = dA; + M Qj + S

-At1Di -

G-

~)(dt1>2 D1

R.1

PM(t) = {-12, -6,

-VSL,

M,(t)
12, 6, -VSL} 4V5L

Lb2(t)
6.2-4. pi:(t) = {O, 9, 2L, 0, 21, -3L}6()

pb(t) = {- 36, 27, IOL , -84, 63, -ISL} Lb2(t)

300
Py,(t)
6.2-5. p;(r) = {O, 9, 2L, 0, 9, - 2L}-9
' r,:
Py,(t)
p,,(t) = {-1 8, -9, 2V5L, -18 , -9, - 2v5L}
V5
9
M,(t)
6.2-6. p,&(t) = {O, -18, -L, 0, 18, -L}~

M,(t)
PM(t) = {18, -18, -V2L, -18, 18, -V2L} 8V2L
Lbi(t)
6.2-7. pi:(t) = {l, 0, 0, 3, 0 , 0}12

p,,(t) = {2, - 1, 0, 6,

Lb2(t)
3, 0} 12VS

At1Nwnr" lo l'r oblumu

,,o,

' , ..I I , p,,' (I )

I'" (I)

,,.J..z.

5/,, IH , 0, 11/ , /Hf 1,/, (I)

192

(20/,,

1>,\(1) ... {O,

44/,, 331., 390} Lb, (!)

960

I5L, 90,

4L , 20, 0, 2L , 7} p;~)

577

6.4-4. p,:(t) = {O, 2, 0, 0, 3, O} Py,(t)

5

PPi (t) = ~PPk(t) = {O, 0.2828, - 0.2828}Py,(t)

6.4-5. ph(t) = {O, -1, 0, 0, l, O} M~(t); L =

V29

PMi(t) = -pMk(t) = {-0.07773, -0.1184, - 0. 1201}M,,(t)

{ 4L , - 12L, 20VW, 2L, 6L, 7VJO} P, (i)

27VJO

1>,,(1)

tt.J.J . ,,;,(,) - {O, 3l, 18, 0, -5l, - 18} My-(t)

16L

l>M(l) ... { 3L, 3l, 18V2, 5l, - SL, - 18V2} My(!)

16V2l
t,..1.4, 1,:1(1)

{12l, 3L, 18 , 4L, - 5L, - 18} My(t)

16V2L

P.11(1) :.:: {9l , 15L, 18V2, 9L - l - l8V2}My{t)

'
'
32l

td-5. 1>J,(1) ""' {12L, -3l, - 18, 4l, 5L, 18} M;)!j_
16 2l
l>M (l)
{1 5l, 9l, - J8V2, - l, 9l, J8V2}Mx(t)
32
,, . .1.(,. J>,;(I) {O - 3L 21 0 2L 9}lbi(t)
'
' ' '
'
60
1>,,(1)

{ 9L, - l2l, 105, - 6l, -8L , 45} Lbi(t)

300

(t,J.7, 1>/,(1) ... {O,

1'1 (t)
(,.J. ff , 1J,(1)

'

'

' '

16

{9L, I 2l, 80, - 9l , - 12l 80} P, (t)

'
80
(0 I O O - I O} My(t)
3
,

'

{2 I O - 2 - I O}M,.-(t)

,,,4- L 1>,;(1)

{O, I, 0, 0, I, O}Lb)"(t) ; L

,I

'( )

'"" 1>,, I

()

t,,4.,1. 1>,\(t)
I>/'/(/)

'

3\/5

= VJ4

1>114(/)

{O , 0, 3, 0, 0 , 2)
fo,,d1)

0.5192}b,,,(t)

- ; L = v 14
6
{ 0.3126, 0.2233, 0.49 12}/J,,(,)
} / ,/),,(/)

I , 0, 0, I -

'

1>,,,(t)

4
l>11A(I) ""' { 0.3043, 0.7 161 ,
{()

_ {
My-(t)
PMk ()
t - 4, 0, 0, 0, - 4, 3}---is-

_ {
M,,(t)
656 PMj ()
t - 0, -16, 12, 5, 0, 0} - 75
_ {
M,,(t)
PMk ()
t - 0, 16, -12, 0, 0, 0}75

'

l>M (I)

' . '

V29

PMi (t) = -pMk(t) = {-0.1332, -0.03805, 0.1327}My-(t)

