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#1•11 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.
I 11,11N•11 1N, l'hlll ll .. (d111c) IL Title.
Ill ~" '"
IAM·I W•IJ 11/H/ 624. 1'7 1 86-12257
l~IIN II I I H1 110H (1

I111l1111 l11l/prnducllon supervision

111111 l111t11l111 deNlgn: Kuthryn Pavelec
Contents
<'11vt11 d1•Nl1111: l~dRnl Enterprises
M111111t 1w1111 l11g huycr: Rhett Conklin

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

W .I 1/111/, 1•11/ror

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

A I >lvlNl1111 ol Sl111on & Schuster IX
I 1111lrw1111d <'llll'H, New Jersey 07632
1 Introduction to Structural Dynamics 1
1// 1/11/111 , , 1,1•1•11w/, No part of this book may be 1.1 Structural Dynamics Concepts
1,111111/11, ,.,1, lt1 1111v fmw or by a11y means, 1. 2 Dynamic Influences 4 ·
11'///,111111•1•1111/1,1•/1111 /11 wrlring from the publisher.
I . 3 Discretization by Finite Elements 6
1.4 Computer Programs 9
References 9
l'il1111•tl h1 1111, l l11il1•d Stutes of America

Ill lj II I (I 5 4 3 2 l
2 Systems with One Degree of Freedom 10
t ~llN n 1, '3- 853508-6 025
2.1 Introduction 10
2.2 Free, Undamped Vibrations 12
I 1ill'N Ill I II Al I. IN I I'll NATIONAL (UK) LIMITED, London 2.3 Harmonic Forcing Functions 19
11/tl 'N II( I " "' ,. ()11 A USTRALIA J:>rv. LIM!TBD , Sydney 2.4 Effects of Damping 25
11111 NIii I II AI I ( 'ANAl)A INC'., Toro1110 2.5 Periodic Forcing Functions 35
11111 N 111 I I IAI I I IISl1ANOAMIHUCANA, S.A., M exico 2 ·6
1'111 N111 I' I IAI I. 011 INlllA P1UVATI! LIMl'l'l'.D , N1•w {)('/!,/
Arbitrary Forcing Functions 38
1'111 N111 I I IAI I <II' JAPA N, IN(',, 'J'ok yn
2·7 Step-by-Step Response Calculations 45
1'11 1 N Iii I II AI I (Il l S0\ 11111.AS'I' AS IA 1•111. I.ii>,, Sfl1111111m r 2.8 Response Spectra 51
I 11111111 ,\ l'IIINIIC 'I' II AII llO IIIIASII , I ll >A, U/11t/1• .lr111r/10 Rl'fl'n.lllt'CN 59
w111 ,111A1, 111,,1,.,~ 1.1~11,11 ,, w,,1111111 11111, N,•,,• 1tl/ll111,,, Prohl1~n,11 59

Ill
 Contents Contents V
Iv
73 6 Framed Structures 241
Finite Elements and Vibrational Analysis
6. 1 Introduction 241
3. 1 Introduction 73 6.2 Plane Frames 244
3.2 Stresses and Strains 75
6.3 Grids 249
3.3 Equations of Motion for Finite Elements 78
6.4 Space Trusses 253
3 .4 One-Dimensional Elements 82
95 6.5 Space Frames 259
3.5 Transformation and Assemblage of Elements 6.6 Programs for Framed Structures 264
3.6 Vibrational Analysis 105
6.7 Guyan Reduction 282
3.7 Symmetric and Antisymmetric Modes 112
6.8 Constraints Against Axial Strains 290
3.8 Program VIB for Vibrational Analysis 118
6.9 Programs DYPFAC and DYSFAC 299
References 124
References 303
Problems 125
Problems 303

4 Normal-Mode Method 7 Two- and Three-Dimensional

of Dynamic Analysis 138
Continua 310
4.1 Introduction 138 7 .1 Introduction 310
4 .2 Principal and Normal Coordinates 139
7. 2 Stresses and Strains in Continua 310
4.3 Normal-Mode Response to Initial Conditions 147
7. 3 N atura1 Coordinates 318
4.4 Normal-Mode Response to Applied Actions 152
7.4 Numerical Integration 326
4 .5 Normal-Mode Response to Support Motions 157
7. 5 Isoparametric Quadrilaterals for Plane Stress and Plane
4 .6 Damping in MDOF Systems 164 Strain 333
4 .7 Damped Response to Periodic Forcing Functions 168
7.6 Program DYNAPS for Plane Stress and Plane Strain 340
4. 8 Damped Response to Arbitrary Forcing Functions 172
7. 7 Isoparametric Hexahedra for General Solids 345
4 ,l) Step-by-Step Response Calculations 175
7 .8 Program DYNASO for General Solids 351
4. I() Program NOMO for Normal-Mode Response 177
7. 9 Isoparametric Elements for Axisymmetric Solids 357
References 192 7. 10 Program DYAXSO for Axisymmetric Solids 365
Problems 192 References 369

Direct Numerical Integration Methods 195

8 Plates and Shells · 370
'i. I Introduction 195
197 8.1 Introduction 370
5.2 Extrapolation with Explicit Formulas 8.2 Element for Plates in Bending 371
5.J Iteration with Implicit Formulas 203
8.3 Program DYNAPB for Plates in Bending 379
5.4 Direct Linear Extrapolation 211
2 17 8.4 Element for General Shells 382
5.5 Ncwmark's Generalized Acceleration Method
8.5 Program DYNASH for General Shells 390
5. Cl Numericul Stahility und Accurucy 223
225 8.6 Element for Axisymmetric Shells 394
5.7 J>rogrulll DYNA for Dynamk Rl·sponSl'
8.7 Program DY AXSH for Axisymmetric Shells 406
Rl'lb l'lll'l'~ 7'<, Rcf'L-n:nccs 410
l '111llh•111~ l I1
 vii
Contents Contents
111
Appendix C Flowchart for Program
') Rigid Bodies within Flexible Structures 411
DYNAPT 519
9. 1 Introduction 411
9. 2 Rigid Bodies in Framed Structures 413
9. 3 Program DYRBPF for Rigid Bodies in Plane Frames 424 Answers to Problems 564
9.4 Rigid Laminae in Multistory Buildings 425
9.5 Rigid Bodies in Finite-Element Networks 434
9.6 Program DYRBPB for Rigid Bodies in Plate-Bending Index 579
Continua 438
References 443

to 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
10.6 Component-Mode Method for Trusses 471
10.7 Programs COMOPT and COMOST 477
References 482

Notation 483

<~< 11wral References 491

Appendix A Systems of Units and Material

Properties 495
A. l Systems of Units 495
A. 2 Material Properties 497

Appendix 8 Eigenvalues and

Eigenvectors 498
n.1 Inverse llcmtion 498
n.2 Tra11sfo1111ntion Methods 505
lh•fo1l'lll'l'H ~ 17
 124
I lnltn I ln111n11t11111111 Vll1111tlw111I A11uly11l11 c:tu1p. 3 1'111hlu11111 125

0 'I . Timoshenk o, S. P., Young, D. II. , nmI W(i:uvc1,. W., Jr ., Vibration Problems in
" I Wiley New York I n4.
--~-- • . .
f•11g111('er111g,
,
4 th c, ·• ' ' E · 2nd ed
IO. Gere, J. M., and Wcaver,.. W., Jr ·• Matrix Algebra for ngmeers, .,
"" II " , Brooks/Cole, Monterey' Calif., 1983. . N Y k 1973
I_ _ _
",--2!,JI_ _ 11 Hohn F E. Elem entary Matrix . Alge bra, 3rd ed . ' Macnullan, ew orD., V I.
' · P.' G., "Symmetry in Structural Meehamcs,
ll.· Glockner, · " ASCE ' J . Struct . 1v., o ·
99, No . ST! , 1973, pp. 71- 89.
(c)

Figure 3.18 (cont.)

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

PROBLEMS
E = 6.9 X 107 kPa p = 2.62 Mg/ m 3
L=5m 1.4-1. . 1 I ment with a linearly distributed load (force per
A = 6 X 10- 3 m 2 Figure P3.4-1 shows an axia e e - b + (b - b1)x/ L. Find the equivalent
unit length) given by the formula bx - .1 • i
where SI units are implied (again, see Appendix A). nodal loads Pb(t) = {pb1, Pb2} due to this mfluence. f"' { f , {, 1= [ i-{ -f. J
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· ~ b, c fo ~ 1,-, xJ
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 •:l J ,Y,u, f't-•:." 0
],'
mode is symmetric with respect to the plane of symmetry, while the second mode is
antisymmetric.
- I-------1 eb/f)=il I ["~"
Po, b,
P b2
l [h1+(kl,1)1 }
L o -
L-)1) ~ b,+[k-l,').,
~'l I
- I ( l [( . ~
REFERENCES
Figure P3.4-1 -T.. ), x t 1, 1 +-( b, - L1}-z::] .;
1.4-2. A parabolically d!stributedd ~oa:. (fo;~\:zer D~~~~::r:~e ~~sui:~e:~:~~~
bx = bz(x/L)2, as illustrate m ig. ·. ·
I. Weaver, W., Jr. , and Johnston, P. R., Finite Elements for Structural Analysis, loads Pb (t) = {pb1, p ,,2} resulting from this body force.
Prentice-Hall, Englewood Cliffs, N.J., 1984.
2. Zienkiewicz, 0. C. , The Finite Element Method, 4th ed., McGraw-Hill, Maiden-
head, 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.
Figure P3.4-2
7. Archer, J. S. , "Consistent Matrix Formulations for Structural Analysis Using Finite-
Element Techniques," AIAA J. , Vol. 3, No. 10, 1965, pp. 1910- 19 18 . I ..I .t . 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
8 . Clough, R. W. , "Analysis of Structural Vibrations and Dynamic Response ," Rec. 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 - '
Adv. M at. Methods Strucr. Anal. Des., ed. R. II. Onll ughe r, Y. Yamada, and J. ·1~ shupc pc1rts
. (I luncllons
i 4 I J) I,ini •ind (1 (2.1 I /,).,·/ /,2 . 0 c, 11v
.· e the 3 x 3 stiffness matrix
Oden, University of Al abama Press, ll1111tsville, Alu ., I C
J7 I , pp . 25 45.
j~ l'w· ;his l'll11m111( 1r't1w' nxinl ii gld lly FA is constant along the length.
 131
Problems
130 Finite Elements and Vibrationa l Analysis Chap. 3
1,5-7. Figure P3.5-7 shows a two-element beam for which the parameters E, I, A, and
y
pare constant along the length. Assemble the stiffness matrix Ss and the consis-
tent-mass matrix Ms (for translational inertias) in rearranged and partitioned
,I CD forms.
0----------0 2

I0.6L
0 ©
y

L 0 4

/
X

/ ,k
1 CD
~
2 0
~
3

.,
X

1· 0.8L ·I l \.
L L

Figure P3.5-4 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 I ~-It Repeat Prob. 3.5-7 for the two-element beam shown in Fig. P3.5-8.
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. y

y
CD 2 0 ~3 X

~
1~
,I
,/ ~, L~

Figure P3.5-8

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

,---0.BL-- --.
i..l

Figure P3.5-5 Figure P3.5-6 1 0 G) 4ii 0 r,-3 -X

3.5-6. Repeat Prob. 3.5- 1 for the plane truss shown in Pig. P3.5-6. In this case let the 1/ .. - -L- -\- -L - - 1
cross-sectional urcus for llH.lmhcrs I ll11011gh 4 tw NJUIII tn /\ , while those fo1
nw111lm1s 'i 111HI <, 11n• l'q1111l to V? A,
134 Finite Elements and Vibrational Analysis Chap.3 Chap. 3 Problems 135

3.6-7. Repeat Prob. 3.6-5 for the two-element continuous beam shown in Fig. P3.6-7. .t7-3. For the symmetric continuous beam shown in Fig. P3.7-3, find the angular
Cross-sectional properties I and A are constant, and the free displacements at frequencies and mode shapes for (a) symmetric and (b) antisymmetric dis-
joints 1 and 2 are both rotational. tortions. Use only half the structure, assuming that each of the four elements has
y the same values of I and A.

, ____(D=,_ _ _.. 2___®=2---~ _3- - x

4
,/ 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 Figure P3.7-3
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
I 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
2 ISym. the (a) symmetric and (b) antisymmetric mode shapes.
1~
CD • 0 3~ X

i-Sym.
I

z
/ 1· e e L= 2C

Figure P3. 7-1

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
y
'l L

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 Sec. A.2 Material Properties 497

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 = 1 N/m 2
(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.

TABLE A.I Conversion of US Units to SI Units

A.2 MATERIAL PROPERTIES
Quantity US Units x Factor = SI Units
To analyze solids and structures composed of various materials, we need to
Length inch (in.) 2.540 X 10-2 meter (m) know certain physical properties. For structural dynamics, the essential material
Force kilopound (kip or k) 4.448 kilonewton (kN) properties are modulus of elasticity E, Poisson's ratio 11, and mass density p.
Moment kip-inch (k-in .) 1.130 X 10- 1 kiloncwto n · meter (kN · m)
Stress kip/inch2 (k/in. 2 or ksi) 6.895 X 103 kilopascal (kPa)
Table A.3 gives these properties in both US and SI units for some commonly
Mass kip-sec2 /inch (k-s 2/in.) 1.751 X 10 2 megagram (Mg) used materials. Note that the shearing modulus G is not listed in the table
because it can be derived from E and 11.

For any numerical problem in structural mechanics, we must use a consis- TABLE A.3 Properties of Materials"
tent 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 Modulus of Elasticity E Mass Density p
Poisson's
Material
units for force, length, and time appear in Table A.2. For instance, in SI(l) w1.· k/in. 2 GPa Ratio v k-s 2/ in. 4 Mg/m3
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 Aluminum 1.0 X 10• 69 0.33 2.45 X 10- 7 2.62
acceleration ii in millimeters per second squared (mm/s2 ), and so on. Brass 1.5 X 104 103 0.34 8. 10 X 10-7 8 .66
Bronze 1.5 X 10• 103 0.34 7.80 X 10- 7 8.34
TABLE A.2 Consistent Systems of Units Cast iron 1.4 X 10• 97 0.25 6.90 X 10- 7 7 .37
Concrete 3.6 X 103 25 0.15 2.25 X 10- 7 2.40
System Force Length Time Magnesium 6.5 X 103 45 0.35 1.71 X 10- 7 1.83
Nickel 3.0 X 10• 207 0.31 8 .25 X 10- 7 8 .82
(I) newton millimeter second Steel 3.0 X 104 207 0.30 7.35 X IQ- 7 7.85
ST (2) kilonewton meter second Titanium 1.7 X 104 117 0.33 4.20 X 10- 7 4.49
(3) meganewton kilometer second Tungsten 5.5 X 104 379 0.20 1.80 X 10 6 19.2

(I) pound foot second "Numbers in this table are taken from J. M. Gere and S. P. Timoshenko, Me chanics of
us (2) kilopound inch second Materials, 2nd ed . , Brooks/Cole, Monterey, Calif., 1984.
(3) megapound yard 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 , thcrl•hy
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• 1m·ll I
 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 astro-
nautical 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 under-
graduate level. Included among the former courses are differential equations,
matrix algebra, and computer programming. Other desirable courses are vi-
bration theory, matrix analysis of framed structures, and finite elements. Al-
though 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 ~e-
11p1111Nl'S to dynamic loads against results for static loads. Next, we describe
11111111u>nly encountered dynamic influences, which are initial conditions, a~-
plil•d 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
11yNll•111s. It also serves as an introduction to the subject for anyone without such
11 h11ckground. Moreover, this book on structural dynamics becomes more self-
l'1111t11i11cd by including material on one-degree systems.
We present derivations for finite elements and vibrational analysis in Chap-
h•1 \, where all of the discretized structures have multiple degrees of freedom.
1~111·1HY consistent stiffness, mass , and load matrices are developed for one-
di11u111sional 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
m•tworks, 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 t11·11sN normal-mode responses to initial conditions, applied actions, and sup-
11111 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~e-
dtlll' 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,,•1•111111• 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
<'011n•pl~ d!'V!'lop1•d i11 p11•vil111 Nl h11ptl•1:, loa pl111ll' la llHHl1H nn: l'Xll•11dl•d 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 elimi-
nate 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 two-
and 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. Pro-
gram 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 struc-
tures 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,~1tlt•1, fol' v111l111111 mntcrinls . In Appendix B we divide
 1
KIi Preface

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
Introduction
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 to Structural Dynamics
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.
1. 1 STRUCTURAL DYNAMICS CONCEPTS
WILLIAM WEAVER, JR.
PAUL R. JOHNSTON 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.
i\(.'knowledgement To examine some of the differences between static and dynamic analysis,
h•t 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
ht•11111 theory is ·
This book was written in collaboration with C. Lawrence Loh, Staff
E11gint!er, Engineering Information Systems, Inc., San Jose, California. Larry (1)
rnmposed some of the computer programs, provided computer examples, and
dll'cked the Answers to Problems.
111 which m is lhc flexural rigidity of the cross section. The response curve
1,ilwll•d 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' lt•11µth 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
l11~1111111111t·11111,ly 111 tl11s l.'ll~l' thl' h l'HIII will not only displace but will also
1111 rlr111lt• 111 l'Vl'IY p111111 111011µ 1h ll•11pth , 'l'ht.•1·dort· , the displacement D(x, t)
 1
Preface

11111rn of' eigenvalue problems into the topics of invcl'Sl' lll'f"lllion (for few
ul1•N) 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,wl•d 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
Introduction
111•m•lil' 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 to Structural Dynamics
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
1·11• 1•11grnsscd in its development. As before, Suzanne M. Dutcher did an
11hl11mltng job of typing the manuscript, and working with her was a great
h•11,111 l.'. Failure Analysis Associates of Palo Alto allowed us free computer
1•1111-11• , 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.

1.1 STRUCTURAL DYNAMICS CONCEPTS

WILLIAM WEAVER, JR.
PAUL R. JOHNSTON
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.
At·knowlcdgement 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 ·
This hook was written in collaboration with C. Lawrence Loh, Staff
I t1Hlnl'l't , Engineering Information Systems, Inc., San Jose, California. Larry PL 3
(v2\1 = (])
111111pw,1•d some of the computer programs, provided computer examples , and 3£/
1h1•1 k1•d tlw Answers to Problems. 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· l1•ngth of the beam is a function of x only .
/\1 lhl' olhl·r l'Xlrl'mc, suppose that the force in Fig. l. l (a) were applied
1111,l1111111111•m1sly 111 1hi1, l'H/'ll' 1h1· lwa111 will not only displace but will also
1111 l'lrtilll' ,,1 rv1•1y p111111 al1111F tis l1•nr11t Thl'tcforc, the displacement v(x, r)
 111111111111 tlon to !itru1,111r1I Uvn•mh 11 <.hnp. 1 !1tr111 t111nl lly111011h1• <.11111 npt1 3
a "" 1.1

lwrn1m•s 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
Y, V 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.
P(t)
A crude discretization of the cantilever beam problem is represented by the
v(x,t) unalytical model in Fig. l.l(b). Appearing at point 2 is a concentrated mass m,

,~ t '2 X
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

1
/ ~ L

(a)
applied load are the elasticity force,

(2a)
y, v
the inertial force,
P(t) P(t)
(2b)

m!2t t
1~--~~~~~~---
ilt • X
and the dissipative force,
PD = cv(t) (2c)

,/ L

(b)
c~
PE PM P0
where c is a damping constant. From D' Alembert's principle we have
P(t) - Pe - PM - PD
which expresses dynamic equilibrium of the mass m. Substituting Eqs. (2) into
=0 (3a)

Eq. (3a) and rearranging yields

(3b)
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
0 physically and mathematically. Therefore, Eq. (3b) is characterized as an ordi-
nary differential equation instead of a partial differential equation.
(c)
Second , we described the response at point 2 due to the applied force for
li'IJ(lll't' 1.1 (u) Beam with distributed mass; (b) lumped-mass approximation; (c) both gradual and instantaneous rise times. In Fig. l. l(c) we can see that the
Nllllll 1111d dynumic rcspon~cs. 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~ l11dl1•11tl' 11•fc1cm·1·1 111 lhl• end of 1hc chapter.
 l1111111h1t lh111 111 hl11111111nl llv11n111l1 11 < l111p I
I'

lht· ~11111l' 11111g1111t11k . <'111w .> 11st ill11h· ~ 111,11111 llw displm•1•d p11111111111 (r 1),1 wllh
11 p,·1111d '/' L'quul lo lhc 11a111ral 11,·,·wrl t!/ 1•i/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
111111 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
1•ll1•l'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:

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
111111• ltl i.. lol'ics, maximum values, or response spectra, depending on the nature
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 1•1111ditions that result in dynamic responses. In the first case, a crane

'
(f )
(g)

(11) P(t)

lb) le)
(11)
Figure 1.2 Dynamic innuences.
l1'l1111r11 1,2 (('(l11 / .)

6
 Introduction to Structural Dynamics Chap. 1 Sec. 1.3 Discretization by Finite Elements
7

1111ddcnly drops its load and rebounds from an initial condition of static-load y
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. 0..
2.,.__ _ _ 2 _ _ _ _3
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 acceler-
ntions. 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. ---x
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, z
/
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 Figure 1.3 Plane frame with subdivided members.
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 masses ~ould be ~ocate~ for the purpose of writing more equations of motion.
d111placcment can be arbitrary. For example, it may vary according to the square- Thus, ~1th these hberahzed definitions for members and joints we are able to use
wnvc pattern labeled 2 in Fig. l.2(g). ~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.
1.3 DISCRETIZATION BY FINITE ELEMENTS 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
To Nt•t up an analytical model for dynamic analysis, we will discretize solids and denve energy-consistent stiffnesses, masses, and equivalent nodal loads for such
NII lll'htrcs using the method of fmite elements. The applications include framed clements and assemble them into an analytical model having a large but finite
11t111l'l11rcs [2J, two- and three-dimensional continua, plates, and shells [3]. The number of degrees of_f~eedom. By so doing, we are able to simulate accurately
1111111· t·lernent method enables us to convert a dynamics problem with an infinite the response of the ongmal structure. The theory of finite elements appears first
1111111lll•1· of degrees of freedom to one with a finite number in order to simplify z
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 re-
11po11Nc11 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
1111•111lwrs 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,•sirt•d , wt• could 1n1hdividt• t•u,·h ancmhcr inlo four purls and view each
flql,llll'III IHI (I llt'W llll'llllll'I . Thon lilt' IIIClll' lt'hm·cl IIIIUlylkul lllOdl•l would have
I) tlll' llllll'r~ und I I 1111111 ~ Sud, 111t'll11t'111t•111 w1111ld p111vid1• IIIOll' po1111~ whl'lt' 1,·1111111• 1.4 I l y p111 ~hl'II tli~l 11'1111•11 hy lilllll' CICllllllllN .
 Introduction to Structural Dy namics
8
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 normal-mode method (Program NOMOPT). Then we further extend the pro-
X
/ in the form of line prints, but the amount of information for structural dynamics
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
~t111t·tmc di vided into five types of substructures that are joined at common
(i11tt·1fun:) nodes. Five groups could analyze the substructures individually while
11 sixth group handles the assembly and solution process for the entire structure.
/ .....
/ I
/ I
/ I
M111111• I .<, !-:v 111111111 l, 111 1111111 il1 vhh d 11 1111 - 1111~1111, 1111,•,
Chap. 1 Sec. 1.4 Computer Programs 9
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 Chap-
ter 4 we extend the vibrations program to include dynamic responses by the
gram in Chapter 5 to calculate dynamic responses by a direct numerical integra-
tion 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 FORTRAN-
oriented 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
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 '1l·ntit·c ll ull , Englewood Cliffs, N . J ., 1984.
 Sec. 1.1 Introduction

2
11

D(t) D(t)

t t
~ em
•
j. L ./
&
). L
2
c_{ 2
~
(a)
(b)

Systems D(t)

with One Degree /1-lom____

of Freedom /
/
/ I
I
.,,
/
/

(c)

2.1 INTRODUCTION
/
1•111 preliminary or approximate analyses, simple structures may often be ideal- (d)
1/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 (e )
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 li'l1turc 2.1 Structures modeled as SDOF systems: (a) cantilever beam; (b) simple
lwum ; (c) plane frame; (d) space frame; (e) fixed plate.
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 wh11·h. is th~ stati~ translation caused by a unit force applied in the positive
tll11plm:cmcnt. dl1n 111:11 of the dis~lace~~nt D ._ In this expression the symbol E represents
Fl,•1/bilitit·.1· for the structures in Fig. 2.1 are obtained as displacements \ 111111p 1:1 modulus of elast1c1ty, I 1s the moment of inertia* of the cross section
(t11111sl111io11s or rotations) due to unit values of the corresponding actions (forces 111 llll' ht•n111 , and t, is its length.
111 1110111l' 11l11) . If thl' canlilcvcr bcum in Fig. 2. l(a) is prismatic, the flexibility 6
S11//i11·.1'.1'('.I' nrc found as static holding actions, corresponding to the dis-
111 lht• 111•1• 1•11<1 Ill plt111 1111• 111s 11how11 , caused by inducing unit values of those displacements. At
I. ' ' I It,· 111111111•111 ut hw11l11 111 1111• l lllHN Hrt'li1111 IH 11 1111N11omcr for the second moment of area
11'/ 1,1 1111 1111~~ ~,·, 111111 Wllh 11·~111·1 I lo 1hr 111•1111111 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 o
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
--·
ku mii

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 u(t),ii(t)
l111ving onl y one degree of freedom. Analytical models having coupled multiple (b)
d1·11H'l'S of' freedom will be discussed in subsequent chapters.
Topics of interest for SDOF systems consist of free and forced harmonic Figure 2.2 (a) Mechanical analogue for SDOF system; (b) partial free -body dia-
grum .
1110111111s with and without damping, response to arbitrary time-varying loads or
,.11ppo, 1 motions, and response spectra for dynamic loads. These subjects are
1111•1tial ~estoring force mu also arises due to the presence of the mass. From the
l'OVl'll' d in the ensuing sections of this chapter and will be used throughout the
po1l111l free-body diagram in Fig. 2.2(b), we see that
hook .
- ku - mu =O (1)
\\ hid, is. an applicatio_n of D' Alembert's principle for dynamic equilibrium.
2.2 FREE, UNDAMPED VIBRATIONS l~111 111111grng Eq . (1) slightly, we have the following equation of motion:
mu+ ku _= O (2)
Figure 2. 2(a) shows a mechanical analogue for any of the SDOF systems
dt•sl·nbcd in the preceding section. The mass m and the spring constant k are I lr h1 Nt'l'l~nd-order differential equation may be simplified further by introducing
1111 IIO(H(I Oll
d1•ll'1111incd from the properties of a given elastic structure , and the symbol u (t)
dl·notes the single displacement coordinate. The figure also indicates the accel- k
1•1at ion ii (t), which is the second derivative of u (t) with respect to time (J) 2 = -
m (3)
(II d 2 u/ dt 2). If the displacement is rotational, then m, k, and u may be
H·plt1l'l'd by /,. k,., and 60 (a smnll angle of rotation). whit h prncluces
We assume that the sys1c111 in Fig. 2.2(a) is initially at rest inside or outside
11 ~rnvi1111 io1111l field I 11. Tht•11 lt•l 11s disturb it in some manner from its position
u + w 2u = 0 (4)
ol .,,111k l'qtdlihr i11111 1>111• lo th,• d1Hpl11n·111cnt II relative to the static position , ',111 Ii 1111 1•q1111tio11 will be satisfied if we take the general solution
1t ll'S1t11111 ~ t1111·t· 1•q1111l 111 fo ill' vrl11p1, 111 tlw s p1111 g, 11s shown in Fig. 2.2(h) . An
11 <'1 cos Ml I C, sin "'t (5)
 Systems with One Degree of Freedom Chap. 2 Sec. 2.2 Free, Undamped Vibrations 15
14

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 u

the constant
(6)
w=~

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

f=..!!!.._=_1 {k (7)
21T 21T "{;;;
and the natural periodJ (sec) becomes

T = = 2: = 21r ~
7 (8)
(a)

The last expression is the time for which the vibration repeats itself.
u
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)

u
I
A cos (wt - a)
Thus, Eq. (5) becomes

u = u0 cos wt + - sm wt
Uo. (9)
w A te T

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
_l_
0 T T
c:un be represented graphically, as shown in Figs. 2.3(a) and (b), by plotting 4
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
(c)
pl,1111•. Let a vector OP of mngnitudc u0 rotatu with constant angular velocity w
11101111<1 thl' flxt•d point O JI at tillll' t O till' wctrn OP coincides with the 11 l•'l1111n• 2 ..\ RC'sponhl' funl'tionN: (n) cosine; (b) sine; (c) combined.
nxi~. tht• 1111gk it 11111kt•s with tht· t11111w nxis nt 1111y 1111111 tinw tis l'qunl to (I)(. The
p11111·l·ti111111I tl11'i v1·rl111 1111thr111a,w1 i'i l''lll1tl l111111 l 1111 ,,,, 1111d u•p1t•st•11ts till' flr~t
 Sec. 2.2 Free, Undamped Vibrations 17
Systems with One Degree of Freedom Chap. 2
10
a
u te = -(J) (13)

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
/ ~ sin wt flexible columns, each having length Land constant bending rigidity El. Neglecting the
/ w
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.
&"/ A COS (wt - a)

p/
y
/
I ~

'o I
_
Uo
w "\ \
i , - - - - - - - m_ _ _ _ _--,1 - u(t)

\
w
\ 1
0

Figure 2.4 Phase-plane .representation.

/ - - -x /
11•1111 in Eq. (9). If we also take another vector OQ of magnitude u0 / w and
1w1 pl•ndicular 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 (a) y
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
(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 11l•hm•d by l\q . ( 11 ). Nott· th11t till' 1t•1mlt1111I vl'l·tor ()It in Fig. 2.4 lags the
1•0111p11111•111 v1•1'llll OP hy tlw pllllMl' 1111)'11• cv Pm thl N11•11s011 th1· tolul vihrntion 11111111 11\ 1,x11111ph·~ (11) 111111Nl111lu1111l ;
tlti 1111,1111111111 I
/ ill)
111 Fi)' J \(1) l11pl'I 1111• 111111prn11·11I 111 tlu· 11111tlllll 111 l'l p , \(11) hy th1· 111111•
 I ln11n1111l1 I 1111 1t1u I 111111111111 19
,y111111111 with 01111 l>nu1nn 111 I 1nnil11111 Cl111p, 'I
1ft
HARMONIC FORCINO FUNCTIONS
lly 1runsl111ing the girder u unit tllHluncc (u I) 111 1h1• 1 dl11•1 liou , we tlnd lhc
Ntlll11w1H l'Onstunt us the following holding force: '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
k = 2 (12EI) = 24£1 (a)
LJ L3 111 q111 m•y ol the funcllon. The first of these functions appears in Fig. 2.6(a) ,
11ppl li•d lo Ih~ undarn.ped ~DOF system of the preceding section. From the partial
' lfa111 tht• 11n~ul11r frequency w from Eq. (6) is Ii, 1 hody d1ngrum JO Fig. 2.6(b), we can write the differential equation of
11111111111 Ior this case, as follows:
w= 'J; f6Ei
ik=IL 'J-;i (b)
mu + ku = P sin Ot (1)
1111!1 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 (c) u+ w2u = Pm sin Ot (2)
21r 1rL 'J -;;;f,
/\IH11 , the notuml period T from Eq. (8) is seen to be
p
Pm= - (3)
T=7= 1rL~ (d)
m

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
k = GI (e)
' L

111 whk h <, is the shearing modulus and

1rr4 (a)
J=-2
(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)
---·
ku mu
- P s i n flt

111111

w
J -=-=-r
21T 21T -21,y/,
2
I-t&G (h) u(t),iJ(t)

(ll)
1111d
I 271 {?t,. /, 1<111111·,, 1.t1 (11) ll111111onlt' fo1d 111-1 f'unl·llon npplicd to SDOF system; (b) partial
(i)
'/ '
f ,.J \/ ,,c; 111 1 liudy dlllHIIIIII,
20 y1111111111 with 01111 1>11u11u1 111111111110111 Chnp. 'J 11111.. 2.3 Harmonic l·orol1111 I 111wll111111
21

Thu total solution for Llq . (l) rnnsists of lhe su111 111 llll' ~1·m•rul solulion /3
IOI' the homogeneous equation, plus the particular solul1011 lhnt :mtislies the I
whole equation . The form of such a particular solution is I
3 - I
u = C3 sin flt (4)
I
Suhstituling this expression and its second derivative with respect to time into I
l\ll , (2), WC have I
2 I
I
'l'hncforc, I
I
Pm _ _ _ _ _ _ J_ _ _
C3 = 2 n2
(5)
W - ,H
I
'l'lwn the total solution becomes I
I
u = C 1 cos wt + C2 sin wt + C3 sin flt (6) o'---~~---1.~~~--L~~~...L_~~____J~~~_J_~-
2 fUw
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 Figure 2.7 Magnification factor.
/on'1•d part (with angular frequency fl).
Let us rewrite the forced part of the solution in Eq. (6) as factor a~d th~ amplitude of forced vibration rapidly increase and theoretically
p,,, . n
(7)
become mfimte for the case of n
= w. This state is known as the condition of
u = w2 - fl2 sm :ut 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
'l'hc11, by using Eq. (3) and w 2 = k/m, we may cast this expression into the form 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
u = [l-(~/w) 2 ]fsinflt (8) 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
'l'hc second factor on the right-hand side of this equation is the displacement of this ratio becomes very large. Thus, when a harmonically varying force of high
llll' mass in Fig. 2.6(a) if the forcing function were applied statically. The frequency acts on a SDOF system, the response is very small; and in such cases
11hsol11tt.: value of the term in brackets is called the magnification factor /3. Thus, the mass may be considered as remaining stationary.
Wl' h11ve 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
(9) v1bratmg mass JS m the same positive sense as the disturbing for.ce. On the other
f3 = \ 1 - (~/w) 2 1
hand, when _fl > w _the expression is negative, meaning that the displacement
whkh reprt.:scnts the ratio of the dynamic response to the static response. The of the mass is opposite to that of the force . In the first case the vibration is said
p111 I of lhe motion in which this factor appears is called steady-state forced to be in phase with the excitation, while in the latter case the response is said
to be out of phase.
1'ihl'lltio11.
Figure 2.7 shows values of the magnification factor f3 plotted against the . It is also possible to cause forced vibrations with harmonic ground motions
t1c.·qucncy ratio fl/ w, which is the ratio of the impressed angular frequency fl instead of applied forcing functions. Two types of ground motions of interest in
ol the disturbing force to the natural angular frequency w of the SDOF system. ~-tructural .dy~amics. are. ground displacement ug (t) and ground acceleration
I 11)1' s,nall values of the frequency ratio , the magnilicut ion factor is approximately ~g(t), as mdicated m Fig. 2.6(a). A harmonic version of the first of these
1111i1y ; und tht.: response is about the same ns t'or the slatic application of the influences may be expressed as
h111111011ic.: load . When this mt io 11pproadws 11111ty , however, the magnification
ux • d sin flt (10)
 y11"1111 with 01111 DM\11 "" 111 I 1m11l11111 C:hnp. ') 111 ~. I I 11111110111! I 011 111111 I 11111 lilll in 23

wht'll' till' symbol d represents thl' 11111plltuuc or grou11d di.~plun•1m·11t. For this 11* (17)
,·1111t• Wl' write the fo llowing differential equation of motio11.
lt1 th1'Nt' 1t·lutionships the symbol ulf< represents the relative displacement of the
mu + k (u - Ug) =0 ( I I) 11111 N with rnspect to the ground, and u* is the relative acceleration. Substituting
111 whkh ,, u is the displacement of the mass m relative to the ground.
,, //• 111HI Ii from Eqs. ( 17) into Eq. (11) and rearranging, we obtain
S11h111it11ti11g Eq.\JO) into Eq. (11) and rearranging the latter gives mu* + ku* = -mug (18)
mu + ku = kd sin flt (12) II \VI' 11 IH0 use Eq . (16) in Eq. (18), the result is
'l'lw11 division of both sides by m yields mu* + ku* = -ma sin Ot (19)
u+ w 2
u = pg sin Ot (13) 111111 di vision of Eq . (19) by m produces
wht'll' u* + w 1 u* = Pi sin Ot (20)
kd \\ 111 11•
Pc= m (14)
Pi = -a (21)
Th INv11 h1c of p11 is an equivalent force per unit of mass, due to ground displace- l ltl11 v11 lm! of pf is an equivalent force per unit of mass, due to ground acceler-
111t•111 . ,11 h 111
( 'olllparing Eq. (12) to Eq. (1), we see that kd has replaced P. Therefore, l(q11111ion (19) is of the same mathematical fonn as Eq. (1), except that u*
1t IH l'Vident that occurrence of the harmonic ground displacement d sin flt is 1111~ 111pluccd u and - ma has replaced P. Thus, the case of harmonic ground
,•q11i v11 1L•1Jl to applying a force kd sin flt directly to the mass m. All previous 11 11 It rnl ion a sin flt is equivalent to applying a force -ma sin Ot directly to the
r111wl11sions regarding the harmonic force may now be applied to the case of 1111111N 111 . Using - ma instead of Pin Eq. (8), we have
111111111111 k grnund displacement. That is , if we replace Pin Eq. (8) with kd and
t 'lllll'l' I k, the result is 1 ]ma."'
u* = - [ 1 - (O/w)2 ksmut (22)
(15)
,, hh Ii ii, lhc displacement of the mass relative to the ground. This relative motion
I IINr t II I for flnding the force in the structure,* which is represented by the spring
'l'ht• tl•1111 d si n Ht in this expression is the motion of the mass when the ground 111 th1• SDOF system. Note that the minus sign before the brackets in Eq. (22)
dl1,pl111..·c11wnt occurs very slowly (or statically). The premultiplier in brackets is 11111111 tlw n.ilati ve response of m out of phase with the ground acceleration.
!Ill' 1,111111· f'uctor as that discussed before [see Eq. (9)]. Thus, we need only
I•,111u11lt• 2.3
1•111111id1•1· 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 '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
Nyt1t1•111.
111 II O ~w/6.
111 1111111y cases it is more convenient to deal with ground accelerations than
l(1•1·11lllng that the stiffness constant for this frame is k = 24£1/L 3 , we substitute
I'' 111u1d dl splucemcnts because a measuring device called an accelerometer [2] 1111 v111i11•s o l' !land k into Eq. (8) to obtain .
hns ht•1111 IIH\:d 10 obtain information about the ground motion. For example,
1•111 tliq1111k1• µround motions are usually measured and reported in terms of three 1 PL3
u = 1 _ (S/ 6) 2 24£! sin Ot
111 lh11prn11tl coinponents of ground acceleration, which are north-south, east-
w1•111, 1111d Vl'l'lical. Therefore, we shall reexamine the ground motion problem by 3PL3
Np1•1 ilyi11g II hurmonic ground acceleration , as follows : = - - sin Ot (a)
22£/
a~- (I sin n, ( 16) N111, th11t lh1• vn lul.! of thl.! mngnifi cution factor is f3 = I36/ LL I, in which the positive sign
wh1111• 1lw symbol <I denotes the un,plitu(k ot gm111ul t1l'l't•l1·rn1ion . In order to use lhl' vi hrntlon is in ph11sl.l wi th the forcing function.
1111pll1 N 111111
lhlNt•xp11•ssio11 in Hq . ( 11 ), Wt' 11111 st d111111•l' 1111111•q1111tlo11 to lhc rt'lmi v,• ,·oor- 1 w, wlll 111111111 lo thlN 1111111<,1 o l 1•111,,11lutl1111 tl11i 111l111ivc displ uccrncnt of the mass w ith

1/11111/1• 11 ~1111 I 111 1hr p1111111d 111 llx11111ph• 1 1.1

:u V•h1t111 with 0110 IJ11111nn 111 I 11111!111111 Chup. 'J '" ~ ,1 I 11111111 111 I >1011pil111 26

y I) pt rnlitl'l'S
I.
I
11(1) p,,,!l
I
w2 - 112
.-------m--------, - P sin !lt I hll N,
C
2
= uo _ p,,,11/w (25)
w w2 - 112
111111 1111• lolnl solution in Eq. (23) becomes

11 + -iio
uo cos wt . wt +
sm Pm l'\ (sm
. I'\ 11 sm
~ £t - -
. wt) (26)
w w 2 - 3.£2 w
11 111,, 11111ial conditions are u0 = u0 = 0, Eq. (26) reduces to

u = w2 P,,,
_ 112 (·('\
sm Ht - 1 1 .wt)
w sm (27)

Figure 2.8 Frame with harmonic forcing functions. I q1111l 11111 (27) represents the response of the SDOF system to the forcing
I 11111 11011 I' sin 11t, and we see that it consists of two parts . The first part is
IC1mm1>lt1 2.4 1111 !ilt•ndy state response discussed previously, and the second part is called
11 lh1• Nt1111c frumc as that in Example 2.3 has a harmonic ground acceleration ii8 = a sin 1111 11,111si1•11t re.sponse. This name refers to the fact that the term dies out in the
0, (Nl'l' Fig. 2. 8), find the steady-state forced vibration for O = 3w. 1111 ~t lll'l' or damping, as do all free vibrations. The sum of the two terms is not
In this case the relative response, as given by Eq. (22) is: 1 h111111011 ic motion even though it is composed of two harmonic functions,
1111 1111N1· lhll components have different frequencies, 11 and w.
l mal3 . When the forcing function is P cos 11t instead of P sin 11t, the term cos 11t
u* = - I - (3)2 24£/ sm Ot
11 11l1111•s sin !lt in Eq. (23). In this case the initial conditions result in the

maL3 • l 11ll11w 111g constants of integration:

= --sm Ot (b)
192£/ p,,,
C I = Uo - 2 l'\2 (28)
lk1,· w1• Nl'l" tlrnt the value of the magnification factor is /3 = l-1/8 1, so that the W - ~L

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
,11liict1t111ion of' these values into the total solution gives
vlh1ntio11 ts ul'lunlly in phase with the ground acceleration.
II
. wt + - Pm
uo cos wt + -Uo sm AZ
(cos I'\
ut - cos wt) (29)
2
W W - H
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 II 1111 111illn l conditions are taken to be u0 = u0 = 0, this equation becomes
1•11111111m• the free vibrations associated with the first two terms, which will be
dilh·ll' ttl frolll those in Sec. 2.2. Substituting Eq. (5) into Eq. (6), we have u = w 2 p,,,
-
(
AZ cos ut -
~£
.I'\
cos wt) (30)
. + p,,, . I'\ (23)
II C I cos WI + C 2 s in wt w2 112 Sll1 Hf 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 >illl•n•111inting this equation with respect to time yields

ti C 1w sin CJJI I C1 cJJ cos"'' I

,,llln
, n l cos
{\
al (24)
n F CTS OF DAMPING
(1/ ~£
111 pH v11111s tllst·ussions or free und fon:etl vibrations for SDOF systems, we did
flu1111 il1l' i11iti11l rn11dilio1111 11 11 111 litm· I 0 , l \q ( .> \) giws C, 110, which 11111 1 c111N1dt•1 lht· l'fft·t·ts ol disslpuliVl' i11lh1t·nct!s, such as friction or air re-
iN illl' hllllll' llh lil•l1>1l' I loWl'Vl' I , H1tl11,li1t1ti1111111 fi ti11 Ill tiilll' I 0 into Eq . l 1111111• <'011st•q111111tl y, wt• l1rn11d 111111 llw n111pllt11dt· of' frllt! vibration remains
 Gy • IMl!lft with 01111 l>MIJtnM 111 I 111Md11111 Chup, 'J HII , :,!.11 I 11111:11 ol l>nt11pl1111 27

1·1111Nl11111 wllh lime; but experiem:c shows that the 11111plit11d1• d1111i11l shes with w, 111• the hon1oµeneous equ ution for free vibration as
ll1w• 1111d that the vibrations arc graduall y damped out. Si111ll111'1 y, for undamped
l111 c1•d vihrutions the theory indicates that the amplitude cun grow without limit ma + cit + ku =0 (1)
111 11•so1111 11cc. 1lowcver, we know that because of damping there is always some 1111• rnc l"ficient c in this equation denotes the damping constant, which has
1111111• 11111plitu<.lc of steady-state response, even at resonance . dl11ll't11,ions of force per unit velocity . Note that in the partial free-body diagram
'l'o bring our analytical discussion of vibrations into better agreement with IN1•t• Fiµ. 2.9(b)] the damping force acts in the direction opposite to the velocity.
11•1d 11y, do111f)i11g f orces must be included. These forces may arise from several I >lviding Eq. (I ) by m produces
11111s1•N, such as fri ction between dry or lubricated sliding surfaces , air or fluid
11·NtNl1111c1·, clectric impedance , internal friction due to imperfect elasticity of ii + 2nu + w 2u = 0 (2)
1111111,1luls, und so on. Among all of these sources of energy dissipation, the case 111 which we define the new constant
wh1•H lhL' dumping force is proportional to velocity, called viscous damping, is
1
C
llw i;it11pkst to deal with mathematically. For this reason resisting forces of a 2n = - (3)
m
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 dissi- 1111 rnnvenience in the ensuing solution.
p111 lo11 of" energy per cycle of vibration to that for viscous damping . For example, To solve Eq. (2), we assume a harmonic function in the following form :
.1·tr 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 u = Ce 51 (4)
tl11111pi11p i11 lhl) form of a dashpot damper , as shown by Fig. 2 .9(a) . Assuming 111 which C and s are constants that satisfy Eq. (2). Substituting Eq. (4) into Eq.
lh111 11 viscous fluid in the dashpot resists motion in proportion to velocity, we I 1) , we find that
s2 + 2ns + w
2
=0
;

1111111 which
s = -n ± \!n 2 - w2 (5)
II lhe value of n is less than that of w, then the quantity
wa= w2 - n2
IM positi ve, and we obtain for s two complex roots, as follows:
~u,U,iJ

(a) wh1•re 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\ddinµ these formulas , we obtain the general solution of Eq. (2) for damped free
1•/11mtio11s as
~ 11, 11, 1/
(6)
(Iii
l'hl· 1111·101 1• "' i11 lhis t'<111111io11 d1•1·u•us1•H wi lh li1IIL', so the vibrations originally
~w111•1111t•d wtll lw HI 11d1111ll y d11111111•d 11111
 Systems with One Degree of Freedom Chap. 2
Sec. 2.4 Effects of Damping 29
28
u
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 (7)

which is called the angular frequency of damped free vibrations. The corre-
sponding 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) Figure 2.10 Damped_free vibration.