6.4-7. PMt) = pbk{t) = {- 1.000, 0.4242, O}b(t)
Pbi (t) = Pbk (t) = {-0.5992, -0.1141, -0.8988}b(t)
6.4-8. pi:)t) = 4Pi:h) = {-0 .3334, 0.6009, O}b2 (t)
Pbi(t) = 4Pbk(t) = {-0.3234, 0.3637, -0.4851}bi(t)
6.5-1. Pbi(t) = {-1.057 , 2.439, -0.4269, - 1.769, - 0.49 13, 1.572}b,,(t)
Pbk (t) = {-1.057, 2.439, -0.4269, 1.769, 0.491 3, - 1.572}/JAI)
6.5-2. Pbi(t) = {- 1.601, - 0.4447, 1.423, 0.6523, - 1.505, 0.26'.'ltlfh,,,(1)
Pbk(t) = {- 0.3695, - 0.1026, 0.3284 , - 0.2965, 0.6842, 0. 1 l<J7fi,,.,(t)
6.5-3. PPj(t) = {-0.3675, 0.4492, 0.6 125, 0.3811, 0. 1429, (), 111, f/ \,(/)
PPk (t) = {-0.06805 , 0.08318, 0.1 134, 0. 1270, 0.047M, 0 111 )} /\,(/)
6.5-4. ppit) = {0.1132, - 0.04244 , 0.09902, 0.07639, (),()() l.1(1, (), I l'/ \f/ 1 , (/)
PPk(t) = {0.6111, - 0.2292, 0.5347 , 0.2292, 0.280 I , 0. IHI C) )/ 1,(/)
_ {
My-(t)
6.5-5. PMj (t) - -4, 0, 0, 0, 0, O}---is-

JL J6 0 3L 16}?,(t)

'

1>1,1(/)

6.4-6. ph(t) = {O, 0, I, 0, 0, -1} M~(t); L =

6.5-7. Pbi(t)
Pbk(t)
6.5-8. Pbi(t)
Phk(t)
6.7-1.

W1

6 , 7 2,

W1 2

6.7-3.

W1

r,-;

'\(I)

(- 0.4899, 0.2449, 0.2449}/>.,(t)

'

{1.5, 0, 0, 0, 0.5, 0.5}bx(t)

= {1.5, 0, 0, 0, -0.5, - 0.S}b,.(t)

{- 0.01111 , 0.02222, 1.028, -0 .3000, - 0.1500, O}b 1(t)

= {0.01111, - 0.02222, 0.4722, 0.2000, 0.1000, O}b 1(t)

15.69
L2

/El,

-V;;;.

= 5.603,

= 9.941

l2

31.19 2l

-v/El,
pA

lfil,
-

pA

; <f>

[ 1.000
0.5435

1.000
-0.9364

1118

t,.7-4,

6.7-5.
6.7-6.

An"wnr11 to l't uhln1111

t,)1,1 . ,

W 1,2 -

W 1.2

0, 0, 22.47 /}1
I
1. 648 , 5.540L2

Jn

pA ; (I>

{iii. ; <I> =
Y"iA

{iii,
= 0.9044, 3.840 L2I Y"iA;

/G]
6.7-7. w, __ 6. 149 Y"iA

u:-

6.18 .

[0.8844
1.000

O
IJ
_ 5.943 /GJ
w, - u:- Y"iA
<I>=

[ 01

Index

Absolute damping, 166

Anisotropic materials, 311
Absolute system of units, 495
Antisymmetric modes , 112
Accelerations:
Applied actions, 4
ground, 21: arbitrary, 42; periodic,
normal-mode response, 152
38; rigid-body, 159
Applied body forces , 80
incremental, 212
Arbitrary forcing functions, 38, 172
relative , 23
Arbitrary ground accelerations, 42,
Accelerometer, 22
159
Accuracy, 208
Assembly:
numerical, 223
of elements, 95
Action equations of motion, 140
of substructures: in componentAction vector:
mode method, 470; in tridiagoapplied, 152
nal method, 448
reduced, 282
Average-acceleration method, 219
Actual nodal loads, 98
direct linear extrapolation, 212
Adjoint matrix, 106
iteration with, 203
Algorithmic damping, 218
Axes:
Amplification matrix, 223
body, principal, 420
Amplitude:
rotation of, 95
forced vibration, 21
stress, principal, 315
free vibration, 16
translation of, 157, 414
suppression, 225, 234
Axial element, 82
Angular frequency:
Axial rigidity, 94
dunircd free vibrnlions, 28
Axinl s1rnins, 83
u111l11111ptd l'itt vih1111 ions, l 11, l O'i
co11strni111 s 11uuim.t, NO

580

Axisymmetric loads, 357

Axisymmetric shells:
element AXSH3 , 394
nonaxisymmetric loads , 401
Program DYAXSH, 406
Axisymmetric solids, 357
element AXQ4, 359
element AXQ8, 359
nonaxisymmetric loads, 361
Program DYAXSO, 365

Backward substitution, 461

Beams:
continuous, 243 ,
frequency coefficients (table), 288
Blast load, 50
Block mass submatrices, 437
Block stiffness submatrices , 437
Body axes, principal, 420
Body forces, 78
Body-oriented method, 434
Boundary nodes, 446
Bracing in tier buildings, 468
Buildings, multistory, 425

Center of mass, 418, 427, 434

Central-difference predictor, 197
Characteristic equation, 105, 498
Characteristic matrix, 106
Characteristic-value matrix, 141
Cholesky method, modified, 498
Cholesky square-root method, 106
Coefficients:
damping, 165
ncxibility, 145
ror Gaussian Quadrature (table) ,
328
frcqullncy, 287 , (t(lb/e), 288
lllUSS , 82
st i ffrH.'SS, 82
<'0111pom11t rnodl' nwthod , 468
IINNl'lllhly of s11h11t1m'llll'l'N, 470

Index

dynamic coupling in, 470

for plane trusses, 471: Program
COMOPT, 477
for space trusses , 477: Program
COMOST, 477
substructure equations of
motion, 468
Computer programs, 9
Condensation, matrix, 282
Conditional stability, 208
Consistent-mass matrix:
for axial element, 84
for axisymmetric shell element
with nonaxisymmetric loads ,
405
for axisymmetric solid element
with nonaxisymmetric loads,
364
for element, 82
for element AXQ4, 359
for element AXSH3 , 401
for element H8, 347
for element PBQ8, 378
for element Q4 , 336
for element SHQ8, 389
for flexural element, 91
for grid member, 251
for plane frame member, 246
for plane truss member, 103: in
component-mode method, 474
for space frame member, 261
for space truss member, 255
for structure, 98
for torsional element, 87
Consistent systems of units (table),
496
Constant-acceleration method, 219
Constraint, modal, 500
Constraint matrix, 291, 372, 382, 394
Constraints:
against axiul strains, 290
rlgid-body, 4 I I , 427
C'ontinuu:
Hll'l'SSl'S

and Hll'HillN in , J 10

Index

two- and three-dimensional,

310-69
Continuous beams, 243
loads , 265
Program DYNACB, 265
Convergence, rate of, 206
Conversion factors, 496
Conversion of US units to SI units
(table), 496
Coordinates:
natural, 318
normal , 139
relative, 22, 159
Corrector, 203
Critical damping , 30
Critical time step, 198, 224
Curvature, 90

D' Alembert's principle, 3, 13

Damped forced vibrations, 30
Damped free vibrations, 27
Damped response of MDOF systems:
to arbitrary forcing functions , 172
to periodic forcing functions, 168
Damping:
absolute, 166
critical, 30
effects of, 25
in MDOF systems, 164
modal , 167
numerical, 218
proportional, 165
relative, 166
Damping coefficients, 165
Damping constant, 3, 27
modal, 165
Damping matrix, 164
Damping ratio, 31
modal , 165
Decomposition:
Fourier, 36 1
spectral , 223
Dl'flution , 500

581

Degrees of freedom, number of, 105

Density, mass, 80, 497
Dependent displacements, 282 , 291
Determinant of Jacobian matrix, 330,
332
Diagonalization, 140
Differential equations of motion, 75
Dimensionless coordinates, 319
Direct iteration, 498
Direct linear extrapolation, 211
average-acceleration method, 212
linear-acceleration method, 213
Direct numerical integration methods ,
195-240
Direct stiffness method, 98
Direction cosines, 96, 101
Discretization of structures, 6
Displacement equations of motion ,
145
Displacements:
dependent, 282 , 291
generic, 75 , 78
ground, 21: rigid-body, 157
incremental, 38, 212 ,
independent, 282, 291
initial, 14, 147
kinematically equivalent, 414
nodal, 75, 79: free, 99; restrained,
99
relative, 23 , 159
Displacement shape functions , 74 , 79
for axial element, 82
for element AXSH3 , 398
for element H8, 345
for element H20 , 349
for element PBQ8, 375
for element Q4, 333
for element Q8, 338
for element SHQ8, 386
for flexural element, 89
for torsional element, 85
Dissipative force, 3
Duhumcl' s integral , 39, 152, 173
Dy11n 111 il' tq11ilibl'i11111 , I

lnda,c

ISB2

I>y11111111l' lrnt'l',
I >y111111m ,ntlucnc.cs, 4
I >y1111111k load data:

lo, 11u11tislory plane frames (table) ,

lh5
trn plane frames (table) , 269
IOI plane trusses (table), 180
IOI Program DYAXSH (table), 407
tor Pmgram DYNAPB (table) , 380
IOI l'rogrum DYNAPS (table), 341
IOI Prngram OYNASH (table) , 391
lor Progrum DYNASO (table), 352
Im :-.puce fnunes (table) , 279
t >y1111111ic loads, 1
hlnst, 50
ICl'llll'Ocul theorem, 154
I >y1111111ic reduction, 282