In the foregoing discussion of Eq. (2), we assumed that n < w. For the
The first harmonic term in this equation depends only on Uo, but the second
opposite case of_n > w, both of the roots in Eq. (5) become real and negative.
dl.lpends on both u0 and uo.
Substituting them into Eq. (4), we obtain two solutions of Eq. (2), and the
Equation (10) can also be written in the equivalent form
general solution becomes
u = Ae-"1
cos (wdt - ad) (11)
(15)
111 this expression the maximum value is In this case the solution is not periodic, so it does not represent a vibratory
2 motion. The viscous resistance is so large that when the mass is displaced from
(uo + nuo)
A= Vcf + c~ = 2
Uo + 2 (12) its equilibrium position, it merely creeps back to that position.Jn such a case the
WJ
system is said to be overdamped, anq the motion is called aperiodic. The
und 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
ar1 = tan - I -C2 = tan _,uo + nuo
(13)
C 1 Wr1Uo
from which
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. (16)
Figure 2. 10 shows this damped free vibration with u lag time of
n,, Thus, Eq. ( 15) becomes
( 14)
,,,,, 1iu ,\', llo I
II t•\I (I 7)
,1' 1
111di1•111t•d Ill 1111' 11ppt' I li•II
"''
10 Systems with One Degree of Freedom Chap. 2 Sec. 2.4 Effects of Damping 31

'l'ht· general appearance of a graph of Eq. (17) depends on the parameters n, uo , from which the solution is
1111d 1.i0 .
- (w2 - fi2)Pm
Between the underdamped and overdamped cases lies the special case of (25)
Q - (w2 - fi2)2 + 4n2fi2
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 By substituting these constants into Eq. (24), we can obtain the particular
Ccr = 2nm = 2wm = 2'Vkm (18) solution of Eq. (23).
The total solution of Eq. (23) consists of the sum of the particular solution
ln which the symbol cc, denotes critical damping . For the critically damped and the general solution derived previously as Eq. (6). Thus , considering only
SDOF system, Eqs. (5) and (7) show that s1 = s2 = -wand wd = 0. Neither subcritical damping, we have
l\q . (6) nor Eq. (17) constitutes the solution, which in this particular case of
11 pealed roots takes the form
u = e- "1(C 1 cos wdt + C 2 sin wdt) + Q cos fit + R sin fit (26)

u = e-w1(C1 + C2t) (19) 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.
Suhst ituling the initial conditions into Eq. ( 19) and its first derivative, we find The harmonic force maintains these latter vibrations indefinitely, so they are of
lhnt great practical importance.
C2 = uo + nuo (20) The steady-state response in Eq. (24) may be written in the equivalent
phase-angle form as
'l'hl·ll the general solution becomes
u = A cos (fit - 0) (27)
u = e-w [u 0 + 1
(uo + nuo)t] (21)
where
which ugain depends on n, u0 , and Uo.
Turning now to damped forced vibrations , we take the applied forcing A = y Q 2 + R2 = Pm
h111ttion in Fig. 2.9(a) to be P cos fit. Then the equation of motion for the mass \!(w 2 - iV) 2 + (2nfi) 2
ht•t'OlllCS Pmf w 2 (28)
mii + cu + ku = P cos fit (22) Y[l - (fi/w)2]2 + (2nfi/w 2)2
I >ivlding this equation by m produces und
ii + 2nu + 2
w u = Pm cos fit (23)
() =
R
tan- 1 -
2nfi
= tan- 1 ---,.----,,
Q w2 - fi2
llll' which all of the notation has been defined previously. A particular solution
111 Hq . (23) can be taken in the form _ _
1
2nfi/w 2
(29)
(24) - tan 1 - (fi/ w )2
u = Q cos fit + R sin fit
wlw11· () und R are constants. To determine these constants, we substitute Eq . Thus, we see that steady-state forced vibration with viscous damping is a simple
()•I) 1111d its derivatives into Eq. (23) and obtain hunnonic motion having constant amplitude A , phase angle (), and period
lj 2n/!l.
( U'() I 211fiR + w 2 Q - Pm) cos fit Using the values of w2 = k/m and Pm = P /m and introducing the symbol
+ (-fi 2R - 2nfiQ + w 2R) sin fit =0 'Y for the damping ratio ,
'l'hiH t·quation can be solved for all values oft only if the expressions in the n C
y =-=- (30)
p11rn1Hhcscs vttnish. Thus , for calculating Q and R we have two linear algebraic w Ccr

,•q1111llo11s, as follows : w1 11111y s11hst itull· l~q. (28) into Eq. (27) to obtain
1
H () '2.11 HU I ,,,i (J fi 111
oi Ii l 11 O(! I 1
1,1 Ii II
" t1 I',. l'Os m, o) (3 1)
 Sec. 2.4 Effects of Damping 33
Systems with One Degree of Freedom Chap. 2
J2
4.0 ,--,----,--,--..--.-rT--ir--.-----,---,
In which the magnification factor f3 for damped forced vibrations is y= O
1 (32) f3
/3 = Y[l - (0/ w)2]2 + (2y0/ w)2
Also , Eq. (29) for the phase angle becomes
_ -I 2y0/w (33)
e- tan 1 - (O/w)2

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
1.0
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. 1.0 1.5 2.0
Sl•tling the derivative of {3 with respect to 0/ w equal to zero, we find that the n
(a) w
111uximum occurs when
o
- = V1 - 2y 2 (34) y=O
w /
Fm small damping ratios ( y :5 0.20) the maximum value of {3 occurs very near
0
to tl sonunce, and taking the value of {3 at resonance as the maximum provides
1

1,111 lldcnl accuracy for engineering analysis. Then, from Eq. (31) we have

P 1 P w P P (35)
Amax = /3,esk = 2y k = 2n w2 m = cw
'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.
2 3
While the phase angle is of less consequence, we also show in Fig. 2.1 l(b) Fll(urc 2.11 (a) Magni fication factors;
D.
th) phnsc unglcs. (b) w
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 of the amount of damping, the phase angle is always equal to 1r/2 at resonance.
c11Sl' of zero dHmping, the forced vibrations are exactly in phase (0 = 0) with the '1'11111 is, ut resonance the response lags the force by a quarter cycle.
d1st11rhi11g force I'm nil vn lues of H/ <u • I , On the other hand , they arc a half ll,x111111,h• 2.5
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 l11dl•t(1111ii11111l' 11! t l'Nlllllllll'l', whl'tl' U ,,,. Wlwn d111nping is I It till' 11•vt11111t11l11r pl11nl' l'ru1m• 111 Fig. 2. 8 hnvc u hnrmonic force P cos D.t in place of
1111 loll'<' 11N1•d 111 l~x11111pl1• l. \, AINo, ln1111tim• lhal lhi: ground has an acceleration
11011 tl' 111 , Wl' t1t1ll' 111 ,1111111111111 h11111w 111 /J11Mtlw 11111110 / ,,1 v11il1•H llut , tl'p11rdkss
 Systems with One Degree of Freedom Chap. 2 Sec. 2.5 Periodic Forcing Functions 35
34

1/M a cos flt instead of the sine function in Example 2.4. Calculate separately the where
11h·111ly -state forced responses of the mass m due to these innuenccs, assuming that
(h)
y 0.02 and fl = 0.9w.
For this example the magnification factor in Eq. (32) becomes and
1
5 17 (a)
_
1
Qn + RO
ad= tan Q (i)
f3 = \/[1 - (0.9)2] + [(2)(0.02)(0.9)]2 = ·
2
Wd

llsing this value and k = 24E// l 3 in Eq. (31), we calculate the response to the force as In the last expression the symbol ad represents the phase angle for the damped system.
3
PL
u = (5.17) 24£/ cos (flt - 8)
2.5 PERIODIC FORCING FUNCTIONS
Pl3 (b)
= 0.215£1 cos (flt - 8)
In previous discussions of forced vibrations , we assumed simple harmonic
<'lmNldcring the ground acceleration, we need only replace Pin Eq. (31) with - ma to functions proportional to sin fit or cos fit. It is also possible to encounter
ol,111111 the relative response as 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
mal learn how to calculate the response of a SDOF system to such a function.
u* = -0.215 El cos (flt - 8) (c)
We can represent a periodic dynamic load of any kind by decomposing it
1,:x111111>IC 2.6 into a trigonometric (or Fourier) series, as follows:
1 IH· 111111sil!11t response of a SDOF system with subcritical damping may be found by P (t) = ao + a 1 cos flt + a 2 cos 2flt + .
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.
+ bi sin flt + b2 sin 2flt + . ..
Using u = u 0 and u = iio (at time t = 0) in Eq. (26) and its first derivative with
11•Np1·r t to time, we evaluate the constants of integration as = ao + L (ai cos iflt + b; sin iflt) (1)
_ uo+ n(uo - Q) - R il
(d) The period of the applied force is Tr = 27r/fl, and the symbols a0 , a; , and b;
C1 = Uo - Q C2 -
Wd
represent constants to be determined. To calculate these constants , we may use
S11hNl1t11t111g these constants into Eq. (26) gives lhc procedure described next.

u= e-"'(uocos wdt + :dnuosin wdt) 10

' p

+ Q[cos flt - e- '"(cos wdt + ;dsin wd1 J (e)

+ R(sin flt - e_,,, ~ sin wd9

II th1· 1111t1ul conditions arc taken to be u0 = 1,0 = 0, the transient portion u1r of the
11·11111111111g tl'sponsc is
Qn I RH . ) (f)
11,, (' "' ( Q cos t,i11 t ~
,,,., sin w,,t

I ll jlll'N~ III!' !Iii \ 1111n ~i1•11t in phUNl' 11111!11• 101111 , Wl' hnw
(.
111 (g)
t1l,,f
l/1,
,,,., l II ~ ( ti ,/ )
lo'IMtt1·1• .2. 1.2 l'ri l11dl1 ltm 11111 f 111w1in11 .
 i,y11111111N wllh U1111 l>11111"'1 11f I II nd11111 Chllft, ') tine , J.l, 111111111111 I 1111:l1111 I 11111Jtl1111• 37
lfl

Assuming thal a 1 in Eq. (I) is dcsir<.!d , post11111ll1ply hnth sidl'S ol' th<.! the i 111t•grnl
1•q1111t 1011 hy t.:os i Ht dt and integral<.! from t O to t ·1;. 'l'hl'II W<.! S<.!C thul

L''/' a 0 cos iOt dt =0 J(1'/ a1 cos jOt cos iHt di = 0

r
·r,
P (t) dt

(2) is simply the area under the sawtooth diagram for one cycle. Since this area is
(1i a· zero, the constant ao in Eq. (3c) is also equal to zero. Next, Eq. (3a) involves
('' h1 sin j!lt cos iOt dt = 0 Jo a; cos2 iOt dt = jI; the multiplication of the forcing function by cos iflt dt and integration from
Ju t = 0 tot = 7; = 277/0. From the antisymmetry of P(t) and the symmetry of
wl11•1t• / und j arc integer numbers I , 2, 3, , oo. By using these results , we cos iOt with respect tot = 'TT/fl, we conclude again that the integral in Eq. (3a)
1111d thnt 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
a; = -2 l1j P(t) cos iflt dt (3a) t = 27T/O it is symmetric about t = 37T/20. However, when i is an even
I; 0
integer, corresponding parts of sin iOt are antisymmetric with respect to
111 11 si111ilar manner, multiplication of Eq. (1) by sin iOt dt and integrating t = 7T/2fl and t = 37T/20. Thus, for all even values of i, we conclude that
pt rnhlt'l'S b; = 0.
When i is an odd integer, both P(t) and sin iOt are antisymmetric with
b; = ~ (1i P (t) sin iOt dt (3b) respect to t = 'TT/0, and Eq. (3b) gives
7; Jo
1111111lly , 111ultiplying Eq. (1) by dt and integrating yields
fi
b; = -
l21r/OP(t) sm. iOt dt = 40
-
l1r/20 P(t) sin iOt dt (5)
7T O 7T 0

a0 = .!. (1i P(t) dt (3c) In the time interval O ::s t ::s 'TT/20, the function in Fig. 2.12 has the formula
7; Jo
2
'l'l111s, l!y using Bqs. (3), we can calculate the Fourier coefficients a;, b;, and ao P(t) = -P1 0t (6)
7T
1111 1111y known periodic function. . .
Assuming that the forcing function has been decomposed mto ~ Fo~ner where P, is the maximum value of the forcing function. Substituting this expres-
H1•iit•N, W<.! t.:1111 now write the equation of motion for damped forced v1brat1ons sion into Eq. (5) yields
IIN
BP, 0 l1r/20 . . 8P1 li1r/2 .
2

111fi I cu + ku = a0 + a 1 cos Ot + + .
a2 cos 2flt
b; =- -2- t sm 1 Ot dt = Tz v sm v dv
7T o 17To
(4)
+ b 1 sin flt + b2 sin 2flt + . . . in which the new variable is v = iOt. Integrating this expression by parts and
'l'ht• 101111 solution of this equation consists of the sum of free and forced substituting the limits produces
vihrnfio11s assot.:iated with initial conditions and the terms on the right-hand side.
ll1•c11111w of damping, the free vibrations will disappear with time, leaving only b'. = ·2BP1z sm
• i'TT = BP, (- )Ci-IJ/2
2 ·2 2 I (7)
l 7T l 7T
thr lort·t·d vibrations. Each harmonic term in the Fourier series contributes a
1,lt•111ly sfnt<.! forced vibration of the type described in the preceding section. In this formula the odd values of i are 1, 3, 5, 7, and so on. Thus, the Fourier
'l'lm1t'1111·t· , we conclude that forced vibrations with large amplitudes can occur series that approximates the periodic function in Fig. 2.12 is
wh1•11t·vcr th<.! period of any term in the series is the same as (or close to) the
8P1 ( sin
P (t) = -:;;i • 1 sm
. 30t + 1 sm
. 50t - ...)
11111111111 pcfiod of the SDOP system. This always happens when the period 7; of Ot - 32
52 (8)
tlw lrnl'ing function is equal to , or an even mu ltiple of, the natural period -r, of
lht• dn111pl'd syst<.!111. und we need only superimpose the sine curves with odd numbers of periods in
I .t·l 11s now apply the 1111: thod of this st·t·lio11 to lht• piccl•wise-linear forcing 1he intervul t O to t 27T/ H .
llrm 11011 ill11slrntl'd i11 Fil\ , 2. 12. To <kt•o111pos1• th1• l1111ctio11 into Fourier com- If we omit do111ping und use II frcq11<.! ncy ratio 0/w = 0.9 , the magni-
pom•11t s, wt• haw tlw 10111111111s in Eqs , ( \) St111 ti11 u wllh l\q . ( k), we st.:c thul firntio11 tal·to, for tlw 111 st lt:1111 in Eq . (H) is
 Sec. 2.6 Arbitrary Forci ng Functions 39
Syste m s with One Degree of Freedom Chap. 2
.Ill
p
1
/3, = 1 - (O/w)2 = 5.26
whereas that for the second term becomes
1
/33 = 1 - (30/ w)2 = -0.159

1<'1111hermore , a multiplier of 1/3 2 appears in the second term ofEq. (8). There-
fm1.·, we conclude that using only the first term for the response p

U
8 -P1 /3 Slll
= -7T2 . :Ht
t
0 (9)
k
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 calcu-
lnlions. For example, the result in Eq. (9) would be changed to

u = - - 8 -ma/3 , sm
. at
0 (10) Figure 2.13 Arbitrary forcing function.
7T2 k

whk h is only the first term of the series. (3)

2,8 ARBITRARY FORCING FUNCTIONS 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
Wl' now consider dynamic loads that have no periodic character and may v~y applied force as the integral
111 1111y nianner with time. Such arbitrary forcing functions must be handled m
11 11pl•ci11I way , as described in this section. , . . u = -e-nt
Wd
l'
O
em' Pm sin wit - t') dt' (4)
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 This mathematical form is known as Duhamel' s integral.
thl• time when the response is to be calculated. If the function is applied to a Equation (4) gives the complete displacement produced by the forcing
tl111111wd SDOF system, the differential equation of motion becomes ~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
mu + cu + ku = P(t') (1)
studying the response of a SDOF system to any kind of short-term (or transient)
/\1 11uy instant o f time t' , we may calculate an fncreme~tal impulse P dt' , load. If the function P (t ') cannot be expressed analytically, we can always
ll'j)ll'Scntcd by the hatched strip in Fig. 2.13. That impulse 1m~arts to the mass evaluate the integral in Eq. (4) by some numerical method. To include initial
11, 1111 instantaneous increase in velocity (or incremental velocity) equal to l'Onditions for the damped system, we need only add their effects to Eq. (4), as
lollows:
p dt'
d,:i =- - = p,,, dt, (2)
m
'l'his l'X pr1.·ssion is vu lid rcgnrdlcss of what other forces (such us the spring f?rcc) (5)
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'H•at\111-1 1lw i11t·n·111l·11t11l Vl'lol'lty us ii 11 Wl'll' 1111 1111t111I velocity
t I
<-11,,
l' e"',
o
p,,, sin w"(t - t') dt' ]
(111 till' lillll' ( 1 ), l'llllrlt1d1• 111111 lhl• /11r11•1111•t1(///
WI' ,//,1'/lftll 't'lllt'III of till' lllllSH Ill
I htN l'q11111io11 11°pll'li1' 11I N till' lol11l 1wlulio11 or l!q . {I) .
IIIIY 11111'1 111111· f wtll lw
40 Systems with One Degree of Freedom Chap. 2 Sec. 2.6 Arbitrary Forcing Functions 41

If damping is neglected, we haven == 0 and wd == w, and Eq. (4) reduces Pm == P1/m is constant and acts for an indefinite time. For the case of zero
to damping, Eq. (6) gives

u == -1
W O
l' . ( ')
Pm sm w t - t dt , (6) u == Pm
-
W O
l' sin w(t - t') dt'

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

u = Uo cos wt+ -u0 sm

.
wt+ -1
W W
l' .
O
Pm sm w(t - t ) dt , , (7)
u = p'"(l
2
w
- COS wt)= Pi(l -
k
COS wt) (8)

'l'l11s briefer equation for the total response can be used whenever damping is not
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
p
From this result we see that a step force P1 produces free vibrations of amplitude
Pi / k superimposed on a static displacement of the same magnitude, as depicted
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 un-
damped 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 = kPi (1 - cos wt) (8)

u = Pi
k[cos w(t - t 1) - cos wt ] (9)
0 t'
(a) 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
u
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,

(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 11,, cos w(I - 11)

u,,
+ - sin w(t - t 1) (11)
w
S111I1-111111ti111111l 1·x pn·Msio11:, ( 10) 11110 Eq. ( 11 ), followed by trigonometric manip-
M111111 .1,1,1 1111 "1t, 11111111111111. 1111 ,, fil"'"" 11111111111, y1rld, thr ~11111r 11·,1111 11, lhnt 111 hi , ( 11)
 v11111111• with 01111 l>11u1nn "' I 1n11!111111 <.111111 'J ,111 , in A1ltl1tury I 111111111I1111111111111 43
/l
d1~pl11t•t·111t·11t 11>1< ol tlw 11111ss relative to grnund . Thus, f'or a step acceleration a,
ol llw ground, l:q. (8) bccomcs

u"' - - ~(
2
I - cos wt) = - ma , (1 - cos wt) (12)
w k
111 addition, the second part of the response to a rectangular function for ground
1wt·eleration is
ma,
u* = - k[cos w(t - t 1) - cos wt] (13)
0
t' t, which is a modification of Eq. (9).
(a)
1r.1mmple 2. 7
l•l11d the undamped response of a SDOF system to the linearly increasing force, known
p mmp function, given in Fig. 2. l 6(a). The rate of increase of the force P (t ') per unit
IIN 11
111 lime is P,/t1.
As indicated in the figure , the forcing function for this example is
/ 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
l111l11,11·11ting this expression by parts yields
/ P= - P,
. wt~
u = -P1 ( t - -l sm (c)
P, - - - - - - - - - -'-------"'------- kt, w

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 ampli-
(b)
111d1• l'1 /kt1 w, as shown in Fig. 2.16(b).
Figure 2.15 (a) Rectangular impulse; (b) step-load simulation. The velocity at any time tis equal to the first derivative of Eq. (c) with respect to
llllll• That is,

Depending on the duration time t 1 for the rectangular impulse, the response
111 11SDOF system may take different forms. However, the maximum response Li = .!'.!_(
kt,
1- cos wt) = .!'kt,.!_ (1 - cos 2T7Tt) (d)
l'II II ,wvcr be more than the value 2P1/ k. Instructions for determining loci of
I 111111 !his 1·xprcssion we conclude that the velocity is zero at times t = 0, T, 2T, 3T, and
rt•1ipo11st· muxima due to various types of forcing functions appear in Sec. 2.8.
11 1111 'l'lll'wf'on.', the slope of the displacement curve in Fig. 2. 16(b) is zero at t = 0,
I I the dumped SDOF system in Fig. 2.9(a) is subjt•ctcd to arbitrary wound 1
I 1, 11; 1111<1 .~o on . Furthermore, the velocity is always positive and has a maximum
,u·,·,•l,•mtio11.1·, we need only rcplucc lhc f11111;tio11 p111 (t ') in liqs. (4) through (7) v11l111 111 1/ '1/At1 11t tinws r T/2, 3'/'/2, 57'/2 , und so on rscc Fig. 2.16(b)].
wilh tht· l11nl·tion fl~ (t ') ii~ (11 ). In imch t'lllWH tht•st· t·quutions yield the II w1• linw 11 11111:p µround 11,t'l•k'rution i11sll•ud of an applied force, it may be
 45
Chap. 2 Sec. 2.7 Step-by-Step Response Calculations
Systems with One Degree of Freedom
44
Similarly, the velocity in Eq. (d) is replaced by
p
u* = - mai (1 - cos wt) (g)
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

(a) p = P,t'
t,
P,

:IP,T 0
kt,
P,t
Figure 2.17 Triangular impulse.
kt,~
/
The function appearing in Fig. 2.17 is called a triangular impulse. For the first
/ '/I 2P, lime interval (O :S t :S t 1), the response is the same as for the ramp function, as given
~ kt,
1
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', I
kt,
U = -A~t - -1 Sill
kt,
. Wt
w
~ (c)

u = .fl [t, cos w(t - t,) + .!. sin w(t - t,) - .!. sin wtl (t, :S t) (h)
kt, w w J
0
2T 3T If the triangular impulse is a ground acceleration with maximum value equal to a,, the
(b)
Il'Sponscs in Eqs. (c) and (h) have u* in place of u and -ma, in place of P1.

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

'J.,7 STEP-BY-STEP RESPONSE CALCULATIONS
l'II J>l\'NNl'd LIS
(c) 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
wh1•11· 11 IH thl' f11'01lnd lll'l'l' lo l'llli11111111lnw 1, . Thon 1l1l' n.isponHC in llq . (c) is converted l11lill•, 111 such cases it may be feasible to replace the data with certain formulas
1 hy l'ttt Vl' titting 111ethods und then lo use those formulas in the Duhamel integral.
tn lh1• 11•l111l w 11111111111
I l11wrw1 11 11101l' l,ll'lll' l'III m~thod fol' t•vnlunting the response consists of using
1

I ~Ill 1111) (I} u111pll· 111ll' 1p11l11lim1 1'1111ctl1111 in II rnpt1titiw 1wdcs of calculations. The
111 11111· 1
1('
,11
 Sec. 2.7 Step-by-Step Response Calculations 47
46 Systems with One Degree of Freedom Chap. 2
2
latter approach is discussed in this section for piecewise-linear interpolation U3 tiI'j [ w2 t' - 2n + e- nr ·(2n cos wdt' - w~ wd
= !itjkw - n sm
.
wdt ' ) ] (Zc)
2
.fimctions.
Figure 2.18 shows a forcing function of general shape that is approximated This formula is available as the solution of Prob. 2.6-3 .
by a series of straight lines. For a particular interpolating line in the time interval By differentiating Eqs. (1) and (2) with respect to time, we can also find
1 -c;; t :s tj+ 1, the response of a damped SDOF system may be written as the sum velocity expressions in three parts. Thus,
1
of three parts, as follows: (3)
U = U1 + U2 + U3 (1)
where
Using the definition t' = t - tj, we have for the first part
(4a)
(2a)
and
'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). (4b)
The formula for this portion of the response is drawn from Eq. (2.4-10).
p 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)
1'11
become
_ 11 !!.,.( " Uj + nuj . " \ (Sa)
( u 1) i+ 1 = e \u.j cos WJutj + WJ sm wdutj)

I 'I (u 2)i+1 = l1
k [1 - e-11!!.1-1 ( cos "
wdutj n sm
+ WJ . wdut
" 1\) J (Sb)

(Sc)

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

(u1 )1+ 1 = e- "l!. 1 [- 1

(ujwd + n ui :dnui) sin wd!iti + u1 cos wd!iti] (6a)
Figure 2.18 Piecewise-linear interpolation of forcing function.
2
The other two parts of the response in Eq. (1) are associated with the (u2)1, 1 = ljw e - "tJ.'1 sin wd!it-1 (6b)
11t111ight line rorcing function in Fig. 2.18. The one caused by the rectangular kwt1

= ~~i [1
l!llpulse of magnitude ~ is
(,i 1)1 1 1 - e ntJ.r1(cos w,,titj + ;d sin wd!itj) J (6c)

"i I
ljk I , , 111' ( -os <tJ,1t' I
II
sin ,,,,,1 ') I (2b)
h111111i1111s (5) and (6) <.·. rnistitute rec111T<'IIC(' formulas that may be used to calcu-
\ '"" 11111' tlw d11111pvd n•s p1111 Nl' tit the cud of stt•p j and to provide initial conditions at
wl11l'11 c1m11·11 1111111 1111· s11l11ti111111t 1'111h ) ,, ) < >11 tht· nth1~1 h1111d , till' t•llt·cl of
th,• II 11111pul111 l111p11l111• 111 11111v111t11d1· Ali 'i 1 1 , ; 1!1·1·111111 11 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 49

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

(u1)i+t = ui cos w!l.ti + !!i. sin w!l.ti (7a)

w
p
(u2)j+I = t(l - COS w!:J..tj) (7b) P,
P, sin flt

(u3\+1 = A!l.Pki (w!l.ti - sin w!l.tj) (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 approx-
i111utc results. u
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, 2
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 expres- 0
T, T,
sions. I lcnce, these coefficients need be computed only once and then used 4 2
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 - 3
tl111l' steps of duration !::.t = 'Fj-/20. Using the method of this section, calculate approxi-
11111tely the response of the system. Assume that the initial conditions are uo = Lio = 0 , 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

u = kP, (.sm a/
r.
-
0. . wt
w sm ){3 (a)
(b)

Figure 2.19 (a) Sinusoidal forcing function; (b) approximate responses.

111 which the nrngnificution fuctor f3 hus the value tl11w stl•ps. 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,

/J • I1 (~l/,,,)J I · -1- !-0-,1-,)l. ~.) <, I (h) till' 11pprnx1111Ull' displut'l' t11entH urc slightly less than their exact counterparts, because
11111·111 111h•1 pul11t1011 ol' llw siiw t·111vc is impt·rf'ect. Decreasi ng the step size would, of
A1111pp111x1111111r ~11!11111111 IN 111111111 hy 11pply111p ht~ ( /) 111111 (H) 11'1 III NIVl'ly Ill I() HJll'l'ifll'd 111111~1·, lrncl to 1•xm·1 v11lm·N (l'Xl'l'J)I Im 1rn11HI ofl l'crms) .
 Response Spectra 51
Sec. 2.8
110 Systems with One Degree of Freedom Chap. 2

u
TABLE 2.1 Solution for Example 2.9

P; Uj Exact j P; Uj Exact
0.8
I 0.309 0.006 0.006 11 -0.309 1.407 1.418
'}, 0.588 0.048 0.049 12 -0.588 1.000 1.001
I 0.809 0.154 0.156 13 -0.809 0.404 0.407
0.6
·1 0. 951 0.338 0.341 14 -0.951 - 0.338 -0.341
~ 1.000 0.593 0.598 15 -1.000 - 1.151 -1.161
(1 0.95 1 0.896 0.903 16 -0.951 -1.945 -1.961
I 0.809 1.203 1.213 17 - 0.809 - 2.616 -2.634
0.4
K 0.588 1.461 1.473 18 -0.588 -3.068 -3.094
1) 0.309 1.613 1.626 19 -0.309 - 3.220 -3.246
10 0 1.607 l.620 20 0 - 3.020 -3.045
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.
0 t,o t,.
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 Figure 2.20 (cont.)
hlnNI lon:c rises quickly to the maximum value P1 and then diminishes more slowly (and
TABLE 2.2 Solution for Example 2.10
1·v1~11 ht•comcs 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 j P;
j P; Uj Uj
11•Npo11st· of the structure. Let the values of both Pi and k be unity, and assume that the
l11it11il rnnditi ons are uo = uo = 0. 1 1.000 0.007 9 0.070 0.780
A I lcr using Eqs. (7) and (8) recursively to obtain the time history of the response , 2 0.850 0.050 10 0 0. 830
wu 1h1111 list the calculated displacements in Table 2.2. For this problem there are no exact 3 0.720 0.127 11 -0.050 0.844
4 0.590 0.230 12 - 0.080 0.819
/l 5 0.475 0.350 13 -0.100 0.755
6 0.360 0.474 14 -0.080 0.654
7 0.250 0.594 15 -0.050 0.521
8 0.155 0.699 16 0 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 1l11Nh1·d curve that gives the effect of using Eqs. (5) and (6) with 'Y = 0.05.

() ,II R SPONSE SPECTRA

1111 dy11u111k louds discussed in this chapter cause vibrational responses in

1 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 (rn Ill' l[lll' lll'Y ) is llw ch11rnc1cristic that dclermines its response
11, ,1 ptv, 11 1111, 1111 h111l·l11111 111 11ddit1l111. !Ill' slrnpl' tllld durntinn of' the forcing
11'1111111• 10 (II) i l1N1 l 1 il11 d iilllNI 11111d , t i,) ll)ll•t11,l11t1tl1 II ~1'1111"1I
 63
62 Systems with One Degree of Freedom Chap. 2 Sec. 2.8 Roaponoo Spoctro

function itself play important roles in the response. Plots (or loci) of maximum
icsponse values against selected parameters of the structure or the forcing func-
lion are called response spectra. Such diagrams are useful in design because they p
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 accom-
P, l - - - - - - - - ,
pnny 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 0
lhc only part for each curve exhibited in the figure . In all such cases , the
(a)
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
u
- = cos w(t - t 1) - cos wt (1) 2 I o.s7
Ust .L,,.-"r---.10.47
whl.lrl.l the static displacement due to P1 is
o.Jr I
I
P,
Ust = k (2)
I
I
l>ilfcrcntiating Eq. (1) with respect to time, we obtain
--.~o.2r I
I
~= w[sin wt - sin w(t - t 1)] I
Ust
I
lly setting the term in brackets equal to zero, we find an expression involving I
lht• time t,,,, at which the maximum response occurs. That is , I
sin wtrn = sin w(t,n - ti) I
I
'l'hl• following relationship satisfies this equation: I
7r Wf1
wt,,, (3) 7 7
2+ 2 0
4 2
(b)
I lt•nl'l' . t,,, is II lillL'Ur fun<: tion of t 1• Substituting lhc formula for wt,,, from Eq. (3)
11110 Eq . (I) yil•lds 1,·1Murl.' 2.21 (11) Rl'Clllnl!uli11 impulse; (b) response curves.

II 111/ 1
l S Iii \ )( l (4)
1
 ll4 fly•1111111 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· Np1•1 l111111 lot the rcct
1111~111111· i111pulsc us follows :
Um
-us, = 2 sm. -7Tf1
T (5a) I
~
t,,, I t
-=-+-1 (5b)
T 4 2T
HIid 1
II

(5c)
t/11 t,
- =-
T 2 (5d) o'---- -~,----~,----3:------:---- T
4 2 4
I >l11gl'll111s of' these dimensionless expressions appear in Fig. 2.22 , where u,,,/us, (b)
1111d t,,,/'J' nre plotted against t 1/T. From Eq. (5a) we see that if the time t 1 is less
Figure 2.22 (cont.)
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. I 1111h,•1111ore, while damping is of great importance in the problem of forced
Al this point we note that plots of magnification factors for forced vi- , 1l1111flo11s, it is often omitted as a consideration in response spectra due to
l>rnllous constitute response spectra, as defined in this section. Figure 2.1 l(a) f111pt1INiVt' excitations. Small values of damping have littl~ ef~ec~ on such re-
1·011t11ins u series of curves for /3 = um/ U 51 plotted against the frequency ratio 111111111• 111nxima, which usually occur before much energy 1s d1ss1pated. How-
U/ ,,,. We recall that these curves represent only the steady-state part of the • ,., 1, 11 group of damped response spectra can always be co~structe? for any
11•sponsc nnd that a different curve is obtained for each level of damping. If the 1t111 l11u function, with a different curve for each level of dampmg. This may be

l11111sk·n1 rurls of the forced vibrations were to be included, the response spectra ,11 111111plishcd for simple cases by deriving the appropriate analytical functions ,
111 Fi1', 2. 11 (ii) would be somewhat higher, but this effect is of little significance. 11111 Im rn1nplicated situations a numerical approach must be used.
u,,, 1,11111ph• 2. 11
u.,
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
v•t• 111 In lhis excitation is
') - --- - - - .;;-------------
u = !:2_ (.!. _ sin wt) (O ::;; t ::;; t,) (a)
k t, wt1

II -
Pi
-
[ I + ---'------
sin w(t - !1) - sin wt]
- (b)
k wt1
11 11 11hji•1•1lvc of this example is to determine the response spectrum and the correspond-
t11i• 111111• l1111c1ion. . .
0
---,----"',----3
- ._____ ....____ t,T ll y i11spcction of' Eqs. (a) and (b), we see tha~ th~ max_1mum response will always
111,111 11l1t•1 111111: 11 • Therefore, only the latter equation 1s of mterest here, and 1t may be
4 2 II , •111, MN1•d 111 llw dinwnsionless form
(11)
1/ I
1 l<ut, I sl u 1,,(r /1) - sin wt] (c)
1'111111·1• 2.22 (11) Rt•NpttONl' Nl)l'L'l111111 fm 11•1•11111~111111 l111p11IN1•: (h) tlim· ol' 11111xl 11111m 11>(1
ll' NPIIIIM' .
1tllh 1, 11111111111{ l\q . (1•) with 11•sp1•1•f lo 1111w, w,• oh111i11
56 Systems with One Degree of Freedom Chap. 2 Soc. 2.7 Response Spectra 57

p
u 1
- = -[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

wt1
wt,,, = 1r +2 (d)
(a)

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

1
Um = l + ~ sin wti = 1 ± - - V 2(1 - cos wti ) (e)
u,. wt, 2 wt,

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

u,,,
Ust =
l
+
T
1Tl1
I sm-T I
. 1rt. (f)

lm = .!_ + ~ (g)
T 2 2T
Figure 2.23(b) and (c) contain dimensionless plots of the response spectrum from
' - - ~ ~~--'-~~~~'--~~~-'-~~ ~~.l.._~~ - t, liq. (f) and the time of maximum response from Eq. (g) . We see that the response
o 2 3 4 r Hpcctrum has its highest value of u,,,/u,, = 2 when t, = 0, for which the input becomes
(b) 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
T 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
Nl1°11dy-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 = -fl2 d cos flt (h)

1
2 I h1·1cfore, the equation of motion in the relative coordinate u* becomes
mu* + cu* + ku* = -mug = m02 d cos flt (i)
I'11111pudng lhis relationship with Eq . (2.4- 22), we see that u* has replaced u and mfl2 d
l t,
0 ----'----- T h11Htl'pluccd P. llence, the steady-state response [see Eq. (2.4-31)) becomes
1 1 3
'1 'J:
(c:)
" 11• /3 111!l2d en
k cos ~£1 -
0) (j)

Mw1111• 1.z.1 ( 111 H11111p Hh'p 111111111111, (hl 11·~p1111 H1· Hiu·, 11,1111 , (, > 111111 " ' 11 11i,d 11111111 111 wh11'11lh1• 11111~111hl'lllior1 Im Im /I 111111 lhl· ph11s1· 11111-1k· 0 111c given hy Eqs. (2 .4- 32) and
ll' ~pttllHI'
( I •I I\)
 68 Systems w ith One Degree of Freedom Chap. 2 Chap. 2 Problems 59

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
Il'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
~ uu = dcosnt u 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
(a)
W becomes the same as {3.
The forcing functions treated in this section lead to explicit expressions for
11111 / us, and tm/T. but it should be mentioned that these are exceptional cases. In
1-tt•ncral, it is difficult to identify the time range within which the maximum
4.0
I ll'Nponse occurs . In addition, the equation containing tmis usually transcendental
1..-- 'I =0
1111tl cannot be solved explicitly. Under these circumstances , values of unJus, and
r,
.-- 0.10 t,,,j'r must be found by exhaustive calculations, using a series of values for the
3.0 tt111c ratio t 1/T. For each ratio the expression for u/u,, in terms of t/ T may be
..- 0.15
plotlcd, and a value of u,./ us, as well as tm/T may be obtained therefrom.
0.25
13* 2.0 ·"'~

-~ \ ~0.375
~~
REFERENCES

Jl'-'
1.0
Ti moshenko, S. P., Young, D. H. , and Weaver, W. , Jr., Vibration Problems in
0.50 _.-,;::
.__.... l:'11Rineering, 4th ed., Wiley, New York, 1974.
1.00 1
Wilson, J. S. , "Performance Characteristics and Selection of Accelerometers," J.
~
~ Sound Vib., Vol. 12, 1978, pp . 24-29 .
0 1.0 2.0 3.0 4.0 C'rnig, R. R., Structural Dynamics, Wiley, New York, 1981.
,!l
w

(b) PROBLEMS
l•'l~ure 2.24 (a) Dam~d SDOF system with harmonic ground displacement; (b) ,i I. For the prismatic cantilever beam in Fig. 2.l(a), find approximate expressions
1cspo11sc spectra fo r spring force.
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
To improve the form of Eq. U), we use w2 = k/ m and define the new ~upportcd prismatic beam in Fig. 2. l(b).
11111g111fkntion factor, {3 *, as follows :
: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.
{3* - {3(~)2 = Vl I
(0/w) 2
(0/w)r lr I (2'Y0 / wP
(k) D(t)

lh1•11 11•w11k Eq . (j) in the n·dut·cd form 1

II*' fl+ ,I l'ON({ lt fl) {{') ...~ - - ~ ~-
~ ----1 , - - ·Im
' ll11· ~h·11dy N( illl' lrnn· Ill 1111· ~1111111• 11111y llllW Ill" tllkl'II IIN ---- ----+ - - -l - j
/ fi A,/ , IIH ( 111 //) (111) 11'111111·1• l'l.J. .I
60 Chap. 2 Problems 61
Systems with One Degree of Freedom Chap. 2

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?

r
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 vi-
bration of the mass m due to this influence?
L
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 rela-
tive 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
tm
l,2•5. A rigi~ b_ar AB of mass density p has cross-sectional area A and length L. At its
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
Figure P2.3-3 1·
L
2
.1. L
2
.1
ol w, J, and T for this problem, assuming that the bar can only rotate about the
uxis of the rod.
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 steady-
state rotational response of the bar to the harmonic moment M cos flt, where
fl= 2w.
A

-2r

2r
-
M cos Ot

I
')
I
') -~ Flflurc P2.2-5
l,l•<,. For lhl'. 11,wd (ll''.1111 Nhow11 111 111H , P.2.} I! , dt1h11111l1w 1lpproxl1111,tl' cxpn:ssions 8
lot 1,1, / , 1111d / , l•w lhl Np11q111H1' ll•I 1111' ll11x 11111I ilHldlty /o'I hr 1•011Nt1111t ulonH the
l1111Hlh ol llu- lu-11111
62 Systems with One Degree of Freedom Chap. 2 Chap. 2 Problems 63

2.3-5. Let the prismatic cantilever beam shown in Fig. P2.3-5 be subjected to a ground 2.3-10. If the left-hand support of the overhanging beam shown in Fig . P2.3-10 has the
translation Da = d sin Ot, as indicated in the figure. Calculate the steady-state harmonic displacement d cos Ot, what is the response of the mass m? Assume
response of the mass m for O = 2w/3 . that the beam has constant flexural rigidity and that O = 6w.

m••- ------------~ f0 9
(t),6, (t)
m
td cos Ot

-L----.l
I• t + - - -
i ~
f - - - - - L - - --
#;,
L-----+i
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.
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 fre-
quency 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

2.J-8. Repeat Prob. 2.3-7 with rotational ground acceleration Da = 88g cos Ot
n = 5w, assuming that the mass moment of inertia of the disk is !,.
2.J-9. Let the right-hand support of the simple beam shown in Fig. P2.3-9 oscillate in
and P,
-
accordance with the harmonic displacement d sin Ot. Determine the steady-
slate response of the mass m, assuming that the beam is prismatic and
n 7w/8. 0 ,r 2,r 3,r 4,r
:n n n n
d sin Ht
t
W,~,- -·-, ----9»}1 m
p1 ~

I~- ~ - ~~ ·I
1<1111111• I' l, I 11 l•'l111m• 1'2.~- 1
 Systems with One Degree of Freedom Chap. 2 Chap. 2 Problems 65
64

2.5-2. In place of the periodic forcing function shown in Fig. P2.5-2, determine a 2.5-5. Derive the general expression for steady-state forced vibrations of a damped
Fourier series. SDOF system due to the forcing function given as the Fourier series in Eq.
p (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.
p,~
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
I I I I SDOF system to the first term in the series from Prob. 2.5-3.
0
'IT 2'lT 3'lT 4'!T
n n n n 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.
- P, I--
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
Figure P2.5-2 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 p
Pourier series.
p

- P, - - - - - - - -'------

Figure P2.6-4

,(1°!i. Assuming that the forcing function shown in Fig. P2.6-5 acts on a SDOF
system, find expressions for the undamped response.
Figure P2.5-3
p
2.!i-4. 11or the sawtooth function shown in Fig. P2.5-4, substitute a Fourier series.
p

P,

() IJ11 1111
/11 1111
II II II II
'111411111 I' L. t, ~
1111
Cht1p. 'I l 1111hl11111 11 67

1..,,.,,. I )llllVC l!Xl)l llSShHIH 1111 till' 1111tl11111pcd response o l HSI)( )I • N)'Nlll lll to the f'o1·ci11g f1 1) , l>l•tlw L' Xptcss l1111s for lht undumpL·d ,csponsc of a SDOF system to the para-
function shown in Fig . l'J. (1 (1 . l>ulil' g1ou11d 11ccclcrntio11 i/H a,(1 1 /1,/ given in Fig. P2.6-9.
p

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.
iig Figure P2.6-9

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
o, - - - - --,,...------------ 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 p
undumped response of a SDOF system.
iig
P,

a,

Qi£-_ _ __L_ _ _ __;ll.----~----~----

t,o

,., --------- -
0 --------~~ ~------( 1,

1,·111111·11 p.1.7-~
 Chap. 2 Problems 69
Systems w ith One Degree of Freedom Chap. 2
""J..7.(1. Calculate the undamped response of a SDOF system to the forcing function
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
shown in Fig. P2.7-6, using 20 equal time steps with l:!.t = T/20 .
duration At = T/30.

p p

P, lo---------. P,

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

t,o t20

P, - - - - - - _JI--------~ o .....""""'~~~~~~~~--''--~~~~~
t,o

Figure P2.7 -8

Figure P2.7-6

p Apex
P,

0 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 J..7-1). The parabolic forcing function shown in Fig. P2.7-9 has the formula
At '/'/ 10 , 1111d ll11d llw 1111cl11111pN I ws pollMl' wh\111 it Is Hpplicd to n SDOF P P 1( 1 12 /do), With 10 equal time steps of duration l:!.t = T/25 , obtain
llw u11d111111wd 1l•Npo11sc of' 11 SDOF system.
syshl111 .
 Chap. 2 Chap.2 Problems 71
10 Systems with One Degree of Freedom

p
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 .

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 p

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, '-- - - - - __,......__ _ _ _ __,

l<'lgm·l' P2.ll· 1

l,H•2. lfop1•11t P, oh ) .H I 101 th,• lw d 111' i111ll't lo11 Nhow11 111 Fl1,1. I' ) .H l. Ukc l'mh.
) '1 "I Im 11·~pn11N1' 111111111111H )
 3
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
P,
and Vibrational Analysis

0 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

In this book we use the method of.finite elements [1-3] to discretize solids and
P,
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
0 t, 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
Figure P2.8-6
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 pro-
gram this method on a digital computer.
Figure 3. I shows various examples of solids and structures that are dis-
rrclized 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 mem-
hL•1s . Figure 3.1 (c) depicts a two-dimensional slice of unit thickness, represent-
111~ tht· rnns lunt slutc known us plnnc strain on the cross section of a long,
p11s11111lk solid . On lht· olht·1 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 lort·t·s npp lil:d in its
74 Finite Elements and Vibrational Analysis Chap. 3 Sec. 3.2 Stresses and Strains 75

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

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 en-
forced.
/
/ 3. Assume displacement shape_functions within each_element so that dis-
/
placements, 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.
/
/
(c) /
/

- 6. Develop differential equations of motion for the nodes of the discretized

continuum by assembling the_finite-element contributions.

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

a vibrational analysis for any linearly elastic structure . This type of analysis
(d)
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
(e)
(/)
shell structures .

Jt'lgure 3.1 Structures modeled by finite elements: (a) beam; (b) space frame; (c)
plunc strain; (d) solid; (e) plate; (f) shell. 3 .2 STRESSES AND STRAINS

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 d1·g1t·l'N of freedom and will be
1l•lc11'l'd lo us MOOF .l'y.1·t,•111s.
Tht· liniw l' k11w111 111l•lhod 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 1•h·1111· 11I ' ('ht•w lt111t•lio11s giVl'
11pp111x111rnlr 11•1111II N whr11 1111' 1•lt•1111•11t 111 111 1111111 - 111· 111111 1•x11l'I 11•111111~ 111
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 right-
l11111d 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'ic1m•nt in Cartesian coordinates, where
llw 1•d)'1'H 1111• ol lt·HHths r/1 . di', ,11ut d·. Normal nnd s/11•r11·/111{ .1·1r<•sse.1· arc
 76 Finite Elements and Vibrational Analysis Chap. 3 Sec. 3.2 St resses and Strains 77

For convenience, the six independent stresses and the corresponding

strains usually will be represented as column matrices (or vectors). Thus,
0-1 erx E1 Ex

er2 ery E2 Ey
I
er3 u, E3 E,
I dy (J' = E =
I U4 Txy E4 Yxy
I ers Tyz Es Yy,
er6 Tzx E6 'Yzx
! )-.,. T,,

where the boldfaced symbols denote the vectors shown.

Iri, Strain-stress relationships for an isotropic material are drawn from the
)----- - -----
y,v /
theory of elasticity [4], as follows:
/ 1
/ = -E (CT - vay - ver,)

f-'·"
EX X
/

dx 1 Ty,
Ey = - ( - Verx + ery - Ver,) Yyz =G
l,W E
Figure 3.2 Stresses on an infinitesimal element. 1 1'zx
E, = E (- verx - very + er,) 'Yzx = G
indicuted by arrows on the faces of the element . The normal stresses are labeled
11 ",rv, and er,, whereas the shearing stresses are named rxy , Ty,, and so on. From where
l'q11ilibrium of the element, the following relationships are known: E
(l) G = 2(1 + v)
'l'hus, only three independent components of shearing stresses need be con- In these expressions Eis Young' s modulus of elasticity, G the shearing modulus,
11idl·1cd. and vis Poisson's ratio. With matrix format, the relationships in Eqs. (5) may
Corresponding to the stresses shown in Fig. 3 .2 are normal and shearing be written as
,1•/ro/11.1·. Normal strains Ex , Ey, and E, are defined as
e=C<J'
au = -
av aw
E =- in which
Ex= - E
ay az (2)
ax y z
1 -v -v 0 0 0
Whl1l'l' 11 , o, and ware translations in the x, y, and z directions. Shearing strains, -v 0 0
-v 0
Y, 1., 'Y1•11 un<.I so on , are given by
-v -v 1 0 0 0
au av C = l_
'Y.<y = ay + ax = Yyx E 0 0 0 2(1 + v) 0 0
0 0 0 0 2(1 + v) 0
av + -aw
yy: =- 0 0 0 0 0 2( 1 + v)
CIZ ily
Matrix C is an operator that relates the strain vector E to the stress vector CJ'. By
<'. IV ! i)11 the process of inversion (or simultaneous solution), we can also obtain stress-
y,
dI ii ' stmin rt'lotio11ships from Eq. (7) , ns follows:
I 1!•1111• , 011l y (h11•1• 111 lh1· ~l11•111 IIIH ~(1111111, Ill 1111l1·111·111l1•11t (9)
78 Finite Elements and Vibrational Analysis Chap. 3 Sec. 3.3 Equations of Motion for Finite Elements 79

where time variation for each component of body force is assumed to be the same
E = c- 1 throughout the element. That is , we may have one time function for bx, another
for by, and a third for b,.
1- J) J) J) 0 0 0 Time-varying nodal displacements g(t) will at first be considered as only
V l-v V 0 0 0 translations in the x, y, and z directions. Thus , if n. 0 = number of element
J) J) 1- J) 0 0 0 ooct.e.s.,
E 1 - 2v q (t) = {q;(t)} (i = l, 2, ... , n n)
0 (3)
0 0 0 -- 0 0 D..Q)
(1 + v)(l - 2v) 2 where
1 - 2v
0 0 0 0 -- 0 q;(t) = {qx;, qy;, q,;} = {u; , V; , w;} (4)
2
However, other types of displacements, such as small rotations (av/ iJx, and so
1 - 2v
0 0 0 0 0 -- on) and curvatures (a 2v/ ax 2, and so on) will be used later.
2 Similarly, time-varying nodal actions p(t) will temporarily be taken as
Matrix E is an operator that relates the stress vector CT to the strain vector e . only forces in the x , y , and z directions at the nodes. That is,
For the elements in this chapter, we will not need the 6 X 6 stress-strain p (t) = {p; (t)} (i = 1, 2, ... , n.n) (5)
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 in which
t:lcments , larger matrices, such as that in Eq. (10) , will be needed. p;(t) = {px; , py;, p,;} (6)
Time functions for Pxi , py;, and p,; at each node may be independent and arbitrary .
3 .3 EQUATIONS OF MOTION FOR FINITE ELEMENTS Other types of nodal actions, such as moments, and so on, will be considered
later.
Wu shall no w introduce definitions and notations that pertain to the finite ele- For the type of finite-element method to which this book is devoted, certain
inunts to be studied throughout the book. By using the principle of virtual work , assumed disvlacement shave functions relate generic displacements to nodal
we can develop equations of motion for any finite element. Such equations displacements, as follows:
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
u (t) = f g (t) ru
next section . ln this expression the symbol f denotes a rectangular matrix containing the
Assume that a three-dimensional finite element with zero damping exists functions that make u(t) completely dependent on q(t).
in Cartesian coordinates x, y, and z. Let the time-varying generic displacements Strain-displacement relationships are obtained by differentiation of the
11(1) at <Illy point within the element be expressed as the column vector 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
u (t) = {u, v , w} (1) multiplication. Thus, ·
whcru u, D, and w are translations in the x, y , and z directions, respectively .* E(t) = d u (t)
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: 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)].
(2) 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 e (t) = B q.(f2_
whurc
*'1'11 NIIVl' Np ill'I', rnl1111111 Vl'l ' IIIIN 11111y lw wt1 11,111 111 11 lll W 11111 l11N1•tl hy h1 111•1•N ( l ,111111 wilh
rUIIIIIIIIN ~11p11111l lt1f1 lh1• 11•1111~ n dr
 110 I l11llu I l11111u1111111111 I Vll11n11111111I A1111ly1ln Chop. 3 m ,1.,1 I q1111llm 111 ul Motlo11 fw 11111111 I 1111111111111 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' .1·rre.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
1,odv LJdv dy

o(t) = E B g (t)
in which the matrix product E B gives stresses at a generic point due to unit
values_of nodal displacements.
(12).
/
-·-
// b,(t) dV

I b,(t) dV
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 J--- piidV - - - - - -
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 /
au.= aw. y, V, ii
dx
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
dt•lllil , we assume a vector aq of small virtual displacements. Thus,
(i = 1, 2, . . . , nen) (14)
)-x"·'
z, w,W

'l'ht111 lhe resulting virtual generic displacements become [see Eq. (7)] Figure 3.3 Applied and inertial body forces.

au= f aq (15) Substitution of Eqs. (17) and (18) into Eq. (13) produces
llsinp lhc strain-displacement relationships in Eq. (9), we obtain
ae = B aq (16) L aeTo(t) dV = aqTp(t) + L 8uTb(t) dV - L 8uTpii dV (19)
Now lhe infernal virtual strain energy can be written as Now assume that
ii= f q
'l'ltcn we can substitute Eqs. (12) and (20) into Eq. (19) and use the transposes
where integration is over the volume of the element. nf Eqs. ( 15) and (16) to obtain
For the <'Xterna/ virtual work we turn to Fig. 3.3, which shows an infini-
lt•si 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
J>ql' LBTE B dV q = 8qTp(t) + aqT L fTb(t) dV - SqT L pfTf dV q

d11t• lo lhc accelerations u, i5, and w. The symbol pin these expressions repre- (21)
Nt•11ts !he 1110.1·.1· density of the material, which is defined as the inertial force per
<'1111ccllution of aq rand rearrangement of the resulting equations ofmotion gives
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 ut·n•ll•f'11lio11s . Thus, we add the Mq+ Kq = p (t) + Pb(t)
t'Xlt111111I vlrtunl work of' nodul nnd clislribu(L·d hody loll't'N us follows: Wltl' ll'

l>W,. h11 'p (I) I ( Im 'h (I) 111'

Iv " r'ln 11111 rlV , ( 18) K r II I 14: n <IV
 Sec. 3.4 One-Dimensional Elements 83
82 Finite Elements and Vibrational Analysis Chap. 3

u
and q, 2 q, -
------,!.--------- ----x,u
M = 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
1 ~
coefficients that are fictitious actions at nodes due to unit values oL nodal dis-
placements. Equation (24) gives the form of the consistent-mass matrix. in (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 1
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 (c)
influences are considered to be statics problems.
Figure 3.4 Axial element.

The first function diminishes linearly from 1 to 0, whereas the second increases
3.4 ONE-DIMENSIONAL ELEMENTS
linearly from O to 1.
From Fig. 3.5 we see that the single strain-displacement relationship
In this section we develop properties of one-dimensional elements subjected to
du / dx for the axial element is constant on the cross section. Thus, Eqs. (3.3-8),
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. (3.3-9), and (3.3-10) yield
du df
'l'hus, from Eq . (3.3-1) we have E(t) = Ex= - = -q(t) = B q(t)
dx dx
u(t) = u
Therefore, we have
'l'hc corresponding body force is a single component bx (force per unit length) ,
B = d f = df = .!. [- 1 1] (2)
11cting in the x direction. Therefore, Eq. (3.3-2) gives dx L
b(t) = bx
y
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) = {q1, q2} = {u1, u2}
( ' oi responding 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
u t' q (t)
wlwn•

I' (I)
Flw1111• I,!! i\~ i11l 1h'lt11t111tlilll1N
84 Finite Elements and Vibrational Analysis Chap. 3 Sec. 3.4 One-Dimensional Elements 85

which expresses the strain Ex.in terms of the nodal displacements. Similarly, the q, 1 -0. 2 q,
single stress-strain relationship [see Eqs. (3 .3-11) and (3 .3-12)] becomes merely - - - - - -.......- - - - - - - - - - - - - X

O'(t) = CTx = E E(t) = EEx = EB q (t)

Hence,
(a)
E=E and EB= 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),
(b)
as follows:

K = Iv BT EB dV = i [-! Jr-1
1

= EA [
L
1
-1 -!J (4)
(c)

Figure 3.6 Torsional element.

assuming that the cross-sectional area A is constant. Similarly, the consistent-
mass matrix M is found from Eq. (3.3-24) to be 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.
M = Jv pfT f dV = { 2 f L[ L : x ]EL - x x] dA dx
llence,
q(t) = {q1, q2} = {()xi, Ox2}
= pAL[2
6 1 2
l] (5) In addition, the corresponding nodal actions at points 1 and 2 are

assuming also that the mass density p is constant. p(t) = {p ,, pz} = {Mx1, Mxi}
We see that the stiffness matrix K and the consistent-mass matrix M are which are moments (or torques) acting in the x direction. As for the axial
unique for a prismatic axial element of uniform mass density . However, an clement, we assume the linear displacement shape functions /1 and !2 shown in
infinite number of equivalent nodal load vectors Pb (t) may be derived, depending Fig. 3.6(b) and (c). Therefore,
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 ex = f q (t)
the axial element. Then Eq. (3.3-25) produces
in which the matrix f is again given by Eq. (1).