li:!1,wnvaluc matrix, 14 l
hiHl'IIVttlue problem, 105, 314
lwnvulucs, 106, 498
ll'Pl'Ull'd, I 07
1'.11\l' nvcclor matrix, 140, 315
1 1),tl'llVCClors, 106, 498
11m111ulization of, 142
l,lusticity , modulus, 497
1'111st it'ily force, 3
llt111l' II( /\XQ4 , 359
i'.h111L11t /\XQ8, 359
lh1m111 /\XSll3 , 394
hlt'llll 1l( /\XSR3, 394
ll1111tnt 118, 345
(<'.ltJlll'II( 11 2(), 349
I ,lt111Lnt PBQ8 , 371
I k1m11l PQR8, J7 I
lk1m11t ()4 , 1 n
1111111111 ()8, JJ8
I l11m11t R4 , 1 U
lh111111t RH , \ lH
lh1111111 RS8, '4.'l
lh1m11t RS l O, W>
lh1m11t SI 1()8, IH)
II11111111111011.
1

1111 WIii d, ,1(,1

Clauss .Jordun, 291

Elimination method, purolld, 4~':t
Equations of motion , J , 11, 7':t, 18
action, 140
displacement, 145
for finite elements, 78
incremental , 2 12
structural, 98
Equivalent nodal loads:
for element, 82
nonproportional, 267
for structure, 98
for surface pressure, 366, 408
Equivalent viscous damping, 26
Euler's extrapolation formu la, 20~
Explicit formulas for extrapolation , 11) I
Extrapolation , direct linear, 2 11

Factorization:
Cholesky, l06, 498
Givens, 514
Factors , conversion, (table), 496
Finite-difference formulations, 195
Flexibilities, I 0
Flexibility coefficients, 145
Flexible-body motions, l 38
Flexural deformations, 90
in plate-bending, 378
Flexural element, 88
Flexural rigidity , 94
Flexural strains, 89
Forced vibrations , 19 , 30
Forces, body , 78
Force systems, types of, 241
forcing functions:
arbitrary, 38, 172
hurmon ic, 19
periodic, 35, 168
piecl'wise lincur, 46, 175, 182
Fonn f1w1or , 177
FORTRAN, 9

Fm wnrd tli11ii11ntio11, 4<> I

F111w111d lll'llllioo, 41>8
111111111 nwll1r1111h, IC,

583

Index

Fourier series , 35, 169, 361, 401

Framed structures, 6, 241-309
Frames:
plane, 243 , 244
space, 244, 259
Free nodal displacements, 99
Free vibrations, 12, 27
Frequency, 14, 28
Frequency coefficients, 287, (table).
288

Gaussian quadrature, 328

coefficients for (table), 328
Gauss-Jordan elimination, 291
Generalized acceleration method,
217
Generalized rotation matrix, 505
Generalized stresses and strains, 92
General shells:
element SHQ8, 382
Program DYNASH, 390
General solids:
element H8, 345
element H20, 349
Program DYNASO, 351
Generic displacements, 75, 78
Geometric center, 318
Geometric interpolation functions ,
320, 324
for element AXSH3, 396
for element HS, 345
for element H20, 350
for element PBQ8, 374
for element Q4, 333
for element Q8, 338
for element SHQ8, 383
Geometric transformations, 431
Givens factorization, 514
Givens method, 509
Global axes, 95
Sram-Schmidt orthogonalization,
107, 500
Grnvitalionnl system of units, 49.5
<liid~. 24 I, 249

Program DYNAGR, 270

Ground accelerations, 21
arbitrary, 42
periodic, 38
rigid-body, 159
Ground displacements, 21
rigid-body, 157
Guyan reduction, 282, 445

Harmonic forcing functions, 19

Hexahedral coordinates, 323
infinitesimal volume in, .no
Hilbert-a method , 222
Householder method, 'l 11

Immobility, 1>ta1 it', II'I

Implicit fo11n11lt1Nfrn ill' lllll1111 , IO\
Incremental ncnkratH111s, 1 1)
Incremental uclions, 21l.
Incremental tlispluccments, J8 , 2 12
Incremental equations of motion, 212
Incremental impulse, 38
Incremental velocities, 38, 212
Independent displacements, 282, 291
Independent motions of support
restraints, 162
Inertia, moment of, 91
Inertial body forces, 80
Inertial force, 3
Inertial moment per unit length, 87
Inertias , translational and rotational
(rotary), 91, 378, 390, 401
Infinitesimal area in quadrilateral coordinates , 329
Infinitesimal volume in hexahedral
coordinates, 330
Initial conditions, 4, 14
normal-mode response, 147
Integration, numerical, 326
Integration points:
for hexahedron , 332
for quadri lateral, 330
I111l1fo1 lllldl'N, 44(1

ll84

h11 hue

Index

585

lllll'I polutioo f'1111ctio11s:

gLonwtlie, J20, 325 (.1<<' olso CJ co

111etric inlerpolation functions)
piecewise-linear, 46, 175, 182
Inverse ilcralion, 498
Inverse of Jacobian matrix , 321 , 325
In verse of normalized modal matrix ,
143
lsoparamctric elements, 310- 69
clement AXQ4, 359
elcmenl AXQ8, 359
hcxahcdron H8 , 345
hexahedron H20, 349
quadrilateral Q4, 333
quadrilateral Q8, 338
Iso tropic materials, 77 , 311
Iteration (for eigenvalues), 498
ltcralion with implicit formulas, 203
1

,lncobian matrix, 321, 324

determinant of, 321 , 325
inverse of, 321, 325
Jacobi method, 505
Joinls, 241

Kilogram (kg), 495

Kilopound (kip) , 497
Kincmalically equivalent displacements, 414
Lu~ time, 17
I,inear-acceleration method, 219

direct Iinear extrapolation, 2 13

iteration wilh , 206
Linear differential operator, 79
I ,incur extrapolation , dirccl, 2 11
I .oud oplll'ltt or, 223
Louds:
11xi sy 11111wtric, 357
rn1 rn11lin11011s hcums, 265
dy11111n il', I
111ip1t w 11l tlwc111111 , 15,1

ct1uivulcnt nodul :
for elcmenl , 82
for structure, 98
nonaxisymmetric, 36 1
normal-mode, 152
on grids, 27 1
on hexahedra, 351
on plane frames, 267
on plane trusses, 179
on quadrilateral s, 340
on space frames , 278
on space trusses, 274
Local axes, 95
Lumped mass matrix, 99

Magnification factor, 20, 32, 169

Main program:
for DYNA (flow chart), 226
for NOMO (flow chart) , 178
for VIB (flow chart), 11 9
Mass
center of, 418 , 427, 434
of a body, 419
Mass coefficients, 82
Mass density, 80 , 497
Mass matrix:
for element, 82
for structure, 98
lumped, 99
principal , 140
reduced, 283
Mass moments of inertia of a rigid
body, 41 9
principal, 420
Mass products of inertia of a rigid
body , 419
Master displacements, 283, I 0.2- 1<)
Material properties (table), 497
Materials:
anisotropic, 311
isotropic, 77, 311
orthotropi<:, J I I
pmpl'I I ils of (/o/,/1') , ,11>7
Mntd \ l 'Clll<ll' IINIIIIClll , 18 )

MDOF systems, 74
Mechanical analogue, 12
Member-oriented approach, 416
Members , 241
Meter (m), 495
Modal constraint, 500
Modal damping , 167
Modal damping constant, 165
Modal damping ratio, 165
Modal matrix, 140, 505
normalized, 142: inverse of, 143
Modal truncation, 139, 143
Modes:
antisymmetric, 112
flexible-body, 138
rigid-body, 138, 148
symmetric, 112
vibrational, 138
Mode shape , 105
Modified Cholesky method, 498
Modified tridiagonal method, 451
for multistory buildings, 457
Modulus of elasticity, 497
Motion, equations of, 3, 13, 75, 78
Moving load:
on a beam, 266
on a plate, 381
Multiplicity (of repeated eigenvalues) , 107
Multistory buildings:
modified tridiagonal method, 457
with rigid laminae , 425
Multistory plane frames:
dynamic load data (table), 465
Program DYMSPF, 463
structural data (table) , 464
Natural coordinates, 318
Natural frequency, 14, 28
Natural period, 4, 14, 28
Newmark-,8 method , 2 18
Newton, 495
Nodal actions, 7<)
Noda I l'i re ks. 15 I

Nodal coordinates:
for element H8 (table), 347
for element H20 (table) , 351
for element Q4 (table), 335
for element Q8 (table) , 340
Nodal displacements , 75 , 79
free , 99
restrained , 99
Nodal loads:
actual, 98
equivalent, 153
for element, 82
for structure, 98
Nodal vectors:
for element AXSH3, 396, 406
for element SHQ8, 385,39 1
Nodes:
boundary, 446
interior, 446
of finite elements, 6, 73
Nodewise solution, 198
Nonaxisymmetric loads:
on axisymmetric shells, 401
on axisymmetric solids, 361
Nonproportional equivalent nodal
loads, 267
Nonna! coordinates, 139
Normalization of eigenvectors, 142
Normal-mode load, 152
Normal-mode method, 138-94
Normal-mode response:
Program NOMO for, 177
to applied actions, 152
to initial conditions, 147
to support motions , 157
Nonna! stresses and strains, 75
Notation, 483
for matrices and vectors, 483
for programs, 487
Greek letters, 486
simple variables, 485
subscripts for matrices and vectors,
484
N111111111 11I lll'l'tll ary,
N 11111e11111l d11111p111p,

.U I
1 18

~88

N11111l'lll'lll 111ttg1allo11, I 2C,

N11111t1 il-111 stabilit y, 22.l
( )uc-dlmcnslonal clements, 82
relationships, 140
<>1 thogon11l ization, Gram-Schmidt,
107, 500
Ot thotropic materials, 311
<>ullinc of' Program DYMSPF, 463
Outline of Programs DYPFAC and
DYSFAC, 299
Output selection data (table), 182
Ovcrdumping, 29
Overlay technique, 453, 461
<)1 tho).\onulily