Pb (t) = f fT bx dx = if[ L : X J dx = bx ~ [ !J (6)

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
which shows that the equivalent nodal loads are equal forces at the two ends. vurics linearly with the radial distance r, as follows:
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 d()x
'Y = r - = rijJ ffi
a double-headed arrow). Thus , dx
u (t) = f), wlwrc 1/1 is the twist , or rate of change of angular displacement. Thus,
Corresponding to this displacement is u single body action dO,
1/1
b(t) Ill,
86 Finite Elements and Vibrational Analysis Chap. 3 Sec. 3.4 One-Dimensional Elements 87

y
(L r21r (R
1] Jo Jo Jo (r 2 )r dr dO dx
= GJ[ 1 (12)
L -l
z where GJ is constant. The polar moment of inertia J for a circular cross section
is defined as
J =
21T 11R
r 3 dr dO = -
LR
4

0 0 L 2
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]
Figure 3.7 Torsional deformations. for J. If a more precise analysis is desired, the theory of nonuniform torsion [6]
may be applied.
Equation (7) shows that the maximum value of the shearing strain occurs at the To obtain the consistent-mass matrix M for a torsional element, we will
surface. That is, first integrate over the cross section and then over the length. Due to the small
'Ymax = Rt/; 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
where R is the radius of the cross section (see Fig. 3. 7). Also, we see from Eq. the cross section, we find the inertial momentper unit length to be -p 18,., where
(7) that the linear differential operator d relating 'Y to Ox is 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
d
d = r-
dx
Thus, the strain-displacement matrix B becomes
which is a specialized version of Eq. (3.3-24). Integration of Eq. (14) over the
r length yields
B = df = L[-1 1] (10)
M = pl2 (L [L - x ][L - x x ] dx
L Jo x
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
= pJL
6
[2 1]
1 2
(15)

r = Gy 'f'his array is the consistent-mass matrix for a torsional element of constant cross
Hcction and uniform mass density.
where the symbol G denotes the shearing modulus of the material. Hence, 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
E G and GU Grl 11 ( 11 b)
/, l1111ding Gq . (3 .3-25) gives

Thcsl' 1l'111tionships llll' 111111Io»ous lo Eqs . ( 1) 101 till' 11x111 l l'll•1rn·111 ,

W,• 1111,y 110w 1111d th,• 1<11si111111l stll l111•11N 11111111, K hy 11pply111» l\q.
(' f"'m, dx
Jn
m1 ('·
,, Ju I1L -, xl dx = m)::.2 [1]
1
(16)

( I I I I), 11~ lollow~ 1'11,·'lt' l'q111v11l,•111 1111d11I londs llll' l'q1111l 1111111w11111 nl lhl' two ends or the clement.
88 Hnlto E:lomontt1 nnd Vll1111tl111111I A1111lynl11 Chup. 3 nc I i1 01111 lll11111111lm111I I 111111111111 89

. F_igure 3.8(a) show~ a ~traii!U}l~ elert1J.'III , for whlt:h the x-y plane is 1111111-1 111 tht· y direction . I loncc,
a principal ~Ian~ of bending. Indicated in the figure is a single generic displacc- b(t) = by
111cnt v, which 1s a translation in they direction . Thus,
I 1111dc I !sec Fig. 3.8(a)] the two nodal displacements q1 and q2 are a trans-
U (t) =V l,1111111 111 the y direction and a small rotation in the z sense. The former is
The corresponding body force is a single component by (force per unit length), h11lk 11tt·d 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 = dv, fJ z = dv2
zl dx ' 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) = {p1, Pi, p3 , p4} = {py1, M,1, Py2 , M,2}
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 .
lb) For the flexural element we assume cubic displacement shape functions in
111111ri x f, as follows:
l/1 Ji f3 J:i]
~ ,- f,

·y=:==-- -----
- =__.~ I
- [2x 3
/}
- 3x 2L + L 3 x 3L - 2x 2 L2 + xL 3 - 2x 3 + 3x 2L

x 3 L-x 2 L 2]
(17)
(c)

l'hl'sc four shape functions appear in Fig. 3.8(b)-(e). They represent the vari-
11t1011s 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 ele-
1m111t by assuming that plane sections remain plane during deformation, as
Id) lll11strutccl in Fig. 3.9. The translation u in the x direction at any point on the
1 rnss section is

dv
u = - y-
dx
1llling thi s relationship , we obtain the following expression for flexural strain:
(11)
y</> (19)
 90
f In Ito Elomont11 nncl Vlhr ntl1111nl A1111lyuls Chaµ. 3 I 111, 3.4 One-Dimensional Elements 91

y,v
" L BTE B dV

-r- ri- J Ey6

Jo A L
2
[ ~:;
-
~ :t
l 2x
2

+ 6L
] [12x - 6L . . . 6xL - 2L 2 ] dA dx
y ........ +---\ 6xL - 2L 2
-_;;;-""--~--
z
~-- Mu ltiplication and integration (with El constant) yields
I M, / \ \

I '-
10
.,........J----
_ dv
dx
\_,,
I
+---
\

-:;,I
\ K = 2El [3:
L3 -6
~~2
-3L
:3:
6 -3L
~~
] (24)

......-" ................ 3L L2 -3L 2L2

----------1~-v,....
-dx~
[= L 2
y dA
Figure 3.9 Flexural defonnations.
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
in which </> represents the curvature: • lll'11ring deformations are given in Ref. 5.
The consistent-mass matrix M for a flexural element will be developed in
d 2v 1wo parts. A typical cross section of this type of member translates in the y
<I> = dx2 (20) dlr1•l'ti on, as indicated in Fig. 3 .8(a). However, the section also rotates about its
I:rom Eq · ( 19) we see that the linear differential operator d relating Ex to v is m·utru 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
d2 I lwm Eq. (17) in Eq. (3.3-24) , we find that
d = -y dx2 (21)
f
l
The n Eq. (3.3-10) gives the strain-displacement matrix B as
M, = fv pfTf dV = pAfTf dx

U d f = _ 1'._ [l 2x - 6L 22L 54
6xL - 4L 2 -l2x + 6L 6xL - 2L 2] [ 156
L3 = pAL 22L 4L 2 I3L -13L
-3£2
(26)
(22) 420 54 13L 156 - 22L
In addition, flexural stress er.. in Fig. 3.9 is related to flexural strain Ex simply by - 13L -3L 2 -22L 4L 2
whkh is the consistent-mass matrix for translational inertia in a prismatic beam.
(23a) Notational inertia (or rotary inertia) terms for a beam can be deduced from
l lcnce, I •IH I . 9, where the translation u in the x direction of a point on the cross section
111
E =E and EB IW (23b) II - y f)z
Element stiffnesses muy now be ohlnirwd 1111111 11•11• (,, l .,l, 21, l ) ' us, ,.() ows:
11 111 thi11 1·xprl1Hsion ,
 93
Sec. 3.4 One-Dimensional Elements
92 Finite Elements and Vibrational Analysis Chap. 3

and flexural elements. Let us reconsider the torsional element in Fig. 3.7 and
(), = V,x = f,xq (t) (28) integrate the moment of the shearing stress T about the x axis. Thus, we generate

fr
where the symbols v,x and f,x represent differentiation with respect to x. Simi- the torque Mx , as follows:
larly, the acceleration of the same point in the x direction is
2
Mx = 1T Tr dr d()
u = -yo, (29)
where Substitution of the stress-strain and strain-displacement relationships from Eqs.

fr
(l la) and (7) produces
0, = V,x = f,xQ (t) (30)
By integrating the moment of inertial force over the cross section, we find the Mx = Gt{, 1r r 3 dr d() = GJ t{, (36)
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 1
lf we take Mx as generalized (or integrated) stress and as generalized strain,
corresponding virtual rotation 80, leads to the formula the generalized stress-strain (or torque-twist) operator G becomes

M, = f p!f;xf,x dx
G = 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. Mx = Gt{, (38)
( 17)] with respect to x, we find that
By this method the operator d in ~. (9) does not include .the mu.ltiplier r.
f,1 = ~3 [6(x 2
- xL) 2
3x L - 4xL + L 2 3
-6(x 2
- xL) 2
3x L - 2xL 2
] Furthermore, the generalized matrix B [Eq. (10) devoid of r] is used mstead of
matrix B. That is,
(32) B = rB (39)
Suhslilulion of this matrix into Eq. (31), followed by integration over the length,

I
Prom this point we can conclude that evaluations of th~ terms in the stiffness
pmduccs matrix K do not require integrations over the cross section. Therefore,
36 3L -36
pl 3L 4L2 -3L -3L
L2 K = JoL BT G B dx
M, = 30L -36 (33)
-3L 36 -3L
[ -L2 4£2 This expression for K is equivalent to Eq. (3 ._3-23) used previously.
3L -3L
Turning now to the flexural element in Fig. 3.9, we integrate the moment
which is the consistent-mass matrix for rotational inertia in a prismatic beam. of the normal stress r:r., about the neutral axis to obtain M ,, as follows:
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
M, = L~ U xY dA

force by (per unit length) applied to a flexural element. Equivalent nodal loads Then substitute the stress-strain and strain-displacement relationships from Eqs.
ul points I and 2 [see Fig. 3.8(a)] may be calculated from Eq. (3.3-25) as (23a) and ( 19) to find
( - Jorl f
p,, t) - T
by dx -- by L { 6, L, 6, - L}
12 (34) L
2
M , = E</> y dA = EI</>
(42)

For this integration the displacement shape functions / 1 through [4 were drawn hu this clement we can take M, as generalized (or integrated) stress and </> as
frolll Eq . ( 17). l' l'nc,·tilizcd strain . Then the generalized stress-strain (or moment-curvature)
By using f./.l't11'mll: l'd ,\•tr,·.1·.w·.1· wul .1·trol11.1·, wt• c1111 uvoid n.ipctitious inlc lljll'l'll(OI' /: is
µ111110111, ow, llll' nrn,1, "l'l'tio11s o l om· d1111l•11si111111l l·lt·111l' nl s Allhou~h this
rom·,•pt 1, 111thl'I t11vi11l l111 111111x111l l·k11u•111, 11 l'llll hr 1111111• II Mt•lul Im 1t11 s101111I
F m (43)
94 Finite Elements and Vibrational Analysis Chap. 3 Sec. 3.5 Transformation and Assemblage of Elements 95

which is the flexural rigidity of the cross section. Thus, from Eq. (42) we have
Thus,
M, = E</> (44) Mx =GB q(t) (50)
With this approach the operator din Eq. (21) is devoid of the multiplier -y. In Then the shearing stress at any point on the cross section can be found with Eq.
addition, the generalized matrix B[Eq. (22) without the factor -y] may be used (49).
in place of matrix B. That is, For a flexural element, the relationship between the nonnal stress CTx and
B = -yB (45) lhe bending moment M, is

Then integration over the cross section for terms in matrix K becomes unneces-
CTx = _ M,y
I
sary. Hence,
'l'his term as well as E = E and Eq. (45) can be substituted into Eq. (3.3-12),
K= fiJTEBdx yie lding

which is analogous to Eq. (40). -My

-' = -
-yEBq(t)
I
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 I knee,
nodes . After those steps, we can find the time-varying stresses C1(t) within each M, = EB q(t) (52)
clement, using the equation I'quation (52) is a formula for the bending moment at any point along the length
C1(t) = E B q (t) (3.3-12) 111 11nexural element in tenns of the nodal displacements q (t). After this calcu-
lulion, we can find the normal stress at any point on the cross section from
This ex pression may be converted to a special fonnula for generalized stress,
l•q (51).
depending on the application. For example, an axial element has the simple If we wish to find actions only at the ends of a member, we can simply
.~lrcss force relationship 111111iiply its stiffness matrix__K and the vector of nodal displacements q(t). While
l'x 1h1H calculation is valid for framed structures, it has no such physical meaning
(T = - (47) liu two- and three-dimensional finite elements, where the resulting actions are
X A
th lllious.
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
,(S TRANSFORMATION AND ASSEMBLAGE OF ELEMENTS
l'x = EB q(t)
A
11 /oni/ axes fo r a finite element are not parallel to global axes for the whole
Therefore, ~,, m lure, rotation-of-axes transformations must be used for nodal loads, dis-
l'x = .EB q(t) (48) 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
where E=EA is the axial rigidity of the cross section. After the axial force has 1,1 l11h11I directions at each node.
been determined from Eq. (48), we can find the normal stress using Eq. (47). The concept of rotation of axes applies to a force. a moment, a translation,
A torsional element with a circular cross section bears the following 11 , 111ull rntatio.n.. velocit~celerations, orthogonal coordinates, and so on.
relutionship between the shearing stress T and the torque M.,: I 1v.1111· \ . IO(a) shows a two-dimensional force vector F and its components E-
M,r 1111111~ 111 the I and y directions. The figure also gives the components E-· and F;,·
T (42.} 111 1111· d11cl'tions of inclined axes x' and y'. The scalar values of the components
.I
111 11111•, 1u111s of the primed axes can be computed from those for the unprimed
Suhslilulinµ lhi s 1l•1111 1" wl'II m, I•~ <; 111ul l\q ( l'J) 11110 l\q . (J .J 12) prnduc:es IKI 11, IIN follows:
M,, /•:, (11: 10 · I ' A11 I·: A11 f ~ (1a)
.I I\• ( I•: I.;,) • .I ' Aq / •; A11 I•: ( I h)
 96 Transformation and Assemblage of Elements 97
Finite Elements and Vibrational Analysis Chap. 3

y
(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

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
(a)
directions. Components E·, F;,·, and F,, in the directions of inclined axes x', y ',
y
nnd z' are given as well. For this case Eqs . (1) can be expanded, as follows:
y'
f'x· = (E + Fy + F,) · i' = A1,f'x + A12F, + Al3F, (5a)
F,· = (E + Fy + F,) . j = A2, f'x + A22F, + A23F,
I (5b)
F,, = (E + F;, + E) · k' = A31f'x + A32f', + A33F, (5c)
where k' is a unit vector in the z' direction. In matrix form, Eqs. (5) become

A,2 Al3]
A22 A23
[F,]
F, (6)
A32 A33 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

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

"
111 this expression Risa rotation-of-axes transformation matrix containing two
z' Identical rotation matrices in diagonal positions. The reverse transformation is
(b)
A= R'A'=RTA'=[RT 0 J[F'J
Figure 3.10 Rotation of axes: (a) two dimensions; (b) three dimensions.
0 RT M'
As menlioned before, displacements and other types of vectors also can be
In_th1cse cxp;cs~ions the symbols I' and J' denote unit \ll'ctor.1· in the directions
l11111 Nl111111t•d to and from local and global di rections. Therefore, we shall convert
of -~ nndy . h .1rllwrn1ort.·, llw cm•f'fkit.•nls A 11 , A 1J, und so on, ore dir,•ction
111,• 1•q11111lo11s or motion for a finite clement from local axes to global axes. For
of
'.'(/,\' /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 111111 p111111>Ml', ll•I us rt·wdtl' Eq . (1 . l ?2) rm incl ined axes, as follows:
IN lhl• dlll'rl1011 l·o.~ 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, M 'c1' I K'c1' p '( t) I p /,(t) (9)
 99
98 Finite Elements and Vibrational Analysis Chap. 3 Sec. 3.5 Transformation and Assemblage of Elements

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
'
MFF MFR]
[ MRF MRR DR
[DF] + [SFF
SRF
SFR]
SRR DR
[DFJ = [AF(t)J
AR (t)
(17)

M q + K q = p (t) + Pb (t) (3.3-22) 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
in which
the subscript R denotes restrained nodal displacements. Writing Eq. (17) in two
K = RTK'R (10) parts gives
.. .. (18a)
and MFFDF + MFRDR + SFFDF + SFRDR = AF(t)
M = RTM'R (11) and
.. ..
Also, MRFDF + MRRDR + SRFDF + SRRDR = AR(t) (18b)
p(t) = RTp'(t) (12) If support motions (at restraints) are zero, Eqs. (18) simplify to
and MFFDF + SFFDF = AF(t) (19a)
Pb(t) = R_TpHt) (13) and
I !ere the matrix Rcontains rotation submatrices for all the nodes of the element. (19b)
After stiffnesses, masses, and nodal.Joads for 'ndividual elements have
These equations will be used in subsequent work for calculating free displace-
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 ments DE and support reactions A 8 (t).
In many problems it is sufficiently accurate merely to lump tributary
all the elements to obtain stiffnesses, masses, and nodal loads for the whole
masses at the nodes of a discretized continuum [8]. When doing so , we form the
structure. Thus , by summation* we have
lumped mass matrix Me for the whole structure as
lie

Ms= LM; (14) M, 0 0 0

i= I
0 M2 0 0
and (20)
n, Me= 0
0 0 MJ
As(t) = L p;(t) (15)
i=I 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 equiv-
alent nodal loads for the whole structure. Then the undamped equations of
, notion for the assembled structure become
MsDs + SsDs = A,(t) + A .,b(t)
i~ wh!ch Ds an~ Ds are vectors of structural displacements and accelerations.
Equation ( 16) gives the structural equations of motion for all nodal displace-
ments, regardless of whether they are free or restrained .
. In pre paration fo r solving Eq . (16) , we can rearrange and partition it, as
lollows:
*
flm till' 111w111tlonNIll 111,~. ( Iii ) 111111 (, , i. 1111' 11111!1 IX 111 Vl'l'llll 1111 lh1• 1l11ht lllll Nl Ill' 1•x.p11111k•d
With /l'IIIN Ill 111111'1• II 1111• Hllllll' Nill IIM till' llllllii~ 111 Vt'I 1111 1111 1111• h II
0 0 0 Mn,,
where n,, is the number of nodes. The typical submatrix M1 in Eq. (20) signifies
11 small diagonal array defined to be
(21)
(16) 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 par-
t ll Hllll'd fo1111s. Ass11llle th111 the cross sL·clional area of member 1 is equal to A,
, 111111 ol ,11L·111lwl' J is l'q111d lo O,Cl/\, 1111d th11l ol' tlll!lllber 3 is equal to 0.8A.
100 Finite Elements and Vibrational Analysis Chap. 3 Sec. 3.5 Transformation and Assemblage of Elements 101

2
r --2
jl = 2} - 1 }2 = 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
(22)

restrained displacements (numbered 2, 5, and 6). Note that the arrows for
y restrained displacements have small diagonal cross lines to distinguish them
j2 from free displacements .
Figure 3. ll(b) shows a typical member i in the plane truss. Its ends are
t
j O - j1 0 0.8L
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
L = \/x;k + Yjk (23)

we find the direction cosines of the inclined axes to be

YJk -
A11 = cos )' = L = Cx
Xjk \ -
/\J2 -
.
sm '}' -
-
L- Cy
(24)
A 21 = -sin )' = -cy A22 = cos )' = Cx

(a)
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,
y the cross-sectional areas, the lengths, and the direction cosines cx and Cy.
x'
y' q~ qk2 /
TABLE 3.1 Member Information for Truss of Fig. 3.ll(a)

----- ---
x,, \tY ~qk1
k
Direction Cosines
Member Joint Joint Area Length
i j k c, Cy

I 1 2 A L 0.6 0.8
2 1 3 0 .6A 0.6L 1.0 0
3 2 3 0.8A 0.8L 0 - 1.0

(b)
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

~i
Figure 3. U Plane truss: (a) structure; (b) member. as follows:

~~
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
K' = EA [ -~ (25)
mc1ubcrs, and joinl numbers arc placed adjacent to the joints. The numbering L -l O 1 0
Hystr111 for joint displuccm<.: nt s is shown hy nu111bcrl·d arrows that depict the 0 0 0 0
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>tul11L·d l'n1111 llll' joint 1111111bcrs, us This matrix is the same as that for the nxial clement in Eq. (3.4-4), except that
lollows : tht zuros arl' insl'lll•d for st iffm·ssl'H 111 the y' direction. Thus, the size of the
102 11111111 f- hm11111t11 111111 Vll11 nll1111nl A1111ly1l1 Cho p. a nl' .u, l 111111,l1111n11!11111 111111 A11111nhln1111 111 I ln1111111lM 103

matrix is expanded from 2 x 2 to 4 x 4. Equation ( I 0) l'nnhks us to trunsl'or111 Noll' that tht\SC indexes uppeur us subscripts for the .t and y translations at joints
the stiffness matrix from local to global axes. !Jenee, / 1111d kin 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
-c; r11d1 mutrix in Eqs. (29). After assembling the structural stiffness matrix using s.
K = RTK'R = EA
L
[c:!Y
-c; -cxCy (26)
1111' first of Eqs. ( 14), we obtain
-0.36 -0.48 -1
-cxCy -c; 1.36
0.48
0.48
0.64 -0.48 -0.64 0
0
0 2
For this type of element the 4 x 4 rotation-of-axes transformation matrix R is Ss = EA - 0.36 -0.48 0.36 0.48 0 0 3
L -0.48
(3 1a)
-0.64 0.48 1.64 0 -1 4
R= [~ ~] (27) -1 0 0 0 1 0 5

where the submatrix R can be written as 0 0 0 -1 0 6

2 3 4 5 6
R =[ l'x Cy] (28) lfr11rranging and partitioning this matrix in the form shown by Eq. (17) produces
-cy Cx

The latter matrix contains only the direction cosines of the member axis. Ss = [SFF
SRF
SFR
SRR
J
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: 1.36 -0.36 -0.48 ' 0.48 -1 0
'
0.36 0.48 -0.36 -0.48J -0.36 0.36 0.48 :-0.48 0 0 3
I

Ki = EA 0.48 0.64 -0.48 -0.64 2 EA -0.48 0.48 1.64 : -0.64 0 -1 4

= - -------------------------~------------------ (31b)
L -0.36 -0 .48 0.36 0.48 3 L 0.48 - 0.48 -0.64 : 0.64 0 0 2
[ '
-0.48 -0.64 0.48 0.64 4 -1 0 0 ' 0 1 0 I 5
0 0 -1 0 0 1 6
2 3 4
3 4 2 5 6

[ I 00 -]0 00 J wltl'rc the indexes down the right side and across the bottom show the rear-
K
2
= EA 0 2
(29) 11111gcment.
L -1 0 1 0 5 As with stiffnesses, we can generate the consistent-mass matrix M ' for
0 0 0 0 6 lrn·al axes of a typical member to be
2 5 6
2 0 1 OJ
0 3 M' = pAL O 2 0 1 (32)
K, -EA[~ L 0
1 0
0 0
-1OJ
0
4

5
6 1
[
0
0
1
2 0
0 2
0 -1 0 1 6 111 lhis instance, the terms in Eq. (3.4-5) for the axial element are repeated for
3 4 5 6 1hr 11' direction, because accelerations in that sense also give rise to inertial
,111ro11s. Thus, the consistent-mass matrix for inclined axes is of size 4 x 4. It
The terms in each of these matrices may be transferred to the appropriate 11111y he lrnnsformcd to global axes using Eq. (11). However, when that equation
locations in the structural stiffness matrix S, by culculating the following joint 1- 11pplil.'d, we !ind the resu lting matrix M to be the same as matrix M ' in
displacement indexes: ht ( \l). Thc1cforc , wt· rnnl'luclt· that for II plnne truss element the consistent-
JI )2 2/ 1111u111 111111tix ii! i11v111 lunl with rotation of uxcs.
J. I AJ. • 1.J. /\pply1111:, Eq . ( \l) 1111d Oil' <111111 ltu111 Tuhlt· \. I, we find the consistent-
104 I lnltn I lnmn1111 nnd Vlhr1tl11r111I A11nly•I• Chnp. J
11 111 Vlhrnllonnl A11nly1l1 105

mass matrices for the members to be

1111'11 111 lh1.· Joi111~ of 11 1111ss, tlwy 111ny IK· mld~·d directly to the diagonaLten:n_s.
M1 =M M2 = 0 .36M M, O.MM (JJ) j11 1·11lw1 M, o, M ,.
Al this point we shou ld observe that considering the masses for the three-
In these expressions the matrix M is the same as M' in Eq. (32). Terms in
1111111lll'1 truss lo be associated only with jQint translations is a poor approxi-
matrices M 1, M 2, and M3 may be transferred to the consistent-mass malrix M,
11111t1011 to the truth . In fcw.L. there often will be flexure in the membeis thl!.t
for the whole structure using again the displacement indexes j 1 through k2 .
hl~11if11.·1111tly influences the dynamic characteristics of the structure. For exam-
Thus , when M, is assembled in accordance with the second of Eqs. ( 14), we find
l'k, ti II time-varying force is applied in the x direction at the middle of member
that
I l'lt't' Jlig. 3. 1 l(a)I, the flexural deformation in that member will dominate the
2.72 0 1 0 0.36 0 11 ~p1111se. We shall deal with this type of problem for trusses by the component-

0 2.72 0 1 0 0.36 2 11111111• method in Sec. 10.6. In the meantime, we will use Ms or Me in the forms
0 3.28 0 0.64 0 3
l111w11 here, even though the numbers of members and joints are small.
M = pAL (34a)
s 6 0 1 0 3.28 0 0.64 4
0.36 0 0.64 0 2 0 5
0 0.36 0 0.64 0 2 6 It VIBRATIONAL ANALYSIS
2 3 4 5 6
111 ~kc. 3.5 we derived undamped equations of motion for free nodal displace-
Rearranging and partitioning this matrix as shown in Eq. (17) gives 111r11l'> of a discretized structure [see Eq. (3.5-19a)]. If there are no applied
,1111011s, these equations can be written in homogeneous form as
M, = [::: ::] MD+SD=O (1)
I

2.72 1 0 : 0 0.36 0 I lt•1l' the subscript F (for free displacements) is omitted to simplify the notation.
I

1 3.28 0 : 0 0.64 0 3 Equation (1) has a known solution [9] that may be stated as follows:
I

pAL O O 3.28 : 1 0 0.64 4 D; = 4>; sin (w;t + a;) (i = I, 2, ... , n) (2)

(34b)
- 6 0 0 1 :I 2.72 0 0.36 2
wh1•1c II is the number of degrees of freedom. In this harmonic expression , 4>;
0.36 0.64 0 : 0 I
2 0 5 1 11 vector of nodal amplitudes (the mode shape) for the ith mode of vibration.
0 0 0.64: 0.36 0 2 6 I h1 symbol w1 represents the angular frequency of mode i, and a; denotes the
3 4 2 5 6 11/t,111• a11gle. By differentiating Eq. (2) twice with respect to time t, we also find
111111
where the indexes at the right and bottom again show the pattern of rear-
rangement. D; = - wT4>; sin (w;t + a 1) (3)
Alternatively, the consistent mass matrix M, in Eq. (34b) could be re-
'111h1,litution of Eqs. (2) and (3) into Eq. (1) allows cancellation of the term
placed by a lumped-mass matrix Me, as follows:
111 (1,,,1 f- a1), which leaves
0.68 0 0 0 0 0
(4)
0 0.82 0 0 0 0 3
0 0 0.82: 0 I Ills ruanipulation has the effect of separating the variable time from those of
0 0 4
Me= pAL ------------ -- -----L----------------- (35) ,pun· , und we are left with a set of n homogeneous algebraic equations.
0 0 0 : 0.68 0 0 2
I
I Ll4uution (4) hus thc..form oLthe algebraic eigenvalue problem From the
0 0 0 ' 0 0.5 0 5 1hr111 y of homogeneous equations 1101, nontrivial solutions exist only if the
'
0 0 0 '' 0 0 0.5 6 d1'11•1111i11unt of the coefficient mutrix is equal to zero. Thus,
'
2
4
" JS cu f MJ 0 (5)
For L'IIL'h muss matrix M, and Ml', the totul muss of llll' trt1ss is 2p/\/, for hoth
I ,p1111:-;i1111 nl this d1•ll'll11i1111111 yi1• ld~ 11 pol y11n111111l of order II called the charac-
llw I and 11 dirL'rtions, us it shou ld hL· . ll 11011til1ur1u1.il nHIL'L'Utrntcd 11rnss1.·:,; also
1,·1/1ft11•1111,1f11111 'lh1• 1111111l~ ,o: ol llm poly1111111i11 l 1111• thl' 1·ht11'<11·1,·ri.1·fic values,
 !11 111 Vll11111l111111I A11nly• I• 107
100 11111111 ~ ln111nnt1 nnd Vll11 nll1111nl Annly•I• Ch11p. 3

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

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
The methods implied by Eqs. (5), (6), and (7) are conducive to hand
calculations for problems having small numbers of degrees of freedom. Exam-
ples at the end of this section demonstrate such calculations. However, a struc-
ture 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
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.~,
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
u - T(UTU - w7M u - u)<I>; = 01
, 1p1•11v11l11t' 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 fre-
1111r11t·yJ 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,.1•11vcctors have been found from Eq. (10), the angular frequencies and mode
h11pt·s (i n the original coordinates) can be determined as
(7)
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)
(8)
,loc~ 1101 necessarily preserve orthogonality among the eigenvectors, Instead, the
, l~l·nvcctors <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-1«1>uj = <l>~;<l>uj = 0 lli2
111 ~L'C. 4.2 we will show that the eigenvectors are also orthogonal with respect
111 111111rix M.
From vibrational analysis of structures, we may find repeated frequencies
(9) 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
1111ll•pendent 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 ,
m m
M u L c;<l>u; = · AmLC;<l>u; (15)
i=I i= I

wht•ic ci is a scalar multiplier of the ith modal vector and Am is the repeated
lkwriting terms in reverse order, we find that 1 lu1•11vulue. ll is always advantageous t0 form a new set of eigenvectors that are
(Mu - A;I)<l>u1 = 0 111 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
Wlll'll' 111 d1oil'es for nonunique orthogonal vcclors corresponding to a repeated eigen-
M11 U 'l'M U 1 \'llh11·. The (iram Sd1111idt 1>1·tlrow malizutio11 procedure [ 11] is a formal mathe-
,,,1 l llJ 11111l 1t·11 l 11pproad1 rn111111011ly usL'd lo l'o11st111c1 1111 orthogonal set of eigenvectors
1111111 11 l11ll'11rly i1uk•p1•11dt•11t Ht•I
Eq11ntio11 ( 10) is now in thl' sln11d111d , ~y 111111l•t1 ll' 1111111 of till' l'igl·11val11l'
p111hll 111 g1vt·11 hy 11q (H) TIil' 1111111h A i'4 ll'Pll'~t·11tt•d hy M11 ltlll' lu st of
0
II llll' 'tllflt1l'"" 11ml11~ ts s1:111tlk·hnill', tt rn1111ol hl' fnl'torl'd as dcslTibcd
108 1111110 1:-lomontN nnd Vlh111ll1111nl Annly•h• Chnp. 3 Soc. 3.0 Vlbrntlu1111I A1111ly11I• 109

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
1..,__ _ _ 0.6L----
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)

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).
i
I
Example 3 .1
I
I
Figure 3. l 2(a) shows a plane truss with two degrees of freedom at joint 1. Assume that I
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
I
structure. I
The 2 x 2 stiffness matrix S for the free displacements in this problem is I
S = EA [ 0.36 - 0.48] (a)
L - 0.48 1.64
(b) (c)
and the consistent-mass matrix has the diago nal form

M pAl, , .1 .2H O
(1 0 1,2 H
I (b)
Figure 3.12 (a) Plane truss; (b) mode l; (c) mode 2.

11 Wl' ll•I .I' !•:/\ / /, 1111d Ill \,} HriA I / 11 , lhl• l111111111•l' lll' IHI N l' lllllllillllN in 1-'.q. ('1) hl l' lH1W
1
110 Finite Elements and Vibrational Analysis Chap. 3 Sec. 3.6 Vibrational Ana lysis 111

y
0. 36s - mw'f -0.48s J[ <l>1;J [OJ (c)
[ -0.48s 1.64s - mw'f <P2; = 0
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 (d) G) 2 t 0
'

1~ • X

The roots of this quadratic equation are

o/.l
wf = 0.2-

from which we find the angular frequencies to be

m
s
w
2
2
= 1.s.:.
m
(e)

z
/ L

(a)

Wi = (oTs =
\j-;;-
0.6049
L
@.
\jp
W
2
= _ 1.815
- --
L
lo
-
p
(f)

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

q>I = [~] (g)

(b)

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

!001.uiue 3..Z
A t·nntilc 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 - 3L] Figure 3.13 (a) Cantilever beam; (b) mode l; (c) mode 2.
(h)
L 3 -3L 2L2
s s
Also, Eq. (3.4-26) gives the consistent-mass matrix as
wr = 0.02971 -
m
W2
2
= 2 .885 -
m
(C)

pAL [ 78
M, = 210 - Ill
- lll]
2L2
(i)
' uhstituting the known values of sand m and taking square roots , we find the angular
li1•quc ncics, as follows: ·

Lettings 2EI /C and m = pAL /2 10, we substitute matrices S and M, into Eq . (4) to _ 3.533 /El Wz = 34.81 /El
uhlul n the ho mogeneous equations W1 - ~ \/pA L2 \jpi4. (m)

r 6(s - 13mw;) -L(3s - t tmwn ] [ <1>11] = [OJ (j)

Wl11111 these formulas arc compared with exact values [9] , the errors are found to be
I L (3.1· - I lmwr) 2L2(s mwr) <1>21 0
, 1 I 0.48% and e 2 = +58%. Thus , the first-mode frequency is a good approxi-
111111 hu1, hut the second-mode frequency is very poor.
whc1l' Ilic 2 x 2 cocfli cicnt matrix is the c hnructeristic matri x 11,. As indicated by In this cxumplc we obtain the mode shapes by using the first column of the adjoint
l\q. (~). Wl' Sl't tlw dctc1111in11nt ol' 11, cqunl to Zl'l'O , produci ng II ,', which is
. l
11111t ilx

( k) 21.2(.1· lll(rJl )
whil'h IK 1111' 1 l1111 m h·iiNll1 1•q1111111111 'lh,•11 1hr 1111 ,tM111 l 1
q (~ ) 111 1• lrn1111l to lw
II '/,
I /, ( \,\ I I 1111,,f )
(n)
112 f lnltu I l11111nn1 11 nncl Vih111tl111111I A11nly1l1 Chnp. 3 !iy111111111tlr 111111 A11t l1ym111nltlt: Mod111 113

Substitution of cu ~ and wl from Eqs . (f) into thi s column yil'ld~

(1)1 = L ] (()) I
[ 1.378
1- sym.
Of course, the second column of Hf would serve equally well. Figures 3. l 3(b) and (<:) I
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
y
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 I
7,i 77, - X 7} 7,7, 7,i 7,7, 7,i 77, 7} 7,7,
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
(a)
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 /
/--sym.
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 (bl
small slashes on their arrows. In general , the component of nodal translati<)Jl 1,·11111n1 J.14 Sym111c1ric i,11 u~·tl11~·s: (u) frumc; (b) plate.
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 p11ltt·1n of distortion.
 114 Finite Elements end Vlbrntlonol Analysis Chap.3 n, 3.7 Symmetric and Antisymmetric Modes 115

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

y \·111111ctry, the nodal displacements, strains, stresses, and reactions will also be

/A
I
t 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
Plane of /
I t ,Ii l111111ation. That is, the translations in the x direction and the rotations in the

k·--
symmetry~
1· 1111d z senses are all in positive directions at both points k and k'. From this we
// I , 1111dude that the same displacements on the plane of symmetry must be allowed
/ +5 I /I h I nl'nir freely. They are the displacements labeled j I, j 5, and j 6 at point j. On
/ -t.1 1/ I
/
( ,· t
1111 olhcr hand , the translations in they and z directions and the rotations in the
Y,k,
j2 I / Y;, , 1•11sc are in opposite directions at point k and k'. Therefore, the same displace-
r----~-~------+--- • ~ -1 111r 11ts on the plane of symmetry need to be set equal to zero. This may be
I/
V z1,,
I
/.
Jl'J3 i1i4)
/
//
V z1• ,11111111plished by introducing restraints corresponding to }2, }3, and }4, as
/ I/ i6 / / 111d1t·ntcd by the slashes on their vectors. In summary, the components of nodal
z /4----x ,,-4
1

, I /
/L:. ---y'
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\ Pl'CI to the plane.
I / 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
(a) \ 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
y ,llvldt it into two equal parts and introduce new nodes on the bisecting plane th.at
llll H'Strained as described aboYe.,

/,1
I
t I , 11111plc 3.3
Plane of ---. /
t
-·- I
7 I 11•1111• \. I 6(a) shows a simply supported beam composed of two flexural elements. This
k'i/ symmetry
// I k•-- 111 11111 hus four degrees of freedom and is symmetric with respect to its centerline, as

/i / t i5 I / I 1111111 ntt·d 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
I Yw
r-----+--r-----~--. -
(/ j tT
j2
1/ I
I I Y;,
~ - ---+----,- -- ~ ·
I 11•1111• 1. 16(b) illustrates the symmetric case, for which the rotation D 3 in the plane of
11111111•1ty 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
/.i1 i4) '/
I ,/ 1 3 / I , z1, ", lluw only two degrees of freedom instead of the four degrees of freedom in the
I , j6
7/:.~ / V 1111.,111111 problem. Now let us find two angular frequencies and mode shapes from each
111 1111• NUhsidiary problems, using translational consistent mass terms.
z 7~/ Fm lhc symmetric case in Fig. 3. 16(b), the 2 x 2 stiffness matrix for the free
/ I / 1ll•11l11n•mcnts (D2 and D4) is
1//
v s=7
2£1 [ 6 3e
3e u 2
J (a)

(b)
\\hlo h INtl111w11 from Eq. (3.4-24). Also, the 2 x 2 consistent-mass matrix becomes
Figure 3.15 Restraints on planes of sym111ct1 y: (11) syn11m·1ric modes; (b) anti·
symmetric mmks.
M,
r>A
420
fr 156
I H'
l.3el
4f2 (b)

" 11v, 11 hy hi ( I •I )(!) Thl' l h11111l 'h•11NI 11 1111111 ix II, tnr 1hi~ L'IISL' hns 1lw I'm m
 Symmetric and Antisymmetric Modes 117
116 Finite Elements and Vibrational Analysis Chap. 3 Sec. 3.7

2 s (e)
WT = 0.007305 !_ W3 = 0.9026 -
m
m
y
Substituting the known values of s and m and taking square roots , we find that
I
Wt = 9.909 (ii W = 110.} (ii (f)
, CD Ja, 0 3
l2 \JpA 3
l2 \JpA
D'/4 DJ/~ D 4/ ~ -
. --x 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%.
/Ji-, e--------~ e~ , j We obtain mode shapes corresponding to w, and W 3 using the first column of the
adjoint matrix Hf, as follows:
z
(a) H1 = [
2
2£ (s - 2mwT) J (g)
' -£(3s + 13mwf)
Substitution of 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

Mode 3 lb)
S 2£/e [2 1]
=
1 2
(i)

1111d the consistent mass matrix is

3
_ pAC [ 4 (j)
M, - 420 -3

Mode 2 I hl.ln the characteristic matrix becomes

2 3
H, = [2(s - 2mwf) s + 3mw f J (k)
s + 3mw f 2(s - 2mwf)
whl·re s and m are the same as before. Expanding the determinant of H, and setting it
-.....::_:::;.. 1 qunl to zero gives the characteristic equation

Mode 4 (c) 1m 2 wt - 22mswf + 3s 2 = 0 (£)

Figure 3.16 (a) Symmetric beam; (b) symmetric modes; (c) antisymmetric modes.

l
11111 11 which the roots are

2 [ 6(s - 26mwn e(3s + I 3mwl) 2 1 s 2

=
s
3- (m)
HI = s- w, M, = (l' ) W2 = - -
1m
W4
m
e(3s + l 3111c,l ) ze 2
(s 2mwn
e nnd 111
1
pA (' / 420 . /'k'lling thl· dctc, mi111111t or 11111tri x 11 1 equal to ~.l'1t1 111m 'l'l 0
ding us before, we find the angular frequencies to be
where .1·
giws
2/i//

•l"l ',111 ' 111,' ,ll ,11111111; I I.I ~ 0 (d)

,,, 1
43. 82
/,J
J/i/
11A
(1) ,1
2001.8
I,
J!J·I
11/\
(n)

Whh h I" 1111 1 h111 111 h il Mih 1·q 1111tl1111 1111 11111(~ 111 l'q (ti) 111r lt11 w li1d1 th1• 1·11111 ~ 111r I 11 'Y., 111111 I J. / 17,,
118 Finite Elem ents and Vibrational Analysis Chap. 3 Program VIB for Vibrational Analysis 119

Mode shapes are given by the first column of H 1, which is Flowchart 3.1 Main program for VIB*

a
Hi; -
_ [ 2(s - 2mwt) J (o)
IOI
I. Read and write
-(s + 3mwt)
SDATt structural data.
Substituting w~ and w~ from Eqs. (m) into this vector yields

(p) 2. Generate structural

stiffness matrix.

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.
3. Generate structural
The error calculated for the first angular frequency w1 is acceptable, while those
consistent mass matrix.
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.
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 5. Solve eigenvalue problem
analysis of any type of linearly elastic framed structure or discretized continuum. by Sec. B.2.
Steps in the main program appear in Flowchart 3. 1, which calls seven sub-
programs indicated by the names in double boxes. Subprogram SDATreads and
6. Transfom1 eigenvectors
writes input data for a particular type of structure and calculates nodal displace- to original coordinates.
ment indexes. Subprogram STIF generates the structural stiffness matrix (for
free nodal displacements only) by assembling contributions from element stiff-
nesses, as indicated by the first of Eqs. (3.5-14). Next, the consistent mass 7. Write angular frequencies
matrix for free nodal accelerations in the structure is assembled by Subprogram and mode shapes.
CMAS, using contributions from individual elements [see the second of Eqs.
(3.5-14)] . The subprogram named STASYM then converts the eigenvalue prob-
lem 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
eEND
Go to IOI and process
another structure.

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 struc-
tures 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'
*Applies to any type of linearly elastic structure.
t Subprograms that differ for every type of structure.
1111 N, 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 Appl•ndix 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 d1•11sily 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
 l'royrnm Vll3 lo, Vlbrn1lo1111I A11ulysl11 121
120 Flnlto El omonts 1111tl Vlhr 11tlo1111I Arw lysls Chop. 3

Type of Data

Problem identification
TABLE 3.2 Structural Data for Phmc Trusses

No. of
Lines

I
Items on Data Lines

Descriptive title
,_T
t4 3

Structural parameters I NN ,NE, NRN, E, RHO

O.Bl
Plane truss data

L
(a) Nodal coordinates NN J, X(J), Y(J)
(b) Element information NE I, JN(T) , KN(I), AX(l)
(c) Nodal restraints NRN J, NRL(2J-l), NRL(2J)

Y(J). The element information (NE lines) consists of the element number I, the
0
• t J - - -- - - - - - 0 ~
~r 5
X

j node JN(I) at one end , the k node KN(I) at the other end , and the cross-sectional ~l~ /
area AX(I) .
Each of the NRN lines in the last block of data contains a node number J \- ~ 0.6l~
and two code numbers which indicate the conditions of restraint at that node. ThL· (b)
(a)
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, _l_
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.
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
(c) (d)
the following numerical values are assigned to the parameters E, p, L, and A:
E = 3.0 x 104 k/ in. 2 p = 7.35 x 10 7
k-s2 / in.4 l•'l1111rc 3. 17 (a) Plane truss; (b) mode l ; (c) mode 2; (d) mode 3.

L = 250 in . A 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'0111p11lt•1 AIN11 co111p11ll'd 1111d prinll'd
lll l' till' t'lt•1t \l'III l1•1111th11 l•l • 1111' d11 1•1 l11 111 l m1 111· ( ·x 111111 ( 'Y, 1111· 11111111w, 111 \lt'j,\ll'\'S 111
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 lh•qucnc: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 ~ hl·t·11 ,1111 1111tli:n·d with H'N(ll'l I 111 its l11rgcs1 term, and the mode shapes
11• d1 ph lrd Ill l•tp \ I l(h) (d )
 123
Program VIB for Vibrational Analysis
122 Finite Elements and Vibrati,Jnal Ana lysis Chap.3
TABLE 3.4 (Continued) ·
TABLE 3.3 Structural Data for Example 3.4
MODE 2
Type of Data Alphanumerical Values ANGULAR FREQUENCY 1.1677E+03
NODE DJl DJ2
1 8 . 6725E-01 O.OOOOE+OO
Problem identification Example 3.4: Three-member plane truss l.OOOOE+OO
300(().0 0.000000735
2 -l.7149E-01
Structural parameters 3 3 2 3 O.OOOOE+OO O.OOOOE+OO
- Plane truss data MODE 3
(a) Nodal coordinates I 0.0 0.0 ANGULAR FREQUENCY l.8618E+03
NODE DJl DJ2
2 150.0 200.0
1 1. OOOOE+OO O.OOOOE+OO
3 150.0 0.0
2 -6.0504E-01 -6.1068E-01
3 O.OOOOE+OO O.OOOOE+OO
(b) Element information I I 2 JO.O
2 I 3 6.0
3 2 3 8.0 K11.11mple 3.5
1<1 urc 3.18(a) shows a plane truss with a larger number of members. The cross-sectional
(c) Nodal restraints I 0 I 11
111• s of diagonal members are equal to I .SA , and those of other members are equal to
3 I I 1 11
\ 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 sym-
1
1111,11y for symmetric and antisymmetric deformations, as indicated in Figs. 3.18(b) and
TABLE 3.4 Computer Output for Example 3.4 ti ), In addition, the cross-sectional area of member 12 is divided by 2 . Assuming that
PROGRAM VIBPT y

*** EXAMPLE 3.4: THREE-MEMBER PLANE TRUSS***

STRUCTURAL PARAMETERS
NN NE NRN E RHO
3 3 2 3.0000E+04 7,3500E-07
NODAL COORDINATES
NODE X y
l 0.000 0 . 000 - - x
2 150.000 200.000
3 150.000 0.000
ELElMENT INFORMATION
ELEM. J K AX EL ex CY
1 1 2 10 . 0000 250.0000 0.6000 0.8000 (al
2 1 3 6.0000 150 .0000 1 .0000 0.0000
3 2 3 8 ..oo 00 200.0000 0. 0000 -1. 0000

NODAL RESTRAINTS
NODE NRl NR2
1 0 1
3 1 1
NUMBER OF DEGREES OF FREEDOM: NDF • 3
NUMBER OF NODAL RESTRAINTS: NNR • 3

MODE 1
ANGU~AR FRIDQUENCY 4.1995m~o2
NODIC OJ l D,12
I 2.J137lt 01 o. 0000,c, oo
2 I • OOOOICi 00 2, 472 21C 0 1 1"1)l1111• ,\. Ill (11) 1'111111• t111NN; (h) 111111!1- 1 (Nyn1111cttlc); (c) mode 2 (antisymmetric).
I 0,000!llCtOO o. ooomc, no
 124 11111111 I l11mn11t N1111d Vlh1ntl111111I A1111IYNh1 126
t hnp I I 11ol!ln111a

11 l lrrirniltl•,iko , S. t>., Yo11111-1, I>. 11. , 111111 Wl' IIVN, W, , J1., , Vl/m 1tion Problems in
-- ~-- 1,:11Hl1w1•rl11H, 4 th ed ., Wil ey, New York , 1974 .
It) <li•ll', .J. M., nnt I W . r W
cave , ., ·• Jr Mlllri.A- Algebra f or
.
Engmeers, 2nd ed.'
"" ll " \. llrooks /Co lc, Monterey, Calif., 1983. . Y k 1973
11 1lohn ' E E. ' W eme11tary Matrix Algebra , 3rd ed .' Macm1llan, New or .' .
l ", I
---~-- 'I ·k ' . p
I l ( t oc 110 1, , • , G "Symmetry
1111 , No . ST! , 1973, pp. 71- 89.
in Structural Mechanics," ASCE, J . Struct. Dzv., Vol.

(c)

Figure 3.18 (cont.)

the truss is aluminum , we give parameters the following numerical values: PROBLEMS
E = 6.9 X 107 kPa p = 2.62 Mg/ m 3
F P3 4-1 shows an axial element with a linearly distribut~d load (fo~ce per
l=5m A = 6 X 10- 3 m2 u~~~~; ngth) given by the formula bx = b_i : (h2 - bi)x/ L. Fmd the eqmvalent
where SI units are implied (again, see Appendix A). nodal loads Pb(t) = {pbi' Pb2} due to this mfluence. f" [
f' {. 1; [ ,-{ {_
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 ~ b , •. f,~ , L-X xJ
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 ' ; [ 1/ '<"'L' ri.•:.,
mode is symmetric with respect to the plane of symmetry, while the second mode is
antisymmetric.
- I--------1 rt [ rh,.
p,,,
.,
b,
L
Pb,
l\,f)"
o
1,~'(

-
(kb,)
"
L [(L-X) l \,,+ li,._ -~')-;:-
Figure P3.4-1 =-1 ). ~ ( 1,i-1-( b, - L,) I:_
REFERENCES , .... 2. 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
I. Weaver, W. , Jr. , and Johnston, P. R. , Finite Elements for Structural Analysi.1·, loads Pb(t) = {pb,, Pb2} resulting from this body force.
Prentice-Hall, Englewood Cliffs, N.J., 1984.
2. Zienkiewicz , 0. C., The Finite Element Method, 4th ed. , McGraw-Hill, Maiden-
head, Berkshire , England, 1987.
3. Cook, R. D., Concepts and Applications of Finite Element Analy sis, 2nd ed., Wiley,
New York, 1981.
1 ~b~--x,u
4. Timoshenko, S. P., and Goodier, J. N., Theory of Elasticity, 3rd ed., McGraw-Hi ll ,
New York , 1970.
5. Weaver, W., Jr., and Gere, J.M., Matrix Analysis of Framed Structures, 2nd ed .,
Van Nostrand Reinhold , New York, 1980.
p:--i.-,,-~---.1 p.,

6. Oden, J. T., Mechanics of Elastic Structures, McGraw-Hill , New York, 1967. Figure P3.4-2
7. Archer, J . S., "Consistent Matrix Formulations for Structural Analysis Using Finite-
Element Techniques," AJAA J., Vol. 3, No. 10, 1965 , pp. 1910- 19 18.
.,.....,. " 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;
8. Clough , R . W. , "Analysis of Structural Vibrations and Dynamic Response," Rel'. .I . · p·irt• (b) (d) of the figure are. f1 -
Adv. Mat. M ethodsStruct . Anal. Dt•s., ed . R. 11. Gullugher, Y. Ynrnadu, a nd J. ·1: Jishape(I} uncti4.,ons
i )/Im ' "
}, and.{, (2.i I /,) ,1· / / '..
2
0 cn·vc thc 3 x 3 stiffness matri
Oden, University of Alabama Press, llunt svillc, Alu ., 197 1, pp. 25 45. 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 1•111hl11111 11 127

--
q,
L
2
2

./
I . ~q,
L
2
3

- Q3
----x, u

~ · ~ - - - - --x,IJ,
/: L .I /1 111 , - m,
'
- - - - -
L------+l
P.,

,K
(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)

-M. 2
-
P11,
-----+--------- - ------ 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
element with three nodes [Fig. P3.4-3(a)]. Find the equivalent nodal loads Figure P3.4-8
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
1.4-1), Figure P3.4-9 depicts a concentrated force Py and a concentrated moment M,
loads p1, (t) {P111, p1,2} caused by a puruholicully dislribulecl moment (per uni I nppl icd to II flexural elcmcnl at the distance x rrom node 1. Obtain the equivalent
length) given ns 111, 111,d I (l'/ 1.)2 J, 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 129
l 111>l>lt1111"

y, V y

t
I,
P", t pr t PoJ
I, 0.8LI

r
/

;·/1:
1 2 2
X
CD
x M,/,1
P-~, II II
CD
0.6L
y

l,
L 2
z X

'/
Figure P3.4-9
I 0.8L
X

3.4-10. In Fig. P3 .4-10 a linearly distributed load by = b1 + (b 2 - b1)x/L (force per Figure P3.5-1 Figure P3.5-2
unit length) acts on a flexural element. Find the equivalent nodal loads
Ph (t) = {Phi, Pb2, P&3, Pb4} due to this influence .
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.
y,v
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.

I, CD

r
2

L
0 0

l
Figure P3.4-10

3.4-11 . Rederive the 2 x 2 stiffness matrix K for a torsional element, using moment M,
3 0 X

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
L
·I
be constant along the length. Figure P3.5-3

3.S-1. The plane truss shown in Fig. P3.5- I has cross-sectional areas of 0.6A and A for l.~-4. Ropcut Prob. 3.5- 1 for the plane truss shown in Fig. P3.5-4. Assume that the
members I and 2. For thi s structu re !incl the stiffness mutri x S.,· and the consistent cmss sectional arcus of mc1nhers I 11nd 2 arc 0.8A, those of members 3 and 4
mass matrix M, in rourranged und pnrtit ioncd f'onns. 111'l' 0 .6/\. und th11t or 111c111lwr ~ Is l'qunl to/\ .
 130 1lnltn ~ lnt11n11t• nnd Vlhr11tl1111nl A11n ly1l11 l'rohlnm• 131
Glinp. 3

y l•'1gtlll' 11 I..~ I NhowNn two d ct11lllll ht•1111 1 Im whk'h lhc pnrnmctcrs E, I , A, and
11111c l'OIINtn11l 11ln11g the length . Asscnthle the stiffness matrix S., and the consis-
1 I CD
o----_..;-----o 2
ll•nt-11111ss mutrix M, (for trunslutionul inertias) in rearranged and partitioned
forms .

I
0.6L
0
y

L 4.
2 0
~
3
X

,. ~

Figure P3.5-4
,/ ~l•- ---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 Repeat Prob. 3.5-7 for the two-element beam shown in Fig . P3.5-8 .
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.
y

y
y
G) 0
I
,I CD 1~

~
2
~3 X

r
0.6L
/,
0 2

G
rL
©
z
/ ~L
Figure P3.5-8
L~
L
j. O.BL
4
- -x L 3~L
- -x
For the two-element beam shown in Fig. P3.5-9, repeat Prob. 3.5-7.

Figure P3.5-5 Figure P3.S-6 1

•
G) 2 0 ~3 X

4
3.5-6. 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.
, / l I
FIMlll'C PJ.5-9
l~
132 Finl to rlum1111t" 111111 Vll11 nllorml A1111ly11l11 Chnp. 3 l'whlntn• 133

3.5-10. Repeat Prob. 3 .5-7 for the two-element beum show11 In Fig. P3.5- 10. Rl•twnt Prnh. \.6· 2 for the pl1tlll' 11 II NN show11 In Fig. P3 .6-4 , but let the cross-

y
,~
Nl'Clionul ureas of mcmbcrs I and 2 hc l'qunl to 0.8A and A.