Pnrnllcl elimination method, 455

1'11rc11t rectangle:
tlcmenl AXSR3, 394
clement PQR8, 371
lllt.lment R4, 333, 336
l'lt.lmcnt R8, 338
1'11rcnt rectangular solid:
clement RS8, 345, 348
clement RS20, 349
Pt11K'td, 496
Ptriod, 4, 14, 28
l't'riod elongation, 225, 234
l't11iodic forcing functions, 35
d11111pcd response of MDOF systems, 168
l',1 iodic ground acceleration, 38
Ph1111c angle, 16, 32, 105 , 169
l'IIIINC plane, 14
I'ittcw isc-1inear forcing functions ,
46 , 175, 182
l'l1111t frumcs, 243, 244
11xi11l constraint s, 290: program
DYPFAC, 299
dynnmic loud duta (table), 269
1'10g111111 DYNAPF, 267
1lgid bodies: Program DYRAPF,
,1}4
Nlllll'tlll'lil d11t11 (tnl>/t), 268

l11rla11

l11clox

587

l'liuw slltss und pln11t sl111i11 , I l I

l'rogri1111 DYNAPS , MO
Plane trusses, 24.l
component mode mcl11od, 471
Program COMOPT, 477
dynamic load data (tabll'), 180
Program DYNAPT, 227
structural data for (tab/<:), 110
Plate-bending:
element PBQ8, 37 1
flexural and shearing dcf'o rmu
tions, 378
Plates, 370-410
Poisson's ratio, 497
Polar moment of inertia, 87
Pound (lb), 495
Predictor, central-difference, 197
Predictor-corrector method , 203
Pressure, surface, 366, 408
Principal body axes, 420
Principal coordinates , 139
Principal damping matrix , 165
Principal flexibility matrix, 145
Principal mass matrix, 140
Principal mass moments of inerliu ,
420

Principal normal strains, 316

Principal normal stresses, 314
Principal planes of bending:
for space frame member, 259
for space lruss member, 255, 477
Principal stiffness matrix, 140
Principal stress axes, 315
Program COMOPT for plane trusstis
by component-mode method,
477
Program COMOST for space trusses
by component-mode method ,
477
Program DYAXSII for axisymmetr il'
shells, 406
dynamic load dnUI (tablt'), 407
structural data (ta/J/1), 406
Progn1111 DYAXSO for 11xisy11111Hitd
cu l solids. l<i',
'

Program OYMSPF for multistory

plane frames , 463
Program DYMSTB for multistory
tier buildings, 463
Program DYNA for dynamic
response, 225
main program (flow chart), 226
Program DYNACB for continuous
beams, 265
Program DYNAGR for grids, 270
Program DYNAPB for plates in
bending, 379
dynamic load data (table), 380
structural data (table), 380
Program DYNAPF for plane frames,
267
dynamic load data (table), 269
structural data (table), 268
Program DYNAPS for plane stress
and plane strain, 340
dynamic load data (table), 341
structural data (table), 341
Program DYNAPT for plane trusses ,
227
flow chart, 519-63
Program DYNASF for space frames,
277
dynamic load data (table), 279
structural data (table), 276
Program DYNASH for general
shells, 390
dynamic load data (table) , 391
structural data (table), 391
Program DYNASO for general
solids, 351
dynamic load data (table), 352
structural data (table), 352
Program DYNAST for space trusses,
273
Program DYPFAC for plane frames
with axial constraints, 299
Pro11ram DYRBPB for rigid bodies
in plate-bending continua, 438
rigid-body data (table), 440
rigid body dynamic-load data

(table), 440

Program DYRBPF for rigid bodies in

plane frames, 424
Program DYSFAC for space frames
with axial constraints, 299
Program NOMO for normal-mode response, 177
main program (flow chart), 178
Program NOMOPT, 179
dynamic load data (table), 180
Program notation, 487
Programs for framed structures, 264
Program VIB for vibrational analysis, 118
main program (flow chart) , 119
Program VIBPT, 11 8
structural dutu (tr1bll') , I 20
Proportional dH111pinH, I (1 ~
Properties of' 11111tt1 l11ls (tali/,) , ,111I
Proportionul du111pi11g, IM
Proport ion11 I lo11ds, I8 J
Pscudostutic prohhm1s, 2 I )
Quadrature, Gaussian, 328
Quadrilateral coordinates , 318
infinitesimal area in, 329
QR algorithm, 514

Radius of gyration:
for cross section of grid member,
251
.for rigid body, 433
Ramp function, 43
Ramp-step function , 55
Rate of convergence, 206
Rayleigh quotient, 500
Reactions at supports , 99
Reciprocal theorem for dynamic
loads, 154
Rectangular impulse, 41
Rectangular parent:
element AXSR3, 394
clement PQR8, 371