0.8L- - --

,/~,
1~ G)
~
2 0 3
• X
"l
3.6-1.
Figure P3.5-10
l~

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
J1'l1111rc P3.6-4

1.,,-s.
L~ I
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
of Eq. (3.6-8). Also, show the back-transformation of eigenvectors to the cross-sectional area equal to I and A, but member 2 has 2/ and 2A for its
original coordinates. properties. Determine the angular frequencies and mode shapes for this beam
3.6-2. Figure P3.6-2 shows a plane truss with two degrees of freedom at joint I. using translational consistent-mass terms.
Assume that the cross-sectional areas of members l and 2 are equal to A and
y
A V2. By the consistent-mass method, find the angular frequencies and mode
shapes for this structure.

CD 2
•
0
[, 1~ ~3 X
/, A 2/, 2A

/ .1.

r
~L L~
z

l Figure P3.6-5

L~_,.
.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

j degrees of freedom (at points 2 and 3).

y

.-- L--•I 1
i-,,, - - -- 0.8L - ---....i

Figure P3.6-2 Figure P3.6-3

3.6-3. 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 .
134 11111111 1111111111111111111 Vll11 11ll1111nl A1111ly1l1 Ch11p J 1:

3.6-7. Repeat Prob. 3.6 5 for the two clement co11tin11011Nlw11111 shown in Fig. PU, l
"''' l'rnhl"'"'

\, I, hu thl• Ny1111m•tric rnnti11uo11s IK·11111 shown in Fig. P3.7-3, find the angul.
Cross-sectional properties I and A arc constunt , 1111d the free displacc111cnh 111 tll'(llll' lldcs and mode shapes for (11) symmetric and (b) antisymmetric di
joints l and 2 are both rotational . hu lions. Use onl y half the structure, assuming that each of the four elements h,
y tht· snn1c v11 lucs of I und A.

, _____;:CD:::;.____2_ _ _ _0.;:;2:;.__ _ _~ _3 - - x
#i ~ I
z / ~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' Figure P3. 7 -3
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 fo
I members I and 2 equal to V2 A , whereas those for members 3, 4, and 5 arc
i-Sym. equal to A. Using only half the structure, calculate the angular frequencies anc
the (a) symmetric and (b) antisymmetric mode shapes.
1~
CD 2• 0 3~ X
I
., r-Sym.
I

z
/1· e
·I·
e L= 2e

3.7-2.
Figure P3.7-l

Figure P3.7-2 shows a prismatic continuous beam composeq of four flexural

y
-1 L
elements. Determine the angular frequencies and mode shapes for (a) symmetric
and (b) antisymmetric deformations, using only half the beam.
y
,-:2@ (Ds_l_
/J ' > o - - - - - . = - - - ~ - - - - o ( ' v ' .
/
X

r-Sym. ~ L - 1 4 L~

,~ CD 2 0 31 ® 4 © ~s
~ ~ ~
Figure P3.7-4
I \. - x
z / l-e ,I, e+e ,J, 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
Figure P3. 7-2 lo/\ ,
 136 l 1111 l>lt11 1111 13,
Fl11 lto l: lomo11t11 1111cl Vll>111l lo1 111I A11uly11lo Chop. 3

r y(Sym .)

"ro6L-j-o6L--::i
CD , 0
2

'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 (a)
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.

rSym.

1 I
y

f
0.6L
0
l.---=----0---=--------G ____.__l- x
3
-
-• I - o.sL- -....,, j,_.--o.sL------j

Figure P3.7-6 (b) (cl

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.: trenlL•d as 111 igid body.)
4 n, , 11.:l l'r lrwlpnl n11d Nor mnl C:111111tl11ntn11 139

111rnl(•S of vihmtion . 'l'hcrcfme, such a model has only a finite number of vi-
l111111011ul 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 coor-
dl11111cs . 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
Normal-Mode Method hr II system with only one degree of freedo~. Superposition of. ~ese SDO~
1, "ults is accomplished through a transformation back t~ the ongmal .coordi-

of Dynamic Analysis uules. By this means we can evaluate time-vW?'ing nodal displacements, mtemal
lilH•sscs, 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 re-
q111rcd 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
trt•hnique, known as modal truncation, can make the normal-i:node.method more
4.1 INTRODUCTION dllcicnt than the numerical integration methods to be descnbed m Chapter 5.

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 4,2 PRINCIPAL AND NORMAL COORDINATES
analysis can be performed by solving the algebraic eigenvalue problem associ-
ated with the homogeneous equations. Extracting the angular frequencies and 111 order to study relationships among the natural modes o~ vibratio~, let us
mode shapes in this manner sets the stage for the normal-mode method of 11 msider modes i and j of the eigenvalue problem for action equat10ns, as
dynamic analysis [l], described in the present chapter.
lollows:
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 s
<I>; = WT M <I>; (la)
(immobile), the response will involve only flexible-body motions. However, if S <1>1 = wJ M <1>1 (lb)
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 l'lwsc expressions are modified versions of the algebraic eigenvalue problem
motions may consist of small or large translations but only .smaluQtations . Thus, µlwn previously in Eq. (3.6-4). Premultiplication ~f Eq. (la) by <l>J and post-
problems involving large rigid-body rotations are beyond the scope of this 11111ltiplication of the transpose of Eq; (lb) by <I>; yields
theory. <l>JS <I>; = wf<l>JM <I>; (2a)
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 ~S~=~~M~ ~
natural modes of vibration. If the structure were analyzed as an elastic con- I hl· lcfl hand sides of Eqs. (2) are equal, so that subtraction of the second
tinuum, its flexible-body response to dynamic loads would consist of the sum of 1·q1111lion from the first produces the relationship
an infinite number of vibrational motions . However, if the structure is dis
cretized by the finite-element method, the resulting analytical model will have (wr - wJ) (l>T M <I>, 0 (3)
only a finite number of nodal degrees of freedom and a finite number of nnturul 1111 the other hand , if Wl' divide holh silks of' Eq . (2a) by wr and both sides of

138
140 Normal-Mode Method of Dynamic Analysis Chap. 4 Sec. 4.2 Principal and Normal Coordinates 141

Eq. (2b) by wJ, the right-hand sides become equal. Then subtraction gives Premultiplication of this equation by <f>T and insertion of I = <I> <1>- 1 before D
and D produces
( _!_ - _l)
Wf WJ
<I> TS <I>1 = 0
J
(4)

*
To satisfy Eqs. (3) and (4) when i j and the eigenvalues are distinct which can be restated as
*
(wr wJ), the following relationships must hold: (13)

<l>JM <I>; = <l>;M <l>i = 0 (5) 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
and
the latter equation are defined to be
<l>JS <I>; = <l>;S <l>i = O (6) DP = <1>- 1 D Op = <1>- 1 I) (14)
These expressions represent orthogonality relationships between the modal vec- Ihe generalized displacements De given by the first of Eqs. (14) are called
tors <I>; and <l>i. From Eq. (5) we see that the eigenvectors are orthogonal with Jl.rincipal coordinates, for which the equations of motion [Eq. (13)] have neither
respect to the mass matrix M. Equation (6) also shows that they are orthogonal inertial nor elasticity coupling.. From Eqs. (14) we find that the displacements
with respect to the stiffness matrix S, as demonstrated in Sec. 3.6. and accelerations in the original coordinates are related to those in principal
For the case when i = j, Eqs. (3) and (4) yield
coordinates as follows:
(7)
.. ..
D=<I> DP (15)
and
Here we see that the generalized displacements in vector DP operate as multi-
(8) pliers 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
in which Mp; and Sp; are constants that depend on how the eigenvector <I>; is
system are its natural modes of vibration.
normalized.
We now restate the eigenvalue problem in Eq. (la) more comprehen-
For operational efficiency, we place all of the eigenvectors columnwise
sively as
into an n x n modal matrix of the form
S <I> = M <I> w 2 (16)
<I>,.] (9) in which the modal matrix <I> is given by Eq. (9). The symbol w 2 in Eq. (16)
where n is the number of degrees of freedom. Then we can state Eqs. (5) and represents a diagonal matrix with values of wf in diagonal positions, as follows:
(7) collectively as
wf 0 0 0
(10) 0 (J)~ 0 0
in which MP is a diagonal array that will be referred to as a ,vrinciDal mass w2 = 0 0 W5 0 Ll1l
matrix. Similarly , Eqs. (6) and (8) are combined into • •
<f>T s <I> = Sp (11) 0 0 0 (J)~

where SP is another diagonal array that will be called a principal stiffness matrix. This array, sometimes caHed the spectral matrix, will be referred to as the
Equations (10) and (11) represent diagonalization of matrices Mand S. If either eigenvalue matrix, or matrix of characteristic values. It postmultiplies the ma-
of them is already diagonal, the operations merely scale the values on the trix <I> in Eq. ( 16), so that a typical modal column <I>; is scaled by the correspond-
diagonal. ing eigenvalue wt. Premultiplying Eq. (16) by <l>T and using the relationships
To take advantage of the diagonalization process, let us reconsider the ( 10) and (11) , we obtain
action equations of motion for free vibrations of an undamped MDOF system, (18a)
as follows:
llcncc,
MDISI> 0 ( 12) ( 18b)
 143
142 Normal-Mode Method of Dynamic Analysis Chap. 4 Sec. 4.2 Principal and Normal Coordinates

Thus, in principal coordinates the ith principal stiffness is equal to the ith normal coordinates. From Eqs. (14) , these vectors are related to their counter-
principal mass multiplied by the ith eigenvalue. parts in physical coordinates by
Because the modal vectors may be scaled arbitrarily, the principal coordi- DN = <l>N1D DN= <l>N D 1
(25)
nates are not unique. In fact, there is an infinite number of sets of such gener-
alized displacements, but the most common choice is that for which the mass and the reverse transformations are
matrix is transformed to the identity matrix. We state this condition by speci- (26)
fying that Mp; in Eq. (7) must be equal to unity, as follows:
The inverse of the normalized modal matrix required in ~~s. (25) ~ay be
(19) easily found . We need only postmultiply Eq. (22) by <l>N to obtam the
where relationship
(27)
(20)
Thus, the desired inverse may always be calculated by this simple ~atrix
Under this condition, the scaled eigenvector <l>N; is said to be normalized with multiplication. If we wish to include only a selected number of modes mm our
respect to the mass matrix. The constant C; in Eq. (20) is computed as
analysis, the modal matrix will be of size n x m. Then Eqs . (25) must be
restated as
C; = ±V<l>TM <I>; = ± i <l>j;(i Mjk<I>)1
j= I k= I
(21a)
(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)
j= I
Example 4.1
When all of the vectors in the modal matrix are normalized in this manner, To demonstrate the use of normal coordinates for action equations, let us reconsider the
we use the subscript N and revise Eq. ( 10) to become 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
<1>iM <l>N = MN = I (22)
Pig. 4. l(a) is
Thus, the principal mass matrix is now the identity matrix. Furthermore, the 1.36 -0.36
EA - 0.48]
principal stiffness matrix, from Eqs. (11) and (18a), is seen to be 0 .36 0.48 (a)
S= - -0.36
(23a) L [ -0.48 0.48 1.64
Or, for the ith mode, In addition the consistent mass matrix from Eq. (3.5-34b) is
'

~J
(23b) 2 2
Therefore, when the eigenvectors are normalized with respect to M, the stiff- M = pAL[
6
·; 3.~8 (b)

nesses in principal coordinates are equal to the eigenvalues. This particular set o O 3.28
of principal coordinates is known as normal coordinates. Rewriting Eq. (13) in hu this structure, three eigenvectors were calculated in Example 3.4 (see Table 3.4).
normal coordinates, we have I hl'NC vectors become the columns of the modal matrix, as follows:
(24a)
0.2314 0.8673 l.OOOOJ
(c)

r
or <I> 1.0000 0 . 171 5 - 0 .6050
• O.2472 I.()()()() 0 .6 107
111 1111111111 11,l· thii. 111 111y wtth ll'S \ll' l' l to thr 11111HH 111nt11x, Wl' compute from Eq. (2 1a) the
144 Nn111111I Mnd11 M1111t111l 11f 11y11111 11lc l\1111lynlt1 Ll111p. 11 l'1l11nlpnl 111111 N,11111111 < 11111dlt111l n• 146

, 1111~llllllN

C2 0.9242\/pAL C3 = 0.8097\/pAL (d)

I ll vldl11g the columns o f' (J> by these values produces
F,, -, ,.
0.2803 0.9384 1.2350]
)( 1
<I>N = - - 1. 2114 - 0.1856 -0.7472
I [ (e)
/ \/pAL - 0.2995 1.0820 -0.7542
/
/ 111hNlilution of Eqs. (a) and (e) into Eq. (23a) yields the result
/
/ 0.2701 0 0 ]
/ SN = <l>;!::S <l>N = :2 [ 0 2.088 0 = w2 (f)
/ p O O 5.308
I
/ " l1 k h contains the values of wr, wi, and w~on the diagonal. Their square roots are the
2.__io----~~~ 111111• 11s 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
- - - 0.6 L - - - F,, -J I-- , ,iii 11lutions in later sections.
(a)
lb)
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

-'.j+I
1111111011 expressed by Eq. (12) may be replaced by displacement equations oJ.
11tr>tiulL, as follows:

I F,, ~ FMD + D=O (29)

/r ----+-1-
r,
F 32

i1 r ,. 111 whi ch

F = s-1 (30)
I
I
I
I I lit•.flexibility matrix F contains values of free nodal displacements due to unit
I
/ I v,tlucs of the corresponding actions. We tr.ansforrn Eq. (29) to principal coordi-
/ I 1111lt•s by substituting Eqs. (15) for D and D. Then premultiplication by <1>- 1 and
I I hlMt' I tion of the identity matrix I = «1>- Tcl>T before M produces
/
/ I «1>- I F «1>- Tcl>T M cl> Op + Op = 0
'/ I
whk h can be rewritten as
(31)
I II
1lw symbol Fp in Eq. (31) represents & orincioal flexibility maJ.Jix., correspond-
(c)
Id)
l11p to S,,, and is defined as

l'.'lgurc 4.1 (a) Plane truss; (b) condition A, = I; (c) condit ion A2 = J; (d) condi- Fp = <I> 'F<I> T = Sr' (32)
l10n 111 I.
111 ,·m11·sc, 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
 Sec. 4.3 Normal-Mode Response to Initial Conditions 147
146 Normal-Mode Method of Dynamic Analysis Chap. 4

replaced by
This diagonal array contains the values of A, = 1/ wf, A2 = l I wt and .\3 = 1/wt as
confirmed by Eq. (f) in Example 4.1.
F M<I> = <I> A (33)
At this point we should remember that working with displacement equa-
The eigenvalue matrix A in this expression consists of a diagonal array contain- tions and flexibilities is feasible only for structures that are statically determinate
wt
ing values of .Ai = 1/ in diagonal positions, as follows: nod restrained against mobilities. Upon reflection, we realize that this is a very
limited class of structures that would be encountered rather infrequently. In
A1 0 0 0
~cneral, the approach using action equations and stiffnesses is much more
0 .A2 0 0
Nuitable for dynamic analysis.
A. = 0 0 .A3 0 = (1) - 2 (34)

0 0 0 An 4.3 NORMAL-MODE RESPONSE TO INITIAL CONDITIONS

Transforming Eq. (33) to principal coordinates as before, we find that For a MDOF system, suppose that w~ know the initial conditions (at time t = 0)
FpMp = A (35) uf displacements Do and velocities Do, as follows:
When the modal matrix is normalized with respect to the mass matrix, the Doi Doi
principal flexibility matrix from Eqs. (32) and (35) takes the form Do2 Do2
Do= D03 (1)
(36) Do= Do3
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) Don Don
Hives the equations of motion in normal coordinates, regardless of the method 111 accordance with Eqs. (4.2-25), these initial values may be transformed to
ol formulation in the original coordinates. 11mmal coordinates by premultiplying them with the inverse of the normalized
11' wmJe U 1110<.lal matrix. Thus, we have
/\s 1111 cx11mplc of the use of normal coordinates for displacement equations, we again DNo = <l>,v 1 Do DNo = <l>,v' Do (2)
rn11sidcr the three-member truss in Fig. 4.1 (a). Because this structure is statically deter-
111l1111l c nnd immobile, we can find terms in the flexibility matrix by the unit-load method I he second relationship in Eqs. (2) is obtained by differentiation of the first with
Il l, Figures 4 . l(b)- (d) show the flexibilities obtained from applying unit values of A,, 11'Hpcct to time. The forms of the displacement and velocity vectors in normal
/\ J, and A.1 (all forces). The resulting flexibility matrix for this structure is 111ordinates are the same as those in physical coordinates given by Eqs. (1).

F = :A
l. 000
1.000
1. 000
5.556
0
-1.333
J (g)
From Eq. (4.2-24b) we see that a typical equation of motion for undamped
l11•t· vibrations in normal coordinates is
[
0 - 1.333 1.000
DNi + wt DNi = 0 (i = 1, 2, ... , n) (3)

S11hslituti11g <1,i and M from Example 4.1 into Eq. (27), we obtain the inverse of<IJN as
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
0.3290 0.7089 - 0.1637] 1111' rnnditions of initial displacementDNo, and velocity DNo;, we find the response
<l)N 1
= <1>i M = VpAf, 0.3945 0.05494 0.5915
(h)
111 11 ll' i th normal coordinate as
[
0.4353 - 0.2026 - 0.41 23
(i = l, 2, ... , n) (4)
With this matri x und the nex ibility matri x from Eq . (g) , we find FN using Eq . (36), as
l11llow11:
1111'4 t•x p1'l' Ssio11 is dl'awn from Eq . (2.2 9) for un undamped one-degree system.

Ii',., •I•,., I J1' •l• N1

,,, /['·70) () 0 ti •I /IN
()
0 (i)
w, 11u111ppl y Hq. (II) rqwlitiwl y to rnlrnl11tl' till' terms in the. vector of normal-
111111lr tl1 hplm·t·111l'1Ils l)N \nN,l 'l'hr N1· 1t•1nilt 'I 111t· thl'II tn111Hlornwd hack to the
I 111lp11111I 111111d11111tr 'I, ll 'l i11p 1hr 11p1· 111111111 p1vrn hy th1• 1111,I ol 11.qs, (11. 2 Ui) .
tl ll II I HH I
 Normal-Mode Response to Initial Conditions 149
148 Normal-Mode Method of Dynamic Analysis Sec. 4.3

...,, _ , } PoL
Thus, DNo = 'YN Do = •vr-;-;-
pAL {4.486, - 0.08872, - 0.1407 -
EA
(b)
D = <l>NDN 1
For this purpose the inverse of the normalized matrix <l>,v is available from Eq. (4.2-h).
This _sequence o~ operatio?s is the same regardless of whether the original According to Eq. (4), we find the terms in the vector of normal mode displacements
equat10ns of motion are wntten as action equations or displacement equations. to be
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
wr
DN = 'VpAf,
4.486 cos w, t
- 0.08872 cos W2t
J ~: (c)
.. [
- 0.1407 COS W3t
DN; =0 (6)
Then with the normalized modal matrix <l>N from Eq. (4.2-e), we transform these results
Integration of this equation twice with respect to time yields back to the original coordinates, using Eq. (5) as follows:
DN; = DNo; + DNo;t (7) D = <l>NDN

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. PoL
1.257 COS W1 t - 0.08325 COS W2t - 0.1738 COS W3t 1
= EA 5.434 cos w,t + 0.01647 cos w2 t + 0.1051 cos w3 t (d)
Example 4..3 [ -1.343 COS Wit - 0.09600 COS W 2t + 0.1061 COS W 3t
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~ In this example the response of the first mode of vibration is about one order of magnitude
response of this structure due to suddenly releasing the load. )lrcater than each of the other two.

y
o,~
D
3

,-1
t Po
11;xample 4.4
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 small-
displacement 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
0.8L

1
3L - 6

3
K-2Et:L 3
-6
3L
2L2
-3L
L2
-3L
6
- 3L
L2
3L
- 3L
2L2
(e)

' - - - 0 . 6L - l

Figure 4.2 Plane truss with initial load.

The.initi td displacements of the joints caused by the force may be calculated as tlw
1111d

M = pAL
420
l 156
22L
54
- l3L
22L
4L2
13L
- 3L2
54
13L
156
-22L
-13l2
- 3L
-22L
4L2
1 (f)

produet of lli nnd the second column of the flex ibility matri x F in Eq. (4.2-g). Thus, 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
Ou ( 1.000, 'i. 'i:'i6, l.:lJ.1 ) 1:.'.~' (u l 1 l1wnvnhu.: prnb lern . Thus, so luti on or Eq. (4.2- l a) yields the e igenvalues

AI No, Ill 111111• I () tilt• 111ltl11l w lodtk•N llll' i'>11 0 IJN it11-1 tlu• tli NI ol l\qN: (l). Wl' ,,,,' ' I io fo.'/ 8400 El (g)
1111 111111111 t1111pli111·1111' llt ~ ht 1111111111l 111111dl1111h N, IIN lnllnw" :
111111Ml111111 '" ' 11Af I pAJ,''
160 No rmal-M odo M othod ul l)yn11mlc Analysis Chap.4 Sec. 4.3 Normal-Mode Response to Initial Conditions 151

y
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
Vl'Ctors in Eq. (h) is scaled so that the first translational displacement is numerically equal
tb03 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
t°' CD D3L l\'Spect to M , we use Eq. (4.2-2la) to find the constants

o./.1
X
D,/ C1 = LVm C2 = L~ C3 = L~ C4 = Li (i)

,/
L
111 which m = pAL. Dividing the columns of Eq. (h) by these values, we obtain the
(a)
11ormalized modal matrix as

L\/3 LVS
-2\/3 - 6V5 LY?]
-12V7
L L «l>N= - l [~
LVm L - L\/3 LVS - LY? U)
(b)
0 -2\/3 6V5 - l2V7

Initial velocities in normal coordinates now may be calculated using the second
1 1qlrcssion in Eqs. (2). Hence,
L ~

(c)~ - L . . 0 .
«l>N Do = «1>iM ~OJ D 03
1
DNo =
[

L~

~---~ (d)
~L

(k)

I q1111tions (4) and (7) give the vibrational and rigid-body responses in normal coordi-
L~ ,,----.____ 1111ll'S as
~ " 1-L (e) ce)
Figure 4.3 (a) Flexural element with initial velocity; (b) mode I; (c) mode 2; N111t· that there is no response of the third normal mode, which has a symmetric vi-
(d) mode 3; (e) mode 4. l1111timuil shape. Finally, with Eq. (5) we transform the results in Eq. (e) back to the
1111pi1111l coordinates to obtain
us wl· ll as the modal matri x

L - tL + (L sin W4t)/ w4]

L

r~
12t - ( 12 sin W4t)/ w4 [)03
<I>
2
/, /,
6
~21
/,
(h) [
I ltl (L sin W4t )/w4 lOL
121 ( 12 Sill W,il)/ W,1
(m)

() 2 I)
" 1111'~1' ll'Npt!llNl'H Ill\' v11 )1d 011ly Im 111111111 111-dd hotly 111t111ltHIN.
152 Sec. 4.4 Normal-Mode Response to Applied Actions 153
Normal-Mode Method of Dynamic Ana lysis Chap. 4

4.4 NORMAL-MODE RESPONSE TO APPLIED ACTIONS 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).
Now we shall consider the case of an undamped MDOF system that is subjected For a normal mode corresponding to a rigid-body motion , the eigenvalue
to applied actions corresponding to the nodal displacements. For this situation, 11Jt is zero. Then Eq. (6) becomes
the action equations of motion are (8)
MD+SD=A (1) 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)
DN; = J: r AN; dt" dt' (9)

This equation replaces Eq. (7) whenever a rigid-body mode is encountered.

and n is the number of degrees of freedom. Equation (1) is transformed to normal In summary, we calculate the dynamic response of an undamped MDO~
coordinates,.by premultiplying both sides with <1>i and substituting Eqs. (4.2-26) ~ystem to applied actions by first transforming those actions to norm~! coord1-
for D and D to produce 11ates 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
<1>iM<1>NriN + <1>is <I>NDN = <1>iA double integral in Eq. (9). Finally, the values of the actual displacements m
This equation may also be written as 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
DN+ w 2 DN = AN (3) 11ppropriate equivalent nodal loads discussed in Sec. 3.3 can always be found as
The symbol AN on the right-hand side of Eq. (3) represents a vector of applied 11 preliminary step (see Example 4.6 at the end of this section).
act ions in normal coordinates, computed by the operation At this point, let us examine the effect of a dynamic load A1(t), correspond-
Ing to the jth nodal displacement, on the response of the kth nodal displacement
AN= <1>iA (4)
" k· From Eq. (5) the ith normal-mode load due to A1 is
In expanded form, the results of this multiplication are (10)

AN!] [<l>N11A1 + <l>N21A2 + , , , + <I>NnlAnl If the system has only vibrational modes, the response of the ith mode is found
AN2 <PN12A1 + <PN22A2 + , , , + <I>Nn2An lr om Eq. (7) to be
(5)

;;ln°~I -~ ;N~n~~ ~- ·· ·· • ~ ·;;,.,.·~n

1

[ ~·N;, = 0
DNi = <l>Nji 1 A1 sin W;(t - t') dt' (11)
W; Jo
'l'nrnsformation of this result back to the original coordinates by Eq. (4.3-5)
The i th equation of motion in normal coordinates is yidds the response of the kth nodal displacement as
(i = 1, 2, .. . , n)
where the normal-mode load AN; is taken from the ith row in Eq. (5).
(6)
(Dk)A ,
J
= i
i= l
[<l>Nki<l>Nji rAj sin W;(t - t') dt']
W; Jo
(12)

Each of the normal-mode equations of motion given by Eq. (6) is un- Sl111ilarly, the response of thejth nodal displacement caused by a dynamic load
coupled from all of the others and has the same form as a SDOF system with unit , \ 1{(), corresponding to the kth nodal displacement, may be written
mass. Therefore, we can calculate the response of a typical vibrational motion
using Du/w111el' s integral, as follows: (0j)A1 -_ ~ [<_l)Njl cpNki
L, lt · (t _ ') d ']
Ak Sill W; t t (13)
I I W/ o
t') rlt' (7) 11 /\, /\ 4, the right hand sides of' Eqs. (12) and (13) are equal, and we may
1 q1111tt· llw kfl 111111d sides to ohtai 11
TIIIH t'Xjlll'HHio11 Is d111w11 lm111 Hq. () <> <,) 1111d w11s dl·riwd for 1111 u11d111111wd
cu11• dl'l-' ll't' 'lYNll' lll 111111 ,~ l11i1l11lly 111 11·~1 W1• 11pply II H' pt•liliwly lo c11ll'lll1111• U>1h, (/l1 ),, (14)
 Normal-Mode Response to Applied Actions 155
154 Normal-Mode Method of Dynamic Analysis Chap. 4 Sec. 4.4

This relationship constitutes a reciprocal theorem.for dynamic Loads [3] that is we obtain the vector of normal-mode displacements, as follows:
similar to Maxwell's reciprocal theorem for static loads [2]. It states that the 1.2114(1 - COS W1 t)/ Wi]
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-lace- DN =
[
- 0.1856(1-cosw2t) / w!
-0.7472(1 - COS W3t)/w3
1 (c)

ment 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 2
Substitutionofwi = 0.2701E/ pL2,w~ = 2 .088E/ pL , andw5 = 5.308E/ pL2intoEq.
by using Eq. (9) in place of Eq. (7) in Eq. (11). (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) DN =
4.485(1 - cos w,t)
- 0.08889(1 - cos Wi t)
JEVm
P2pL2 (d)
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.
- 0.1408(1 - COS W3t)
The vector of applied actions for this case is Transforming this solution back to the original coordinates using Eq. (4 .3-5), we find that
A= {O, Pi, O} (a) 1.000 - 1.257c, + 0.08341 c2 + 0. l 739c3]
AL
As required by Eq. (4), we transform vector A to normal coordinates with the operation: D = <l>NDN = 5.556 - 5.433c 1 - 0.01650c2 - 0.1052c3 EA (e)
[
-1.333 + 1.343c, + 0.09618c2 - 0.1062c3
T
1.2114] P2 where c 1 = cos w 1t, c2 = cos w 2t, and c3 = cos W3t . Inspection of these results shows
AN= ((>NA= -0 .1856 Vm (b) 1hat the joints of the truss vibrate about the displaced positions:
[
-0.7472
P2L
D .. = {l.000, 5.556, -1.333} EA (f)
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), 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 + 0.3108c, - 0.4863c2 + O. l 755c3J
P = P2 AL
D = - 1.333 + l.343c 1 + 0 .09618c2 - 0. l062c3 EA (g)
[ 1.000 - 0.3321c, - 0.5607c2 - 0.1072c3
P.quation (g) shows that the joints of the truss vibrate about the displaced positions:

y 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
11•sponsc 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' mpl~4.6
/ ANsumc that the unrestrained flexural element in Example 4.4 is subjected to a ramp force
(b) /' P,t/ t, applied in they direction at its center, as indicated in Fig. 4.5. We shall
~ 0.6 l ~ 111ll·11late the response at the nodes due to this influence, beginning with the element at
(n) 11•NI
lJs111p I I /'2 111 ll w displnl'l'llll'lll sh11pl' lunctions f from Eq. (3.4-17) and
ll ll(lll'l' " ·" l'llllll' IIIIN~ (11) Wllh Nh' Jl 11111 I' ,.,. (h) wllh ~h·p lmn• ,., npply, 1114 Fq ( I I )~) without inlrp,1111011 , w1• hntl th1• 1•quiv11knt nmlul actions to be
 Chap.4 Sec. 4.5 Normal-Mode Response to Support Motions 157
156 Normal-Mode Method of Dynamic Analysis

y Lt 3 - ISL~ - ~ sin
3
W3t) / w5
90~ - ~ sin
3
W3~/ w~
(n)
ISL~ - ~ sin
3
w3t) / w5
- 90~ - ~ sin
3
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
- (i) 4.5 NORMAL-MODE RESPONSE TO SUPPORT MOTIONS
8t1
which consists of a force and a moment at each end . Premultiplication of this vector with We are often interested in the response of structures to support motions instead
the transpose of the normalized modal matrix from Eq. (4. 3-j) yields of applied actions. In this section we discuss problems in which either rigid-body
ground displacements or accelerations are specified. In addition, independent
T { • Ir } Pit (j ) 111otions of multiple restraints will be treated.
AN = <l>NAP = 2, 0, -vs, 0 •1
2t1 vm Figure 4.6 shows six possible displacement components Dg,, Dg2, . . . ,
wlwre 111 = pAL. The vector in Eq. (j) contains normal-mode loads of types 1 and 3 for nH 6 for a point g that is assumed to be a reference point on ground. The figure
th is example. Now we integrate the first term with Eq. (9) to obtain 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 , . . . ,
P1t 3
DNI = - -
6t1Vm
(k) 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
whil'h is a symmetric rigid-body motion [see Fig. 4.3(b)]. Similarly, evaluation of the t lf,!,id-body displacements of the ground at point g using the concept of trans-
thinl term in Eq. (j) in accordance with Eq. (2.6-c) for a ramp function produces lotion 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
DN3 = - P1VS•
(t - -1 .
sm ~
W 3t ce) ti•, ms of those at g are
2 1
2t1W3V m W3

whldt is a sy mmetric vibrational response [see Fig. 4.3(d)]. The antisymmetric rigid- (1)
hody and flexural modes 2 and 4 do not respond to the centrally-placed load. Altogether,
where
till' vector of normal-mode responses is
(2)

0 1111d
(m)

-3vs( -~ 1
sin w,,) / w5
I II Hq. ( I ) the transformalion matrix is
(3)

0
l,
(4)
1'1 1111Nlrn11111tlo11 o l thl'Nl' dlsplttL'l'lllL'lllN bnL·k to phy11k-11t L'ttonll1111tL'N using llq . (4 .J ~)
fLIV!'N
r
0
 159
Normal-Mode Method of Dynamic Analysis Chap. 4 Sec. 4.5 Normal-Mode Response to Support Motions
168

y
To include the effects of ground displacements in the action equations of
motion [see Eq. (4.2-12)), we write them as
M D + S (D - ..:1g) = 0 (8)
to,~ This matrix equation has the same form as Eq. (2.3-11) and can be restated
.to,, MD + SD = A8 .:;\bSo \ ..t.Q.. Coo-rd,.,,..fe.i (9)
I .

o,~,I
---
D,, D14 in which
(10)
/ o,6 Thus, the vector A on the right-hand side of Eq. (9) contains equivalent nodal
I 8
actions due to rigid-body ground displacements .
tD 95
:
Similarly, the accelerations at a typical joint j may be expressed in terms

1~ :~ of rigid-body ground accelerations at point g , as follows:

(11)
g - -----------r---------,,---X
/ o., o••
/ Dg3 I // where
(12)
Dg6 1I /:
/ 91

I / llnd
--- - -- . __ _____ __J/ X gj
(13)

Then the vector of accelerations ..:i8 at all free joints becomes

I
/ ..
..:18
-
-
T ••
T 8 D8 (14)
Figure 4.6 Rigid-body ground displacements.
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 relative coordinates:
D* = D - ..:18 D* = o - ..:i (15)

-:.;J
g

cL = -c8i = C18 = [-~, 1 ~i (5) In these expressions the symbot D* denotes a vector of displacements relative
Y g; Xgj 0 lo the ground, and the vector D~. contains the corresponding relative acceler-
This skcw-sy~metric submatrix contains positive and negative values of the 11tions. Substituting D - ..:18 and D from Eqs. (15) into Eq. (8) and rearranging,
l'0111poncnls of vector r81• These components are arranged in a manner that wc find that
p, od 11t·cs Ihe cross product of a small rotation vector at g and the location vector MD* + SD* = A[ re.t"-t, ve Cao'<~; nt1.ti>{ l6)
l'NJ · Then the time-varying displacements ..:18 at all free joints may be written as
111 which
(6) (17)
Whl' H'
'l'lll'rcforc, the vector At on the right-hand side of Eq. (16) consists of equivalent

TT _
T
TI,1T
112 (7)
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.
/I -
[,- , 1'
I NII/
Afkr tlw l'q11 ivule11t nmlul actions AHor At have been found, the response
l llil'11l 11tions in nhsolull' m 1t·l11l iVl' l'OOl'di1111ll'S proceed as described in Sec. 4.4
1111 11pplil·d m·lillllH 11111 IIH' 111111·1 l 11sr, lht• 11h~11l11tt• disph1l'l' 1Hl' nl s at free nodes
1111d 111 I~ IIH• 11111111 11'1 ol NIil ii 10111111
160 Normal-Mode Method of Dynamic Analysis Chap.4 Sec. 4.5 Normal-Mode Response to Support Motions 161

may be calculated from the first of Eqs. (15) as The rigid-body rotation 8z, centered at point 2, causes the followi ng step-translations at
point 1:

[-0~8L]
D = D* + 4g (18)
which is the sum of the relative displacements and the effects of the rigid-body 48 = TJD8 = 8, (c)
ground displacements.
as given by Eq. (6). The 2 x 1 operator TJ in Eq. (c) contains only terms from the first
Example 4.7
and second rows and the sixth column of the general operator in Eq . (4). Equivalent nodal
Figure 4. 7 shows the plane truss used previously in Example 3 .1. In this truss the loads (forces) at joint 1 are
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. -0.288]
A8 = S 4 8 = [ sL8, (d)
From Example 3 .1, the stiffness and mass matrices for the free displacements at 0.384
point I are
which result from applying Eq. (10). From this point we may proceed to calculate the
0.36 response by the method shown in Sec. 4.4. Thus, Eq. (4.4-4) gives the normal-mode
S=s [ -0.48]
(a) loads as
-0.48 1.64
in which s = EA/Land m = 3.28 pAL/6 . We also know that A= <l>T A= [-1]
N
3
0.48sL8,
'\!'To,;;
N g
(e)

2
w, = 0.2-
m
s
w~ = 1.8..::.
m
1 [3
<l>N= ~ 1 (b) Then the normal-mode responses to these step loads become

_[-(1 - cos w,t) / wf] 0.48sUJz

y DN -
3(1 - W2t)/w~ '\!'To,;;
COS
(f)

These expressions may be simplified by substituting wf and w~ from Eqs. (b), as follows:

DN -
_[-3(1l -- cos
cos w,t) ] 0.8mL8,
'\!'To,;;
W2f (g)

Transformation of this vector back to physical coordinates with Eq. (4.3-5) produces

_ _ [-10
D - <l>NDN -
+ 9c1 + c2 ]0.8L(J,
--
(h)
3c1 - 3c2 10
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
'l'hc prismatic cantilever beam analyzed in Example 3.2 is shown again in Fig.
M hnll determine the steady-state response of node 2 caused by a rigid-body rotational
ll('tClcration Dg6 = e, sin Ot of ground at node 1.
Stiffness and mass matrices for node 2 are

S = s[ 6
- 3L
-3L]
2L 2
M=m [
78
- llL
-lll]
2L2
(i)
/

Dno l _ whllll' .1· 2/:'/ / 1,1 and m = pAL/2 l 0. Normalization of the modal matrix (from Exam-
~

l
0.6l ph 1,2) with respect lo the mass mulrix yields
/
I I 0. 1,\()4/, 0. 1943/, l
(l,N (j)
I. Ill O I ()21 I .•181
 Normal-Mode Response to Support Motions 163
162 Normal-Mode Method of Dynamic Analysis Chap.4 Sec. 4.5

y
MFF MFR][~F] + [SFF SFRJ[DFJ [ AF] (3.5-17)
[ MRF MRR DR SRF SRR DR AR
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,
1~
CD • X
produces
(3.5-18a)

7
/
/
Dg6 L <I and
MRFDF + MRRDR + SRFDF + SRRDR = AR (3.5-18b)

Rearranging Eq. (3.5-18a) gives

Figure 4.8 Cantilever beam with rigid-body ground acceleration. (19)
MFFDF + SFFDF = AF - SFRDR - MFR DR
Due to the ground acceleration at node 1, the accelerations at node 2 are In this form we can see that the terms

.
"18 = T;'D8 = . [L] 0, 1
sin fit (k)
AFR = -SFR DR - MFR DR
on the right-hand side of Eq. (19) are equivalent nodal loads. They are caused
(20)

ns given by Eq. (14). In this case, the 2 x 1 operator T; in Eq. (k) has only terms from by the independent restrai1:.t displacements in the vector DR and their ~or-
rows 2 and 6 and column 6 of the general operator in Eq. (4). Equivalent nodal loads (a responding accelerations in DR .. After the displacements DE and the accelerations
force and a moment) at node 2 become DE 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
Ai = .. = - [ 67 ] mLO, sin fit
-M .1 8
-9L
(C) motions may be represented with zeros in vectors D,uind D.R (or omitted al-
Logether).
whkh is dictated by Eq. (17). Then the normal-mode loads are
Example 4.9
* <l>NAg
T *= - [ 7.613 ] L ,v1 .. .
We shall now reconsider the plane truss in Example 4.7 , which appears in Fig. 4.7. Let
AN= me, sm fit (m)
the support point 3 have a sudden independent step translation DR 3 = din the x direction.
-0.3109
Then determine the response of the free displacements at joint 1.
F,om these actions we find the steady-state normal-mode responses to be By including terms for the restraint displacement at point 3, we can extend the
Nliffness matrix in Eqs. (a) to become
7.613{3,/wf] . , · · .
D,t = -
[ - 0. 3109/32/ W22 L v m (}, sm fit
(n)
0.36 -0.48 i -0.36J
SFFSFR I
llpon transforming this vector back to physical coordinates, we find the relative displace-
lltLlllls:
s = [sRF sRJ = s [ ~~::: --~::: i-~::: (p)

D
* -_ <I>N D N* -_ -
[ (l.06 lb , - 0.0604lb2
l.46 lb1 - 0.4604b2
)L] . .
(}, Sill fit (o)
Slil'fnesses for unmoving restraints are omitted from Eq. (p). Similarly, the mass matrix
11 0111 Eqs. (a) is extended to

MFF
3.28 0 i lJ
MFR I
M [M . MRR] m, [- ~1_ _ }_:_~l~
= =
i2 Q
(q)
When u stru cture has multiple connections to ground , it is also possible to RI

l'nku lalc the respon se to i11rl<'fJ<'tult•11t 11101io11s t?f'support rl'.1·1rai11t.1· by generntinp

w hi•il' 111 ,,M,/6. Thon l'rom Eq . (20), the equivalent nodal loads due to the step
lltl' 11ppropri11tL· Ntiff11n 1s 1111d 11111ss l'ocffkiL'llts I 1, 'i I, 111 such II t'llsc, thl' rcl11tiVl' 1

diMpli1cl'llll' llls 111 thl' ~11pp111 h 11111st Ill' s 11111II i11 111dl•1 to H•t11i11 liiwur lwlwvlrn
I 1•t II~ 11•w1it1• 1111• 11rnl 111 11p1•d l'q11111lo11~ ol 1111,111111 1111 1111 prn11, tl>l1• 11od11l displ11n•
llll" III N Ill 11·11 1111111•1•d 1111t l pill titlo111•d 1111111 , 11~ 111111 ,w~
dl~p(lll'l ' llll' lli J>R I r/ lll l '
() \/, I
O ·IH
rl
f
0(),,tH
.361,I'(/ (r)
 164
Normal-Mode Method of Dynamic Analysis Chap. 4
Sec. 4.6 Damping i n MDOF Systems 165

Next, the normal-mode loads may be calculated as

This matrix contains damping coefficients that are defined as actions required for
T [ 0.6] -sd- unit velocities. That is, any term Cik in an array of viscous damping coefficients
AN= cf>NAFR = (s)
- 1.8 V10m 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
From these actions, we find the normal-mode responses to be
implies that the damping matrix is also symmetric.
- [
DN- 0.6(1 -cosw,t) / wT] -sd - - [3(I - cosw 1 t)] md
(t)
To form the damping matrix, we consider first the systems for which this
- 1.8(1 - cos W2t)/ w~ V10m - - (1 - cos w 2t) V10m urray is assumed to be linearly related to the mass and stiffness matrices. That
is , we take
Finally, the transformation back to physical coordinates yields
C = aM + bS (3)
D = cf>NDN = [10 -- 9c, -
3c1 +
J!!...
c2
10
3c2
(u) 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
In this case, joint 1 vibrates about the displaced position: 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,
(v) in principal coordinates we have
obtained by static analysis. (4)
where

4.6 DAMPING IN MDOF SYSTEMS (5)

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
'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
(6)
'l'hc diagonal matrix w in this expression contains the characteristic values w?
2

function when the frequency of the excitation is not near a resonance. However, for the undamped case [see Eq. (4.2-17)]. Therefore, the ith equation of motion
l'or u periodic function with a frequency at or near a natural frequency, damping in normal coordinates is
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 DN; + (a+ bw;2)DNi
' + w;2DN; = AN; (i = l, 2, ... , n) (7)
vibrational analysis until its importance is ascertained. To make this expression analogous to that for a SDOF system (see Chapter 2) ,
When a discretized solid or structure is assumed to have viscous damping, we introduce the notations
lhc uction equations of motion may be written as
n;
MD+Ci>+SD = A ( I) C Ni = 2n; = a + bwr '}'; = - (8)
W;
whkh upplics only to free nod~! displacements. The damping matrix C pre
111ultiplying lhe velocity vector Din Eq. (l) has the general form
111 these relationships the term CN; = 2n; is defined as the modal damping
,·1111sta111 for the ith normal mode, and '}'; represents the corresponding modal
c,, C,2 Cn c,,, rlfl111pi11R ratio. Using the first of these definitions in Eq. (7), we obtain
C2, C22 C21
(' .
C211 (i = 1, 2, . .. , n) (9)
( ' 11 (' IJ (' II ( '111 (2)
l•m·lt ol th1· 11 l'q1111tio11 s 1l' iWl'Sl'lllt'd by lhis l"Xprcss ion is uncoupled from all of
0

1 1111' 111111'1 ~ 1'11l'11·lo11•, Wl' l'IIII d 1•ti•1111i111• lli1• ll'M iH>IINI' ol' Jill' /th IIH >(k in llw sum1..•
( 11 1 ( '11 •
( "' ( ',,,, 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
. From the definitions in Eqs. (8), we may express the modal damping ratio
Y1 m terms of the constants a and b, as follows:
a + bw7
Yi=
2wi
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 coordi-
nates. Thus, under the condition that a = 0, Eq. (10) becomes
bwi
'Yi = 2
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 veloc-
ities of displacement coordinates. In this case Eq. (10) simplifies to
a Now we rewrite Eq. (9) in terms of 'Yi as
'Yi=-
2wi
.~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'otllplt·x c igcnval.ues ~ccur as conj ugate pairs, and the correspondin g cigcn
Vl'l'lms uls.o t·o~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
111d1•1 l' <llllllll!II~ ol IIH!ltOll lillo >11 IIIH'llll)lh•d ilt,1 111d1•1 1•q1111IIOIIH
Chap. 4 Sec. 4.6 Damping in MDOF Systems 167
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
(10)
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
(11)
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:
W·)e1 (0.5 e1 0. 7) (14)
'Yi = 'YI ( W; ::5 ::5
This expression suppresses the higher modes in accordance with damping ex-
periments [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.
(12)
(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
•I>/ • <l>iM 1sec Eq. (4.2-27)) and rewrite Eq. (16) as
(17)
l'hii; f'o nu of' lhl' ll'll11Hlo111111lio11 is t•spl'd11 ll y appropri ate when not all of the
11111d1•, 1111• ind11dl•d 111 till' 1111ul y1. is (111od11 I t111t1l'1ttio11).
11111111 111
168 Normal-Mode Method of Dynamic Ana lysis Chap. 4 Sec. 4.7 Damped Response to Periodic Forcin g Functions 169

Example 4.10 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
P = {P1 , P2, A, ... , Pn} (2)
is made of steel and that the damping ratio for the first mode of vibration is y 1 = 0.02. In Eq. (1) the terms in Pact as scale factors on the function cos flt. Transfor-
Prom Eq. (4.2-f) the three angular frequencies are mation of the action equations of motion to normal coordinates produces the
wi = 0.5197c1 w2 = I.445c1 W3 = 2 .304c1 (a) typical modal equation
.. . 2 -
where C1 = (VE/p)/L. Applying Eq. (14) with e1 = 0.6., we find that DN; + 2n;DN; + W; DN; - Pmi cos flt (i = 1, 2, ... , n) (3)
0 6
1.445 ) · in which p 111; is a constant. This equation has the same form as Eq . (2.4-23), so
,'2 = 0.02 ( 0.5197 = 0.03694 we can take the damped steady-state response of the ith mode to be
0 6
(b)
2.304 ) ·
'}'3 = 0.02 ( 0.5197 = 0.04887 DNi = Pm; a (,....
2 JJi cos ~ Lt - O;) (4)
W;

Then the normal-mode damping constants are The magnification factor {3; in this expression is
CN1 = 2(0.02)(0.5197)c1 = 0.02079c 1
1
(5)
CN2 = 2(0.03694)(1.445)c1 = O. I068c1 (c) {3; = Y[ l - fl/w;) 2] 2 + (2y;fl/ w;)2
CN3 = 2(0.04887)(2.304)c1 = 0.2252c 1
and the phase angle O; is
Using these values in Eq. (16) along with <l>iv 1 from Eq. (4.2-h), we obtain _ - I[ 2y;fl/ W; J
C = <l>ivTCN<l>,v 1
O; - tan 1 _ (fl/ w;)2 (6)

0.06154
Equations (4) , (5) , and (6) are drawn from Eqs . (2.4-31),.(2.4-32), and (2.4-33) ,
- 0.01 270 - 0.01662]
respectively. The response given by Eq. (4) may then be transformed back to the
= A YpE -0.01270 0.02002 0.01987 (d)
[ original coordinates in the usual manner, using Eq. (4.3-5).
-0 .01662 0.01987 0.07620 To determine the response of the mode having its angular frequency w;
wllich is the symmetrical damping matrix in physical coordinates. 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
Pm;= <1>1;P (7)
4 .7 DAMPED RESPONSE TO PERIODIC FORCING FUNCTIONS
and the back-transformation in Eq. (4.3-5) becomes
As mentioned in Sec. 4.6, damping is of greatest importance when a periodic D = <l>N;DN; (8)
1•xcitalion 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 If desired, this process can be repeated for other modes with frequencies in the
l'llit'uluting steady-state responses of discretized structures to periodic forcing vicinity of fl.
l1111l'tions. Knowing the imposed frequency of such a function and the natural Now we shall consider a lightly damped structure subjected to a set of
lil·qucncies or the system, we can obtain in a direct manner the steady-state actions that are all proportional to the general periodic functionf(t). In this case
ll'Sponses of the modes having frequencies in the vicinity of the imposed frc the applied action vector A may be written as
'(lll'llt'Y . Both simple harmonic and general periodic forcing functions will bl' A = F(t) = Pf(t) (9)
di.~l·11 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 where the vector Pis given by Eq . (2). Proceeding as described in Sec. 2.5 , we
p1op111tio11ul to tlw siinpk harmonic fun<:tion <.'OS n,.
tlw action vector A lllny i:xpn:ss f(r) in the form of' a Fourier series, as follows:
,.
h1• WI iltl'II HS ,
/'(I) <In I ~~ ("1 rnNjUi I h1 sin jHI) ( 10)
A p !' ON 01 (Il I I
 Damped Response to Periodic Forcing Functions 171
1'70 Normal-Mode Method of Dynamic Analysis Chap.4 Sec. 4.7

which is drawn from Eq. (2.5-1) with i replaced by j. The coefficients a1, b1 , and
110in Eq. (10) may be evaluated as indicated by Eqs. (2.5-3).
Transformation of the action equations of motion to normal coordinates
p., - "'"· p ~ vk[0.2803 1.2114 -0. 2995{ ~J
(c)
produces the typical modal equation
l.2114Pi
(i = 1, 2, ... , n) (11) - VpiL
where p,,.; is again the constant given by Eq. (7). From the solution of Prob. The magnification factor for the first mode is obtained from Eq. (5) as
2..5 5, we talce the damped steady-state response of the ith mode to be
11.94 (d)
DN; = Pm;{ao + f {3u[a1 cos (}flt - Ou) + b1 sin (}flt - 8;J]1 (12)
~1=--;========:=~=====(~=:=:='
0 .25 ) 4 0.25 7) 2
2

w, j= l J ( l - 0 .2701 + (0.0 ) 0.2701

in which the magnification factor f3u is From Eq. (4) we find the damped steady-state response of the first mode to be
1 2114
(13) DNI = 1. ~~(11.94) sin (flt - 81)
f3u = Y[l - (jfl/w;)2] + (2y;Jfl/w;)2 2 0.2701£ V pAL
2
und lhe phase angle Ou is 53 .55P2L p . CA 8) (e)
= E\,lpAf, Sill ut - 1

_ - I[ 2y;jfl/ W; ]
(14)
eij - tan l _ (}fl/ w;)2 where

l
(0.04)(0.5) ]
Because a multiplicity of terms contribute to the ith mode in Eq. (12), the = - 1 (0.5197) = 270 21' (f)
81 tan _
possibility of resonance (JD, ""' w;) is much greater for a general periodic func- 0 25
1i1111 than for a simple harmonic function . Therefore, it becomes more difficult l - 0.2701
lo predict in advance which of the natural modes will be strongly affected.
as given by Eq. (6). Transfonning the response of the first mode back to the original
l lt1wl~vcr, after the forcing function has been expressed as a Fourier series, each
coordinates with Eq. (8), we obtain
of the JO frequencies can be compared with thew; frequencies for the purpose
ol p1'l•dicting large-amplitude forced vibrations. D = <l>NIDNI = {15.01, 64.87, -16.04}~~ sin (flt - (Ji) (g)

I~ nnmk 4.1 t 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 second mode as
/', Ni ll Ht, applied in the x direction at joint 2 (corresponding to displacement D 2). The
D = {-0.09478, 0 .01875, -0.1093}~~ sin (flt - 82) (h)
11111111l11r frequency of the forcing function is n = (0.5/L)VE/p. Calculate the steady-
Hinh· 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). 1111d that due to the third mode is found to be
Till' squurc of the impressed angular frequency is
D = {- 0.1824, 0.1104, 0 .1114}~~ sin (flt - 83) (i)
n2 = 0.25£ (a)
L2p
The amplitudes in both of these vectors are small compared to those i~ ~q. (g). Further-
'l'hlN v11h1l' is t'nirly close to the square of the first angular frequency 111ore, the innucnce of damping is significant for the first mode but neghg1ble for the other
IWO.
0 .270 IH
/,, ,, (h)
11:,mm1>lt• 4.12
l•li\tlll' 'I .') Mhows 11 1wriodic forci11A 1'1111l'tio11 in lhe sh'.1pc..of' a square wave. lf this
11H 1'iw11111 l'11hlt• I •I 'I h1•H•lo11•, w1• Nho11ld 1•xp1•1•11h1• tl1 NI 11111111· ol 1hl' s1n1vtl11l' 10.lw 0

llua·tlmi ls 11 pplli•d 111 llw I tlh1•1•1io11111 jnl111 I ol 1111' IIIINN 1111·1~. 4. l(a), find the damped
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
Hh'illl y Hlilh' ll'H Jlllll/'11' 1111 l 'lll hot lh1• IICllllllll 111111h•H
 172 Normal-Mode Method of Dynamic Analysis Chap.4 Damped Response to Arbitrary Forcing Functions 173
Soc. 4.8

p
nre relatively high ( y; > 0.05) , the effects of damping could have some con-
Nc4uence. Therefore, we shall now modify the formulations in Secs. 4.3, 4.4 ,
p1
- 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 struc-
0 111re to initial conditions of displacement and velocity at time t = 0. In the
'lT 2'!T 3'!T 4'lT
TI n n n presence of damping, the free-vibrational response of the ith mode, given by Eq.
(•1. 3-4), must be changed to

- P, t-- D Ni = e
-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
Figure 4.9 Periodic forcing function. l~q. ( 1) is
WJ; = V wJ - nT = w; V 1 - YT (2)
Expanding the square wave as a Fourier series (see Prob. 2.5-1), we obtain
4 1 111 which w; is the undamped angular frequency. Transformation of the initial-
F(t) = Pif(t) = p (sin fit+~3 sin 3fit
7T
+ . . .) (j) vectors D 0 and D0 to normal coordinates remains the same as in Eqs.
1 lllldition
(•U -2), and back-transformation of the response is still given by Eq. (4 .3-5).
Trunsformation of the load vector to normal coordinates produces Similarly, the calculation of normal-mode responses to applied actions, as
d1•scribed in Sec . 4.4, requires only a few modifications associated with modal
F(t)] [cf>N11] dumping. Transformation of applied actions to normal coordinates is the same
AN = <I>J.; A = <l>J,; 0 = <I>Nl2 F(t) (k)
[ 11N in Eq. (4.4-4), but Duhamel's integral in Eq. (4.4-7) must now be written as
0 <l>m 3
111 which the load corresponds to displacement D 1 . In accordance with Eq. (12) the
nw mul-modc responses become '
DN;
e - n;r
=- - lt e 'tH'
AN;
'
sm WJ;(t -
I
t ) dt
I
(3)
WJ; O

3 which is taken from Eq. (2.6-4).

<I>m1~11 sin (fit - 811) + /3~ sin (30t - 8 13) + .. -}/ wf
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
l)N = 4P1
7T m 2{/321 sin (fit - 821) + /323 sin (3fit - 823 ) + . . .} /w~
'l'N1
3 ,111·l'lcr~tions, there is neither displacement coupling nor velocity coupling be-
1w1•1•n the masses and the ground in relative coordinates. There exists only
<I>N13~ 31 sin (fit - 831) +
13;3 sin (3fit - 833 ) + . . -} / wl 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
Wlil'tC _the magnification factors and phase angles are given by Eqs. (13) and ( 14), th1• 1•quivalent nodal actions in the vector At, as given by Eq. (4.5-17). Trans-
I l'N IWt"t1vc ly, li11111111ion of these actions to normal coordinates yields the equivalent modal
11111\ lh
(4)
4 .8 DAMPED RESPONSE TO ARBITRARY FORCING FUNCTIONS
1 lu•u the relative response D~1 in each normal coordinate is obtained from Eq.
1>11111ping sl_1ou_l~ be included in transient response calculation s whenever ii 1 \ J. wi th Ai 1 replacing ;\ Nf, Finally , these displacements are back-transformed
111l1,1l~t Ill' s1gn1licnnt. For t•x111nplc, if the durntinn or u forc ing function is
11· l11p
ll' l11t1v1·ly 10111• t·o111p1111.·d to illl· 1111turnl pt·1'iods ol II st111 ctun·, dninpiug rn\ild lw
1111po1 l1111t. Also , wlll'11 1111• t11111• of i11t1•1t•sl ii.. i..hrn I h111 till' 111rnlnl d11111pi11g ,·nti wi (5)
 175
174 Normal-Mode Method of Dynamic Analysis Chap.4 Sec. 4.9 Step-by-Step Response Calculations

which_ gives the relative responses in physical coordinates. As before, the abso- (e)
lute d1splace~ents at free nodes may be found with Eq. (4.5- 18).
. In ~ertam cases where support displacements are specified, there is veloc-
•? co_uplmg bet~een fre~ displ~~ement coordinates and support restraints. This
where i = 1, 2, 3. The expression in Eq. (e) is a modified version of Eq. (2.7-2c).
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 lorcing functions are not necessarily analytical expressions. The basic approach
dt11pmg. Rec~II that ~e three-member truss in Fig. 4.4(a) is subjected to a step force 111 that section was to approximate the forcing function (or data points) using
l i, correspondmg to d isplacement D 2 • 1>iecewise-linear interpolation and then to use the Duhamel integral within small
Symbolically transforming the vector of applied actions to normal coordinates lime steps. We shall now incorporate this technique into the normal-mode
pwduces 111cthod for calculating transient responses of MDOF structures. As in the pre-
1·cding sections, modal damping will be assumed throughout. Because of the
(a) 1•xtensive calculations required, it is implied that the method of this section is to
AN= ~U = ~ { ; ] = [ :: } he program.med for a digital computer. Such a program is described in Sec. 4.10,
where numerical examples are also presented.
I >uc lo the step function, the damped normal-mode responses are 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)
<l>N21{1 - e-"ir(cos Wd1 t + ;~ sin wd1
1
9}/ wf will be handled at a time, and the piecewise-linear action vector Ae, (or M;) may
Ill· expressed as
DN = P2 <l>N2+ - e- "2 ' (cos wd2t + ;: sin wd2t)} / w1 (b) (j + 1 = 1, 2, ... , n1) (1)

<l>N2+ - 3
e- " '(cos W113t + ;: sin wd3 9}/ w5 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).
'lhl·sc expressions are drawn from Eq. (2.7-2b). 11 more than one such function is applied simultaneously, the responses for each
11:x11m1>tc 4.14 111 them handled separately can be superimposed.
Transformation of the action equations of motion to normal coordinates
S11.p1iosc l!lat t~e gr:~und in Fig. 4.4(a) accelerates in they direction in accordance with
pt oduces the typical modal equation
lhl 111111p_fun ct1on D82 = a2t/t2. Formulate the damped responses of the normal modes
Nl111t111g from rest. ·' t'
(2)
1:m this problem we work in relative coordinates, where the equivalent nodal toad DN,· + 2n,· DN,· + w,2 DN·I = ANI,)· + M N·I ,).At·
-
J
Vl'l'IOI IS
(i=l,2, ... ,n and j + 1 = I, 2, ... , n1)

Mm~:,- mM<:' (<:)

wh1•1e t' = t - t . The symbol AN;,J in Eq. (2) represents the ith normal-mode
1
11111d at time t1. Thus,
(3)
IIN Jt lVl'n hy Eq . (4.5- 17). Then the cqu ivulcnt nonnnl mode loads arc
111 11dditio11, we have the change in the ith modal load during the time step At1,
//, I ,11 li111•d 11s follow s:
/1111 (d)
(4)
•I• II

hll 1(11 lllllljl ltllll llllll, 1(11 d11lll(II ii 1ltlllli11' llitllh II ~(11111~1 I, (111 lllllt
wl11·11· lh1· 11y111l111I A,,, 111 dl•111111•11 ihl' m ltnll 111 111111• 1111 •
 176 Normal-Mode Method of Dynamic Analysis Chap. 4 'foe. 4.10 Program NOMO for Normal-Mode Response 177

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
D Ni,j+I = (DNI + DN2
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
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
+ DNJ)i,j+ I (5) 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

(D NI) i,j+ I = e -n·tJ.t·(D

II Ni ,j cos
A
Wd;utj
+ DNi,j + n;DNi,j .
sm A
Wd;Utj
)
(6a)
DNi,j+I = DNi,j + /)Ni,j Atj + 1(ANi,j + 1AAN;)(AtY (9)
Wd;
1111d the velocities in Eqs. (7) and (8) are supplanted by
This fo rmula represents an extension of Eq. (2.7-5a). (10)
The other two parts of the response in Eq. (5) are due to the linear forcing
•
DNi,j+ 1 = DNi,j
'
+ (ANi,J + 2I AANi,j ) ut1
A

function within the time step. The rectangular portion of this impulse yields 11.quations (9) and (10) pertain to rigid-body motions with no absolute damping.

(D N2 ) i,j+ 1 = -ANi,j[
2- 1
W;
- e - n·tJ.t·(
' 1 cos wd;Atj + -n; sm
.
Wd;
wd;Atj) J (6b)
4.10 PROGRAM NOMO FOR NORMAL-MODE RESPONSE
which is taken from Eq. (2. 7-5b); and that associated with the triangular portion
becomes 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
(6c)
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
whkh is drawn from Eq. (2.7-Sc). l•lowchart 3.1. However, as the last step in Subprogram RESl, the eigenvectors
We can also write the velocity of the ith mode at the end of the }th tinw 111 c normalized with respect to the mass matrix. Then the number of loadin_g
Ntl•p in three parts, as follows: 111,1·tems NLS is read, the loading number LN is initialized to zero, and LN 1s

l)Ni,j+ l = (DNI + D N2 + DNJ)i,j + I (7) 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
'l'lu:sc lhrec contributions may be formed by extending the notation in Eqs. luutling systems NLS. This is followed by Subprogram TRANOR, which tran~-
(J. .7 6) to obtain lmms initial conditions and actual or equivalent nodal loads to normal coord1-

(/)• NI ) /,/1 I = e
n•tJ.t·[
1 1
-
(v Ni,jWd;
+ n;l)Ni,j +Wdin;DNi,j) sm
.
Wd;Utj
A
1111h.: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 trun-
(811) 1111 ion.
Subprogram TIHIST calculates time histories of normal-mode displace-
+ DN;,J cos w,11 AtJ 1m111ts and velocities with the step-by-step method described in .sec. 4.9. Be-
11111se uniform time steps are to be used, the coefficients of DN;,J , DN;,J, ANi,j, and
(8h) ,\ANt,J in Eqs. (4.9-5) through (4.9-10) becom~ consta~ts t~at need ~e deter-
11il1ll'd only once at the beginning of the analysis. For v1brat10nal motions, the
AA.Nl,/1I
'"/ At1
,.
111 11
t 1(cos w,11 At1 I
111

'",11
sin w,11 6.1)
1
l (He)
11•Nponscs in Eqs . (4.9-5) through (4 .9 -8) can be written in eight parts , as
lt1llows:

l\q1111tlo11s (.,) 1h11111)'h (H) t·rn1 stil11lt• ll'l'lll ll'lll't' 111111111111.~ lon·11kul11ti~11• th1· n N1.i1 1 (', nNI., t (' , 1>N1. 1 t c,A.N1. 1 -1 C4 AAN1 .1 (I )
d111111wd ll'N p1111 ~1· ol 1·11d11111111111l 11111tl1• 1111h1• 1•11d 1111111• /lh lii11t• slt•p 'l'lwy 11bo /,NI/II ( \ /)NI,/ I ( ',/>NII I (' 1A.N/,f I ('K/J.A.N/,/ (2)
178
Nurmnl Moctn M111h111l 11l l>y1111111h Annlyala
Flowchart 4.1 Main program for NOMO•
IOI
l. Prngram VIB from Sec. 3.8,
plus nom1aliza1ion of eigenvcciors
with respcc1 to mass mairix.
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 and rrocc,, 111101hcr
,truclurc
l·ND
•Applies lO uny type of lincnrly ch1Ntlc Nlluctu11,
IS11hp1ow11111~111111 dllkr for cvc1y ty1w ol Hl1m•1111,•
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
n~(
C1 = e ' cos A +-~
wd;ut
Wd;
sm
• A~
wd;ut
C2 = - 1- e -n;t!.t sin wd;!:J.t C3
1
= -(1 - C1)
Wd; Wf
(3)
1111 Nl' coefficients, as well as simpler constants for rigid-body motions [see Eqs.
1 I 1l lJ) 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
'" 111•s and/or plots resulting time histories of nodal displacements and axial
l111n·s 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
l11•sses, 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 struc-
11111•, 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
,1ppt•11rn 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 n11111h1•1 lll 1i111t• st1·ps NTS, the duration of the
111111111111 liit1l' sll'P In: und llw d11111p111p 1ut111 I >AMPR. pcrtaining to nll modes.
 Normal-Mode Method of Dynamic Ana lysis Sec. 4.10 Program NOMO for Normal-Mode Response 181
180

TABLE 4.1 Dynamic Load Data for Plane Trusses

No. of
Type of Data Items on Data Lines
Lines

Dynamic parameters 1 NTS, DT, DAMPR

Initial conditions
(a) Condition parameters 1 NNID, NNIV
(b) Displacements NNID 1, D0(2J-l) , D0(2J)
(c) Velocities NNIV J , V0(2J-l), V0(2J)

Applied actions
(a) Load parameters 1 NLN,NEL
(b) Nodal loads NLN J, AS(2J-l ), AS(2J)
(c) Line loads NEL I, BL!, BL2, BL3, BIA

Ground accelerations
(a) Acceleration parameter 1 IGA
(b) Acceleration factors• 1 GAX,GAY

Forcing function
(a) Function parameter 1 NFO
(b) Function ordinates NFO K, T(K), FO(K)

"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 (b)
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 Figure 4.10 Loads on plane truss: (a) nodal loads; (b) line loads on element.
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 AS(Jl) = (2BL1 + BL3)L/6
1 rnmponcnl of force AS(2J-1) and they-component AS(2J). Figure 4.10(11)
AS(J2) = (2BL2 + BL4)L/6
dl•pit•1s lhcse components of the applied force acting at node j of a plane truss. (4)
Also, Fig. 4.1 O(b) shows linearly varying line loads (force per unit length) AS(Kl) = (BLl + 2BL3)L/6
upplied in the x and y directions along the length of a typical plane-truss elemt'lll AS(K2) = (BL2 + 2BL4)L/6
1. 1>11111 for this condition of loading appear as part (c) under applied actions,
wht•1l' L't1c:h of NEL lines contains an clement number I and scale factors for lh1· where subscripts JI through K2 are obtained from Eqs . (3.5-30). Of course, an
load i11tt·11sitics 13LI throuµh BI A shown in Fig . 4. IO(h) . ll can easil y he show11 infinity of other load sets could be applied to the element and their equivalent
lhul 1h1• t•q11i v11k·11l 11od11I loud~ i11 slruclunil din•ctiom1 al joints j Hnd k 111'L' uodu I londs dcri vcd .
 Program NOMO for Normal-Mode Response 183
182 Normal-Mode Method of Dynamic Analysis Chap.4 Soc. 4.10

Data for ground accelerations consist of two lines, the first of which gives 11 usses we calculate and write only the axial force at the k end, which has the
11 parameter IGA for ground accelerations. If this number is nonzero, ground M11me sign as the axial force in an unloaded member. On the other hand, with
uccelerations exist; otherwise, a zero indicates nonexistence. On the second line lwo- and three-dimensional finite elements, the stresses are calculated at sam-
arc scale factors GAX and GAY for ground accelerations in the x and y direc- pling points for numerical integration (see Chapters 7 and 8). If desired, ~e
tions. The possibility of rotational ground acceleration is omitted from this ,•ould specialize the variables JNO( ) and IEO( ) to have a second subscnpt
program. cli,noting a particular type of nodal displacement or element stress.
For either applied actions or ground accelerations, the load data must l~xnmple 4.15
include information defining a dimensionless piecewise-linear forcing function.
l'o illustrate using Program NOMOPT, we shall analyze the three-member pla~e truss
Such a function is given as the last block of data in Table 4.1. On line (a) we IINCd previously in Example 3. 4. Dynamic responses of this structure will be obtamed for
have the number of function ordinates NFO. Ordinates of the forcing function
the following influences:
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) I. Initial displacements of 0.1 in. at all nodal degrees of freedom, with
al that time. For simplicity, we restrict the time T(K) to be equal to an even DAMPR = 0.0
number of time steps DT. If the forcing function has a discontinuity at time T(K), 2. Piecewise-linear force in x direction at node 2, with DAMPR = 0.0
lwo lines are required to define FO(K) on both sides of the discontinuity (for the '- Same as case 1, but with DAMPR = 0.1
sume time). Note again that the function ordinates will receive dimensions only
4. Same as case 2, but with DAMPR = 0.1
when they are multiplied within the program by either applied-action or ground-
11ecclcration scale factors, thus creating time-varying proportional loads. 11, 11 11 cases we take the values of E, p, L, and A as given in Example 3.4, where the
ln all computer programs for dynamic analysis we write and plot selec- 11111tcrial is steel and US units apply.
livcly to limit the volume of output. Table 4.2 shows our method for selecting Table 4.3 contains a partial listing of the output from Program NOMOPT for the
nodal displacements and element stresses for writing and/or plotting in Sub- 111111, analyses considered. With the formats coded in the program, this output becomes
pmgrnm RES2. Line (a) of the table contains output parameters that have the
TABLE 4.3 Computer Output for Example 4.15
lollowing meanings:
IWR = indicator for writing (0 or 1) PROGRAM NOMOPT

IPL = indicator for plotting* (0 or 1) *** EXAMPLE 4.15: THREE-MEMBER PLANE TRUSS***

NNO = number of nodes for output STRUCTURAL PARAMETERS

NN NE NRN E RHO
NEO = number of elements for output 3 3 2 3.0000E+04 7.3500E-07
lt1 line (b) the node numbers JNO( ) for output of displacements are listed, and NODAL COORDINATES
X y
t·kmcnt numbers IEO( ) for output of stresses appear in line (c) of the table. J-101 NODE
1 0.000 0.000
l 1·11111l'd struclures, this output usually consists of generalized (or integrated) 2 150.000 200.000
Nltt•sses al the j and k ends of members (see Chapter 6). However, for plant• 3 150.000 0.000

TABLE 4.2 Output Selection Data ELEMENT INFORMATION EL ex CY

ELEM. J K AX
1 2 10.0000 250.0000 0.6000 0.8000
1 1.0000 0.0000
No. of 2 1 3 6 . 0000 150.0000
Type of' Data Items on Data Lines 8 . 0000 200.0000 0. 0000 -1. 0000
Lines 3 2 3

Output Selection NODAL RESTRAINTS

NODE NRl NR 2
(11) Output pnrnmctcrs I IWR, IPL, NNO, NEO 1 0 l
(h) Nodnl displnccmcnts NNO JNO(l), JN0(2) , .. , , JNO(NNO) 3 l l
(l') l\lt•lll('III Ntll'SN('N NEO 1110 ( I ) , 11 \0(2) , , , , , IEO(NEO)
NlJMl\lOH 01r PMOH!ilVlS 0 1r J<'Rl'l llJDOM I NOF • 3
NUMIIMII OJI NOO/\L lll9!l'rltl\ I N'l'fl I NNH 3
+111•1·111m• plilllillp 1·11p11hlll1 y v11d1•N lllllllllll, Pllll'IIIIII II Nl' I " , 11!1• p1111um•11, 1 11'1. 11111y ht• "'''
I q1111I 111 11•111 111 1111 I 11•1. II "' . ,,," •1•11r1rN1r.OI Ml\'1'111 X IJl1t'OMl'O dr.11
184 Normal-Mode Method of Dynamic Analysis
Program NOMO for Normal-Mode Response 185

TABLE 4.3 (Continued)

TABLE 4.3 (Continued)
MODE 1
ANGULAR FREQUENCY 4.1995E+02
NODE DJl DJ2
1 2 . 3137E-Ol O.OOOOE+OO 5 5.0000E-03 9.7853E-02 O. OOOOE+OO
2 l.OOOOE+OO -2 .4 722E-Ol 6 6.0000E-03 4.9458E-02 O. OOOOE+OO
3 O.OOOOE+OO O.OO OOE+OO 7 7 . 0000E-03 -7 . 4544E-02 O.OOOOE+OO
8 8 . 0000E-03 -l.0603E-Ol O. OOOOE+OO
MODE 2 9 9 . 0000E-03 -5.4252E-02 O.OOOOE+OO
ANGULAR FREQUENCY 1.1677E+03 10 1. OOOOE-02 2.7924E-02 O. OOOOE+OO
NODE DJl DJ2 11 l,lOOOE-02 9.2962E-02 O. OOOOE+OO
1 8, 6725E-Ol O.OOOOE+OO 12 l.2000E-02 4.0849E-02 O.OOOOE+OO
2 -l.7149E-Ol l.OOOOE+OO 13 l.3000E-02 -8.1017E-02 O. OOOOE+OO
3 O.OOOOE+OO O.OOOOE+OO 14 l.4000E-02 -6 . 9040E-02 O. OOOOE+OO
15 l.SOOOE-02 6.8279E-02 O.OOOOE+OO
MODE 3 16 l.6000E-02 l.1979E-Ol O. OOOOE+OO
ANGULAR FREQUENCY l.86i8E+03 17 l.7000E-02 4 . 7138E-02 O.OOOOE+OO
NODE DJl DJ2 18 l.SOOOE-02 -3 . 6778E-02 O.OOOOE+OO
1 l.OOOOE+OO O.OOOOE+OO 19 1. 9000E-02 -8.3692E-02 O.OOOOE+OO
2 -6.0504E-Ol -6 .10 68E-Ol 20 2.0000E-02 -5 , 2695E-02 O.OOOOE+OO
3 O.OOOOE+OO O.OOOOE+OO
MAXIMUM l.1979E-Ol O.OOOOE+OO
*** LOADING NUMBER 1 OF 4 *** TIME OF MAXIMUM l . 6000E-02 2.0000E-02
MINIMUM -l.0603E-Ol O.OOOOE+OO
DYNAMIC PARAMETERS TIME OF MI NI MUM 8 . 0000E-03 2.0000E-02
NTS OT DAMPR
20 l.OOOOE-03 O.OOOOE+OO DI SPLACEMENT TIME HISTORY FOR NODE 2
STEP TIME DJl DJ2
INITIAL CONDITIONS 0 O.OOOOE+OO 1. OOOOE-01 1. OOOOE-01
NNID NNIV 1 1.0000E-03 8 . 5266E-02 l.6390E-02
2 0 2 2.0000E-03 7.2848E-02 -l.0676E-0 1
3 3.0000E-03 6.0750E-02 -1. 0298E-Ol
INITIAL DISPLACEMENTS 4 4 . 0000E-03 -5.4078E-03 3 . 5295E-03
NODE 001 D02 5 5.0000E-03 -8 . 4207E-02 1.0144E-Ol
1 l.OOOOE-01 O. OOOOE+OO 6 6.0000E-03 -9 . 8212E-02 l.0809E-Ol
2 l.OOOOE-01 l .OOOOE-01 7 7.0000E-03 -8 . 5678E-02 2.3188E-03
8 8 . 0000E-03 -9.3358E-02 -9.5989E-02
APPLIED ACTIONS 9 9.0000E-03 -8.2739E-02 -3.8360E-02
NLN NEL 10 l . OOOOE-02 -5 , 1067E-02 9 . 6986E-02
0 0 11 l.lOOOE-02 -2.9195E-02 l.0992E-Ol
12 l.2000E-02 l.9009E-02 -7,1293E-03
GROUND ACCELERATIONS 13 l.JOOOE-02 9.6665E-02 -l.0699E-Ol
IGA 14 l.4000E-02 l.2087E-Ol -l.0634E-Ol
0 15 1. SOOOE-02 8 . 872BE-02 -l.2480E-02
16 1. 6000E-02 7.6180E-02 8.6587E-02
NORMAL MODE SOLUTION 17 l.7000E-02 7 . 2053E-02 5.6728E-02
NMODES 3 18 1.SOOOE-02 3.4921E-02 -7 . 7855E-02
19 l.9000E-02 -3 . 4530E-03 -l . 1650E-Ol
OUTPUT SELECTION 20 2.0000E-02 -3.8910E-02 2 . 4600E-03
IWR IPL NNO NEO
1 1 2 1 MAXIMUM l.2087E-Ol l . 0992E-Ol
TIME OF MAXIMUM l.4000E-02 l . lOOOE-02
NODES: 1 2 MINIMUM -9.8212E-02 -l,1650E-Ol
TIME OF MINIMUM 6.0000E-03 l.9000E-02
ELEMENTS: 1
MEMBER FORCE TIME HISTORY FOR ELEMENT 1
DISPLACEMENT TIME HISTORY FOR NODE 1 STEP TIME AMl
STEP TIMID DJ1 DJ2 0 O.OOOOE+OO 9.6000E+Ol
0 O.OOOOE+OO l.000010 01 o.oooom•oo 1 1. OOOOE-03 2 .8848E+Ol
1 1 . OOOOIC 03 6 , 70b lM Oil O.OOOOJlltOO 2 2.0000E-03 -2.6 47 2E+O l
1. 1,000CJM O'I !, 'J. ., ltil~ () i CJ, 00QCJJ~ I QQ 3 3.0000E- 03 l. 7503E+Ol
I I, OCJO(IIC !l I 1.uo11m 01 () • (l{l(l()lq I ()(1 4 4.0000ID 03 1. O<,bbIC101
4 4, OOOOIC O I I • !\Ii 1111f1l (I ,I o. 11110111q, no b b. oooorn o 1 I, I'/0410 t OI
h I, , OOOOltl ll I • 'J. hht1 1 m, no
188
No111111J Mocln M11tl1111I 11l l>y1111111 lo An11lyel11
Program NOMO for Normal- Mode Response 187
TAJJLI~ 4.J (C'or11J11111•dJ
TABLE 4.3 (Continued)
7 7 , 0000E - 03 -5 .790 21!l+ OO
8 B.OOOOE-03 -8. 3 0 2 7E+Ol NODES : 2
9 9,0000E-03 -5 . 7337E +Ol
10 l.OOOOE-02 3.6233E+Ol ELEMENTS: 1
11 l . lOOOE-'02 l . 7570E+Ol
12 l.2000E-02 -2.2569E+Ol DISPLACEMENT TIME HISTORY FOR NODE 2
13 l . 3000E-02 2.5220E+Ol STEP TIME DJl DJ2
14 l.4000E-02 3.4642E+Ol 0 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO
15 l. 5000E-02 2 . 7426E+OO 1 1.0000E-03 5.3221E-04 -2,0013E-05
16 l.6000E-02 5,l 722E+Ol 2 2.0000E-03 4.4538E-03 -5.1199E-04
17 l.7000E-02 7.2398E+Ol 3 3.0000E-03 l,5965E-02 -3.0147E-03
18 l.BOOOE-02 -2. 3117E+Ol 4 4.0000E-03 3.7705E-02 -8.9069E-03
19 l. 9000E-02 -S.4068E+Ol 5 5 , 0000E-03 6 . 4468E-02 -l.6887E-02
20 2.0000E-02 l . 2287E+Ol 6 6.0000E-03 8.7960E-02 -2.2416E-02
7 7 . 0000E-03 9 , 9425E-02 -2 . 3223E-02
MAXIMUM 8 8 . 0000E-03 9.4028E-02 -2 . 1405E-02
9 , 6000E+Ol
TIME OF MAXIMUM O.OOOOE+OO 9 9 . 0000E-03 7 , 1941E-02 -l,8146E-02
MINIMUM 10 1. OOOOE- 02 3.7520E-02 -l.1247E-02
-B.3027E+Ol
TIME OF MINIMUM 8.0000E-03 11 l.lOOOE-02 -i . 2319E-03 -2 . 5762E-05
12 l.2000E-02 -4 . 4130E-02 l . 1875E-02
*** LOADING NUMBER
2 OF 4 *** 13 l . 3000E-02 -7 . 7419E-02 2.0566E-02
14 1.4000E-02 - 9.6296E-02 2 . 4408E-02
DYNAMIC PARAMETERS 15 l.5000E-02 -9 . 8483E-02 2 . 3290E-02
NTS OT DAMPR 16 l . 6000E-02 -8.4354E-02 l . 9033E-02
20 l , OOOOE-03 O. OOOOE+OO 17 l.7000E-02 -5 . 5403E-02 1. 3650E_-p2
18 1.BOOOE-02 -l.6520E-02 6 . 0714E-03
INITIAL CONDITIONS 19 1.9000E-02 2.4923E-02 -4.9980E-03
NNIO NNIV
20 2.0000E-02 6 . 2330E-02 -l . 6433E-02
0 0
APPLIED ACTIONS MAXIMUM 9 . 9425E-0 2 2 . 4408E-02
NLN NEL TIME OF MAXIMUM 7 . 0000E-03 l . 4000E-02
l 0 MINIMUM -9 . 8483E-02 -2 . 3223E-02
TIME OF MINIMUM l . SOOOE-02 7.0000E-03
NODAL LOADS
NODE AJl AJ2 MEMBER FORCE TIME ·HISTORY FOR ELEMENT 1
2 2.0000E+Ol 0.0000E+OO STEP TIME AMl
0 O. OOOOE+OO O. OOOOE+OO
GROUND ACCELERATIONS 1 l.OOOOE-03 4.8184E-01
IGA
0 2 2 . 0000E-03 3 . 2844E+OO
3 3.0000E-03 9 . 0341E+OO
4 4 . 0000E-03 l . 647BE+Ol
FORCING FUNCTION
NFO 5 5 . 0000E-03 2.1947E+Ol
7 6 6 . 0000E-03 2 .7 113E+Ol
7 7 . 0000E-03 3 . 1667E+Ol
FUNCTION ORDINATES 8 8 . 0000E-03 3 . 08l4E+Ol
K TIME FACTOR 9 9 . 0000E-03 2 . 2442E+Ol
1 0 . 0000E+OO O. OOOOE+OO 1 0 l . OOOOE-02 l.0842E+Ol
2 l . OOOOE - 031. 5000E-Ol 11 l.lOOOE-02 -7.8519E-01
3 3.0000E- 0 3 8 . 5000E- Ol 12 l .2 000E-02 - l . 3479E+Ol
4 4.0000E-03 1. OOOOE+OO 1 3 l .3000E-02 -2 . 4655E+Ol
5 5. 0000E-03 8 , 5000E- Ol 14 l . 4000E-0 2 - 3. 0138E+Ol
6 7.0000E- 0 3 l . 5000E- Ol 1 5 l .5000E - 02 -3 . 0917E+Ol
7 B.OOOOE- 0 3 0.0000E+ OO 16 l. 6 000E-0 2 -2. 7 5 91E+Ol
17 l .7 000E-02 - l .77 88 E+O l
NORMAL MODE SOLUTION 1 8 l. BOOOE - 0 2 - 4 . 09 57 E+O O
NMODES • 3
19 l.9 000 E- 0 2 8 .l366E+OO
OUTPUT SMl ,11CTl ON
2 0 2. ooo om- 02 1. oee1m+o1
I WU I Pl, NNO Nl!O MMC IMUM l, l (1t1 7 1QI 01
I I I I 'I' I MIQ 0 1~ MAX I MUM I, 00 0 01~ 0~
MINIMUM I. II 111 /If, I O I
'l' IMI~ OI' MINIMIIM I . hfllllll~ ll ~
188 189
Normal-Mode Method of Dynamic Analysis Chap.4 Sec. 4.10 Program NOMO for Normal-Mode Respon se

self-explanatory. Note that the first page of the listing repeats the vibrational analysis 20
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 -"0.
plotted (by computer) in Figs. 4. 11 and 4. 12. Figure 4.11 shows that the initial displace- - 10
ments 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. (\J

w O 10 15 20
5
DAMPR-o.o 0
D
T (ms)
z
- 0.10 ......
C <: -I 0

o.os -,
<:
-20
~ o.oo
D
z
,_ -o.os (a)
<:

, - 0 . 10 -
0
DAMPR-0-0

(a) 0. 10

DAMP'<=O.O
- 0 . 10
C
.: o.os
o.os
(\J

~ o.oo
~ o.oo 0
0 z
z ......
.....
<:
-o .os <: -o.os
-,
0

-0. 10

( b)

DAMPR-o.o ( b)
- 0 . 10
C Figure 4.12 Three -member plane truss: (a) applied force; (b) responses.
- o.os
(\J

~ o.oo 11.,mmplc 4.16 · 1

0 . NOMOPT let us reconsider the symmetnc p ane
z AN n ~ccond example usmg Programs). For this.structure dynamic responses due to the
I • 0 . 05 1111KN trom Sec . 3.8 (see Example 3.
< lnlluwing innucnccs arc desired:
N
~ 0 , 10 -
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
, Jlk·t·t·wi1H: 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 191
Chap.4 Sec. 4.10 Program NOMO for Normal-Mode Response

For these cases we use the values of E, p, L, and A stated in Example 3.5 , where the to half the force in part (c). For each load case the number of time steps NTS "'.' 40, and
material is aluminum and units are SI. lhc duration of time steps DT = 5 ms. The responses in Fig. 4.13(~) consist of the
To take advantage of symmetry, we must decompose the load for case 1 into
1,.translation at node 6 for all symmetric modes as well as truncation to mode. l .
symmetric and antisymmetric components and then analyze half the structure twice. Similarly, Fig. 4.13(c) shows the responses from all antisymmetric modes ~nd truncat10n
Thus, half the force must be applied using symmetric restraints on the plane of symmetry; lo mode 2 (for the same nodal translation). We see that the first (symmetric) and second
nnd the other half of the force must be applied with antisymmetric restraints. Of course, (untisymmetric) modes produce the major con~ributi_ons _to the responses. . .
the results of these two analyses must be added to find the total solution for the left-hand Figure 4.14(a) is a computer plot of the p1ecew1se~h~ear ~round acceleration m the
pa rt of the truss.
1, direction. Because this influence induces symmetnc mertial loads, we_ need only
Figure 4 .13 consists of computer plots showing the total step force of 100 kN in ,•nnsider symmetric responses of the structure. Time histories ~f the resultmg y-trans-
purl (a), symmetric responses to half the force in part (b), and antisymmetric responses liuion 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.
z 100
80
,.i.,, I. 00
w 60
'E
0
0
Z 40
z o.so
I- 0
< 20 I-
N <
et:
~ 0 w o.oo
so 100 ISO 200 .....J
(a) T (ms J w
(.)
(.)
<
-o.so
0
ALL MODES z
=i
ee 4
0
et:
<:!> -1 . 00 -
I
>-
<J) 3
w (a)
§ 2
z
'< I
N
~ 0
200 1 .oo
T (ms)

e
..= 0. so
0

::!5 0. 00
0
7
I •

..r - 0 ,50
('J
-.,
N n
~ 0
1 00

(b)
l1'l111m• 4.11 l'l1111r IIIINN (II) 11ppllt•d lor n •: (h) NY lllllll•ltl1 IIINjlOII Nl'N; (t' ) 111111
hy1111111•1rl1 111~p1111 \ 1 -
1!'11111111 4, 14 1'111111" IIII NN 111) ~11 1111111 111 l't h•111lio11; (h) ll'NJ1UII Nt'N
 192
Normal-Mode Method of Dynamic Analysis Chap.4
Chap. 4 Problems 193
REFERENCES
4.3-9. If joint 1 of the symmetric plane truss in Prob. 3.7-7 has an initial velocity Doi
I. Tim?she~ko, S. P. , Young, D. H. , and Weaver, W. , Jr. , Vibration Problems in in the x direction, determine the antisymmetric response of the structure.
Engineering, 4th ed., Wiley, New York, 1974 .
2 · Weaver, W., Jr., and Gere, J. M. , Matrix Analysis of Framed Structures 2nd ed 4.4-1. (a) Calculate the response of the plane truss in Fig. 3.12(a) caused by a step force
Yan Nostrand Reinhold , New York, 1980. ' ., 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
3. Rayleigh , J. W. S., Theory of Sound, Vol. I, Dover, New York , 1945.
4 fo rcing function applied in the y direction at joint 1. Example 3. l gives the
· ;ea~er, W., Jr., and Johnston, P. R. , Finite Elements for Structural Analysis properties, angular frequencies, and mode shapes for this truss.
rent1ce-Hall , Englewood Cliffs, N.J. , 1984. '
4.4-2. Assume that a hannonically-varying force P = Pi sin Ot acts in the y direction
5. Weaver'. W. , Jr., "Dynamics of Discrete-Parameter Structures," in Developments in
Theoretical and Applied Mechanics Vol 2 ed w A Sh p p at node 2 of the cantilever beam in Fig. 3. l 3(a). (a) Find the steady-state response
York, 1965, pp. 629- 651. ' . ' . . . aw, crgamon ress, New at the free end of the beam due to this loading. (b) Confirm the reciprocal theorem
6. Caughy , T· K ·, "Class·1caI N ormaJ M odes m
· Damped Linear Systems " J A / 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
; ech., Yo!. 27, 1960, pp. 269-271; also Vol. 32, 1965, pp. 583-588'. · PP· properties, angular frequencies and mode shapes for this problem are all given in
7. ?ss, K. A., '_'Coordinates which Uncouple the Equations of Motion of D d Example 3.2.
Lm~ar Dynamic Systems," J. Appl. Mech., Vol. 25, 1958, pp. 361-364. ampe 4.4-3. For the plane truss in Prob. 3.6-2, determine the response caused by a ramp force
8. Loui~,. J. J. ~- , "~amping in Structures-a Review," Engineer thesis, Department P = Pit/ ti applied in the negative y direction at joint 1.
of Civil Engmeenng, Stanford University, June 1976.
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
PROBLEMS 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.J-1. 4.4-6. Apply a step moment of magnitude M = M 2 in the z sense at node 2 of the beam
~~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~ in Prob. 3.6-5, and find the response of the structure.
'.rcc-v1biat1onal ~esponsc of the structure due to this impact. The properties 4.4-7. Suppose that a hannonically-varying force P = Pi sin Ot is applied in the y
angular frequencies, and mode shapes for this problem are all given in Exam~ direction at the midpoint of element 2 in Prob. 3.6-6. Determine the steady-state
pie 3. 1. nodal responses of the beam caused by this loading.
4.J-2.
Calc_ulate the free-vibration~! response of the cantilever beam in Fig. 3.13(a) to •1.4-8. Let the beam in Prob. 3 .6-7 be subjected to a ramp moment M = M2t/t1 in the
t~c sudden release of a static force Po in the y direction at node 2 Exa I 3 2 z sense at node 2, and calculate the response of the structure.
~1vcs the properties, angular frequencies , and mode shapes for this pr:e~. · •t4-9. For the symmetric plane truss in Prob. 3.7-7, obtain the antisymmetric response
l·o:· t~e pla~c truss in Pro?·. 3.6-2, find the response caused by initial displace, to a ramp force P = Pit/ t I applied in the negative x direction at joint 1.
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 · · · 4,.!l• 1. For the plane truss in Fig. 4. 7, calculate the relative response of joint 1 to a
the y direction at joint I. . . o an m1t1a ve oc1ty Do2 in rigid-body ramp acceleration Dg1 = a 1t/ t1 of ground in the x direction.
4..1-5.
J,'t t~corcc~'.anePc, m_trussthe innegative
o ,t
Pro?. 3.6-~, determine the response to the sudden release
y direction at joint I.
,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
4..1-6. this influence.
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. ,I. ~.J. Suppose that the ground at point 3 in Prob. 3.6-2 has a small rigid-body step
4..1-7. Let the bca'.11 in Prob. 3 .6-6 initially have a small positive rotation < at node 2 o rotation DR6 = 8, . Determine the response of joint 1 in the plane truss caused by
and a negative rotation O,o at node 3. Calculate the response of the s(/ucturc th1tl thi s motion.
,·esults. ,I,~ 4. For the plane truss in Prob. 3.6-3, evaluate the relative steady-state response at
4••1-H. AsN111ne. that 1111 initial posi1iv1; mClmcnt M O ut node I of the beum in Proh ..U, joint I due to a rigid-body ground acceleration Dg2 = a cos Ot in they direction.
1 ,I,~-~. Ass11n1t· th11t the support ut point 3 of the plane truss in Prob. 3.6-4 translates
Ih h<Iddutl y H·k·11,l•d Oh1111n till' l l'hJ)llllNl' o l till' ~II Ul lllll' l'ltll\l'd h y this l'Oll<li
lion ' · 111dl'Pl'lldl•n1ly in tlw y dircl·tion lll'l'mdin8 to the fun ction D R3 = d sin Ot. Calcu-
lnh· thr 1(·~11ltlnµ Ntl•1tdy 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 1l8fd•hody step rotulion " •n
•t ~-7. For the continuous b, .
cam 111 l'roh I c, c, f 1
rcspons0 due to a rigid-body , t I' I
' 111 llu iw1pons0 of the beam ut 11od1•
,
1 .. relative steady ~1111
• i ti 1111111.(lt.l tie
5
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 calcu-
lnling dynamic responses of structures. These techniques are usually more
c•I fl cient than the normal-mode method, unless modal truncation is used. Thus,
1•i1hcr approach involves approximations of different types. While the normal-
11wdc method applies only to linearly elastic structures, direct numerical integra-
lIon 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 formu-
l"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
11H <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

196
 10/
f><l1np11l11ll1111 with I . ,,11, 1111111111111111
1U8 1111 1111 N11111n11t nl ltllM11fnllo11 M111hod11
I 111111Hl' 11!ti111c 1 110111 pH·viow, rcsponscH by 11si11g 11pprox.imulion formu las.
11 1
11t SI
II t11m·
>OFlo system as these qunnlllll II w, 11111v, 11h 11l11tl• lhc luitiul act:clllrnli1111 l11
0. Fmn1 ,1il1111H111uthmls for doing this tll'i.) tk:sct'ib1.:d in thll fo llowing sections for both
111 ,1111nd MDOF problems .
.. J (
uo = -; /{l kllu ,·,io) I I)
:l • XTRAPOLATION WITH EXPLICIT FORMULAS
(wltil'h is obtained from Eq . (1) at f ime to. Similarly'
l ) yilllds . for a MDOF structure ' I q
11 1•xplicilextrapolation procedure consists of expressing the displacement at
Do= M-1(Ao - S Do - C Do) ( I1 111111• 1 in 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
If
Vl' i.' lor.muss matrix Mis singular ' the vec tor OO .m Eq. (4) may be taken as u 111111
till' 111 llll' displacement at time ti as the acceleration:
1 (1)
i•igure 5. 1 shows a graph of the nu . 1 . Uj = (At)/ui- 1 - 2uj + Ui+ 1)
S l)OF system. This plot is represent d menca solution for the response ol n
Ihough it may actually have slight d.:
ast. a s.n:iooth curve in the u-t plane, i.'Vl II
i con mu1t1es. The symbols Uo hklt implies a SDOF system with a unifonn time step At. Similarly, the
"1 i, "1, UJ' i, 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,. 1 l11l'ily at time tj may be approximated as
t11kl.ln to be of uniform duration t:,,.r altho jh etween tim~s !j ~d tr: 1 is usuul lv . 1 (2)
HI ii y. 'l'hll objective of the '. 1 . ug su~h a specialization is not ne,·1•" Uj = !:,,.t (ui+l - Uj-1)
numenca integration process is to calculate tlu 2
',11hstituting iij and Uj from Eqs. (1) and (2) into Eq. (5 .1-1) and rearranging the
/I
11 ~ult produces
(3)

",, ,
P'(t;) ~ 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
t' 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, .. 1
D1 ""' (iit) 2 (Di- , - 2Di + Di+1)

11'(11111•1• ~. I S11l111lt111 hy tllil'l'I 111111 111111111 l11h' f''11 1l1111

 199
198 lllr!I I N111111111t nl lt1h1111ntlon Muthoda Chnp I x 1111pol11tlu11 wllh I xplh It I 111111111111

111 th is exu111ple we c 1111 cxp1li1111 tlw 111111111111 111

. , r ti , sti t'fness k and the period
f\l1llls o 1e
I(~r)2 M t 2 ~( cJ01, 1 • A+(I/) f I
2

1 m = }5_
w2
= k(.!_)
21r
(a)

A*(r1) = A(tJ - [s - (;t)2 MJ01 l(A 1

)2 M - 2~rc]Dj,
_ Li - o, so the initial acceleration is
/I.I 111111• 10 0 , we have uo - o -
In addition , the generalized form of Eq. (5) is Po P, (b)
' l .. 2 iio=-=-
D- , = Do - Do At + 2 Do(At) m m
which may be used to start the procedure. If there is no damping and the n111 NN - ) with R =Pi.To start the solution, we apply Eq. (5)
hh h Is drawn from Eq. (5 . 1 3 o
matrix is diagonal , a so-called nodewise solution may be devised. That is, lho 1o, 11h111in
solution for a particular displacement D;,i+ I in Eq. (8) becomes
(At)2 "'
1
U - 1 = 2iio(A.t)2 = 2m 20
2
Pi ( T ) _ Pi (21r)
- 2k T
2

20
2
= O 04935~
. k
(!_)
(c)

D;,i+ I = M; A;' (ti) (i = 1, 2, ... , n)

. . Eq ( ) ·ves the displacement at time ti as
1111 II the central-difference predictor m . 3 g1

0- (:)2U-1] (;0)2(2;)2 ~ -
which is simply a multiplication of the effective action Af (ti) by the scul111 2
(At) 2/ M;. Therefore, high-speed computer core storage need only contain info, u, = (t:,.: ~1 - = U -I
mation for one displacement (or more conveniently, one node) at a time.
The central-difference predictor is probably the most widely used explkll Pi P1 (d)
= (0.09870 - 0.04935)k = 0.04935k
formula for solving structural dynamics problems. However, all expressions ol
this type have a critical time ste[l.., above which the solution becomes numerically . . Eq (c) A second application of the predictor at time t2 yields
"·hkh 1s the same as U- 1 m . .
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
u2 = C~
2f, -[k - c: 2Ju, - o}
value of the critical time step for this technique is

(At)c, =2 =~ (12) = [ (; )2(2;)2o

0
_ 0.04935) + c2)(0.04935)]~
w,, 7T

The symbol w,, in this expression denotes the largest angular frequency in the P, (e)
= 0 . 1925-k
analytical model, and T,, is the smallest period.
Key and Beisinger [3] applied the central-difference predictor to the linem . (3) is a lied in the same manner, and Eqs. (1) and
lu i•uch subsequent time step, Eq. p~ I .. 11
dynamic analysis of thin shells. In addition, Krieg and Key [4] showed that us in~ 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,
11 diagonal mass matrix improves the accuracy of the procedure. Successful use
The approximate values o u! .o ta1~e y ws lots of the approximate results
of this method in nonlinear analysis has also been reported by Key [5], who USl'd ,1l11ng with the exact values. In add1.t1on, ~1g. i'2 ~~ 1 PFor the scale used in Fig. 5.2,
111ti!lcial damping to control the inherent instability. Morino et al. [61, searchl•d 1111 both zero damping and a dampmg ratio o r. ..
Im the optimal predictor for systems of second-order differential equations and 1111· plotted curves are indistinguishable from exact responses.
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
I' uauplc 5· 2 . · Exam le 3 1 Recall that the
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.
l'luurc 5.3(a) shows the plane truss analyzed prev1ous~\m O 8A ~nd A.. 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
ltxuma,ll• 5.1 1
1llH'l'llOII Ill jnint I. The time histordy ~; th~.d~n:::~ ~~~~~1~Xrr~=:~~e ~~; di~tor ~ith 20
SuppoNl' thut an undnmpcd SDOF linear system is suhj1.·1.·t1.•<111> 11 step force of mugnit11<1,· , 11k11l11tr the displacements I), an i IISlllK , .
, 1 t:,. 1 '/'1/ J..O st111 tlllij lrnm rest.
/'1, Ht111 ll11g fw111 n•st. l.1·1 us rnkulute the 11pp1oxi 111111l• 11•sponse using lhl' l'1.•111rnl 1111ltoi 111 111111· Ht0ps ol durnt on . . ' ll'l'H for the free displacements at
1•111111 E x11111pll• \ I llw st1llm•ss 11111 1 11111N~ II 1111 i
1lllh·H·m·,· p11•dii101 with 1w,•11ty unifrnm 1i11w ~h·pN 111 d111111in11 t\/ '/'/'20
 201
200 oc. 5.2 Extra polation w ith Explicit Formulas
Direct Numerica l Integration Methods Chap. 8

TABLE 5.1 Response for Example 5.1

Using Central-Difference Predictor"
y
j Approx. u Exact u j Approx. u Exact u

1 0.04935 0.04894 11 1.947 1.951

2 0.1925 0.1910 12 1.800 1.809
3 0.4154 0.4 122 13 1.574 1.588
4 0.6960 P(t)
0.69 10 14 1.292 1.309
5 1.007 1.000 15 0.9804 1.000
6 1.317 1.309 16 0.6712 0.6910
7 1.595 1.588 17 0. 3944 0.4 122
8 1.815 1.809 18 0.1774 0. 1910
9 1.955 J.951 19 0.04 157 0.04894
JO 2.000 2.000 20 0.00034 0

"Tabulated values to be multiplied by P1/ k.

i------ - - -0.6L

(a)

P, 1.----------,

I' I
k Ol__---- - -- -+:-----~--- Tt.t2:o-- -
t,o

/ '

P,
(b)

()
I I
Fi~urc 5 .3 (a) Plane trnss; (b) applied force.

2
 Iteration w ith Implicit Formulas 203
202 Direct Numerical Integration Methods Chap. 5 Sec. 5.3

joint 1 are TABLE 5.2 Responses for Example 5.2 Using Central-Difference Predictor•

Exact D2
S= s[ - 0.36
0.48
-0.48]
1.64
M = m[ol 01] (f) j Approx. D1 Exact D , Approx. D2

0.2467 0.2432 0 0.0047 1

1
in which s = EA/Land m = 3.28 pAL/6. In addition, we have 0.9322 0.05845 0.06831
2 0.943 1
1.963 0.2930 0.2932
3 1.979
0.7593 0.7350
wf = 0.2 .:_ w~ = 1.8 .:. (g) 4 3.227 3.210
1.333
m m 5 4.574 4.556 1.375
5.928 5.901 1.962 1.932
Using the first expression in Eqs. (g), we can relate the mass constant m to the stiffness 6
7.188 7.148 2.365 2.374
7
constant s and the fundamental period 7; , as follows: ' 8.223 8.179 2.557 2.598
8
8.868 2.627 2.662
2 9 8.898
2.667
m = 0.2s ( 7i71") (h) 10 9.107 9.111 2.679
2 11 7.839 8.382 2.720 2.653
5.330 6.315 2.416 2.462
~.l tim~ to = 0 the initial displace~ents an~ :~locities are (D 1) 0 = (D2) 0 = (D 1) 0 = 12
1.952 3.221 1.420 1.787
( >2)0 0. Therefore, Eq. (5.1-4) gives the m1t1al accelerations as 13
- 1.929 - 0.51 87 -0.2636 0.4618
14
- 2.191 -1.333
Do= M - 'Ao = ;[~ ~][~] = UJ ~ (i)
15
16
17
-6.024
-10.06
-13.70
- 4.556
- 8.592
-12.33
- 3.772
-4.683
- 3.128
-4.454
- 16.50 - 15.43 -5.049 -5 .128
Before beginning the step-by-step procedure, we apply Eq. (10) to find 18
-17.49 - 5.238 -5.319
19 - 18.07
D- 1 = I2 Do(At) 2
= I[ ~ (1i)
1
]
2 0 m 20
2
20 -18. 15 -18.22 -5.467 - 5.333

• Tabulated values to be multiplied by P, / s.