688

lndox

Rectangular parent (cont.)

clement R4, 333, 336
clement R8, 338
Rectangular solid parent:
clement RS8 , 345, 348
clement RS20 , 349
Reduced action vector, 282
Reduced equations of motion, 448
Reduced mass matrix, 283
Reduced stiffness matrix, 282
Reduction:
dynamic, 282
Guyan, 282, 445
static, 282, 420
Redundant constraints, 290
Reference point on a rigid body,
4 13,4 16
Relative acceleration , 23
Relative coordinates, 22, 159
Relative damping, 166
Relative displacement, 23
Repeated eigenvalues, 107
Resonance, 21
Response, 4
steady-state, 20, 31
transient , 25, 34
Response calculations , step-by-step ,
45, 175
Response to harmonic forcing function:
l'orced part, 20
free part, 20
Rtsponse spectra, 5 1
Rtstrai necl nodal displacements, 99
Retai ned modes , 469, 472
l{tvcrse iteration, 498
l{igid-boclies, 8, 4 11 - 43
equations of motion, 4 18,437
III finite-clement networks, 434
in frumed structures, 4 13
In plane fr u111cs, Program

l>YRBPJI, 424
111 pink lwnding rnnt i111111 , Prn
g111111 DYH ll l'B, 1118
IIIIIHN

1111d

ll ll l 'iN lllllllH11ll

of im11111

Index

589

Rigid-body center of mass, 41 8

Rigid-body constraints, 411 , 427
Rigid-body data for Program
DYRBPB (table), 440
Rigid-body dynamic-load data for
Program DYRBPB (table), 440
Rigid-body ground accelerations, I!'! '>
Rigid-body ground displacements,
157
Rigid-body modes, 148
Rigid-body motions , 138, 153
Rigid-body reference point, 4 J 3, 4l /
Rigidity:
axial, 94
flexural, 94
torsional , 93
Rigid laminae in multistory build
ings, 427
Ring element, 357
Rotary inertias (see Rotational incr
tias)
Rotating vectors, 14
Rotation of axes:
for grid member, 252
for plane frame member, 24 7
for plane truss member, J02: i11
component-mode method , 47~
for space frame member, 263
for space truss member, 256
for stresses and strains, 3 14
Rotation-of-axes transformations, 1>
.~
Rotation matrix , 97
generalized , 505
Rotational (or rotary) inertias, 91,
378, 390, 40 1

SDOF systems, 10
Second moment of area, 9 1
Second (s), 495
Selection data, output (/ahfr) . IH.l
Series:
1:ou1 kr, JS
s11h~l111l'll11 t11 111 ,14c,
Sllhncks 111 1llt h111!dl11~11. ,l(,H
1

Shape functions, displacement, 74,

numerical, 223
79 (see also Displacement shape
unconditional, 208
functions)
Stability criterion, 224
Shape functions and derivatives for
Standard , symmetric form (of eigenelement Q8 (table), 340
value problem) , 106, 505
Shear cores in tier buildings, 468
Static determinacy, 145
Shearing deformations:
Static equilibrium, 1
in plate-bending, 378
Static immobility, 145
Shearing stresses and strains, 75
Static reduction , 282, 420
Shell element SHQ8, 382
Statically
equivalent actions, 41 4
Shells, 370-410
Steady-state forced vibration, 20
axisymmetric: element AXSH3,
Step function , 40
394; Program DYAXSH, 406
Step-by-step response calculutions,
general: element SHQ8, 382;
45 , 175
Program DYNASH, 390
Stiffness coefficients, 82
Shifting , spectral, 503
Stiffness matri x:
SI units , 495
for axial clement , 84
Slave displacements, 283, 457
for axisymmctric shell clement
Slug, 495
with nonaxisymmetric loads,
Solid of revolution (see Axisymmet405
ric solids)
for axisymmetric solid element
Solids:
with nonaxisymmetric loads,
axisymmetric: isoparametric ele364
ments , 357; Program DYAXSO,
for element, 82
365
for element AXQ4 , 359
general: isoparametric elements ,
for element AXSH3, 400
345; Program DYNASO, 351
for element H8, 347
Space frames, 244, 259
for element PBQ8, 377
axial constraints, Program DYSfor element Q4 , 335
FAC, 299
for element SHQ8 , 389
dynamic load data (table), 279
for flexural element, 91
Program DYNASF, 277
for grid member, 249
structural data (table), 276
for plane frame member, 244
Space trusses, 243 , 253
for plane truss member, 101: in
component-mode method, 477:
component-mode method, 474
Program COMOST, 477
for space frame member, 259
Program DYNAST, 273
for space truss member, 255
Spectra, response, 51
for structure, 98
Spectral decomposition, 223
for substructure in multistory
Spectral matrix, 141, 315, 470, 505
building, 460
S::- ctral radius, 223
for torsional element, 87
Spectral shifting, 503
for x-beam, 428
Square-root method , Cholcsky, I 06
for y-beam, 428
Stability:
fill' z l'llhlltlll , ,1W
('Olldi1101111l , ' OH
pli11l'1p11l , 110
0