= ~ [~] /;s (2;)2(~~)2 = [ o.2;67] 9 (j)

exact values are also listed. Figure 5.4 depicts plots of these responses for both the
lJsl ng this result in the matrix form of the predictor [Eq. (8)] at time ti produces 11pproximate and the exact analyses. Here we see that the curves for the approximate
responses deviate slightly from those for the exact responses.
D, = (~
2

[~ ~]{[~] - (:;)2 [~ ~Jn-J

= {[o.~34] _ [ o.2;67]} 9 = [ o.2;67]
9 (kl
6.3 ITERATION WITH IMPLICIT FORMULAS

The technique to be explained in this section is called the predictor-corrector

which is lhc same as D- , in Eq. (j). Applying Eq. (8) again at time t2 yields method [6, 7]. In each time step an explicit formula (a predictor) is first used to
02_ (Llt)
m
2
[1 O]{[P']
O 1 0
_ s[-0.48
0.36 -0.48] ~ [1
1.64 D, + (Llt) 2 o I D,
o] } 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
.. (2!.20.) _!!_ ( 271") [1 - 0.08883] + 2[0 .2467] ~
2 2
l'o r use in nonlinear problems, where physical properties can change in each
0 .2s Ti O + 0.1184 o s

, 0.05845 o
J
0.4496 -I [0.4935]} ~ = ro.9431 p 1
s Io.os845 -;
l ({')
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
{'
In .suhsl·~ru.e nt ~·n lculutions Wl' upply Eq . (8) rcpctili vllly 10 find rhc response at cnch ol
(1)
ihl tl111111111111g 11ml' st111lo11s. Noll' ih nl lhc 11111ss n1111il x 111 thi s l'x11111pll• is diugonnl so 11q
(K) dm•s not 1t•q11i1 t• sol vl11i.: sl 11111l11111l·t111 s l'q1 111th111s •' , · in whic h 1i1 is the ve lt)cily Ht the preceding time station 11 (see Fig. 5 .1). This
1<1·~1111~ 1111111 lhl~ 11pp1 wd111111t• 111111lys is l111 /l 111111I /l , 111r 11 rv,•11 111 'f'nhk' 'i l, wlwtl' lrn nnlla , know n lo nu,rn·ri rnl 111111lysts as till' t1'<tf>l'zoidul rttfr, implies that the
 205
Sec. 5.3 Iteration with Implicit Formulas
204 Direct Numerical Integration Methods Chap. 6
acceleration in the stepJ s taken to be the average of ii1 and ii1+,. Similarly , the
D,
displacement u + at the end of the step is approximated by the trapezoidal rule
1 1

10 ~s Approx.
with the expression
(2)

Exact
where the velocity in the step is taken to be the average of u1 and u1+1· Substi-
5~ tution of Eq. (1) into Eq. (2) yields
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~ (i > 1) (4)
s (u1+ 1); = Q1 + Hu1+1);- 1At1
(i 2: 1) (5)
(u1+1); = R1 + Hu1+i); At1
l (i 2: 1) (6)
10 ~s (ii1+1); = -(~+1 - ku1+1 - cu1+1);
m
where
(7)
15 ~s

and
R1 = Uj + ! Uj At1 (8)
20 "1
s This iterative procedure is not self-starting because it requires a supple-
mentary formula for determining the first estimate of u1+ 1 in each time step. After
(a)
evaluating the initial acceleration from Eq. (5.1 -3) , we may start the iteration for
u
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) (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,
fl'l1411n1~.•I we l:HI\ 11se till' slil'htl y IIHll'l' cl,,borntc fo1 mul11 thul is valid only for uniform time
(h) /l ,
 207
206 Direct Numerical Integration Methods 111 6.3 Iteration with Implicit I ormul tJII

steps: I q1111tion (16) is the same as Eq_. ( l) of the aver.age-acceleration method, but Eq.

(i = 1) 1 l /) is slightly different_from its clount~erpart

We will apply the lmear-acce era ion me .
:~o~f~
~3!anner analogous to that
E (l)
. h Because Eq ( 16) 1s the same as q · ,
This expression spans two equal time steps from tj- I to tj+i (see Fig. 5.1) and li11 the average-acceleration appro~c : . . . is the same as that in Eq.
utilizes the midpoint acceleration at time tj. 1111 recurrence expression for the ith iteration of u;+ i . I for ii· in
Equations ( 10) and ( 11) are called explicit predictors because they provido 1 Il To obtain a direct relationship between U1+ J an~ Uj+ '.' we so ve J+,
estimates of Uj+I in terms of previous values of u and u. On the other hand, Eq I q ( 16) and substitute the result into Eq. (17), which yields
(1) is referred to as an implicit corrector that yields an improved value of u11 1 .
U1+1
. ) /::,. + I .. (f:.t )2
= u1 + H2u1 + Uj+I ( 18)
t1 -gu1 ,
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' l lt11s, we form the recurrence equation for the ith iteration of u1+1 as:
corrector. * I ( • ) D,.t (i ~ 1) (19)
An iterative type of solution requires some criterion for stopping or chani,t
(u1+1); = Rj + 'i U1+1 ; ,

ing the step size, such as a limit on the number of iterations. A convenient " lll'l'C
method for measuring the rate of convergence is to control the number of (20)
R *=
J
Uj + 32 Uj• ut;
A
+ oI Uj.. ( uA(j )2 •
significant figures in uj+ 1, as follows:
I Ill' formulas given earlier [see Eqs. (9) and (10) or (11)] may be used agam to
I(uj+1); - (uj+1);-1 I < E,,I (uj+1)d (12)
1111 t the iteration in each step. . the recurrence equatio.ns for
where € 11 is some small number selected by the analyst. For example, an accuracy F MDOF structures, we can genera1ize .
of approximately four digits may be specified by talcing €,, = 0.0001. That lewl II• rntio~rinto matrix formats. Considering first the average-acceleration method,
of accuracy is used in the numerical examples of this section. For a MDOI• 1w replace Eqs. (4) through (8) by
structure, we use the length of a translational or rotational vector in Eq. (12), (i > 1) (21)
(Dj+1); = Q1 + Hi>1+1);- 1 D.tj
which is equal to the square root of the sum of the squares of its components . (22)
(i ~ 1)
Another implicit approach for approximating responses is known as tht• (D1+1); = Rj + HDj+l)i D.tj .
fl11ear-acceleration method. As indicated by its name, this technique has thl' (i ~ 1) (23)
(D1+1); = M - l(Aj+I - s Dj+l - C D1+1);
ussumption that the acceleration varies linearly within each time step. Thus , nn
expression for a during the step f:.tj may be written Wlll'l'C
(24)
f I

a (t ') = aj + (iij+ 1 - uj) f:.t. CIJ >

J ,1 1111
(25)
where t' is measured from the beginning of the step (see Fig. 5.1). If till' Rj = Dj + !Dj D.tj
uccelcration varies linearly, the corresponding velocity and displacement wil I
vary quadratically and cubically with time. Therefore, I\IHo, Eqs. (9), (10), and (11) are supplanted by the matrix expressions
2 (D1)1 = Do + Do f:.to (j = O; i = 1) (26)
•( ')
U t
. + Ujt
= U;
.. + (.,Uj+ I
I
-
.. ) (t f:.t·
Uj
') ( 14) • o" A (i = 1) (27)
2 J (D;+1)1 = D1 + i utj
und . o·· A (i = l) (28)
(D·+ 1)1 = Dj-1 + 2 j t
') ' method' Eqs. ( 19) and (20) generarize to
U (t = Uj +•I+"
Ujl Uj
(t')2
2
+ ("Uj+ 1 -
") (t')J
Uj f:.tj
6
( 15)
•
'lli 11 IIurl y' for the linear-acceleration
* 1 (D. ) 1::,.
(D;+1)1 = R1 + !! 1+1 ; t1
(i > 1)
-
(29)
At the end of' the step the velocity and displacement become
,;, ,, 1i1 I 1Ui1 I il1,,)At1 ( 1(1)
Rj D1 I ~ I>, A.11 I ~ i>1 (Atj}2 (30)
1111d
. ( I) (l\) ( tl) ('JC,) (J'f), uml (2K) apply to both methods.
( I '/l I II (' IHll'Ht', I•.qH. 2 t • t 2 t t
 208
1>1111 I N1111111rl1 11l l111n111n1lu 11 Mulhod1 1111111111111 wi th lmpllc It I 1111111,I H 201

II is we ll known thar !ht· 11111111 11111 I« 111111111 111t•lhod is somewhur 1111 / '1 I ,., '/'
uccurure than the average-accch.-111111111111r 1I111d I / l llowever, it has been shu 1:q. (•I ): ( Ii iJ I (I () I/ ~ I H) I() O Oi! HHO
• Ill Ill
181that the former technique is onl y , ,111,/11/1111,tl/\ .1·tahle. Therefore, as in th
1

= 0 .04816 ~
2
central-difference procedure, the solut ion diwqit·11 if the time step is too 111111
Eq . (5): (111 ),1 0 I 0 .!MHXO
k ( ' '120/' ' ) ( 21r)
T k
On the other hand, the average-accelcnition nicthod is unconditionally .1·tuli/ 2
although less accurate. As with the central-difference predictor, a nodt·wiM
iterati ve solution of a MDOF problem is feasible if the mass matrix is diago1111I J',q. (6)·. m ( I - 0 .048 16) = 0.95 18 ~
(11 1)4 = ~
Example 5.3
It 1111 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
We shall now repeat Example 5. 1 using the iteration methods described in this sect,1111 II~ II l(csults for 20 time steps are given m a . '
Recall that m = k(T / 211')2, as given by Eq . (5.2-a), and that the in itial acceleration 111 II i,1!11111~ "1 required in each step .
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 TABLE 5.3 Respons e for Example S 3 Using Iteration Methodsa
time step us ing Eq . (9) to estimate the velocity at time t 1 = At= T/ 20, as followh -
Average-Acceleration Method Linear-Acceleration Method
(u1)1 = uo + f1o tlt = 0 + (~) (:0) = 0 .05 P~T
II; Approx. u n; Approx . u
Then the displacement at time t 1 is fou nd from Eq. (5) to be
4 0.04816 4 0.04855
2 4 0. 1880 4 0.1895
(u1) 1 = Ro + I (u. 1)1 At= 0
2 + 2k
0.05 (Pi T (2T1r) )
2
k
= 0. 04935 P1 4 0.4061 4 0.409 1
20 •I 4 0.6813 4 0.6861
1 isasthe same as u1 in Eq . (5 .2-d). Next, we obtain the acceleration at time t1 from
wh1d(6) '> 4 0.9873 3 0. 9936
11q, 11 4 1.294 3 1.302
4 1.573 3 1.581

(ui)1
1
= -(P1 - ku1)1
Pi
= -(I - 0 .04935) = 0 .9507 -Pi
'
H
1/
3
3
1.797
1.944
3
3
1. 803
1.948
m m m 10 3 2.000 3 2.000
Por the second iteration in the first time step, Eqs. (4), (5), and (6) yield II 3 1.959 3 1.955
1:2 3 1.827 3 l.8 18
Eq. (4): 11 3 1.6 14 3 1.60 1
(ii1h = (I + 0.9507)P1T P1T
- = 0 .04877- 1,1 4 1.343 3 1.326
40m m
l:i 4 1.038 3 1.0 19
( ) =O+ = O.O48 l 3 ~k
2 JI,
Eq. (5): 2 4 0.7300 4 0.7 105
2 0 .04877 (Pi T ) ( 21r)
Ui 2k 20 T )/ 4 0.4478 4 0.4299
IH 4 0.2 187 4 0.2047
Eq. (6): - Pi .
(u1)i = -(I - 0.04813) =
Pi
0 .95 19 -
11) 4 0.06497 4 0. 05665
)0 5 0.00 126 5 0.00025
m m
Third iteration :
• l'uhulatcd values to be multiplied by P1/k.

Eq. (4): P1T P1T

(ti1 h = ( 1 + 0.95 19) 40 m = 0 .04880-;- . . ~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
Eq. (5): ,111 1111111.: ly and with few er iterations.
(u1 )1 = 0 + 0.04880(P1T2)(21r)2 = 0.048 16~ 1,,hlt• '> . L
2k 20 T k
Eq . (6) :
0 .048 16) 0,9 5 18 Pi I• \llllljlh.• 5.4 truss in Example 5 .2 which was shown
Ill 1, 1 11~ 111111ly1.c hy_ itcra~ive methods thct ancthe relationships m =· 0 .2s(Ti / 21r )2 and
l•o1111h lll' rn lio n: 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 211

t b For t_he a~erage-acceleration method, we begin the first iteration in the first timo TABLE 5.4 Responses for Example 5.4 Using Iteration Methods•
s ep Y estimatmg the velocities at time t 1 with Eq. (26), as follows:

(D,), =Do + D0 At= o + [~](:) (~) = [ois] P; j

Average-Acceleration Method

D; Approx. D 1 Approx. D2 D;
Linear-Acceleration Method

Approx. D, Approx. D2
Next, we find the displacements at time t 1 from Eq. (22) to be
I 6 0.2369 0.01168 5 0.2400 0.00834
(D,), =Ro+ i(D,) 1 At= O+ [0.05]
0 0.PiT..2s (T..)(27T)
40 T..
= [0.2467] [l
O s
2 2
3
6
5
0.9121
l.932
0.08378
0.2961
5
4
0.9220
1.948
0.0761 8
0.2941
4 4 3.173 0.7005 4 3.192 0. 7165
. (cl 5 4 4.515 1.263 3 4.535 1.298
;;_ic(~;; i::: same as D1 in Eq. (5.2-k). Then the acceleration at time t, is calculated from 6 4 5.851 1.864 3 5.875 1.902
7 4 7.081 2.357 3 7.112 2.373
8 3 8. 102 2.648 3 8.139 2.629
- 0.48] [0.2467] !j_} 9 3 8.807 2.734 3 8.838 2.697
1.64 0 s
10 3 9.101 2.693 3 9.110 2.670
11 4 8.691 2.596 4 8.742 2.618
(f)
12 4 7.150 2.409 3 7.146 2.482
For the second iteration in the first time step, Eqs. (21) , (22), and (23) produt·t• 13 4 4.451 1.950 3 4.394 2.031
14 6 0.9388 1.014 5 0.8498 1.024
Eq. (21): (D,h = Q0 + icoAt={[~]+ [o.9;12]} ~ (~t)
1) 1
15
16
4
4
-3.005
-7.028
-0.4398
-2.193
3
4
-3.106
-7 .142
- 0.5508
-2.386
- 4.002
[1.911 2] !j_ (Ti) = [0.04778 JP, I;
17 4 -10.82 -3.841 4 - 10.97
= 18 4 -14.09 -5.009 3 -14.27 -5.034
0 . 11 84 m 40 0.002960 m
19 3 -16.53 -5.537 3 - 16.73 - 5.425
Eq. (22): (D,)2 =Ro+ ~(D,h
2
At= 0 +[0.04778 J P,Ti (Ti)(21r)2
0.002960 0.2s 40 Ti
20 3 - 17.87 -5.533 3 - 18.02 -5.395

"Tabulated values to be multiplied by P, / s.

= [0.2358 J!j_
0.01461 s As a second analysis, we apply the linear-acceleration method, using Eqs. (29) and
Eq . (23): (D,h = M- 1(A1 - s D,h t \ti) 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

= _!_
m O 1
[1 O]{[P']
0
_ s[ -0.48
0.36 - 0.48] [0.2358
1.64 0.01461
J!j_}s 1111111• llCCurate than those for the average-acceleration method.

0.9221 JPi
-_ [ 0.08922 ;;; 11 ,4 DIRECT LINEAR EXTRAPOLATION

Such ,7ul~ul~tions arc re~eated u_ntil convergence of displacements is obtained to within 11 lht· equations of motion for a MDOF structure are linear, it is possible to avoid
lour s1gn1f1cant figures (m the sixth iteration). At that stage the values of the displ .. lti 11t1ion of implicit formulas for numerical solutions. Instead, direct linear
111t·11ts cnlculatcd from Eq. (22) are · au
, l111polation procedures may be devised for both the average- and linear-
(D i)t> _ f0.2369
0.01 168
l P1
,\' (g)
11111•lt·r11tion 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
'l'11hil• "i.il t:011t11i11s tilt• displ11t•1•111<'11ts /) 1 1111<11) 1 Im· lwi•uty linw stt•ps , 11s wl'II Its fill'
1 h111111l' lo dl•vl'lop lhl' i1
wn.•111l•nt11l tt•chniqm• , which upplies to both linear and
1111111lw1 ol 1tt•111t101111 11, 101 1•111 h ~ti•p
111111l1rn•111 p1ohlt·1n~ F111 lht·1111111t•, 11 ,., ft•11s1hlt• to st·l up simultuncous equations
 213
212 Direct Numerical Integration Methods !loc. 5.4 Direct Linear Extrapolation

for incremental accelerations, velocities, or displacements. Here we will use .. 4 t::,.D Q (11)
incremental displacements as unknowns and solve a pseudostatic problem for !::,.Di = (Cl.ti )2 i- i
each time step. To save space, only MDOF structures will be considered; and
2 - (12)
any SDOF system becomes merely a special case. 6.Dj = 6.t 6.Dj - Rj
At time ti (see Fig. 5.1), the damped equations of motion for a MDOI• 1

linearly elastic structure are Now substitute Eqs. (11) and (12) into Eq. (3) to obtain
M Di + C Di + S Di = Ai ( I)
_ 4_ 6.D- _
M [ (!::,,.tj)2
QljJ + C [~
l:::,,.tj
!::,,.D- - R} + S 6.Di
;J = Cl.Ai
= ti + D.tj, the equations of motion become
Similarly, at time ti+ 1 J 1

M (Dj + D.Dj) + C (i>j + D.Dj) + S (Dj + 6.Dj) = Aj + D.Aj (2 l t 'ollecting terms, we rewrite this equation in the form
(13)
Subtraction of Eq. (1) from Eq. (2) produces the incremental equations of S 6.Di = 6.Ai
motion as 111 which
(3) (14)

These equations will be used for both the average- and linear-acceleration
algorithms developed in the following discussion. 1111d
6.Ai = 6.Ai + M Qi + C R1 (15)
For the average-acceleration method, the incremental velocities obtained
by the trapezoidal rule at the end of time step D.ti are thus, the pseudostatic equations represented by Eq. (1~) are to be solved fo~ the
• 1 .• •• •• 1 •• lnl'l'emental displacements 6.Di in each step. Then the mcrement~l accelerations
6.Di = 2 (Di + Di+1) D.ti = Di D.ti + 2 6.Di D.ti .\i) and velociti~s ai>i mar. be found using Eqs. (11) and (12) . Finally, the total
1
Similarly, the incremental displacements at the end of the step become vulues of Di+1, Dj+l, and Dj+l are
_ 1 • • _ . I ' (16)
D.Di - 2 (Di + Di+ 1) !::,.ti - Di D.ti + 2 6.Di !::,.ti (5) D1+1 = Di + 6.Di
Substitution of Eq. (4) into Eq. (5) yields i>j+1 = i>j + !::,,.l)j (17)

D.Di
• 1 ..
= Di b.ti + 2 Di(b.ti) + 4 b.Di (!::,.ti)
2 I .. 2
(6)
Dj+l = i>j + anj (18)

Solving for the incremental accelerations in Eq. (6) gives In summary , the procedure for obtaining ~ynamic ~esponses of a MDOF
11 ucture consists of the following calculations m each time step:
.. 4 4 . ..
!::,.Di = (!::,,.ti )2 !::,.Di - !::,.ti Di - 2Di (7)
Ulrcct Linear Extrapolation by the Average-Acceleration Method
Substituting this expression into Eq. (4) produces
Determine Qi and Ri from Eqs. (9) ·and (10) .
. 2 . i 1:ind Sand !::,.Ai using Eqs. (14) and (15) .
!::,,.D = - tJ.D- - 2D-
1 Lltj J J
(8)
Solve Eq. (13) for t::,.D1.
Now we define vectors Qi and R1 that contain only combinations of i>i and i>1• Ca lculate 6.Di and t::,.D1 with Eqs. (11) and (12). .
as follows: \ Add the incremental displacements , velocities, and accelerations to pre-
n·ding values using Eqs. (16) , (17), and (18).
4
~ - A
u.11
D1+ 2D1 ('))
'l'lll lling now to the linear-acceleration ~ethod , we may again derive the
R1 2f>1 ( IOI 1111 1ri 111• 111ol vc lod ty vector froni the trnpc1.01dal rule to be

11~1111-1 llw~t· 1k•f1111tio11~. Wl' ll'wiitl' Eqi. (7) 1111d (K) il !j

6.1>, 1>1 6.11 I ; 6.1>1 6.11 (4)
 215
214 Direct Numerical Integration Methods 110. 5.4 Direct Linear Extrapolation

llowever, the incremental displacements in the step must be found f E 11 ~nmple 5.5
(5 .3- 17), as follows: rom ll 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
ADj = Dj Atj+ H2Dj + Dj+1)(Atj)2 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
= i>j Atj + !Dj(AtY + !ADj(AtJ2 ( 19)
1111 first time step. When doing so, we use notation for the undamped SDOF system, as
Solving for the incremental accelerations in Eq. (19) gives l11llows:
.. 6 6 - 4. .. P, Pi (a)
ADj = (Atj) 2 ADj - Atj Dj - 3Dj (20) Eq. (9): Qo = -At Uo + 2uo = 0 + 2- = 2-
m m
(b)
Now substitute Eq. (20) into Eq . (4) to obtain Eq. (10): Ro = 2uo = o
2 2
AD. · = -3 1.1AD ·
1.1 - 3D· 1 - -At1 D·
.. (2 1I -k = k + -4m- = k + 4k( - T ) (20)
- = 41.53k (c)
1 1 1 Eq. (14):
Atj 2 (At)2 21r T
(d)
Then define Mo = Mo + mQo = 0 + 2P, = 2Pi
Eq. (15):

Qj
* = A6 .
Dj + 3Dj
.. (22) Mo 2A A (e)
1.1tj Eq. (13): Auo = T = _ k = 0.04816,;
41 53

R j* -- 30· 2
j + 2Atj Dj.. (23)
Eq. (11):
.. = - 4 - Auo - -Qo = 4 (20)
Auo - A
P ) - 2-
( 0.04816-
1

(At)2 T k m
1111d rewrite Eqs. (20) and (21) as
.. 6 ( 27T m
2
= 4 -20) -k ( 0.04816-
k
A) - Pi = - 0.04815-
2-
m m
A (f)

ADj = (Atj )2 ADj - Q/ (24)

AD = l_ AD - R* (25)
Eq. (12): Auo = tt 2
ti..uo - Ro = 2( ~) (o.04816~) - 0
J At·
J
I J
= ( r
40) (,
\0.04816;;; Pi) (2T )
7T
2
AT
= 0.04880--;;
(g)
Nl'Xl, we substitute Eqs. (24) and (25) into Eq. (3) and collect terms to find
S* AD1 = AA1* (26) A A (h)
Eq. (16): U1 = Uo + Auo = 0 + 0.04816,; = 0.04816,;
i11 wllkh
. . . PiT AT (i)
S* = S + -
6 - M+-C
3 (27) Eq. (17): u, = uo + Auo = 0 + 0.04880-
m = 0 .04880-
m
(Atj)2 Atj
p p p (j)
1111d ii., = ii. +Mo=_!. - 0.04815_!. = 0.9518_!.
Eq. (18): 0 m m m
AA/ = AAi + M Q/ + CR/ (28)
l \quutll)IIS (~6), (27), and (28) are analogous to Eqs. ( 13) (14) and 15 t~11h• lhllt the values in Eqs. (h), (i), and (j) are the same as those obtained in Example
dt•11wd previou sly for the average-acceleration method Thus 'th , I ( ·1·'· ~ I hy the method of iteration. This procedure is repeated for each of 20 time steps.
Ii •. I' . . . . . · , , c proceu,rc tu V11lucs of the response from direct linear cx.trapolation by the average-acceleration
1 Ht 1 111~ 111 cxtrnpolat 1on hy the linear accclcralion method foll , th • , •
H(t•p H HS h •I'll' , , I • OWS C SH llll 11111hod 111l' listl·d in Table 5.5. Also given in the tublc arc slightly more accurate values
, t t c, tXl'cpt t 1111 hqs . (9) through (I~) nrc rc1
, lnccd by Fqs ("'JJ
lh1011µ,lt ( )8) . ' '. ~ , 11111p11tl•d hy tlw 1111~111 lll'l'l' krntion ,m,thml.
lln 1111 •1 I N1111111h ul l11t11urntlo11 Motl1111l1
•I
J t.\1>11 ()11
(,\)( l())J\ (),) \<,I)
l
J'-'
l•.q , ( 11 ), t.\1>11 (t.\/) '/' I () ,() 11 67 S

- j Avg.-Acccl.
-
Lin . At·t·t•I I Av11 At·wl. Li n.-Acccl.
1
== (4)(20) (0.2s) l0.2369 J!2 _[ 2J!l
0.01167 s O m
- I 0.048 16 0 .04855
=
11 l ,IJ;W 1.955
(27T)2m
(p)
2 0. 1880 0 .1895 12 1.827 1.81 8 = [ - 0.07976J!l
0.09459 m
3 0.4061 0.4091 13 1.614 1.601
4 0.68 13 0.6861 14 1.343 1.326 · 2 - (40)(0 .2369 JP,
0 .9936 15 1.038 1.019 Eq . (12): ilDo = -Lit ilDo - Ro= -Ti 0.01167 -s - 0
5 0.9873
6 1.295 1.302 16 0.7300 0 .7105 2

7 J.573 1.581 17 0.4478 0.4299 = (40)f0.2369 J0.2Pi(!l)

8 1.797 1.803 18 0.2188 0.2047 T, L
0.01167 m 27T
9 1.944 1.947 19 0.06500 0.05673
= I 0.04801 JP, Ti (q)
10 2.000 2.000 20 0.00130 0.00033
L0.002365 m
"Tabulated values to be multiphed by P,/k. 0.2369 ) P,
Eq. ( 16): D, =Do+ L\Do = 0 + [ 0.01167 -s
11:xnmplc 5.6
Now we will calculate the responses of the Ian .
= ro.2369 }!l (r)
t•xtrnpolution methods. From Eqs (5 2 h) pd (; ~ss m Example 5.2 by direct lim·111 L0.01161 s
nnd l>o {l, O}Pi/m. . . . an .2-1) we know that m = 0.2s(7; /211 )
. . . (0.04801 J PiTi
For the average-acceleration method in mat . f Eq. (17): D, = Do + ilDo = 0 + 0.002365 - m
llN shown. Thus, for the first time step we have nx orm, we use Eqs. (9) through ( l Kl
= [0.04801 )Ar.. (s)
Eq . (9): ~· = :/'· + 20, = 0+ 2[:J; = [~]; (kl

. .
0.002365 m
. [l]A ( -0.07976JA
Eq. (10): Ro = 2D0 = O (/) Eq. (18): D1 = Do + ilDo = 0 -m + 0.09459 -m
2
13q. (14): S= S + _ 4_ M = S + (4)(20)
CM2 ~mo
[1
~] = f 0.9202
L
}!l (t)

= s[ 0.36
- 0.48
-0.48] + (4)(20)
1.64 n
2[1
0 ~Jo.2s({~ )2
0.09459 m
1li•ll' the values in Eqs. (r), (s) , and (t) are the same as those found previously by iteration
1111\,rnmple 5.4. Following this time step, 19 other sets of calculations are performed in
= s[ 8.466 - 0.48 J 1111 ~111nc manner, and Table 5 .6 gives the results.
(Ill) ror the linear-acceleration method, we apply Eqs. (22) through (28) in place of
- 0.48 9.746 I q~. (9) through ( 15). Responses for this second analysis are also listed in Table 5. 6.
Eq . ( 15): LlAo = ilAo + M Qo 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, (11)
• NEWMARK'S GENERALIZED ACCELERATION METHOD

l!q . ( 13): L\Do = S , L\Ao = _ l_[9.746 0.48 J [ 2JA

82.28 0.48 8.466 0 ~ 111 this section we study a family of methods that were described by Nathan M.

,0 .2'.\69
().() 1167 ,\'
y, (o)
NL•wmark in a renowned ASCE paper. Two other variants, derived by Wilson
, I 111 . tmd Hilber et al., will also be discussed.
 219
N11w111n1k'1 Un11n1nll1111I /1,111111111111011 Mnlhoil
2.11 lllr1 I N111111111t •I l111nu,nllol1 Mnthocl•

rhl' p11111111t'll'I /I 1111'11 I 'l 1111111111, 1hr vc1111111011 of acceleration within the
1111 h·p 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 ,
·- Avcragc-Accclc111111111 I 1111•111 Al'CCICnllion
-
Method Mr lhod
I •I I 1) becomes (3)
j
Approx. D, Approx . /Ji ApjllOX. D, Approx. D2 I hi l111111ula is known as the constant-acceleration method, because the accel-
I 0.2369 0.01167 0 .2400 0.00835 1,1111111 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
2 0.9121 0.08374 0.9220 0.0762 1
3 1.932 0.2961 1.948 0.2941 11 1111' Huler's formula [see Eq. (5.3-10)] for velocity and the trapezoidal rule for
4 3.173 0.7006 3.192 0.7 164 II pl11l'cmcnt.
5 4.515 1.264 4.536 1.297 II we let f3 = L Eq. (2) yields
(4)
6 5.851 1.865 5.875 1.900
7 7.080 2.359 7. 113 2.370
8 8.101 2.649 8.139 2.627 lhlMa prcssion is the same as that in Eq. (5.3-3) for the average-acceleration
9 8.806 2.737 8.838 2.699 m, r/111rl . When we take f3 = i, Eq. (2) produces
2
10 9.099 2.697 9.108 2.675 Uj+I = Uj + Uj Di.tj + H2iij + U1+1)(Di.tj) (5)
11 8.689 2.600 8.741 2.623
12 7.149 2.409 7.145 2.482 1111111s case the formula is identical to Eq. (5.3-17) for the linear-acceleration
13 4.452 1.948 4.397 2.022
'"' 1!,11d . let us consider a MDOF structure and cast the Newark-/3 method into
Now
14 0.9392 1.013 0.8543 1.011
15 - 3.005 -0.4404 -3.104 -0.5576 111,,tii x l'ormat. In each time step we shall solve a pseudostatic problem for
16 -7.028 - 2.193 -7.143 -2.379 11111,•111cntal displacements by direct linear extrapolation , as in Sec. 5.4. For this
17 - 10.89 - 3.843 -10.97 -3.988 p111pnsc, Eq. (1) is restated in incremental matrix form, as follows:
18 - 14.09 -5.011 - 14.28 - 5.024 t:i.i>j = [(l - y)Dj + yDj+ tJ t:i.tj
19 - 16.53 -5.541 - 16.73 -5.428 (6)
20 - 17.86 - 5.544 - 18.01 5.412 = oj t:i.tj + 'Y t:i.Dj t:i.tj

"Tabulated values to be multiplied by P,/ s. 111 11ddition, Eq. (2) is restated in incremental matrix fonn as
• 1 " " 2
Di.Di = Di !::i.tj + [( i - {3)Dj + /3Dj+1](Atj)
Newmerk-/J Method • 1 " 2 .. 2 (7)
= Dj Di.tj + 2 D/!::i.tj) + f3 t:i.Dj(t:i.tj)
In hi s 1959 paper Newmark [8] . '-i11lving for Di.Di in Eq. (7) produces
intcgrut ion methods that h' ad b . generalized certain direct numcrirnl
een m use up to that t' H .. 1 1 . 1 .. (8)
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• ADj = f3 (Ati )2 t:i.Dj - .f3 Ati Dj - 2/3 Di
"

'l'I
Ujt I

u111
= + [(l - y)uj + 'Yii1+ 1] !::.tj
Uj

= u1 + u1 t:i.tj + [0 - f3)u · + 13 ... ](A )2

11 ll· p11ru111etcr 'Yin Eq . (I) produ ces numerical
u, + t ut1
'
. (or algor 'th . ) d
:t· t1111c step !::.11, If 'Y is taken to be less than ,
( 'l
-
. l .mic ar~1pi11g withi n
(Il ',111l11tilution of Eq. (8) into Eq. (6) yields

t,:01 = ~ ~
hn rnnvcnicncc, we define the vectors
11
6.01 - 7i 01 - u~ - 1) t.t f>
1 1
(9)

u 11ult1-1. On the other hand ·r . . i , an arttfic1al negative dampin):

. , 1 'Y is g1eater than ! such th · · .. (10)
uvoll 1 n111m·ncul damping ·1llogethcr ti I '. . 'mpmg ,s pos1t1vc. To
(I) l>t•t·o111t·s lhc trnpczoid~d rule. , 1c vu uc ol 'Y must he equal to j; und Hq
 221
220 Direct Numerical Integration Methods Newmark's Generalized Acceleration Method
Sec. 5.5

Rj = j 1\ + ( 2~ - 1) Ati f>i ii

Now rewrite Eqs. (8) and (9) in the forms

•• 1 A

ADi = /3 (Ati )2 ADi - Qi (1 2)

' 'Y A

ADi = /3 Ati ADi - Ri (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
A

S
1
= S + /3 (At)2 M + /3
'Y
Ati C (15)
0
LM,_JI;+, r·
and ~M,~

(16) Linear-acceleration method extended to Wilson-6 method .

Figure 5.5
We solve the pseudostatic problem in Eq. (14) for the incremental displacements
We solve Eq. (18) for AD 8 to obtain
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, .. 6 6 . - .. . (19)
AD = - A D e - - Di 3D1
velocities , and accelerations at time ti+ I are determined using Eqs. (5.4-16), e (At8 )2 Ate
(5.4-17), and (5.4-18). . . · t Eq (17) which results in
l'hen substitute this expression m o . '
Wilson-0 Method . 3 . Ate .. (20)
ADe = - ADe - 3Di - 2 Di
Ate
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 Now define
that the acceleration ii varies linearly over an extended time step Ate = 0 Atj, as Q* = ~ i).1 + 3:0j (21)
e b.te
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 . · b.te .. (22)
0 is 1.420815, which can be rounded to 1.42. Rt= 3Di + 2 Di
As with the Newmark-,B method, we shall construct the Wilson-0 variant
11 11d restate Eqs. (19) and (20) as
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 .. 6 AD Q* (23)
ADe = (b.te)2 e- e
at the end of the extended time step At0 , as follows:
, .. I ..
AD o = Di Ato + ~ AD o Ato ( 17) ADo -
3
AOo - R!
(24)
b.to
where the symbol AD 0 denotes a vector of incremental accelerations. Simi larly,
Eq . (5.4- 19) gives the incrl•mentul displucellll'llls us . • ( c; 4 1) nnd collect terms to find
'i11hHt11utl' l~qs. (23) 11ml (24) into 1~q. ·
• I .. ) I . l (25)
Al>11 D1 At11 I 21>1(A111) I ,.Al>11(A111l ( 18) st M>,1 AAt
 'l'J'J. ll!t It N11111n1 h ,11 l1111111111tlo11 Mnthod11 1 It ll N11111111 le 111 t.tnhlllty 111111 /\1, 111 111 y 223

'l'hl• st11111.! d ll!1111s in this l''lll1tfhtt1 1111 n NUMERICAL STABILITY ANO ACCURACY
st ~
,"l I A
(ul,i)
(1
) M I 11 ~tucly the stability and accuracy of various one-step direct numerical
11111 p,111tio n procedures, we may cast them into operator form [12], as follows:
und
U1+1 = A U1 + L Pj+I (1)
(J. I) 1ltlNl'xpression pertains to a SOOP system that might equally well be considered
where , 1111c natural mode of vibration for a MDOF structure. The symbol U1 in
I 'I (I) represents a column vector containing the three response quantities u1,
AAo = e AA1 (.'HI ,,, 1111d a1 at the time station t1. That is,
The
AD pseudostatic problem in
. Eq . (25) · so1ved ior
.: the mcremental
.
II·
Tl . is disp1acenw111 -
1en we ca~ determme the mcremental accelerations AD 8 from Eq (2 II
U1 = {uj, u1, u1} (2)

HiltI reduce them lmear1y by the formula · 111111 the vector U1+ 1 is similarly defined at time tj+ t to be
.. 1 .. U1+1 = {u1+1, u1+ 1, ii1+1} (3)
ADJ = -AD
e o (2'))
1111' t·oefficient matrix A in Eq. (1) is a 3 x 3 array called the amplification
Next, the incremental velocities AD1. and displacements AD are obt · d f 1t1tt11·/, that we shall examine to answer questions about stability and accuracy.
Pei · (5 4 4) d (5 L.l J ll.me ro11 1 I l1111lly, the symbol L denotes a column vector called the load operator, which
· 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 ' I 11111ltiplied by the load lj+ 1 at time t1+ i· If there is no loading, Eq. (1) simpli-
lll'l'l'lcrntions at t' t· p acements, veloc1t1es, and
1me 1+ 1 . 111 H to
(4)
Hllber-a Method
lt11 free-vibrational response.
To investigate the stability of a numerical algorithm, we apply spectral
. To imp~·ove control of numerical damping, Hilber et al. [11 ] introduced 11
p111 amctcr a into the equations of motion at time t· .: IIows.. ,l,•,·0111position [13] to the amplification matrix A, as follows:
J+ l , as 10
MD1+1 + C D,+1
. + (1 + a)S D1+1 - (5)
aS D1 = A,+ i (30)
ltt this equation A is the spectral matrix of A, containing eigenvalues A1, A2 , and
~11btracting similar equations of motion at time t from Eq. (30) prod ti .
tru.:rnmcntal equations , uces 1( \
1
In diagonal positions; and <I> is the 3 x 3 modal matrix of A, with
1 l~l·nvcctors <1> 1, <1>2 , and <1> 3 listed columnwise. If we start at time t0 == 0 and
M AD1 + C AD1 + (1 + a)S AD1 - aS AD,- ,= AA1 (3 1) l ll~ l' 11 time steps using Eq. (4) , we have
1
Ncl>lw substitute Eqs. (12) and (13) from Newmark's method into Eq (31) and Vn = AnjUo (6)
1
l'O cct terms to find ·
wlwrc the vector U0 contains initial conditions, and vector Un gives the response )

Sa ADJ= AA.a} (32) v11 l11cs at time t,.J' Raising the decomposed form of matrix A in Eq. (5) to the
111 which powe r n1 yields
(7)
(33)
1111d N11w let us define the spectral radius of matrix A as
,.,
,., AA<\i = AA1 + aS AD1 (J4 )
(r)A = max IA,\ (i = 1, 2, 3) (8)
1
Expressi<~ns for S and A~ were derived previously us Eqs. ( 15) and ( 16) F . I hl'II Eq . (7) shows that we must hnvc
lhl• llrst time step, we luke A Di 1 _ in Eq . ( ,,I). o · or (1 )A (9)
 224
Direct Numerical Integration Methods 1 11c. 5.7 Program DYNA for Dynamic Response 225

in o~d~r to keep the numerical solution from row· . , lhnn unity. Thus, one important type of error to be considered is amplitude
c~nd1_hon is known as the stability criterion for a !ven:!t:~h°;t boun~. Th~H
~ntenon to the constant acceleration method (see Sec 5 5) . ~p~ym~ ~111
time step for this conditionally stable approach to be. . ' we n t e cntical
J 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 negli-
(At)c, = _!!T.1T = 0.318Tn (J 0) 1,11ble by using sufficiently small time steps. Newmark [8] recommended a time
M i
l<.:p of duration equal to! or of Tn, which is the smallest period of a MDOF
This value is the same as that given b E (S 2 . Nltucture. However, a more commonly used time step is lit= Tn/10.
predictor For th r . Y q. · -12) for the central-difference
. 1
e mear-acce erat1on method, the critical time step is At first glance, the average-acceleration method appears to be the best
1 hoice among the implicit approaches, because it has no amplitude suppression

(At)c, = V3 '7T
[;_ = 0.551T.,, (JI l
1111d the least period elongation [12]. However, a small amount of amplitude
~uppression is desirable to reduce or eliminate unwanted responses of higher
On the other hand the spectral d' f; h 111odes in an undamped MDOF structure. But if the spectral radius of an
always unity. The;efore it has ra IU~f o~ t_e average-ac~eler~tion method is 11111plification matrix is too small, the response of the structure will be unduly
dilionally stable. ' no en ica time step and is said to be uncon
~uppressed, as in the Houbolt method [14]. Although the optimized Wilson-8
The matter of accuracy of a n · 1· . lmmulation produces a reasonable level of amplitude suppression, it also ex-
re lated to that of stability Figure 5 6 u:enc\ mteration procedure is closely lllhits an undesirable tendency to overshoot the true response in the first few time
system to an initial dis la~e . s ows t e un amped response of a SDOJ• ~1,·ps [15]. Thus, we conclude that the best choice of algorithm is the New-
Ihose labeled 2 3 ~ 4 ment uo. The_ curve labeled 1 is the exact result, and
11111rk-/3 method with the Hilber-a modification, as described in Sec. 5 .5 . Proba-
' ' an
Nlrat<.:s an amplitude increase (Al) that · r
shows no amplitude change and curve
Ut•cuuse of the stabTt 1 1
-' . .
i:~
represent various approximations Curve 2 d
· . emon
ie~ an unstabl~ algonthm. Curve .l
epicts an amplitude decrease (AD)
hl y, the optimum selection of parameters for this approach is to let a = -0.1,
/I 0.3025, and 'Y = 0.6.
· 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 ,7 PROGRAM DYNA FOR DYNAMIC RESPONSE
siblc algorithms are represen~e: b1yc cu~/;pech~ahl ~adml_s of umty. Other_admis
' w IC imp 1es a spectral radms Jess
LJ Now we shall describe a general-purpose program named DYNA that calculates
dy11umic responses of linearly elastic structures. This program includes vi-
--.. PE
111111 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
'
Al 1111• main program, responses may be obtained using either the normal-mode
1111•1hod or direct numerical integration. For the latter approach, we select the
AD Nt•wmark-/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 nor-
111111izution 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
1111(1111• ~.<, S 111h1l11 y 1111d un111m y 111 11111111,111·111 lr11t•p1111i1111 llll'lhrnlN 11111111111 111e111l' 1tll'lhod (ISOLVE 0) 1111d dirl'L'l numcricnl integration
 M11w1'11111 I 'I, I 1\,111111 1111111111111 t111 U\' NA•
IOI
V IIJ1 J 1'11111111111 VIII 11 11111
JIiii , lllllllllllilllllllll Oi Cl)IC ll Vl'('l(ll N
II 111 1 1,•,1w1'I Iii 11111,s IIIHIII X.
R1·ud 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 l1111•11oly 1'111M 11 ,,11111'lll ll'
IS11 hp111~111111~ 111111 dilll•1 1111 l'Vl'IY ' Yl'l' 111 , 1111l lilll'
~. I 227
S1'1' I H,
I) . SubNt'qlll'lltl y. 1lll' p111g1 11111 l'tllllpt1r0s !SOLVE with unity to
I , hh• whk h ly pl! of solulion prnn•d111t• 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, Sub-
1111111111111 NUMINT computes dynamic responses by direct numerical integra-
111111 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 applica-
llt 1111 us indicated by the second footnote below Flowchart 5.1. Note that Sub-
p111grnms 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. Vari-
1111111 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 three-
1111111hcr 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
,11111lys1•s. 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) m·c prnctically indistinguishuhlc from those in Fig. 4.12(b). This good
111111•l111lon INdm• to thl' facl that the ti11w s11·p 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 229

TABLE S.7 Computer Output for Example S.7

TABLE S.7 (Continued)
PROGRAM DYNAPT
INITIAL CONDITIONS
*** EXAMPLE 5.7: ~HREE-MEMBER PLANE TRUSS*** NNID NNIV
2 0
STRUCTURAL PARAMETERS INITIAL DISPLACEMENTS
NN NE NRN E NODE DOl D02
RHO
3 3 2 3.0000E+04 7.3500E-07 1 l.OOOOE-01 O.OOOOE+OO
2 l.OOOOE-01 l . OOOOE-01
NODAL COORDINATES
NODE X y APPLIED ACTIONS
l 0.000 NLN NEL
0.000
2 150.000 200 .0 00 0 0
3 150.000 0.000
GROUND ACCELERATIONS
ELEMENT I NFORMATI ON IGA
ELEM. J K 0
AX EL
l l 2 10.0000 ex CY
2 l 3 250.0000 0.6000 0.8000
6.0000 150.0000 DIRECT NUMERICAL INTEGRATION
3 2 3 8.0000 1.0000 0.0000 ALPHA= -0 . 1000 BETA 0.3025 GAMMA 0.6000
200.0000 0. 0000 -1. 0000
NODAL RESTRAINTS OUTPUT SELECTION
NODE NRl NR2 IWR IPL NNO NEO
l O l 1 1 2 1
3 l l
NODES: 1 2
NUMBER OF DEGREES OF FREEDOM • NDF
NUMBER OF NODAL RESTRAINTS: • 3
NNR ELEMENTS: 1
3
STIFFNESS MATRIX DECOMPOSED DISPLACEMENT TIME HISTORY FOR NODE 1
STEP TIME DJl DJ2
MODE 1 0 O.OOOOE+OO l.OOOOE-01 O.O OOOE+OO
ANGULAR FREQUENCY 4.1995E+02 1 l.OOOOE-03 6.9151E-02 O.OOOOE+OO
NODE DJl DJ2 2 2.0000E-03 -9. 7574E-03 O. OOOOE+OO
l 2.3137E-Ol O.OOOOE +OO 3 3.0000E-03 -8. 0921E- 02 O.OOOOE+OO
2 1. OOOOE+OO -2.4722E-Ol 4 4.0000E-03 -6 . 9989E-02 O. OOOOE+OO
3 0.0000E+OO O.OOOOE+OO 5 5.0000E-03 2.1718E-02 O. OOOOE+OO
6 6.0000E-03 8.8292E-02 O.OOOOE+OO
MODE 2 7 7.0000E-03 3.9844E-02 O. OOOOE+OO
ANGULAR FREQUENCY l.1677E+03 8 8 . 0000E-03 -7.5170E-02 O.OOOOE+OO
NODE DJ l DJ2 9 9 .000 0E-03 -l.2500E-Ol O.OOOOE+OO
1 8.6725E- Ol O. OOOOE+OO 10 l . OOOOE-02 -5.7660E-02 O.OOOOE+OO
2 -L 7149E-Ol l .OOO OE+OO 11 l . lOOOE-02 4.7566E-02 O.OOOOE+OO
3 0.0000 E+O O 0.0000E+OO 12 l .2000E-02 8 . 9567E-02 O.OOOOE+OO
13 l .3000E-02 5.1576E-02 O.OOOOE+OO
MODE 3 14 l.4000E-02 -l . 2272E-02 O.OOOOE+OO
ANGULAR FREQUENCY l.8 618E+03 15 l.5000E-02 -4 . 6102E-Oi O.OOOOE+OO
NODE DJl DJ 2 16 l.6000E-02 -2.5622E-02 O.OOOOE+OO
l l.OOOOE+OO O.OOOOE+OO 17 l.7000E-02 3.6195E-02 O.OOOOE+OO
2 -6 .0504E-Ol -6.1068E-Ol 18 l.8000E-02 8.5913E-02 O.OOOOE+OO
3 O.OOOOE+OO O.OOOOE+OO 19 l.9000E-02 6 . 2893E-02 O.OOOOE+OO
20 2.0000E-02 -2.9158E-02 O.OOOOE+OO
LOADING NUMBER l OF 4 ***
MAXIMUM l.OOOOE-01 O. OOOOE+OO
DYNAMIC PARAMETERS TIME OF MAXIMUM O.OOOOE+OO 2.0000E-02
JSOLVE NTS DT DAMPR MINIMUM -l.2500E-Ol O.OOOOE+OO
l 20 l.OOOOE - 03 O.OOOOE+OO TIME OF MINIMUM 9.0000E-03 2.0000E-02
 231
Program DYNA for Dynamic Response
230 Direct Numerical Integration Methods

TABLE 5.7 (Continued)

TABLE 5.7 (Continued)

DISPLACEMENT TIME HISTORY FOR NODE 2 *** LOADING NUMBER 2 OF 4 ***

STEP TIME DJl DJ2
0 O.OOOOE+OO l.OOOOE-01 l.OOOOE-01 DYNAMIC PARAMETERS
1 l.OOOOE-03 8.8727E-02 3.4123E-02 !SOLVENTS DT DAMPR
1 20 l.OOOOE-03 O.OOOOE+OO
2 2.0000E-03 6.8467E-02 -B.5095E-02
3 3.0000E-03 4,9368E-02 -1. 2309E-Ol
4 4.0000E-03 l.2394E-02 -4.4408E-02 INITIAL CONDITIONS
5 5.0000E-03 -5,2366E-02 6.8993E-02 NNID NNIV
6 6.0000E-03 -l,1040E-01 l.1863E-01 0 0
7 7 ,0 000E-03 -l,2015E-01 7.5778E-02
8 8 . 0000E-03 -8,9808E-02 -l,3763E-02 APPLIED ACTIONS
9 9.0000E-03 -6,2312E-02 -7 ,2918E-02 NLN NEL
10 l.OOOOE-02 -5.2629E-02 -4,8878E-02 1 0
11 l . lOOOE-02 -3,4703E-02 3.7619E-02
12 l.2000E-02 9.6841E-03 9.6294E-02 NODAL LOADS
13 l . 3000E-02 6,2708E-02 5.2826E-02 NODE AJl AJ2
14 l.4000E-02 9 .8915E-02 2 2.0000E+Ol O.OOOOE+OO
-6.1248E-02
15 l,5000E-02 l.1326E-01 -l.3161E-01
16 l . 6000E-02 l.0852E-01 -B.6674E-02 GROUND ACCELERATIONS
17 l.7000E-02 8.0000E-02 2.3038E-02 !GA
18 l.BOOOE-02 3.1105E-02 8.6592E-02 0
19 l.9000E-02 -l,48 6BE-02 5.3559E-02
20 2.0000E-02 -3 ,9867E-02 -2.6064E-02 FORCING FUNCTION
NFO
MAXIMUM l . 1326E-01 l . 1863E-Ol 7
TIME OF MAXIMUM l.5000E-02 6.0000E-03
MINIMUM -l.2015E-01 -l,3161E-Ol FUNCTION ORDINATES
K TIME FACTOR
TIME OF MINIMUM 7.0000E-03 l.5000E-02 O.OOOOE+OO
1 O.OOOOE+OO
2 1. OOOOE-03 1. 5000E-01
MEMBER FORCE TIME HISTORY FOR ELEMENT 1 8.5000E-01
STEP TIME AMl 3 3.0000E-03
4 4,0000E-03 1. OOOOE+OO
0 O.OOOOE+OO 9 . 6000E+Ol B,5000E-01
1 l.OOOOE-03 4.6852E+Ol 5 5.0000E-03
6 7.0000E-03 l,5000E-01
2 2.0000E-03 -2.5370E+Ol O,OOOOE+OO
3 3.0000E-03 -2,4355E+Ol 7 8.0000E-03
4 4,0000E-03 1. 6683E+Ol
5 5.0000E-03 1. 2893E+Ol DIRECT NUMERICAL INTEGRATION 0.6000
6 6.0000E-03 -2,9174E+Ol ALPHA= -0 .1000 BETA 0,3025 GAMMA
7 7.0000E-03 -4.2452E+Ol
8 8.0000E-03 -2,3751E+Ol OUTPUT SELECTION
9 9.0000E-03 -2.4867E+Ol !WR IPL NNO NEO
10 l.OOOOE-02 -4 .330 1E+Ol 1 1 1 1
11 l. lOOOE-02 -2. 3119E+Ol
12 l.2000E-02 3.4926E+Ol NODES: 2
13 l . 3000E-02 5.8727E+Ol
14 l,4000E-02 2.1256E+Ol ELEMENTS: 1
15 l . 5000E-02 -l.159BE+Ol
16 1. 6000E-02 l.3378E+Ol DISPLACEMENT TIME HISTORY FOR NODE 2
STEP TIME DJl DJ2
17 l,7000E-02 5.3656E+Ol
18 l.BOOOE-02 4.3667E+Ol 0 O.OOOOE+OO O,OOOOE+OO O.OOOOE+OO
19 l.9000E- 02 -4.5712E+OO 1 1. OOOOE-03 B,4579E-04 -l.OOOBE-04
20 2.0000E-02 -3. 2732E+Ol 2 2,00 0 0E-03 5.5055E-03 -8.5533E-04
3 3.0000E-03 l,7942E-02 -3,4897E-03
MAXIMUM 9.6000E+Ol 4 4.00 0 0E-03 3.8969E-02 -8,9831E-03
TIME OF MAXIMUM O.OOOOE+OO 5 5.0000E-03 6,3694E-02 -l,6261E-02
MINIMUM - 4.3301E+Ol 6 6.0000E-03 8.4309E- 02 -2,1959E-02
TIME OF MINIMUM l.OOOOE - 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
10 l.0000~ 02 1,07/ bE 01 1. 0521E-02
 232 Direct Numerical Integration Methods 233
Soc. 5.7 Program DYNA for Dynamic Response

TABLE 5.7 (Continued) DA'1PR=O.O

11 l.lOOOE-02 l . 3673E-04 - 2.3257E-03 - 0. 10

12 l. 2000E-02 -3 . 7755E-02 8 . 5393E-03
13 l . 3000E-02 -6 . 9338E-02 l . 8614E-02
14 l.4000E-02 -9 . 0054E- 02 2 . 3972E-02 - o.os
15 l.SOOOE-02 -9.5688E-02 2. 3788E-02
16 l . 6000E-02 -8 . 4382E-02 l . 9802E-02 ~ o.oo
17 l.7000E-02 -5 . 8635E-02 l . 354 2E-02 0
18 l. SOOOE-02 -2.3925E-0 2 5 . 5924E-03 z
19 l.9000E-02 l.4150E-02 - 3. 2004E-03 1- -o . 05
20 2 . 0000E-02 5 . 0421E-02 -l.1543E-02 <

MAXIMUM 9.4437E-02 2 . 3972E-02 , -o.