Index

590

Stiffness matrix (cont.)

reduced, 282
reduced and assembled, 448
tridiagonal, 446
Stiffnesses, 11
Stiffness method, direct, 98
Strain-displacement relationships, 79,
310
Strain energy, virtual, 80
Strain energy density, virtual, 316
Strain-stress relationships , 77
Strain transformation matrix, 316
Stress-strain relationships, 77, 80,
311
Stress transformation matrix, 316
Stresses and strains, 75, 310
generalized, 92
Structural damping, 26
Structural data:
for multistory plane frames (table),
464
for plane frames (table), 268
for plane trusses (table), 120
for Program DYAXSH (table),
406
for Program DYNAPB (table),
380
for Program DYNAPS (table), 341
for Program DYNASH (table),
39 1
for Program DYNASO (table),
352
for space frames (table), 276
Slructural equations of motion, 98
Slructural mass matrix, 98
Sii uctural stiffness matrix , 98
S11hpt1rametric elements, 333
Suhprogrnm CMAS, 118
S11hpmgra111 DHCOMP (flow chart),
C\59
Suhprngrnm DYLO, 177, 225
S11hpm~rn111 DYi.OPT, (.flow chart),

., 1-,

S111tp10~111111 lil<ll1N2, 11 H

Subprogram FACTOR (flow chart) ,

561
Subprogram INVERU (flow chart),
560
Subprogram NORMOD, 227
Subprogram NUMINT, 227, (flow
chart), 547
Subprogram RESl , 118
Subprogram RESlPT (flow chart),
532
Subprogram RES2, 179, 227
Subprogram RES2PT (flow chart),
553
Subprogram SDAT, 118
Subprogram SDATPT (flow chart),
519
Subprogram SOLVER (flow chart),
562
Subprogram STASYM, 118, (flow
chart), 525
Subprogram STIF, 118
Subprogram STIFPT (flow chart),
523
Subprogram TIHIST, 177, (flow
chart), 545
Subprogram TRABAC, 179, (flow
chart), 546
Subprogram TRANOR, 177, (flow
chart), 543
Subprogram TRAVEC, 118, (flow
chart), 529
Subprogram VIB, 177, 225
Substitution, backward, 461
Substructure equations of motion,
446
in component-mode method, 4C,H
Substructures, 8, 444- 82
Substructures in parallel, 455
Substrnctures in series, 445
modified tridiagonal method, I~ I
tridingonal method, 445

Supc1 purnmllric ckml11ts, \ \ \

S11pp1111 111111,ons, I
i111l1p11ul1111 , I <,J

Index

normal-mode response, 157

Support reactions, 99
Support restraints, independent motions of, 162
Surlace pressure, equivalent nodal
loads for, 366, 408
Symmetric, standard form (of eigenvalue problem), 106, 505
Symmetric modes, 112
Systems of units:
absolute and gravitational, 495
consistent (table), 496

Three-dimensional continua,
310-69
Tier buildings, 425, 461
bracing, setbacks, and shear cores,
468
Program DYMSTB, 463
Time step, critical, 198, 224
Torsional element, 84
Torsional rigidity, 93
Torsional strains, 85
Transformation matrices for framed
structures (table), 415
Transformation methods for eigenvalue problems, 505
Transformations, geometric, 431
Transient response:
with damping, 34
without damping, 25
Translational inertias, 91, 378, 390,
401
Translation of axes, 157, 414
Trapezoidal rule, 203
Triangular impulse, 45
Tridiagonal matrix, 509
Tridiagonal method, 445
modified, 451
Trigonometric series, 35

591

Truncation, modal, 139, 143

Trusses:
component-mode method, 471
plane, 243, 471
space, 243, 253, 477
Twist, 85
Two-dimensional continua, 310-69
Types of force systems, 241
Types of framed structures, 241

Unconditional stability, 208

Undamped forced vibrations, 19
Undamped free vibrations, 12
Underconstrained frames, 290
Unit-load method, 146
Units, systems of (SI and US), 495
Unit vectors, 96

Velocities:
incremental, 38, 212
initial, 14, 147
Vibrational analysis, 105
Program VIB , 118
Vibrational motion, 152
Vibrations:
forced, damped, 30
forced, undamped, 19
free, damped, 27
free, undamped, 12
Virtual strain energy, 80
density, 316
Virtual work, 80
Viscous damping, 26, 164

Wilson-8 method, 220

Work, virtual, 80
Working points, 416