0
TIME OF MAXIMUM 7 . 0000E-03 l.4000E-0 2
MINIMUM -9 . 568BE-0 2 -2 . 3226E-02 ( a)
TIME OF MINIMUM l . SOOOE- 02 7 . 0000E- 03
MEMBER FORCE TIME HISTORY FOR ELEMENT l
STEP TIME AMl
0 O. OOOOE+OO O. OOOOE+OO
1 l. OOO OE-03 6 . 1168E-Ol
2 2 . 0000E- 03 3 . 4981E+OO
3 3.0000E-03 9 . 63 72E+OO - 0.10
4 4.0000E-03 l. 7047E+Ol C
5 5 . 0000E-03 2 . 2466E+O l
6 6 . 0000E-03 2 . 5649E+Ol - o.os
7 7 . 000 0E-03 2 . 8511E+Ol N
B B. OOO OE-03 2 . 9596E+ Ol
9 9 . 0000E-03 2.4238E+Ol ~ o.oo
0
10 l . OOOOE-02 l.1821E+Ol z
11 l.lOOOE-02 -1. 9340E+OO I- -o .os
12 l . 2000E-02 -l.2285E+Ol <
13 l . 3000E-02 -2 . 0281E+Ol
14 l . 4000E-02 -2 . 7599E+Ol , -o .
15 l . SOOOE-02 -3 . 1270E+Ol 0
16 l . 6000E-02 -2 .7 564E+Ol
17 l . 7000E-02 -1. 7935E+O l (b)
18 l. BOOOE-02 -6.9027E+OO
19 l.9000E-02 3 . 9111E+OO
20 2 . 0000E-02 l.5351E+Ol
DAMPR=O . O
MAXIMUM 2.9596E+Ol
TIME OF MAXIMUM 8 . 0000E-03
MINIMUM -3.1270E+Ol - 0 . 10
TIME OF MINIMUM l .SOOOE- 02 C

- o.os
N

~ o.oo
0
l r.101111l)IC 5.8 z
,_ -o.os
N1•x1, we repeat Example 4. 16 (the symmetric pl ane truss), with the same number 1111d <
Ni11• of time steps. llowcvcr , the first load case in that example will be treated diffcrcnl ly N
ci " 0, 10 -
l11Nll•11d 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
<c 1
Nt1t'1>11tl wt• t11k1• ALPHA 0 . 1, BETA 0. 102'>, n11d ClAMMA 0.6. 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 Program DYNA for Dynamic Response
235
Direct Numerical Integration Methods Sec. 5.7

20 z 100
80
"'C.
- 10 w 60
D
0
(\/ z 40
I--
w O 20
0
0
s 10 15 20
<(

z T (ms J (\/

I--
~ 0 so I 00 l so 200
0 T (ms)
<( -; (a )
-,
<(

-20

NEIIMARK

~ 4
(a) E

DAMPR =O.O
Q. J 0 '::;: I

(\/
-, 0 200
D JOO
_': 0 OS ( b)
T (ms J

(\/

~ o. oo
0
z
I-- E 4
E
< - o.os NE I/MARK
..,
0
lD 3
w
- 0. I 0 8z 2
~l
(\/

( b) 6 0

Figure 5.8 Three-member plane truss: (a) applied force; (b) responses.
Figure 5.9 Plane truss: (a) applied force; (b) symmetric responses; (c) antisym-
Figures 5.9(b) and (c) show computer plots of the symmetric and anti-symmcl r rt metric responses.
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 Responses to the piecewise-linear ground acceleration in Fig. 5.lO~a) .are/~ot!~d
1111d (l') . They all mulch fairly well , even though the lime step is rather long . Nole th111 I11 Fi 5 ·, O(b.) i-or the two cases of DAM PR = 0.0 and 0.1. They were o tame y e
lhl• 11pprniti111011· rl·sponscs in Fig. 5.9(c) huvc II signilirnnl f)Ni()(/ 1'lo111,1ar/011 rclutiw lo !:I· •n · tll)JWOttch using the same .intcgrn11on
I lilhcr · parameters as before .hThese plots
l are
d
lhut i11 Fjg, 4. 1 l(l') , Also, llw llilhl•r plol in Jllg , 'i .')(c) hus II rmlkl·11hk (//ll/1llr11tl1• · ' , J" 4 14(1
li11IINti11guiNht1hlc from those 111 ·tg. , · 1 , w ) hich were found by t e norma -mo e
1·111i1111·.1·.1'io11 rd111iv,· tu llw N1•w111111k plot , 11s it should 11111(hotl ,
 23 1
2311 1111 ,11 1 N1111111lt11l li1ln11111tlm1 M11thod11
...
('l I 00
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,
"
-'
z
l

o. so
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
0 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
<C
n::
w o.oo 11, Nponse of Axisymmetric Solids," Comp. Methods Appl. Mech. Eng., Vol. 4,
...J
w t~11 2, 1974, pp. 195-218.
(.)
(.)
<C I, f\1111 !110, L., Leech, J. W , and Witmer, E. A., "Optimal Predictor-Corrector Method
0
-o.so 1111 Systems of Second-Order Differential Equations," AJAA J., Vol. 12, No. 10,
z
::i 111/4, pp. 1343-1347.
0
n:: ' l l111oshcnko, S. P. , Young, D. H ., and Weaver, W., Jr., Vibration Problems in
'f' - 1 . 00 -
>-
I 1111i11eering, 4th ed., Wiley, New York, 1974.
N1•wmark, N. M. , "A Method of Computation for Structural Dynamics," ASCE J.
/•1111. Mech. Div. , Vol. 85, No. EM3, 1959, pp. 67- 94.
(a ) '' 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 Integra-
I . 00 tl1111 Methods," Earthquake Eng. Struct. Dyn., Vol. 1, No. 3, 1973, pp. 283-291.

6
~ o.so
[ II
llllbcr, H. M., Hughes, T. J. R., and Taylor, R. L. , "Improved Numerical Dissi-
pulion for Time Integration Algorithms in Structural Mechanics," Earthquake Eng.
,\'Imel. Dyn. , Vol. 5, No. 3, 1977, pp. 283-292 .
0
llnthc , K. J., Finite Element Procedures in Engineering Analysis , Prentice-Hall,
I'
~ o.oo l~n~lewood Cliffs, N. J. , 1982.
0
7 11 lll1rc, J. M., and Weaver, W., Jr., Matrix Algebra for Engineers, 2nd ed.,
.... ll1 ooks/Cole, Monterey, Calif., 1983.
< -o.so I loubolt, J. C. , "A Recurrence Matrix Solution for the Dynamic Response of Elastic
11
..,
N
Aircraft," J. Aero. Sci. , Vol. 17, 1950, pp. 540- 550.
Cl

- 1 . 00 -
,, 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.

( b)
PROBLEMS*
Figure 5.10 Pla ne t russ : ( a) ground acceleration; (b) responses.
I.• I. Confirm the approximate results of Example 5 .1 in Table 5 .1.
1•2. Repeat Prob. 2.7-2 using the central-difference predictor.
1.•.\. Repeat Prob. 2.7-3 using the central-difference predictor.
REFERENCES Repeal Prob. 2. 7-4 using the central-difference predictor.
i-~. Ri;peat Prob. 2.7-5 using the central-difference predictor.
I. Dnhlqulst, G. G., "A Special Stabilit Pr · l (1, Repeat Pmb . 2.7-6 using the central-difference predictor.
Nord. 'l'lrl.1·kr. !11.f. Bl'hw1d/i11g Vol 3 YI 9;1blcm ~2or L111car Multi step Methods,"
). K i , ' · • · , pp. 7 43 + Solutions for pmhlc111s in this ch11p1c1 til l' , 111hcr tedious and should be handled using a
· rill,\ , I,. I)., " Unconditional Stnbilit · N ,: ·
.I. Ap11!, M1•c ·lt Vo l 110 N11 2 I 117; 111 111111,111•111 1111w ln11;11n1tlon MulhodN" jt• 1-1111111 Vlllllplltl'I,
'' · · · , '/ ., , pp. •I I I ,111 '
 238
Direct Numerical Integration Methods l l111p. 5 Problems 239

5.2-7. Repeat Prob. 2.7-7 using the central-difference predictor.

t 4-4. Repeat Prob. 2.7-4 using direct linear extrapolation by the average- and linear-
5.2-8. Repeat Prob. 2.7-8 using the central-difference predictor. acceleration methods.
5.2-9. Repeat Prob. 2.7-9 using the central-difference predictor. ~.4-5. Repeat Prob. 2.7-5 using direct linear extrapolation by the average- and linear-
5.2-10. Repeat Prob. 2.7-10 using the central-difference predictor. acceleration methods.
5.2-11. Confinn the approximate results for the plane truss of Example 5.2 in Table 5.2 ~.4-6. Repeat Prob . 2.7-6 using direct linear extrapolation by the average- and linear-
2 acceleration methods.
5, -12. Rep~at Pro~. 4.4-3 (plane truss) for 20 time steps, using the central-differenl·c
predictor with !::.t = Ti /20 = r1 • ~.4-7. Repeat Prob . 2.7-7 using direct linear extrapolation by the average- and linear-
5.2-13. Re~at Prob. 4.4-4 (plane truss) for 20 time steps, using the central-d·rn . acceleration methods.
1
predictor with !::.t = Ti./20. erenlr ~.4-8. Repeat Prob. 2.7-8 using direct linear extrapolation by the average- and linear-
acceleration methods.
5.3-1. Confi_nn the iterativ~ results of Example 5.3 in Table 5.3 for both the avera 'l' ~.4-9. Repeat Prob. 2.7-9 using direct linear extrapolation by the average- and linear-
and hnear-accelerat1on methods. g acceleration methods.
5.3-2. Repeat Prob. 2.7-2 using iteration by the average- and linear-acceleratiori ~.4-10. Repeat Prob. 2. 7-10 using direct linear extrapolation by the average- and linear-
methods. acceleration methods.
5.3-3. Repeat Prob. 2 ·7· 3 us1·n g I·terat·10n by the average- and linear-acceleration .4, 11. Confirm the approximate results for the plane truss of Example 5. 6 in Table 5. 6,
methods. using both the average- and linear-acceleration methods.
5.3-4. Repeat Prob. 2 .7-4 us1·ng 1·terat1on
· by the average- and 1· J ~.4-12. Repeat Prob. 4.4-3 (plane truss) for 20 time steps, using direct linear extrapo-
methods. mear-aece eration
lation by the average- and linear-acceleration methods with !::.t = Ti / 20 = t1 •
S.3-5. Repeat Prob. 2.7-5 using iteration by the average- and linear-acceleration .4-13. Repeat Prob. 4.4-4 (plane truss) for 20 time steps, using direct linear extrapo-
methods. lation by the average- and linear-acceleration methods with !::.t = Ti. / 20.
5 ..1-6. Repeat Prob. 2. 7-6 using iteration by the average- and linear-acceleration ,4, 14. The equations of motion for the average-acceleration method may be converted
methods. to the form: S Di+ t = Ai+•· Derive e.:pressions for the matrices Sand Ai+t to
5.3-7. Repeat Prob. 2.7-7 be used in this approach.
using iteration by the average- and linear-acceleration
methods. ,•l-15. The equations of motion for the average-acceleration method may be converted
5.J-8. Repeat Prob. 2.7-8 using iteration by the average- to the form: C Di+ 1 = Ai+1 . Derive expressions for the matrices C and Ai+ 1 to
methods. and linear-acceleration be used in this approach.
5.3-9. Repeat Prob. 2.7-9 using iteration by the average- 8,4- 16. The equations of motion for the average-acceleration method may be converted
methods. and linear-accelerat ion to the form: M Di+1 = A1+ 1 . Derive expressions for the matrices M and Ai+ 1
to be used in this approach.
5.3-10. Repeat Prob. 2. 7-10 using iteration by the average- and linear-accelerat ion
methods. 4 17. 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
5.3-11. Confirm the iterative results for the plane truss of Example 5 4 · T bl 5 4
.· bh h . .mae . , and !::.Ai to be used in this approach .
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
4 Iti. The incremental equations of motion for the average-acceleration method may
be converted to the form: M !::.D1 = !::.Ai · Derive expressions for the matrices
average- and ltnear-aceeleration methods with !::.t = Ti./20 = ti. y
M and !::.Ai to be used in this approach.
!i.J-1.l Re~eat Prob. _4.4-4 (plane truss) for 20 time steps , using iteration by !ht·
·I 19. The equations of motion for the linear-acceleration method may be converted to
average- and linear-acceleration methods with !::.t = Ti./20.
the form: S * D1+ 1 = A *1+1 . Derive expressions for the matrices S* and AJ+1 to
5.4- 1. Confirm the a_pproximate results of Example 5.5 in Table 5.5 ~ be used in this approach.
b h !ht·
average- and hnear-aeceleration methods. or ot •I lO. T he equations of motion for the linear-acceleration method may be converted to
!i.4-2. Rlipcat P.rob. 2.7-2 using direct linear extrapolation by the avcrnge- and li tll'llt the form: C * D 1 , 1 = Aj, •· Derive expressions for the matrices C* and Aj+ 1 to
Ut'l't•lerut 1011 met hods. he used in this approach .
5.4-J , ,,t,l l . 'f'hc cqu utinns of motion for the linear-acceleration method may be converted to
Rl' Pl' HI Jlmh . '2.7 l usinR cli1vt·t lint•11r cxtrnpol111ion hy tlw uwrugc- und lint•ru
1111'1 hrnJi,
lll'l'l' ll'l llt IOII 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

an I
Direct Numerical Integration Methods

5.4-22. The incremental equations of_ motion for the linear-acceleration method may h
d AA* to the form:. c * .AD·J = AA*
converted J . Denve
to be used m this approach.
. expressions
. for the matrices ( ••
6
5.4-23. The incremental equations of ~otion for the linear-acceleration method may h
converted
M* d AA* to the form·
b . AD
. M* . J -- AA*
J •
D enve
· expressions
· for the matri<.'l'"'
an I to e used m th1s approach.

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
Framed
m th1s approach. 1+1 usu
5.5-2. iThe equations
. CD. _of' motion for
. the Newmark-/3 method may be converted to 1hr Structures
. 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+ ,. Denve express10ns for the matrices M and A. t I
used m this approach. 1+ 1 o ll
5.5-4. The incremental equ,atio~s of motion for the Newmark-/3 method b
verted to the form· C AD · - AA D . . may econ
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
vcrtcd to the form· MAD - AA D . may e con
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)
1 rnnpared 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 curvi-
linear) 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
1•1l'ments [2] for discretizing and analyzing continua. However, a framed struc-
1111 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~l·d 101 1•x111Hpll·i, 1111d pmhll'IIIN iu pn•vio11s (' haptcrs .
 242
Framed Structures I oc. 6.1 Introduction 243

y
Figure 6. l(a) shows a straight continuous beam, subjected to a parallel-
1•oplanar 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
/ /-x
z
/
2 /
1
.tI
.
(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
z
/ moment in the z sense. If a force applied to a beam has a component in the x
(a) i direction , the resulting internal axial stress must be combined with the flexural
(b)
~tresses due to bending, even though the analyses are uncoupled .
y
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

-I
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.
--x T he rigidly connected plane frame in Fig. 6. l(c) carries a general-coplanar
/ /
t2 (d)
•r( of forces in the x-y plane, while applied moments act in the z sense. Resulting
z
(c)
i
/
3
·- 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. Displace-
1111•11ts 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 1·1~s at its hinged (or universal) joints that are concurrent in space. The forces
z . t2 ell t111g on members and joints may be in arbitrary directions, but any moment
I e -... ..... uppl k·d to a member must have its vector normal to the axis of that member. The
/,.-; 1 ,t 1r11No11 f'or this restricti on is that a truss member is incapable of resisting torque.
(e)
(I) 6 I llNpl111·1·mcnts at u free j o int ure three components o f trans lation in the x, y, and
d111·l'lio11s. Dlll: to npplicd louds, u 111c1ulwr of II space tru ss may have local
l<'l1t11n• 6.1 Typl·~ ul 1111 11 , 1 ,
pl11111• 11111111•; (cl) wlcl (1•) :,1:,:./',',',',lH·~ur,1,· ~, (11) rn,111i11111111- 111·11111, Ch) pl11111· hllNN, (l') lln1111· 111 two pr111c1p11I plum·:, of lwmlinA "" Wl'II "" 1111 uxiul for1·c from the truss
• "• " P1111· 1111111•
11111tly1i1N
 245
244 Framed Structures Chap. 8 !Jue. 6.2 Plane Frames

Figure 6.l(f) shows a space frame, which is the most versatile and com- r,
plicated type of framed structure. Locations of joints and directions of members O 12 I Sym.
4L l
2
are completely arbitrary, and the structure is designed to carry forces that are O 6L (1)
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
K' = [~t !t] -------------~----
-,, O O l r,
o -12 -6L i O 12
consist of an axial force, two shearing forces in principal planes of bending, two
2L2 I O - 6L 4L
2
bending moments in the same principal planes, and a torsional moment. 0 6L
A typical prismatic member in each type of framed structure has stiff-
nesses, masses, and equivalent nodal loads drawn from those of the axial,
torsional and flexural elements described in Sec. 3.4. Such member properties y'
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 clirec-
tions. 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 Z, z'

flexural deformations. Because continuous beams and plane trusses have been
discussed before, no separate sections are devoted to them. However, Secs. 6.2 y
through 6.5 give further information about plane frames , grids, space trusses, k2
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.
x,, t kl
--- - - - --- - 7 k
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 k3 I
specialized computer programs that use these reduction methods in the dynamic IY,.
analyses of plane and space frames. I
I

6.2 PLANE FRAMES i3/ j1

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•'
/ lb)
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 l•'IMUl'I' (i,l Phuw I111111l' w, (1t) lm·nl
111r111I
dill'l 11 nnN, (h) ~lohnl dh cc1io11s.
 247
246 Framed Structures uc. 6.2 Plane Frames

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 where f4 =g Is = -2g3 + 3g2 /6 = (g3 - g2)L (7b)
section with respect to the z axis. We form the stiffness matrix for a plane frame . . ·n Eqs . (6b) and (7b) are drawn from Eqs. (3.4-1) and (3.4-17)
member by adding contributions from an axial element and a flexural element , l•1111ct1ons given 1

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.
lllr ax:~~::e
d
~=~: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 .
Similarly, we can form the 6 x 6 consistent-mass matrix in local direc lnrces Px· an y' as we as a . :Uanner similar to that for distributed
tions, as follows: , 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
140 x
Ille moment vecyt,or M , we need first derivatives with respect to x off2,hJs, a
0 156
I
I Sym. /ri to determine the following equivalent nodal loads:
I
pAL 0 22L
----- - --------, - -----
4L 2
I , _ dfi = M, dfi = 6M, (g2 _ g) (8a)
420 70 0 0 I 140
(2) PM2 - M, dx' L dg L
I
0 54 13L I O 156 , _ df3 = M, df3 = M (3g2 _ 4g + 1) (8b)
0 - 13L -3L 2
I O -22L 4L2 PM3 - M, dx' L dg '

Again, this matrix is obtained by adding terms from those in Sec. 3.4 for an axial , _ d/5 = M, dfs = _ 6M,(g2 _ g) = -pM2 (8c)
clement and a flexural element [see Eqs. (3.4-5) and (3.4-26)). PMs - M, dx' L dg L
1n addition, equivalent nodal loads in local directions due to distributed
forces on a plane frame member may be calculated from , _ df6 = M, df6 =M (3g2 _ 2g) (8d)
PM6 - M, dx' L dg z

Pl (t) = L f fTb'(t) dt (3) 111 which the subscript M replaces the subscript b. Note that no integrations are
11•quired for evaluating the~e formulas. d l d atrices to global (or struf-
In this expression, the vector of time-varying body forces b'(t) is ln order to convert stiffness, mass, an oa m . . R
1111111) directions, we compose a 6 x 6 rotation-of-axes transformation matnx '
b '(t) = [bx']
by•
(4 ) 1111 follows:
(9)
which contains forces (per unit length) in the x' and y' directions, as indicatt•d
in Fig. 6.2(a). The 2 x 6 matrix of displacement shape functions in Eq. (3) ha~
111 which
the partitioned form
Cx Cy~l
where
(5)
R = [ -;Y 6 ~J (10)

f. =
J
[/1O !20 hOJ (611)
..
. h
I 1·1111s m t e rows o
1
f th' 3 x 3 rotation matrix R consist of direction cosines
is · . S 35
andz' withrespecttox,y, andz. Asm ec. · , weap
ply matrix
I I lllWS X ' y '
'

'l'ht !'unctions in this submatrix for the j end are

u 11 11d its transpose to ob:i: RTK'R = [Kii Kik] (ll)
(6h) Kk1 Kkk
io whkh the dimensionless coordinate is ~ 1 ' // , . Also , for the k end we huvr
/, () 0 I M
(12)
I /, I,,
()
(711,
 248

and
Framed Structures Chap, 8 1
10c. 6.3 Grids

II' such forces happen to be in structural directions instead, a conversion to

249
l
111ember directions must precede use of the formulas in Eqs. (16).
Pb(t) = RTpb(t) = [Pb/t)J After stiffnesses, masses , and equivalent nodal loads for individual mem-
Pbk(t) (Bl bers 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
Perfonning the multiplications indicated in Eq
submatnces · (11) yields the stiffness the global indexes j l through k3 shown in Fig. 6.2(b) , as follows :
jl = 3j - 2 j2 = 3j - 1 j3 = 3j
(17)
EI [r1c; + 12c; Sym.J
Kii = L3z (r1 - 12)cxCy r1c; + 12c; kl = 3k - 2 k2 = 3k - 1 k3 = 3k
(14a)
-6Lcy 6Lcx 4L2 These indexes serve as subscripts for placing terms into the stiffness, mass, and
load matrices for the whole structure.
El [-r1c;- 12c; -(r1 -12)cxcy 6Lcy ]
Kkj = L3z -(r 1 - 12)cxcy -r1C Y2 - 12c2x -6LCx (14b) ltxample 6.1
-6Lcy 6Lcx 2L2 (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
EI, [r1c; + 12c; Sym.J lnnuence.
Kkk = -L3 (r I - 12)CxCy -r1Cy2 + 12c; (14c) First, we transform the body force to member directions, using a 2 X 2 rotation
6Lcy -6Lcx 4L2 111atrix, as follows:
Also, multiplications given in Eq . (1 2) produce the consistent-mass
. submatrices b' (t)=Rb(t)=[-:: : J[~]bysinOt=[: } ysinOt (a)
M = pAL [140c_~ + 156c; Sym.J
11 -16cxCy 140c; + 156c; (l 5 a) Then we apply Eq. (3) to obtain
420
-22Lcy 22Lcx 4L2

pAL [70c; + 54c; 16cxcy - I3Lcy]

(b)
Mkj = 420 l6CxCy 70c; + 54c; 13Lcx (15b)
I3Lcy - 13Lcx - 3L2
which are forces and moments in local directions. As the last step, Eqs. (16) give
2
M AL [140c; +
= L_ 156cY Sym.J (c)
420 -16cxcy 140c~ + 156c; Pb (t) = {O, 6, Lex, 0, 6, -Lex}~~ sin Ot
kk (15c)
. . 22Lcy -22Lcx 4L2
lly inspection, we can see that these results are correct.
OI course, K;k = K t and Mk = MT bee .
Fl11ully' Eq. ( l 3) results in ~he foni ' . aus; matnces Kan? Mare symmetric .

Pbi(t) = [wi~~;;~ ~t;,Jrs of equtvalent nodal 1;·:::

1
0 .3 GRIDS

p 1,3 'l'o make the analysis of a grid similar to that of a plane frame, we place the
p /14C', p/,~(\,l 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
P1ik(t) -
r p/,1<\· 1 1,;,,,-.
1
/1 t,r,
( 16h)

11~·11· Wl' i111ply 111111 hud y lon'l'S lll'l' >,:iv1•11 111 tlw d l11•1 lh111Nol lm·11l 1111·111lwr ;llWS
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 lwro1111•1-1
250
Framed Structures 110. 6.3 Grids 251
I
I
111 this matrix the dimensionless ratio r 2 is Glx/Ely, where Ix is the torsion
0 4L2 / Sym. , ,mstant of the cross section and l y is its second moment of area about the y ' axis .
K' = Ki KJk] = g_ __Q_ ___:-.§f___l]_ I I krc we combine stiffnesses from a torsional element and a flexural element, as
[
K;j K;k L3 -r2L2 0 0 Tr~L-2 __ _ 11ivcn by Eqs. (3.4-12) and (3.4-24). In this case the torsional and flexural terms
I
0 2L2 -6L I 0 ,ut· uncoupled for local directions.
4L 2
0 We can also write the 6 x 6 consistent-mass matrix in local directions as
6L -12 / O 6L 12
y 140r~
y' 0 4L2 I Sym.
M' = [M» MJk ] = pAL
Mkj Mkk 420
0
?Or:
-22L
0
156 1
---------- ---~-------
0 J 140ri
(2)
-----
x,. 0 - 3L2 13L I O 4L2
0 - l3L 54 J O 22L 156
where ri = J / A is the radius of gyration squared. Terms in matrix M' come
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

y
b' (t) = [ ::-] (3)

x,, 1'11is vector contains a distributed axial torque mx· (per unit length) in the x'
------ -- tk2. k1
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)
11111 now the functionsf2 and f 3 in the submatrix f; are switched, and the sign of
reversed. Similarly, functionsf5 and/;; in submatrix fk are interchanged, and
/1 ls
i3/ i 11 - - - - - -- - - - L - - - - -x tlw sign of f 6 is reversed. Thus , the submatrix tj becomes

z
/ 0
~]
(b) (4)
Odd llll'1t1hc r: (11 ) lm·ul dh n 111111N, Ch) 11luh11l dlt1•l'l in11~
,11111 !he s 11h111n1rix 1·4 is
 21>2 253
I rumod Struoturoe
1111, 0.11 Speoo Trusses

.I, 0 (JI
I {) ./11 h (8a)

O O
= log - (f' g )L
2
- 2g3 + 3g2
]
l
~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 w·th
1 t · .
the functions in Eqs. (4) and (5 ) H x " ou mtegrat11111 ut Sym.]
negatives of first derivatives with re~:~~~~ ;?r the mome~t My'. we need lh l40r;c; + 4L2 c; (8c)
rows of Eqs. (4) and ( ) b
. 5
P. of the functions m the scl'oiul
' ecause positive rotation . th , 22Lcx 156
negative slopes. Thus, we have s m e Y sense are equnl 111
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.
-M, dfi __ My' d/2
y dx' - T dg = --t-
6M ,
ce - g) (C,h) 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
-M ,d(-/5) = My, df5 - 2 used as subscripts for this purpose appear in Fig. 6.3(b), and we may calculate
PMS= Y dx' L dg - My,(3g - 2g) their values with Eqs. (6.2-17).

P M6 = - M , dfs = - My, dfs - 6My, ( 2 ICxample 6.2

Y dx' L dg - T g - g) = -p;,,,3 (C,d) 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
/\gain, these terms ~o ~ot require integrations.
111cmber.
Because the gnd hes in the x 1 For this example we can integrate in accordance with Eq. (6.2-3), as follows:
the sequence x-y-z, rotation-of-axes-~rfn:~i:d. the nodal displacements art• 111
plane frame. The resulting terms in stiff
become
~hons_ are the same as those fo1 n
ness su matnces for structural directio1111 pi,(t) = L Jr[~J #
1
0
cos Ot

2 Lb, (a)
Ely[(r2c; + 4c;)L Sym.J = {0, -L, 6 , 0 , L, 6} l2 cos Ot
2
Kg = U (r2 - 4)L cxcy (r2 c; + 4c;)L2 (711)
6Lcy -6Lcx 12 'l'hcn Eqs. (6.2-16) produce

K - E1y (- r2c; + 2c;)L2

kj-u -(r2+2)L2CxCy
-Cri + 2)L2c cy
(-r2c;+2c;)L2
X
6LCy
-6Lcx
J (7h)
pb(t) = {Ley, -Lex, 6, -Ley, Lex, 6} ~~ cos Ot (b)

[ It is easy to see that these values are correct.

-6Lcy 6Lcx -12

Kkk -
- ElY
U
he;+ 4c;)L
(r2 - 4)L2cxCy
2

(r2c; + 4c;)L2
S ym. J
Ciel 6 .4 SPACE TRUSSES
[
,. . - 6Lcy 6L<\ 12
A typical member of a space truss appears in Fig. 6.4(a), having universal hinges
S11111 larly, th1.; consistent-mass subrnal . • . f
lions lll'l' • • rices or 11 1•, Id riwmhcr in slructun1l din•c 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 8p11co I r11u n1 266

1
1-1111 11 and z ', in conjum;tion with axis x ' o f the member , de fine two principal
1ln111•H 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
' 1
0 0 I
Sym.

K' [K»
Kkj
J
K]k =EA
K kk
0 0 0 1
- --- --- --+- ----- -
L - 1 0 0 11
(1)
I
0 0 O I' O 0
0 0 olo 0 0
N111lll:ro terms in this matrix, pertaining to translations 1 and 4 in the x 1
ill11•l·tion, are drawn from Eq. (3.4-4) for an axial element. However, most of
1111' lt•ims in matrix K' are zero, because a truss member has no joint stiffnesses
1(
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 2
(a)
0 2 I Sym .
I
O O 2 1

y M' -_ 6pAL 1--0-01_

2______ (2)
I
k2 0 1 0 10 2
p o o 1 Io o 2
- k1 I Ir I l' we use terms for the x ' direction that were derived previously for the axial
-7' k
k3 / 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.
I
I Y;, Equivalent nodal loads in local directions caused by distributed body
l 111 l·l\S on a space truss member are calculated as
I
X (6.2- 3)

h11 this type of member, we can accommodate three time-varying components

111 lt11e loads,
z
(b) b'(t) = {bx• , by' , b,,} (3)
Figure 6.4 S
pace truss member: (a) local directions; (b) "lobal direct· I r 1111s in th is column vector are forces (per unit length) in the x', y 1 , and z 1
c, IOl!S .
,lhc•cti 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
r Iii f1 I (6.2-5)
 257
256 Framed Structures Chap. 11 Sec. 6.4 Space Trusses

However, we now have two 3 x 3 submatrices, In these expressions the symbol e represents a unit vector in the direction
Indicated by its subscripts. In particular, the unit vector ex· is

f; - I,!, - [ i !~} (4)

where
(8a)

Y·k ZJ
und C = _L C = ..J._
L
(8b)
Y L z

fk = l)/4 =
1 01 0OJ f4
0 (5) und
[0 0 1 L = Y xJk + yJk + z}k (8c)

where Ji = 1 - tandf4 = t,
as before. A similar description may be given for the unit vector ejp, using the coordinates
A space truss member may al b b· d . of points j and p. Collecting the three required unit vectors into a rotation matrix
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 R, we have
hccause a truss me b · · g. 6 .4(a). (Note that Mx' is omitted
m er 1s mcapable of re · t' · 1 (9)
!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 In which the direction cosines of axes x', y', and z' are listed in rows 1, 2, and 3.
. . . us,
For the purpose of transforming stiffnesses and equivalent nodal loads to
PM2 = M,· df, = M, , df, = - M,, structural directions , we form a 6 x 6 rotation-of-axes operator R as in Eq.
dx' L dg L
(6a) (6.2-9). Using matrix R as indicated in Eq. (6.2-11) , we find the member
stiffness matrix in structural coordinates to be
PM3 = -My• df1 =- My• df1 - My,
dx' L dg -y (6b)
CxCy 2
Cy
I
I
Sym.
I
PM5 = M,, di4 = M,, df4 = M,, - I
(6c) Cxe, eyCz e,
2 I
I (10)
dx' L dg L - -p M2 - - - - - - - - - - - - - ---,-- - - - - -
-e; - cxcy -exc, 1 c~
I
PM6 = -M , df4
Y dx '
= _ My•df4 _ _ My' =
L dg - L - PM3 (6d)
-exey -ey
2
-eye, i CxCy Cy2
_ 2 I c2
-cxe, -eye, e, I exC, Cye, z
As w '.' )1 c?~centrated. forces' these terms need no integrations.
Notice that only the direction cosines of axis x' (the axis of the member)
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 Influence these terms. Applying matrix R in accordance with Eq. (6.2-13)
<>A(n) shows such a point lying in\hne a,Pr~nctal plane of bending. Figurt- pmduces the equivalent nodal loads
Wht•ncvcr 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 , - A~ '() - [RTPbj(t)J (11)
Pb t - R Pb t -
()
RTPbk(t)
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~ :
which are now in structural directions . Of course, the detailed forms of the terms
_ Ce X C,,, 111 1>1i (t) depend on the nature of the loads o n the member and the location of the
Cr• - - ,, ('/a)
I
c,, X ('1,, J 1hlrd point p. Finally, the consistent-mass matrix for a space truss member
11 11d IN l11vniin11t with rotation of axes, jl1Sl HS it is for ll plane truss member (see
(',,• t•,· )( ,.. ('/I)) S1•1-. I , 'i ),
 259
258 Framed Structures Sec. 6.5 Space Frames

As with plane frames in Sec. 6.2 and grids in Sec. 6.3, we can assemble Px·] [cos if!] [-0.6889] (h)
stiffnesses, masses, and equivalent nodal loads by assessing these properties P'(t) = Z,: = sinO if! P (t) = 0.7;49 P(t)
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
Then the equivalent nodal loads in local directions become
using Eqs. (6.2-17).
At the end of Sec. 3.5, we mentioned that the analytical model for a truss, pplt) = fJP'(t) = PPk(t) = fIP' (t)
as described here, is not really suitable for a structure having only a few (i)
= {-0.3445, 0.3625, O}P(t)
members. For better accuracy in the dynamic analysis of such trusses, we
recommend the component-mode method discussed in Sec. 10.6. l'or which the subscript p replaces the subscript b. Also, their counterparts in global
directions are
Example 6.3
PPi(t) = RTPPi(t) = PPk(t) = RTPPk(t)
In Fig. 6.4(b) a time-varying force vector P(t) acts at the midpoint m of member i. This (j)
force is directed toward point p, which is located on they axis. As in Fig. 6.4(a), the = {-0.4365, 0.1091, - 0.2182}P(t)
third point p, along with points j and k, defines the x '-y' principal plane of bending. . b E (11) Because of the central location of P(t), the equivalent nodal loads
Dimensions appearing in the figure are ns given y q. ·
nrc the same at both ends of the member.
Xj k = Xk - Xj = 4 Yik = Yk - Yi = 3
(a)
Zjk = Zk - Zj = 2 YiP = YP - Yi =2
6.5 SPACE FRAMES
Find the equivalent nodal loads for both local and global directions due to the concen-
trated force. . · f
frame which is similar to
1:i ure 6.5(a) depicts a typical member z o a space . . ' .
From the given dimensions, the length of the member is ~ truss member- but now rotations are included at Joints J and k. As befor~,
II space ' . h , , lane and the x, -z, plane. Six
L = \!xJk + Y% + zJk = V29 = 5.385 (b) the principal planes of bendmg are t e x -y p . t f tr s-
and the direction cosines for its axis are 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
X·k Z ·k the 12 X 12 stiffness matrix for local axes is composed of the o11owmg
Cx = { = 0.7428 cY = Yik
L = 0 .5571 c, = { = 0.3714 (c)
1mbmatrices:
r1I,
For this example the unit vector eiP is Sym.
0 12/,
eip = [O I OJ (d)
12/y
E 0 0 (la)
Applying Eq. (7a), we obtain KJ; = L3 0 0 0 r2 L2ly
e,, = [ -0.3714 0 0.7428) = [ _ _ ) 0 0 -6Lly 0 4L2 ly
0 4472 0 0 8944 (c)
V(0.3714) 2 + (0.7428)2 · 6Lfz 0 0 0 4L2 I,
0
and Eq . (7b) gives
1111d
e y• = [-0.4983 0.8305 - 0.2491) (f)
0 0 0 0 0
-r1lz
Altogether, the rotation matrix becomes 0 0 0 - 6Lfz
0 - 12/,
0.5571 - 12/y 0 6Lly 0
[ 0.7428 0.37141 E 0 0 (lb)
K'kJ -
- -
2
- r2L ly 0 0
R = - 0.4983 0.8305 - 0.2491 (g) LJ 0 0 0
0.4472 0 0 .8944 6Uy 0
2
2L ly 0
0 0
{) 0 0 2L2 l ,
which contnins thl.! uni! v1.·1.·1ors c,,, c,,., and c , lis11.·d 10w wlst•. {) 6U
llsin11 1111.· 1tt·omt·II y of II innglc jmp, w1.• rnn llncl th,• nnitll· 1/1 lo lw 2 , 11 111d
'l'h1.•1l'11111•, 1111' 1111111H1111•11t N111 tlw 1111\' l' wrtm P(I) In 1111111cll11•1.•11111111 1111.• 111111
 261
260 Framed Structures Chap. 6 Sec. 6.5 Space Frames

Nonzero terms in these submatrices come from Eqs. (3.4-4) , (3.4-12) , and
r 1 I,
(3 .4-24) for axial , torsional, and (two) flexural elements. All of the structural
0 12/, Sym. parameters in Eqs. (1) have been defined previously.
0 0 12/y Similarly, the 12 x 12 consistent-mass matrix M' for local directions
Kkk = E ( le)
LJ 0 0 0 r2L2f y contains the 6 x 6 submatrices
0 0 6Lly 0 4L2I y 140
0 -6LI, 0 0 0 4L2/, 0 156 Sym.

, pAL 0 0 156 (2a)

Mji = 420 0 0 0 l40r~
0 0 -22L 0 4L2
0 22L 0 0 0 4L2

und
70 0 0 0 0 0
54 0 0 0 l3L
0
0 54 0 -l3L 0
M' pAL 0 (2b)
kj = 420 0 0 0 70r 2g 0 0
0 0 l3L 0 -3L2 0
- 13L 0 0 0 -3L2
0

140
z (a) 0 156 Sym.
0 156
y M' pAL 0 (2c)
kk = 420 0 0 0 140r}
0 0 22L 0 4L2
0 - 22L 0 0 0 4L2
p
k1 k4
Nonzero entries in these submatrices are taken from Eqs . (3.4-5), (3.4-15), and
/ , k3 I ( L4-26), and all of the structural parameters were given earlier.
k6 I
I Y;k Distributed body forces applied in local directions to a space frame mem-
hcr cause the equivalent nodal loads
. t i2 I
I I
/J3 /1 /4 ______ :_I_/__,_ _ _ x (6.2-3)

,1e I / z,k lhnt urc also in local directions. As for the space truss member, the three

/ x,. ----- ~ / 1•11111puncnls of force (per unit length) arc

(6.4-3)
I
Iii)
whk h 11pp1•1 11 nl 11 1-f,l'111~1k p11l 11I on 1111• 1111•111hl' I' 111 Fl1 . (1 , 'i(n) . In Eq . (6 .2-3) the
11'1111111• ,,. ~ l.]' 11111 I I1111111 1111"111 Illl l (II) I111111 illt1·1 ll1tt1II, (It) 11l11h1tl 1ll11•1•IIIIIIN
 262
Framed Structures Chap. 8 Sec. 6.5 Space Frames 263

}J
matrix of displacement shape functions is once more
0 0 f4 0
f = [~ fd (6.2-5) 0 - Js,x· 0 /6,x' (10a)
t, - [~
But now the two 3 x 6 submatrices have the forms Ax· 0 0 0

and
~-G !2
0

0 ~J!2
0
0
0
0
0
0
0
-/3
(3)
where

I' '
J 6,x
= df6 = 3 J:2 -
dx' .,
2 I:
., (10b)

Functions / 1 and / 4 appearing in Eqs. (9a) and (1 Oa) are not differentiated with

1 ~ ~ ~ ~J ( 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
ck = [fio~ 4) of first derivatives that multiply the moment My'· Finally, nonzero terms in the
o ls o -A o third row consist of positive values of first derivatives that multiply the moment
M,,.
Functions Ji throughA in these arrays were defined previously in Eqs (6 2-6b)
and (6.2-7b). · · 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,
. Figur~ 6.5(a) also .shows the possibility of six time-varying concentrated
us follows:
~ctlons ap?he~ at any pomt on the space frame member. Three force components
m local d!fecttons are given in the column vector ·

P'(t) = {f,·, P y', P,,}

Due to these forces, the equivalent nodal loads become
pf,(t) = fTP'(t)
(5)
R- [[ f ; ~] (11)

(6) The 3 x 3 rotation matrix R , appearing four times in matrix R, is identical to

w.hich. require no integrations. Furthermore, three moment components in local lhat in Eq. (6.4-9) for the space truss member. Again, the third point p is
<11rcct1ons are 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
M'(t) = {Mx·, My', M,,} (7) matrices K, M, and pb(t) for structural directions.
To cal~ulate e9uivalent nodal loads caused by these moments, we set up the As for the other types of framed structures in previous sections, the stiff-
following matnx-vector multiplication: nesses, masses, and equivalent nodal loads are assembled from member con-
ldbutions. To transfer terms from the member arrays to the structural matrices,
pl,(t) = f:,.,M'(t) (8) we need the global indexesj 1 through k6 shown in Fig. 6.5(b). Their values are
~hi~~ ~ecds no in~egrations . However, Eq. (8) does require the matrix f , l'alculated by the formulas
rnntammg appropriate first derivatives of displacement shape functions wi;,; j l = 6j - 5 j2 = 6j - 4 j 3 = 6j - 3
respect to x '. The first submatrix in f ,x' is
j4 = 6j - 2 j5 = 6j - 1 j6 = 6j

~..,· = [~ ~
Ii 0 (12)
-1.x' 0 h.x· (9u)
kl = 6k - 5 k2 = 6k - 4 k3 = 6k - 3
0 fi.x• 0 0 0 k4 = 6k - 2 k5 = 6k - 1 k6 = 6k
wlwrc which arc used as subscripts in the assembly process.
d/i ll:xum1>lc 6.4
h ,,'
"I I
.h ,· l "I I
., I I
.,A/: . (%) 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 t·1·t 1t111 1111 tht• N(ltll 1' I 1111111• 1111•111h1•1 I .111•111 io11Nol pnints .I, k, nnd p arc the same
111 rnld1lto11 , 1111· ~1'1'111111 , 111111111111 11 IM
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 Chap nc. 6.6 Programs for Framed Structures 265
0.7428 0.5571
0.3714] I >YNAPF for plane frames, DYNAGR for grids , DYNAST for space trusses,
R = - 0.4983 0.8305 -0.2491 ,111d DYNASF for space frames . They represent the dynamic counterparts of the
[
-0.4472 O
0.8944 1111ic analysis programs described in Ref. 1.
still pertains. We shall detennine equivalent n d 1 1 First, let us consider the task of generating Program DYNACB for con-
directions caused by the distributed force. o a oads for both local and gloh11I 1l11uous beams , using Program DYNAPT for plane trusses as a guide. Both types
li ons a~sing the rotation matrix R , we can find the components of by(t) in local dirl'l ,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.

b'(t) - R b(t)

Then apply Eq. (6.2-3) to obtain

-+:,J ~ [~::~:},(,) lul
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 be-
1111ne: NRL(2J- 1) = restraint against y translation and NRL(2J) = restraint
IIHllinst z rotation.
p;,ir) = L f fJ b'(t) d( 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,
= {0.2786, 0.4153, 0, 0, 0, 0.0692Il}lby{t) u~ shown in Fig. 6.6(a). Because forces applied to a flexural element act only
(h)
Hild
111 the y direction, the linearly varying line loading depicted in Fig. 6.6(b)

pJ,k(t) = l f rlb'(t) d(
tt•quires only two parameters (BLl and BL2) for its definition. The equivalent
y

,, = {0.2786, 0.4153, 0, O, O, -0.0692Il}lby{t) (l I (A, )2, ,

; .h ~~x~~~~r:~~~~i.n (~:~~~~~e: nfio:dal loads in member-oriented directions at joints} uml t - - - - x

RT OJ
p,,J(t) = [ 0 R T p;,j(t)
/ i
(A, )2,

= {O, 0.5000, 0, -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•)

y
I 1•1111N in these vectors consist of equivalent n0 d I I . . . . b,,
tu l1111f1·d dlrct:tions. a oads at Joints J and km struchUl'

8.8 PROGRAMS FOR FRAMED STRUCTURES

In a 1111111111•1· r11wlopo11s lo lhut fi>r Proorar11 l>YNt\l'T ( • •

d1•v •I • " sn
I op w1111n111•r ()1C1frn 1111, for llrl' dyr1111111r u1111ly<11 ~ ol rlw
~· • . ., 7
•"'t t ,., . ) we c1111
li v1• oflll'r • I , I I
/
J_ ii 11111111111
0
l l 11 I IO
_ x
Jill
k

lr111111•d ~r,11r r1111·~ 'llr1·N1· profr11111N1111• n1111wd l>\'Nt\< 'II I • . YPt'\ c,

or 11u1t111110111, lit·a rttN, I 111111• 1111 1·11111 l11111111N 111111111 (II ) 111111111 l11111IN , (h) 11111• 11111d 1111 1•1!111111111
 266
Framed Structures Chap. 9 'loc. 6.6 Programs for Framed Structures 267

fod;l loads at points j and k may be taken as the solution for Prob 3 4 _10 AI l'hc tenn x;(t) in this formula is the distance that the load has traveled along element i
;~Y.rograthm Df YNACB the only scale_ factor needed for ground ·ac~eler~tio:<:~ 111 time t, as illustrated in Fig. 6.7(b) . Because the functions in matrix fare cubic, the
' so e actor GAX must be omitted. 1•quivalent nodal loads given by Eq. (a) will also be at least cubic in time, depending on
Example 6.5 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 · . t2
+ aop
2-
elements of equal lengths A . po e pnsmattc beam, divided into four flexural x;(t) = Vopt (i - 1)£ (b)
from left to right so that .. t mo~~ng load _P of constant magnitude traverses the span
~~=~ :~ ::2e 1;
1
translational resp~nses oft~: ;t:~nc~~on _of)time. We will calculate the In this expression Vop is the velocity of the load when it first contacts the beam (at time
at various speeds [3]. llll pomt caused by the load moving 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
Equivalent nodal loads for this prob) b i . .
through p1,4 indicated in Fig. 6. 7(b).
ment shape functions for a flexural eleme~~r:?Eq~ ~~~~l~t:i~t:,1!tfo~~:;~placc-
Of course, we need to extend Program DYNACB to accommodate a moving load.
P&;(t) = Pr[x;(t)] The load parameters NLN and NEL (see Table 4.1) are augmented with a third param-
(a)
titcr, 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:
y (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
1 2 4 5 physical properties:
• © ~-/----x 7 2 4
E = 3.6 X 103 k/in. 2 p = 2.25 X 10- k-s / in. L = 4£ = 240 in.
2 4
A= 12 x 15 = 180 in. /, = 3375 in. P = 10 k
where the units are US. Figure 6.8 shows computer plots of Di, at node 3, obtained by
1unning the modified version of Program DYNACB and using Subprogram NUMINT for
(a) 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 acceler-
ntlon (AOP = 1.071 x 105 in./s2 , with zero initial velocity) gives 0.3625 in. Their
y
111tios to the static deflection of PL 3/ 48£/, = 0.2370 in. (due to the load applied grad-

~x(t)----1., i + - - - X;(t) ----..i~.

1111lly 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.

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
P., t
/-. t
__________ t- - -- p
___,;,
Pb3
)(
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

z
p"' 1-/ - - 0 / /' t•ontuining 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
(I )) 1t•strnints against x trunslution, y trnnslntion , and z rotation, respectively.
l•'l111m, ,,.7 'l'nhk' c, .) shows dyn11111ic lond dntu Im pln,w r,·11111l'S. In this cusc, the lines
(11) Sl111ph• lw11111 wifh n111vl11n lune! , Ch) rq11lv111«,n1 11ucl11J lrnulN lor 11ii1l11I dlNplm•t•111t·11t., 1111d w locitlt·~ rn11t11i11 tllll't' q1111111i1it•f'I i11slt•nd or two .
268
Framed Structures Sec. 6.6 Programs for Framed Structures 269

VOP CONSr . TABLE 6.2 Dynamic Load Data for Plane Frames
0.40
No. of
AOP CONST. Type of Data Items on Data Lines
Lines

Dynamic parameters l !SOLVE, NTS, OT, DAMPR

Initial conditions
0.30 (a) Condition parameters l NNID, NNIV
(b) Displacements NNID J , D0(3J-2), D0(3J-1 ), 00(3J)
C: (c) Velocities NNIV J, V0(3J-2), V0(3J- l ), V0(3J)

Applied actions
n
(a) Load parameters I NLN,NEL
~ o. 20 (b) Nodal loads NLN J, AS(3J-2), AS(3J-l), AS(3J)
0
z (c) Line loads NEL I, BLl, BL2, BL3, BI.A
Ground accelerations
(a) Acceleration parameter 1 IGA
}
a (b) Acceleration factors• I GAX,GAY
0.10
Forcing function
(a) Function parameter 1 NFO
(b) Function ordinates NFO K , T(K), FO(K)

•omit when IGA = 0.

o.oo flexural elements), we find equivalent nodal loads in member directions. Then
(ms)
the equivalent nodal loads in structural directions can be computed from
Eqs. (6.2- 16).
Figure 6.8 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 displace-
ments; 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 = 207 X 106 kN/ m2 p = ·7.85 Mg/m3 l=2m
2 2 4 4
TABLE 6.1 Structural Data for Plane Frames A = 1.61 x 10- m l , = 3.01 X 10- m

Type of Data
Problem identification
Structurul pnrurnctcrs
l'h111t· lrn1m• <111111
(11) Nodul rn111cli1111tr N
Ch) 1' 11 11111•111 1111111111111111 11
(1 l N111l11I 11·~11,11111~
No. of for which the units are SI. From static analysis , initial displacements due to the load
Lines Items on Data Lines 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,
I Descriptive title nnd (D0)13 = 6.598 x 10- 5 . We used these values as input data for Program DYNAPF,
- I
- NN,NE,NRN, E, RHO 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;
NN nnd in part (b) we have the bending moments AM3 and AM6 for the j and k ends (nodes
I, :\CI), Y( I)
NI! I 1111d 2) of clement I . For all of these responses, the maximum (or minimum) values
I, INCi), KN(I), AXC I). './,1(1) '
Nl(N m 'l ' III' at time I O und diminish thcrcuftcr bccnusc of damping.
~
I, NIU ( II JI, NIU ( 11 I), NIU .( II)
·-
 2/0
I 11111111«1 i,1111111111111 C l111p l'roi,110111 1111 I 11011ntl 111111111 111 271

y (I

/
I 1
• --,, 2

i3 L

z
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
1t1t•r tia ZI(I) on each line containing element information. Furthermore, the
in the nodal restraint list become restraints against rotations in the x and
lr 11ns
y senses and translation in the z direction.
1•
Similarly, the dynamic load data for grids are symbolically the same as
i bl1
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
/ - - - -- x 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
l•1g. 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
(b)
1•111,ily converted to structural directions using Eqs. (6.2-16) withpb 1 = pb4 = 0.
lt'lgurc 6.9 Load I I •inally, the scale factor GAZ for ground acceleration in the z direction replaces
s on p ane frame: (a) nodal loads; (b) line loads on element.
lht· factors GAX and GAY.
_Third, let us brie fly examine Pro ram DYN .
of ~11ds. Because this type of structu! ha h A~R for the dynamic analysis
11,)(umplc 6. 7
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 , I ill· grid illustrated in Fig. 6.1 3 is made of brass and has rather small dimensions and
1 11111! \' senses und ~ lranslatio~ in th : P'c,~ . ~~~ l' J ~ gr!d are rotations in tlw
111 l111uls. The cross section of each prismatic clement is a solid square with side 0 .25 in. ,
T 0 I 1111 c z ' l l t l 11011 H~ indicated in F'1 6 I ( I ,11ul lhc applied actions consist of step nodal und line loads. Physical parameters in this
l ' 1 Al' llw plu1ll' f'rume structural d11111 1 ', ·· • ?' · ').
wt· n1us1 11<ld lht• :..ht•111111n n1odulw, <, I II , n l 11hlt (1 . I lo that for a !!rid , I lllllltpJC UI~'
' ' o II ~111111111111 p11ra11tL'lt·rs . Also', lht• I•' 1.5 X 10 1 k/111 1
(i () 'i (1 " 101 k/111.' f' R.10 X 10- 7 k-s 2/in .4
 212
Framed Structures 'loc. 6.6 Programs for Framed Structures 273

~ 0 , 60
e z
e
0,40

w
25 0.20
z
~ o.oo
(/J

6 -o . 20
I- X
<
_, -o. 40
(/J
z
<
a:: -0,60
I-

(a)

(a)

z
AN3
E
10
z
-"

I- s
z
w
I: b ,,
w
~o X

z
(/J
z -s
0
I-
(.)
<
I -10
0
z
w

( b)
1 k
1 1gure 6.11 Responses of plane frame: (a) displacements; (b) bending moments.
(bl
I, 4 in. A = 6.25 X 10- 2 in. 2 Ix= 2!y = 6.510 X 10- 4 in. 4 Figure 6.12 Loads on grid: (a) nodal loads; (b) line load on element.
,. 0.02 k b1.. = 0.004 k/in.
111 llie member Jscc Fig . 6.3(a)], and the second is a bending mome~t about they ' a~is.
where the units arc US. Mu idmum (or minimum) values of the nodal translations are 0 .1 157 m. and 0.08451 m.;
~ilh lhis ~al a we ran Program DYNAOR, 11.~ing DA MPR = 0.05 and rcspousi• 111111 lhosc for the moments arc 9.333 x 10
3 2
k-in . and -8. 181 x 10- k-in.
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'dcnil·III J/o11 1lh, we shull disl'uss Progrum DYNJ\ST for spucc trusses, a?ai.n using
I 11pp1•111 111 p11,1 (h) ol 1hl· ll1111 rl'. T lw ll1sl ol lht•Nr 11111111r11I N iN111t,1qm• nhout llil' ·1 ' lllliN 1•1,1w11111 J>YNJ\PF 1111· l'o111p11iiso11s. 111 till s l' IINl', tlw tllrt'l' types o f displace-
 275
Sec. 6.6 Programs for Framed Structures
274 Framed Structures Chap. 6
NOOE 2
z

0 -10

j3 C
::; 0. 05
t
•-i2
4
(f)
z
/i 0
~ o.oo
,_
T
j1 <(
~
(f)
~ -0 ,05
a::
,_
I
N -0 . 10

(a)

Figure 6.13 Grid with step loads. C

-;- 0 .10
Q.
ments at a typical node j are translations in the x, y, and z directions, as shown
, ..:,f,

in Fig. 6. l(e).
n 0,05 AMS
In order to make the plane-frame structural data in Table 6.1 apply to a AH4
1-
space truss, we must add Z(J) to each line of nodal coordinates. On a line of z
w
clement information, the moment of inertia ZI(I) is replaced by an identifier IP I:
w o.oo 4 8
that indicates whether a third point p is necessary for locating principal planes ~
w
T Cm s l
of bending. If IP = 0 , the x '-y' principal plane in Fig. 6.4(a) is taken to be z
parallel to they axis. However, if IP = 1, the next line of data must contain the UJ -0-05
coordinates XP, YP, and ZP of the third point p. (This type of data is shown in z
0
Table 6.3 for space frames.) Last, the terms in a line of nodal restraints denote I-

restraints against x, y, and z translations . ~ -0 . 10

I
Dynamic load data for space trusses are similar to those for plane frames, 0
z
hut the meanings are different. For a space truss the initial displacements and w
wlocitics at a joint refer to translations in the x, y, and z directions. The thrcl' (b)
nodal loads at a joint are forces in the x, y, and z directions, as depicted in Fig.
Figure 6.14 Responses of grid: (a) displacements; (b) moments.
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
nlt•nt nodal loads due to these line loads are easily obtained by extending Eqs. llu cc sets of double triangular impulses P(t) at locations indicated in the figure. For this
(•I . IO 4). To accomodate three components of ground acceleration , we must also plllhlcm the physical parameters are
mid the scale factor GAZ to the data table. f 117 X 106 kN/m
2 p = 4.49 Mg/m3 L = lm
1,:xn1111>lt• 6.8 4 2 P.,m = 100 kN
A 9 X 10 m
t•tpllH' C, IC, dt•pil ts II spun• 11 IINN with nine p11s11111t11 !'lt' llll'IIIN , h11v111g l'(lllll l •l ' IONN
hl'I ti111111I llll'IIS 111111 r11111t111s1·d 111 hf1-•h sti1•11pt h 111,111111111 I h111 ~11111 tllll' i~ s11hj1·rtt·il to
111 whtl'h SI units 111t· impli1•d
276
Framed Structures Chap.a Sec. 6.6 Programs for Framed Structures 277

TABLE 6.3 Structural Data for Space Frames y

Type of Data No. of

Lines Items on Data Lines

Problem identification 1 Descriptive title

Structural parameters 1 NN,NE,NRN,E,G,RHO
j2
Space frame data
t
1
(a) Nodal coordinates NN J, X(J), Y(J), Z(J)
(b) Element information NE I, JN(I), KN(l), AX(I), o -i1
Xl(I), YI(l), ZI(I), IP /i
Coordinates• of point p I XP, YP, ZP j3
(c) Nodal restraints NRN J, NRL(6J-5), NRL(6J-4), NRL(6J-3) L
NRL(6J-2), NRL(6J-l), NRL(6J)
a Requued when IP = I. """

y
~7'---X

(a)
z

Figure 6.16 Space truss with triangular impulses.

We used this data in Program DYNAST, with IP = 0 for all elements,

b,2
DAMPR = 0.02, and called Subprogram NUMINT for numerical evaluation of re-
sponses . 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.

Fifth, let us build upon all of the previous programs to describe Program
y 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

----x 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'Oordinn11·H of II lhird poiul ,, 1111 lornli111• ptinclpul pl111ws of'lx·nding. In thc lines
I
{111
UI
 278
Framed Structures
! f!C. 6.6 Programs for Framed Structures 279
100
TABLE 6 .4 Dynamic Load Data for Space Frames

~ so Type of Data
No. of
Items on Data Lines
Lines
a..
~ 0 Dynamic parameters 1 ISOLVE , NTS, DT, DAMPR
c:: s 10 15 20
0 Initial conditions
I.L. T Cms)
(a) Condition parameters 1 NNID, NNIV
8 -so (b) Displacements NNID J, D0(6J-5), D0(6J-4), ... , D0(6J)
_J (c) Velocities NNIV J , V0(6J-5), V0(6J-4), ... , V0(6J)
a..
a..
< - 100
Applied actions
(a) Load parameters 1 NLN,NEL
(b) Nodal loads NLN J , AS(6J-5), AS(6J-4) , . .. , AS(6J)
(a) (c) Line loads NEL I, BL!, BL2, .. . , BL6
Ground accelerations
(a) Acceleration parameter I IGA
(b) Acceleration factors• 1 GAX, GAY, GAZ
Forcing function
(a) Function parameter l NFO
7 200
K, T(K), FO(K)
(b) Function ordinates NFO
.:!
V)
IJ.J
"Omit when IGA = 0.
~ 0
0
IL
,,~ those for a space truss element. However, the equivalent nodal loads for a
_J ~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
~ -200
1111nponents and the solutions from Probs . 3.4-1 and 3.4-10 (for axial and
tl1•1lural 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.
( b)

Figure 6.17 Space truss: (a) triangular impulses; (b) responses.

y

'i°' nodal ~estraints, the terms indicate restraints against three translational and
1 llt·c 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
1 . . . mponcnts an thrl'l'
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
Sec. 6.6 Programs for Framed Structures
4
G = 2.4 x 103 k /in.2 p = 1.71 x 10- 7 k-s2 /in.
E = 6.5 X 10 3 k /in. 2
5 4
A = 0.01 in .2 Ix = 2/y = 21, = 1.667 X 10- in.
L = 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
y 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
' C
~----x
s
z
D
I-
z <(
0:: 0 400 60 800 l 0 00 l 00
w 200
(b) T ms l
-'
w
u
Figure 6.18 (cont.) u
<( -s
Cl
K11.11mplc 6.9 z
:::>
The space frame in Fig. 6.19 has three rismatic . . 0
0:: -1 0
st•cli<lns that are solid squares with side!o 1 . A~agnesm~ elements with equal cross 0
I
14round acceleration Dg1(t) in the x directi~nmPh ~o stown m the figur~ is a rigid-body X
· ys1ca constants for this example arc
(a)
y

DJ3 DJl
0 .1 0
C
2

('\J
o.os
w
Cl
D
z
1- 0. 0 0
<(

Cf)
z
0
~
I-
-o. os
<
-'
(/)
%
;;5 - 0 , IO -

:11
( b)
,
l' ll,IIIII' II jll 'lp 1111 1111 1111 wli lt }• 111111111 11,, i Ii l lllltlll
 282
'rnmncl Structurn 1 Ch
nI C111yn11 Unclw 11011 213
6.7 GUYAN REDUCTION

The concept of · p1 111ll· t1l 011 ll1<1Sl' ol typ1• II , llli l111l11w11,

. matrix co11de111·01/011 11 , I -
mg the number of unknown ·d .. . •' I~ II Wl'll known procedure for rrd U, • 'l',w U11 (7a)
~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 (7b)
. s1~, a s1mtlar type of condensatio . a x onn. For dy11u111
bnngs m ~ ad~itional approximation. n was introduced by Guyan [41, whl
I "' 11 lor static analysis, this relationship is correct only when actions of type A
Startmg with static reduction we w . . 1111 11111 exist [see Eq. (3)]. However , Eq. (7a) follows the finite-element theme
free displacements in the partition~d fo;te act10n equations of equilibrium fi
111 "~lnve" and "master" displacements. Differentiating Eq. (7a) twice with
11 , p1'l'l to time produces

[:: ::][!;] -[~]

Here the subscript A denotes the d' 1
IIJ (8)
subsc · t B ti tsp acements that are t be 1· · I 111 tI1e purpose of reducing the equations of motion to a smaller set, we can form
np_ re ers to those that will be r . o . e immated, and th 1111 transformation operator
of equations, as follows: etained. Now rewnte Eq. (I) as two ~or
(9)
SMDA + SAaD8 = AA
SaADA + SanDo = A8 111 which 18 is an identity matrix of the same order as S 88 . Substituting Eqs. (7a)
Solving for th (8) into Eq. (4.4-1) and premultiplying the latter by Tl gives
111111
e vector of dependent displacements DA in Eq. (2a) yields
(10)
. D;1 = S,U (AA - SABDo)
Substitute Eq (3) · t Eq 111 this equation the matrices S18 and Al still have the definitions given in Eqs.
. mo . (2b) and collect terms to obtain
1,1 nnd (6) . However, the reduced mass matrix M%8 is found to be

in which (4) 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
and (.,) c;
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
Al= Ao - SBA M 1 A As- 111t111ions at the joints of beams , plane frames, grids, and space frames as the
From Eq .. (4) we see that Eqs . (2) have been (ll) ,lqit·ndent set of displacements. However, the method can be used in a much
~~ly_the independent displacements in vecto r~uc~ to a smaller set involvi111l 111111c general manner for various discretized continua. That is, any arbitrarily
as m Eq . (5) is a modified version oft r .n: e reduced stiffness matri, -rll'l'lcd set of displacements may be referred to as type A, while the remaining
reduced action vector A* 8
in Eq (6) ?e ongma1 submatrix Saa. Also 1/1" ,li~placements become type B. The trouble with this generality is that a good
th t · contams tenns d'f · ' "
a are considered to be equivalent loads f mo t ymg t~e subvector A,, , hoice of "slave" and "master" displacements is not always obvious. Even with
Fu~ennore, Eq. (3) may now be view~ ?pe B due to actions of type A. l111111ec.J structures there are cases when joint rotations are more important than
required to find vector DA exactly fro s the back-substitution formulu t11111slations and should not be eliminated.
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 l1 11111nple 6.10
tsp acemcnts arc 1111' fixed beam in Fig. 6.21(a) is divided into three flexural elements, each of which has
Ml) -ISD A 1lw ~111ne properties£, I , p, and A. By Guyan reduction , we shall eliminate the rotations
Then assume m, 11
(4 .4 I) 111 11mb 2 and 3 and retain the translations. The reduction will be followed by a
lll'W approxi murion lhur 1hr <li1;pl11cl·111t· 111~ of ,1l111111onnl annlysis that is compared against exact results.
ry pe A llll'
For 1h is cx111upk, the assembled structural stiffness matrix (without consideration
 284
Fromod Structures Sec. 6.7 Guyan Reduction 285
of restraints) is
y
,----------- -------,
/ 6 /
, 3e ze2 CD / Sym.
/ I 2
I r--------4---- -----
/ -6 -3e I 12 I I
I I I I 3
s = 2£/ I 3e 1
I
e 22 ,
I
'2" o
1
I
4e
s e3 '-t:J
I --------+---------1------- ---l----------1 4
(a)
I I I I
0 0 I -6 -3€ I 12 I I
I I I I 5
0 0 I
I
3€ e 2
I
I
O 4€2 I
I
'3"
'-:!.I
I
I 6
L------- - ~--- ---- - I I
0 0 o o I - 6 -3e 6 i 7
0 0 I I
0 o I, 3e e2 - 3e 2e2 1
I 8 (a)
L - - - - - - - - - - - - - - - - -- - 1
2 3 4 5 6 7 8

Similarly, the assembled structural mass matrix becomes

r ------ - ---------- ---~

/ 156 l I (b)
I 22€ 4€
2
CD I Sym.
2
I ,- - - - - -- - - --l- -- - -- - - -- ,
I I I I
I 54 13€ I 312 I I
I I 1 @ I 3
pM /- Be - 3€ l 2
o se l 2
l 4
Ms = 420 L--·-- - -----f----- ------/--- --- ----f-- --- ---- I
o o / 54 13e l 312 / /
0 0 / -13f -3€2 / 0 8€2 / Q) /
'----------4----------1 I - 1
O O O O / 54 13€ 156 / 7
o o o o !l___- _____
13e -3e 2
-22e
_ _ _ ___ _ ___ _4€2
__ / ! 8
(c)

2 3 4 5 6 Figure 6.21 (a) Three-element fixed beam; (b) mode I; (c) mode 2.
7 8
(h)

Dushcd boxes in Eqs. (a) and (b) enclose the contributions of elements 1, 2, and 3, which
4e 2
Sym. 4
urc drawn from Eqs. (3.4-24) and (3.4-26). Moreover, the joint displacement indexes for SAA SAD] 2EI
e2 4e2 i 6
the problem [sec Fig. 6.2 l(a)] are listed at the right side and below the matrices. S= [ SaA Sall = 7 --------j------- (c)
As the first step , we remove the first, second , seventh , and eighth rows und 0 3e I 12 3
t·olumns from matrices S, and M,,, because displacements I , 2, 7, and 8 arc rcstrai,wd - 3e o i -6 12 5
hy supports. Then the remaining 4 x 4 arrays arc rcarrnngcd to put the rotationul torn1N
heforc the trans lutionul terms, as fol lows:
4 6 3 5
 lU/
Fromod Struoturoe ti. / Cl11y1111 llcuhmtlo11
186 Ill ,

"'h1•1l' L 3e. These values arc in error by I O.h 1% 1111d I 2.6%, respectively; and they
Sym. 4 1111111titute upper bounds of the exact angular f'n:qucncics \51. The corresponding mode

- pA I'
420
K('J I 6 h11p1·s arc
(k)
()
IJe I 312 3
13e 0 I 54 312 5
wh1l'11 appear in Figs. 6.21(b) and (c) .
4 6 3 5
Frequency coefficients /.Li for prismatic beams with various end conditions
1111llll Eq . (c) the inverse of SAA is 1111• summarized in Table 6.5. In each case the beam is modeled by four flexural

8
- 1 _
AA -
e [
30£/ -1
4 -1] 4
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
Substituting this array and the other submatrices
. of S from Eq. (c) into Eq. (5) prod • • p1oduces much better accuracy than the lumped-mass model in beam analysis.
J2£/ [
2 -lJ _ [4 l] Ull ~
St11 = & _
1 2
6
E!
5e 1 4
= 6El [
5e3 - 11 16 5 <, l
16 -11] 3 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
3 5
ln·qucncy for this frame with and without elimination of the rotations at joints 1 and 2.
whk h,i:~ the reduced stiffness matrix . l'o Himplify the analysis, we shall omit axial strains in the members, leaving only the
I o reduce the mass matrix we form the tr . 1111 cc degrees of freedom indicated in the figure.
'I\,, lo l!q. (7b) and substituting it into Eq. (9). ;~~~~rmat10n matrix Tn by evaluati11p The 3 x 3 stiffness matrix for unit values of D 1, D2, and DJ has the form

TAB -- s-AA1SAB= -1 [ 1 4] (g)

-
5e -4 _1 (C)

l
1111d

11ml the accompanying 3 X 3 mass matrix is

I =-~e----0
1 4 2

1 l
Tn = TAB = I_ 8L Sym.J
[ In J 5€ 5
1 (h)
MAB
M
= pAL
420
- 3L 2 8L 2
---------l----
I (m)

o 5e 88 22L 22L ! 732

111 this cnsc the submatrix 1/J in the lower artition of . . 111 which M includes pAL for the mass of member 1. With these arrays we can set up
1111· two rnmaining translational <lisp! p ( matnx 'Is is of order 2 because then· 08
M lrom llq. (d) and T from Eq (ha)c~menEqt s numbe~s 3 and 5). Substitution of matrii<. 1111tl Nolve the eigenvalue problem, which yields the angular frequencies
11 • mto . (11) yields l [El (n)
Wi , 2, 3 = 3.201, 15.14, 32.68 L 2 \JpA
Mlo = pAC
2100
[1696 319] 3
319 1696 5
(i)
Ahm , the corresponding mode shapes are
3 5

l -oLJ
- 0.5528
(o)
wh11·h is the reduced mass matrix. <I> = f<1>, <1>2 <J>3) = - 0 .5528 - 1
111 th1· Using theEqterms
lrn111 or Tl . S111I and
. (3 .6in4)matrices . f M* . .
1111 , ~e ~an set up the eigenvalue problen, I. 0
. 1c angu a1 rcquenc,cs found hy this method arc

"11 ,J 22.~l ,(,l)/1 I I J Jn .

'"' ' ( I)
I hi• Hl111p1· ol 11111dr I I~ ill11 HfrlllCd in Fig . (i.l.2(h).
 - - ---------,

TABLE 6.5 Frequency C oefficients µ ; for Pr ism atic Beams Mod eled by Four Elements

Support Conditions Mode Exact (5) CM- TR % Error CM- TO % Error LM- TO % Error

I 9.870 9.872 +0.020 9.873 +0.030 9 .867 -0.030

• • •
~ ~ ~ 2 39.48 39.63 +0.38 39.76 +0.7 1 I 39.19 I -0.73
L I I 3 I 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

f=
I ~ .67
I
.9 12 1.9 +0.83
I 122.4
I + 1. 2
I 86.84 -28

I
22. 37 22.40 + 0 . 13 I 22.4 1 I + 0 . 18 I 22. 30 I - 0 .3 1

frurl I
~ • • • ~ I
I
2

3
61.67

120.9
62. 24

123.5
+ 0.92

+ 2.2
62.77

124.8
+ 1.8

+3 .2
59.25

97.40
-3.9

-19

3.5 16 3.5 16 +0.00 3.5 16 +0.00 3.418 -2.8

I
J.."IO I
~ • • • • II 2 22. 03 22.06 +0. 14 22.09 +0.27 20.09 I - 8.8

3 6 1.70 62. 18 + 0.83 62. 97 + 2. 1 53. 20

I - 14

I I
15.42
I 15.43
I +0.065
I 15.43 + 0 .065 15 .40 I - 0. 13

~
~ • • •
~
I 2 I 49.97 I 50.28 I +0.62 I 50.56 I + 1.2

I + 6.0
I 49.05

I 91.53
I - 1.8
I -12
I 3
I 104.2 I 106.6 I +2.3 I 110.5

w =~
L1
ff. pA
CM: Consistent Masses
LM: Lumped Masses
TR: Translations and Rotations
TO Trm,.,IM.-,. Onl~

"',,
a;

'X m
:....i
-:

~ C ::; G)
': g "Q C
-<
::. ('";
I»
:::. -0 :,
o;

aci' :,:,
a,
a.
"':.

-:
:,

o'
'"' ..,
N~

"'
SJ"'~ '<
C:
$l.
o·
:,

f'
-: 0=: O'Q
- ·

"':.
3
- -
~
!;
r>
0\
__l G
::.. ,z c. N
,z
-
:,.. ::s
(IQ

s- ~
tJ
l t 1

~
-
z
a
-- ~
Cl>
...,
O
~
-
,- 8
~ ~
~
5-
l
§: ~
- o· ...,
:, ol
,.
-
-
-
-

•
~

-
.;"
t
(>
00
3
0
-·
o'
~
g
~ -' 8- N' .i "'

s:i i
.::__i 3 '-,.'1-1>
"O C.
\
1----
C:

r,
0
., 0 -
:i &
0 ,- .\
- l..,
t
JC g,
"'

"'g.
3
I»
~
,--. 5.
.._,
"O ;,<
 290
Framed Structures Chap. 8
Sec. 6.8 Constraints Against Axial Strains 291

Sto = 24EI _
3
36E/ = 84E/ In this equation the symbols D j t and D j 2 represent the x and y translations of joint
L 5L3 5L3 (q)
j, and Dki and Dk2 are those at joint k. As before, the x and y direction cosines
which is just a single term. Next, the arrays TAB and To are found to be of the axis of the member are denoted by c., and Cy.
1 Similarly, we write the axial constraint condition for zero elongation of a
TAB = -s;J sAB = _ _l_ [ ]
(r) space frame member [see Fig. 6.5(b)J in the following manner:
5L 1
and (Dit - Dk1)cx + (Di2 - Dk2)cy + (Di3 - Dk3)c, =0 (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.
(s) 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 C D1 = 0 (3)
*
M BB _ ToT M
-
569
To = - pAL The matrix C in this expression is called the constraint matrix, which contains
350 (1)
only positive and negative values of direction cosines for the constrained mem-
Now the eigenvalue problem gives bers. 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
w, = 3.215-
]~-/ rotations are involved at this stage.
2
L pA (u)
Due to the constraint conditions, some of the joint translations in the vector
Cornpari_ng this angular frequency against w1 for the unreduced 3-DOF problem we find 1)1 will be linearly dependent on others. To determine which translations are
the relative error to be +0.16%. ' 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
6.8 CONSTRAINTS AGAINST AXIAL STRAINS
translations is arbitrary , and pivoting automatically produces the best selection.
Thus, it becomes possible (in retrospect) to partition the matrices in Eq. (3) as
In Example 6.11 we omitted axial strains without difficulty because members of lollows:
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
[C 11 C,2{~:] = 0 (4)

plane and space fra?1es, pri~arily for the purpose of reducing the number of 111 this expanded form the vector D 1 represents r dependent translations, and D2
d~grees of freedom m dynamic analysis. The number of constraints introduced rnntains the remaining n; independent translations. Because no redundant con-
will be equal to the number of members m for which axial strains are to bl' ~t1nints are included, the rank r of matrix C will always be equal to the number
neglected. By automatically selecti~g m of the joint translations to be dependent ol members m. Therefore, submatrix C 11 in Eq. (4) is a square array of size
?n the rest of th~m, we ca? devise a method for reducing the number ol 111 X m, and submatrix C 12 is of size m x n;.
mdepend~nt_eq~ahons of motion. This reduction method may be combined with Multiplying the matrices on the left-hand side of Eq. (4) produces
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 (5)
no c~mphcat1ons from redundant constraints (or superfluous members) need tw l, 1mwing that matrix C 11 is square and nonsingular, we can solve for vector D1
considered r6J .
111 tl't ms of vector D2 . Thus,
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 (6)
111 whil' h llw OJll'l ator T1 2 is
(n,, n11 k, 1 (n,. n1 ,l1\ o (1l
(7)
 293
lluc. 6.8 Constraints Aga inst Axial Strains
292 Framed Structures Chap. 8
TABLE 6.6 Member Information for
During the Gauss-Jordan elimination h . Three-Member Plane Frame
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 Joint k Cx Cy
Joint j
this inverse to be the ~perator ' i wi e replaced by C ;-t We now define Member
0.7071 0.7071
1 1 2
1 0 1.000
T11 = C 11 (8) 2 3 1
0.4472 -0.8944
3 2 4
and the augmented constraint matrix is
. Ci = [C Im] (()) ~hown in Fig. 6.18. After the Gauss-Jordan procedure is applied, the matrix in
To confirm ideas regarding the constraint . . 11.q. (a) becomes
frame in Fig. 6.23 for which th b . & ma~x: let us consider the plane

-Lit~]
' e mem er m1ormat10n is given i T bl 6 6 T 1 2 4 5
augmented constraint matrix [see Eq · (9)] ior
&
th"1s case is
. n a e · . ho
-1.500 0 II 1.414 1.000
0

c,
1

= [o.~001
2

o.101
-1.000
-0 .101
0
4 5

-0.101 1.000
0 : 0
! o
1.000
o
0
J (a)
c;
-[f 1.000
0
0
-0.500
0 II 0
1.000 l 0
-1.000
0
0
-1.118
(b)

0 0.447 -0.894 : 0
The numbers above the columns of C indicate the translational displacements
0 1.000
l'hcrefore,

T,1 = C 1/ =
[.414
~
1.000
-1.000
-1~18] (c)

0 -1.118
0.5P(t)
y
1111<.I
4

l T12 = -C;-/ C,2 =

[500]
0
(d)

P(t) t 0.500

Thus, the dependent translations are found to be 1, 2, and 5, whereas the

t2 P(t) l111h.1pcndent translation is automatically chosen to be 4. Note that there is no
/ 1
, - - ----t--- - \ - - - - - -1. 111•1,d to rearrange the augmented constraint matrix during this procedure.
3 In preparation for a coordinate transformation associated with axial con-
~1111 lnts, we shall define a generalized displacement vector D, as follows:

D = [~J (10)

1111' llrst part of D consists of a null vector, representing member elongations

2
3/ _ _ _ _ _ _ __J__ _ I whk h urc z.cro) ; and the second part contains the independent translations D .
I 11 11•l111c the vector D to D,, we write
(l la)
D = T,D,
'" (l lb)
11·1111111• t,,1.1 '1111,•u u111111h1•1 pl11111· 11 111111•
 295
294 Framed Structures I nc. 6.8 Constraints Against Axial Strains

In Eq. ( 11 b) the upper part represents the constraint conditions, and the lOWl'f ~tiffnesses of members from the structural stiffness matrix. Then to satisfy
part merely re~roduces D2. Because the generalized displacements are indepcn ,•quilibrium at the joints, we must revise Eq. (15) to become
dent and constitute a complete set, there is also an inverse relationship in th~ M* ·· s*rr D,--
11 D, +
A* + cTQ
1
(21)
form
111 which Q is a vector of axial forces in the constrained members.
D, = TcD ( I.? I To perform the transformation, we substitute Eq. (12) and its second
where dl.lrivative with respect to time into Eq . (21) and premultiply by Tt to obtain
1 1 M D+ S D = A + C Q (22)
Tc= T, 1 = [C,1 -C,1 C12J = [T11 Ti2J ( I II
0 12 0 12 Writing this equation in expanded form, we have
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 M 11
[ M 21
M,2] [
M 22 D 2
~ J+ [~11
S21
~12] [ 0
S22 D2
J [Ai]
A2
+ [l']Q
0
(23)
for _axial ~onstraints. Notice that submatrices T11 and T12 are generated auto
mat1cally m the Gauss-Jordan procedure described previously. Multiplying terms and rearranging the results produces
. Let us now restate the action equations of undamped motion for fi'l'l' (24)
displacements at the joints of a plane or space frame as
1111d
MD+SD=A (4.4-1)
Q = M12D2 + S12D2 - A, (25)
By segregating rotations from translations, we can write this equation in thr
cxpunded form I he barred matrices in these equations have the following definitions:

[M,M,,, M,M,, [D'ti, ] + [S,,S,, S,Su [D'Dr ] = [A']

1 1 S12 = TT1 st\ T,2 + TT1 sf2 (26)
] ]
( 14)
Ar S22 = TL S i1 T12 + TT2 S i2 + Si'i T12 + Si'2 (27)
in :Which the subscr~pts rand t denote rotations and translations , respectively A1 = TT1Ai (28)
lJs1~g Guyan reduct10n from Sec. 6. 7, we eliminate the rotations in Eq. (14) and
1·cta111 the translations, yielding A2 = TT2Ai + Ai (29)

M12 = TT1 M~ T,2 + TT, Mf2 (30)

M 1~D1 + S!Dr = A; ( I ~)
where T * T,2 + TT12 M*12
-M22 = T12M11 + M*21 T12 + M*22 (31)

s: = Sa + T~Srr ( I (1) 1•q1111tion (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
A t* -- Ar + TTrrA, ( 17) vn ·tor D2 and its second derivative D 2 may be substituted into Eq. (25) to
M! = Mu + T~M,r + M,, T,.r + T~M,, T,., ( 181 ,h ll•1mine the vector of axial forces Q in the members. Next, dependent trans-
l111 l1111s D I can be obtained from Eq. ( 6), and the rotations D, are found using Eq.
'l'itl' transformation operator T,., in these expressions relates the rotations D to tht•
, 11)) . Finally, other internal actions and support reactions may be calculated
trnnslntions Dr, as follows: '
l1rn11 known relationships.
D, = T,.rD, ( Ill I If a damping matrix is included in the equations of motion, its reduction
Whl'l'll 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
S,, 1 S, 1 ()OJ 111 111l' lllbcrs. I lowcvcr, these subjects arc more complicated and of less interest
Tlw opt·1·11to1 'I~ lst•c Eq . ( 11)1 cun now ht· IISl'd to trun sforn1 the rcdut•t•d 1111111 thl· 111atlur or Hxinl c:onstrnints, us discussed here.
1•q1111tH111s of 11101l1111 i11 llq . ( I Ci) to lht• /'t'1wrnll1t•d dlspl11t·t·1rn·111 rnordinhti•1, I) Wht•11t·vt·r 11xi11l st rnins 11n· 0111ittl·d from :urnlyscs of plane or space frames,
Ill J1q ( I()) 1!1•11111~1· 11~i11l l'OIISl111int~ JIil' Ill h1· 1111pc1s1•d , Wl' i,lwll 11111it 11x111l ,1 h1~11 11l 11l'l'll1'11l'Y ,~ h111111d t11 m·t·111 'l'hr 11l1 nilll·1111l'l' or such discrepancies will
 296 Framed Structures Constraints Against Axial Strains 297
Sec. 6.8

vary from one problem to another. However, for most practical under- 2 4 5
constrained 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
AL [3~ - 8 288
Sym 11
structures. Moreover, the numerical problem of ill conditioning due to com-
~ ~20
(j)
bining large axial stiffnesses with small flexural and torsional stiffnesses is M,, 6: 8 300.8
completely avoided. Of course, when the members in a frame are perpendicular 62 - 1.6 291.2 :
to each other (as in Example 6.11), omission of axial strains is easily accom-
plished without the formal procedure of this section. , . p·1g .6. 23 , we see that the parts of the rearranged load vector are
I•1om
Example 6.12 A, = {O, O} A, = {l , 1, 0.5, O} P(t) (k)

For the three-member plane frame in Fig. 6.23, let us first set up the stiffness, mass, and For the purpose of applying Guyan reduction , we calculate the inverse of sub-
load matrices for the six degrees of freedom shown. Second, we shall use Guyan 11111lrix S,, as
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 -I L [ 0.1564 - 0 .04790] (f)
best single independent translation, which was found to be displacement number 4. Last, S,, = El, -0.04790 0.2313
we shall calculate the response of the reduced system to a particular set of forcin1,1
functions. t'hlln from Eq. (20) the operator T,, becomes
From Eq. (6.2-14), we determine member stiffnesses (without axial terms), as T,, == - s;,. 1 s,,
scmble them, and rearrange the results to produce the following submatrices of thl·
structural stiffness matrix: 2 4 5

3 6 == .!.[-0.7083 -0.2301 -0.1787 0.2558] 3 (m)

L 0.6764 - 0.3890 -0.6374 0.2650 6
_ El, [ 6.828 1.414] 3
s - - (c)
" L 1.414 4.617 6 I !Hing Eq. (16), we find the reduced stiffness matrix si to be
2 4 5 2 4 5
S = El, [ 3.879 2.121 2.121 -2.121 J 3
(f)
9.940
Sym.11
" L2 -2.121 2.121 3.195 -1.585 6 0.808
"' El, -2.189 (n)
1 2 4 5 S,; == U [
-1.463 0.391 0.565
-1.017 - 0.303 1.374 :
[ 1412
_ El, -2.121 2.121
Syml 111 11ddition, Eq. (17) yields
2.552

S - -3 (g)
" L - 2.121 2.121 2.980 A: == {1, 1, 0.5, O} P (t) (o)
2.121 -2. 121 -1.692 2.336 :
,111tl llq. ( 18) gives
Si mil arly, consistent mass matrices for the members are drawn from Eqs. (6.2- 15), 5
2 4
nNNllmbled, and rearranged to give

3 6 [ 317 9
* _ pAL
-3 1.46 289.1
Syml
(p)
_ pAL3 [ 8 -3] 3 M" - 420 81.91
M,, - 420 -3 (h) 1.152 262.0
8 6
1.733 62.84 6.302 293.6 :
2 4 5

M,, p/\l}
•I W
r I 92
6 .444
<) ,
15 .56 9. 192 9. 192] 3
(ll
Now we shall furth er reduce the stiffness , load, an~ '.11a~s matric~s to accoun~ f?r
,1, 111 111111 strniuts 0 11 the
mcmhllrs. By rcnrrunpiog 1111d p,1rt1t1omng the stiffness matnx m
9. 11) } \\2•1 5.7 12 6 I q (II) Wt' fi11d thnl
 299
Sec. 6.9 Programs DYPFAC and DYSFAC
298 Framed Structures Chap, e
We now have all the matrices needed to find the response in Eq. (24) and the axial
2 5 lorces in Eq. (25) for any forcing function. If the function P(t) is a step force A, Eq.
9.940 (24) takes the form
Sym.] I
*
S11 El,
= L 3 -2.189 0.808 (q) 1305pAL .. 22.41E/, O (a')
[ D4 + L3 D4 = 2. A
2.552 -1.017 1.374 : 420
I 2 5 l'he eigenvalue problem associated with the homogeneous version of Eq. (a') yields the
* = El,
S 21 L3 [ -1.463 0.391 - O. 303] 4 = S *1/ (1) 11ngular frequency:
(b')
4
* = El,
S22 L 3 [0.565] 4
11nd the response to the step force Pi is
the last of which is just a :ingle term. Similarly, for the load vector in Eq. (o) we haw P, L 3 (c')
D4 = 0.08925-(1 - cos wt)
EI,
A, = {l, 1, O}P(t) A; = 0.5P(t) (I)
Also ,
Furthermore, submatrices of the mass matrix in Eq. (p) are .. AL3w2
D4 = 0.08925-COSWt
1 2 5 El,

~
A (d')
Mi,= ~fa [~~;::6 289.1 Sym.] (u)
= 0.6437- cos wt
pAL
1.733 62.84 293.6 5 Substitution of D4, D4 , and the appropriate barred matrices into Eq. (25) produces
l 2 5
0.4442] ( 0.6454] (e')
M*21 -- pAL[
420 81.91 1.152 6.302] 4 = Mi2T (v) Q= 1.617 A - 0.7370 P, cos wt
( -0.7714 - 0.6638
4 which are the time-varying axial forces in the members. To complete the example, we
M;2 = ~fo[262.0] 4 (w) 1 ould also compute the dependent joint translations, the joint rotations, other internal
1111;mber actions, and support reactions. However, these tasks are straightforward and
Then the barred matrices in Eqs. (26) through (31) become will be left as exercises for the reader.
4

~
4
- EI 11.9 PROGRAMS DVPFAC AND DVSFAC
s, 2 = ~, [ ~::~~] S22 = e'[22.41] 4 (x l
-21.17 3 111 this section we briefly discuss programs named DYPFAC and DYSFAC for
dynumic analyses of plane frames and space frames with axial constraints. These

~
4 p1 ograms use Guyan reduction (see Sec. 6.7) to eliminate the joint rotations and
A, = P(t) [ 1.:l ] A2 = P(t)[2 .0] 4 (y) lhl'll further reduce the equations of motion by imposing axial constraints (see
- 1.118 3 '{l'l'. 6 . 8) . Response calculations for the reduced system are carried out using the
11m nrnl-rnodc method , as described in Chapter 4.
4 Programs DYPFAC and DYSFAC both have the same outline, as follows:
4
79 1.31 1 pAI,
574.3 2 470 11 105 1 4 (/) Outline of P1·0~1·nms DYPF AC and DYSF AC
799,7 1
lfrud 1111d wil h· Nl111l· tw11I tl11t11
h11 1111'10' 111111 yN 1111' i111h- 11t•N I ' ') • 11111 I I I Il' llt lh' llll' lllhl'I 1111111!1111~. w h111l'IIH ll1th•x ,1 II S t1111 t11111l p 111 11111t• h· 1P1
II1(11t'Nl' III ~ 1hr ltllll'( II 111h Ill 1111111 IIIIIINl11ll1111
 301
:too I 11111111<1 m, 1101111111 1•111111111111 llYl'I Al irnit 11V I (\I

b. .loi111 l'111111ll1111t,
C. Joint l'estnli11I N
d. Member 11110111111111111
2. Porm constraint lransf'o1111111io11 1111111 kc.:s
a. Assemble constrnint 11111td x
r
L 1 1' (,)
3.
b. Find dependent translations
c. Generate transformation matrices
Generate, condense, and transform stiffness and mass matrict•s
a. Generate member matrices
b. Assemble structural matrices
I'/.
3L G)

c. Eliminate rotational displacements

d. Reduce matrices (due to constraints)
4. Determine frequencies and mode shapes 12
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
-l---..+----X /.
e. Normalize modal vectors with respect to mass matrix
5. Read and write dynamic load data
2L~
a. Dynamic parameters
b. Initial conditions
c. Applied actions z (al
d. Ground accelerations
e. Forcing function
6. Calculate time histories of displacements P(t)
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
P,
7. 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
t,
0
Both of these programs perform the reductions optionally. Therefore, tw11
(b)
m•w 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 Figure 6.24 (a) S
ix-member plane frame; (b) dynamic load.
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 rnh·N 1111 1111tll•1co11s1rnl111•d pl11111• I 1111m• with I wl•lw cl1·H1l'l' N·of I 11•1
 aoa
1'1111!1111111
.102 I rru1111d St1uotur 1111 c;h111
1111 1 tll'h of the th1 cc l'IISl'S W1 Hl I 1\1111 1111 II MpOIIHl' C\IIVC for Case (b) is indistinguish-
dom ( 12 DOF) . 'l'I1i.l. I1111111•
. rn11111l~l• 11I I• ti tldl . .. • . . 1lih 110111that fo r Cus1: (11) . 111 11dd l1h111, till' H'HJlllt1Se curve for Case (c) differs by only
which have the sa me vah1us ol /• ' I I Y i onn~ctcd pr1s mt1t1c me111h1:1N' 1111 11 11111 umount from thu olhl'IH (1•11nt Is 11pprnxl mate ly - 3.3% at the first peak). Thus,
I ';"''
111 thu .1 di rection at jo ints 2 u11d ,'1' , ' ~ I >y111111 tl c lorces p (1) and 2P (1) !Ill' nppll 1111 h•vcl o f accuracy shown hy 1hls L'xumple seems sufficient for practical purposes.
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

E = 3·0 X 104 k/'m. 1 fJ 7 .35 X 10- 1 k-s 2 /in. 4 REFERENCES

4
A = 30 in. 2 I, '"' 1.0 X 103 in.
Weaver, W. , Jr. , and Gere, J. M. , Matrix Analysis of Framed Structures, 2nd ed.,
L = 50 in. 10 k t2 = 2t1 = 35 ms
Vnn Nostrand Reinhold , New York , 1980.
whure US units are implied. 1 Weaver , W., Jr., and Johnston , P. R., Finite Elements for Structural Analysis,
. T hi s structure was analyzed b Pro ra . Prentice-Hall, Englewood Cliffs, N. J., 1984.
wtlhout reduction (12 DOF)· (b) ·tyh 1· g_ m_ DYPFAC three times, as foll ows t111
~ m1nut1on of rotations, followed b reduc .
•Ii . . ' w1 e immat10n of r t t'
o_a 10ns
(8 '
DOF); and (c) wllh
Yoshida, D . M. , and Weaver, W. , Jr. , "Finite-Element Analysis of Beams and
6.25 shows time histories of respofse t :~on_
due to axial constraints (2 DOF). Fip111
or e independent translation D4, as comptth ii
Plates with Moving Loads," Int. Assoc. Bridge Struct. Eng., Vol. 31, 1971, pp.

179- 195 .
<luyan, R. J., "Reduction of Stiffness and Mass Matrices," AIAA J., Vol. 3, No.
D,(i n.)
l, 1965, p. 380.
, 'l'imoshenko , S. P., Young, D. H ., and Weaver , W., Jr , Vibration Problems in
0.2 G) and@ /1'11gineering, 4th ed. , Wiley, New York, 1974.
11 Weaver , W. , Jr. , and Eisenberger, M. , "Dynamics of Frames with Axial Con-
Htraints," 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 equiv-
alent 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
equivalent nodal loads at points j and k, with Xjk = Yi k·
II l, 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,L•4, 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 \rl·CtillllH rm 111 1 und Y/4 4'
r, ,,l,,!I. 'l'w11 cum•1•11ti11l1•cl l11H'l'S P ~·(I) 1111· 11pplk d ut the third points of the plane frame
 304
I t 11111ml Str 1101111111 Prnhl111111t 306
r11crnhc1· In Fig h '(II) c ,111 1
directions usii1u I
'
1
JI
' "
11 ti ''," •
11111 I ,1
q111 1 ,tit Ill 1111d11l louds in memhcr and ~,1111f
I
lk peat Prnh. 6.•I ~ wllh 1111 11111111,•111 111·tl11g in the y' direction at the point
6.2-6. Assume that two 1111 ,, 11 • 1 111 I ' /, /5.
member in Fi l 11 1
1I) ' " I ut tlw q1111rter points of the pl11111 Ir
Ass ume that a unifo1111ly diNtilhuted force b (t) acts on a space truss member in
. g . 6.2(u). lktt•1111111r i·q11l v111l-11t nodal loads in Joe· I I I
coord mates for x1k = YI*· u u11t 11 • the di rection shown by Fig. 6.4(b). Using the relative coordinates Cj k = {2, 2, 3}
6. 2·7. A parabolic distribution of force /J (t)
frame member in Fig 6 2(a) F
e . .
.. 21s applied m the x ' direction on 1hr pl,11
nrtd c1,, = {O, 5, O}, delermine equivalent nodal loads in member and structural
di rections.
I d . . . . or XJA and YJk = - I find th . I
oa s m member and structural directions. ' e equ1vu t•11t 111 I fl ~ H. Repeat Prob. 6.4-7 for a triangular distribution of force [bi(t).

6.3-1. Let a uniformly d'st 'b t d f; fl ~ t. Let the space frame member in Fig. 6.5(a) be subjected to a uniformly distributed
the grid member /n ;i u e orce b, (t) act on ~alf the length (L/2 s ,\ ' I I
force by·(t) over its whole length. Find the equivalent nodal loads for both local
and global directions, ti!:.~:)~ ~al:~a:e e~u~valent nodal loads for both I, and global directions, assuming the relative coordinates c1k = {3 , 2, 4} and Cjp =
6.3-2. Assume that a concentrated f; P. ( ) . 1k • • {- 2, 4, - I}.
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 · fl ~ L. Repeat Prob. 6.5-1 for a uniformly distributed force b,·(t) over half the length
alent nodal loads in both me b d j~ - -: , etermme tht· ,·qui
6 3. . . m er an structural direct10ns. (O s x ' s L/ 2).
,. 3. The gnd member m Fig. 6.3(a) has a . fl I. Suppose that a concentrated force Py,(t) is applied atx ' = L / 4 on the space frame
x ' = L/4. Find the equivalent nodal 10:~~entbMb(:) applied at the point wh member in Fig. 6.5(a). Using the relative coordinates c1k = {2, 3, - I} and
using x jk = YJk· s or ot ocal and global coordi11111 c1,, = {-1 , 2, 2}, calculate the equivalent nodal loads in member and structural
6.J -4. Repeat Prob. 6.3-3 for a moment My(t) in they direction directions.
6 ..l•S. Rep~at Prob. 6.3-3 for a moment Mx(t) in the x direction. fl ~ 4. Repeat Prob. 6.5-3 for a force in the z' direction at the point where x' = 3L/ 4.
<,.J .6. A tnangular distribution of force (I - 1:)b () . . fl 8 ~. A moment My•(t) acts on the space frame member in Fig. 6.5(a) at the point where
1 1
member in Fig 6 3( ) F _ !> acts m the z direction on tlw "• Id t' = L/3 . Determine the equivalent nodal loads for both local and global direc-
· · a · or XJk - -4 and y - 3 b · "
loads in both member and struct 1 d' I* - ' o tam the equivalent 111111,II lions, assuming the relative coordinates Cjk = {O, 3, 4} and Cjp = {O, 0, -2}.
ura coor mates.
<d-7. Suppose that two concentrated forces P. (t) ar . fl ~ <,. Repeat Prob. 6.5-5 for a moment in the z ' direction applied at the point where
g~id member in Fig. 6.3(a). Calculate; uiv ~ applted at the q~arter points ol th r' = 2L/ 3.
directions for X · = 1 a d - 3 q a ent nodal loads m local and ••luh1tl fl II 7. Assume that a uniformly distributed force bx(t) acts in the x direction on the space
1* n YJk - • "
6.J -8. The grid member in Fig 6 3( ) · b' frame member in Fig. 6.5(a). With the relative coordinates Cjk = {I, -2, 2} and
a moment - M ·(t) at x'. =. 2Lj;s ~u ~ecte~ to a moment My-(t) at x ' = l / J 111111 c1,, = {- 3, 4, O}, obtain the equivalent nodal loads for member and structural
nodal loads in Y member and struc.t olr xd~k - _I and YJk = -2, find the equiv11lr111 directions.
ura 1rect1ons.
fl , H. Repeat Prob. 6.5-7 for a triangular distribution of force (I - t )b1(t ) in the z
c,.4- 1. Suppose that the space truss member in F . direction.
fi
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 fl I I. The two-element prismatic beam shown in Fig. P6.7-1 is fixed at point 3 but free
CJA = {3, 2, 1} and CJµ = {- 1 3 -2} rect1ons, usmg the relative coordi1111trN
to rotate at point I. Construct the stiffness and consistent mass matrices S and M
(1.4-2. Repeat Prob. 6.4-1 for a un;fo~ly ~istributed ~ for the three unrestrained displacements. Reduce these matrices by eliminating
segment where L/ 3 s x ' s 2L / 3 . orce b,·(t) applied ove1 flu lhe rotations and retaining the translation. Then find the angular frequency of
(1,4-J. Assume thut a concentrated force P,,(t) act ' - vibration for the remaining system, which has only one degree of freedom.
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 ly
(1.4-4. Hl•peut Prob. 6.4-3 if the force acts in the '
I' 1/,/5 . Y direction al lhe point wh,•11 3L 5+
,,.4-!i. I i·t II moment M ·(t ) be upplied at . t '
1~ 1
CD • 0
6/~ 3
l /
111 Fig . (>.'1 ( 11) . C'11lc111t111· the l'q11iv111~:~:11,m:l11I lt,4/1 on !he SJ)lll'C truss .nll·111h1·1
101 th,· 1l'i11fl w 1,11111 11 1111111.~ 1• /l ' ) 111 N 11 lornl 11ncl 1,1 loh11l d111·ctlt11iN
~/4i 1- 4/·1-
X

• 11 , • 1, 1111d 1• ,
11 { I , l, 2),
/ 1:
' 11'1111111• l't,,7, I
 307
f'rohlnm1
3015 Frnmnd Struoturna 0
y
6.7-2. Figure P6.7 2 shows u two 1•lrml'lll pliN11111tl1· lw11111 thnt is free to trunNhlll
not rotate) at point I nnd is llx1•d 111 polut I Ass1·111hle the stiffness anti rnn•I
mass matrices Sand M for the th11·1· 11n11•s1111incd 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
the members. Then reduce an
J6b liminating the rotations D1and D2, while
e . the angular frequencies and mode
ietaining the translations D3 and D4. etermme
shapes of the remaining system .
z

Figure P6.7-2
03 \ r.:;--,
!..--,_.....,;\V::::::3:__--1~ 4
6.7-3. The simply supported beam shown in Fig. P6.7-3 is divided into two flc,iu /3
elements with equal properties. Set up the stiffness and consistent mass mutrl D,
y
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 .

,+, CD 3L 0 5+ 3
• X

~J;; e Y.,. 6/~

/
z
I: L :1 F igure P6.7-S

. F
Figure P6.7-3 fl 1 ,,. Assume that the plane frame shown m
I und 2; whereas, elements 3 and 4 hav~ ' S
1i P6 7~6
has 2El and 2pA for elements
and . A for their properties . Construct
M· and eliminate the rotations
the stiffness and consistent mass matnc~\. anD a~d D Calculate the angular
6.7-4. A two-element prismatic beam has no restraints whatsoever , as implied in liill U, , 0 2, and D,, while keeping thehtra~s a •.o~~g ;ystcm s~eglecting axial strains
P6. 7-4. Construct the stiffness and consistent mass matrices S and M for the Hix li cquencies and mode shapes for t e ,cmaim ,
unrestrained displacements, and eliminate the three rotations. For the thrc1· '"
111 the members.
maining translations, calculate the angular frequencies and mode shntK'S.
 JOB
308 I 11111111<1 H1111011111111 hnp n 1rnhl1111111
1

n I n·111H in !lig. P6.7-8. l)ctcnnine the

y U
A 1•,1id with tlll l'l' i•q1111l 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,
"·
lrcqucncy of lhL· i1·1h11•1•d ~YH l'II\1 p ,r each o t e m
~O and E = 5G/2.
•
structure, \
1
(1) "· 1, G)
• - - -- -- - - - -
3 o ..
ussume that 21, 2 /1, ,/ , ,.N
z
,/ '

03/
o, / - - L ~ 1- -
l
031
0
\ CD - D2
0 3 I'
y
,__/
4
-----x 5
'/

/--' L

/
.,
Figure P6.7-6
X

4
/
Figure P6.7 -8

(1.7-7. Figure P6.7-7 shows a grid consisting of two prismatic members with 1•q1111I
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

/
Figure P6.7-7
 311

7 wl IIIL·
tlll llNIIOII

displacement
111ul t.111111 111 Irr I 1111111111n

Vl'l'lo1 IH
u - {u, v, w} (3)

I 11t lltl•1more , the 6 X 3 Stra in-displacement operator d has the form

a 0 0
ax
Two- and 0
a 0
ay
Three-Dimensional 0 0
a
az
d = (4)
Continua a a
ay ax
0

a a
-
0
az ay
a 0 a
az ax
. . ak from Eqs. (3.2-2) and (3.2-3).
7 .1 INTRODUCTION 111 which the derivatives are t den t ·n relationships for an isotropic ma-
In addition we develope stress-s rm .
As before' such relations
· h"ips can be wntten-.
,, , /' 1/ .
1 >I<
Tht framed structures discussed in the preceding chapter consist of only sh.-1111 ' (5)
(or one-dimensional) members. However, we also wish to analyze solids 1111d CT=EE.
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 111111 rc the stress vector is
plnnc strain LI] as well as general and axisymmetric solids. Our emphasis will CT = {CTx , <Ty, <Tz, 1"xy, Tyz, Tzx }
(6)
h(.) upon the isoparametric formulations leading to the most commonly l1Nl1d · E was given by Eq. (3.2-lO).
·
t•lc:mcnls 121. Computer programs cover dynamic analyses of plane-stress 1111J ,1 11d the 6 x 6 stress-strain matrix . l' our discussion to problems in two
ph11w ·Strain problems, general solids, and axisymmetric solids . In all cases lh For the moment, Let us specia ize. ly those terms depicted in Fig.
,l1 111onsions, where the stress vector contams on
di screti zed continua are assumed to be linearly elastic with small strains 1111'1
displacements. I l(u). Thus, in the x-y plane we have (7)
CT = {<Tx, <Ty , 1"xy}
· p·1g. 7 · 1(b) ' which
re the strains indicated m
7 ,2 STRESSES AND STRAINS IN CONTINUA ' ' 111 t'cspondingto these stressesda. fi .t simal eiement in the x-y plane. Hence'
hows a displaced and deforme m m e
111 Sec . 3.2 we introduced the topic of stresses and strains in three dimension 11 11 i,train vector is
(8)
1111<1 dclinedstrain-displacement relationships with respect to Cartesian corndl E. = {Ex , Ey, ,'xy}
1111 ll' s. Now
these relationships will be expressed in matrix form as
111 which
E =d U ( 11 au av au av
E =ax
- eY =- --+
,'xy -ay -ax
(9)
which applies lo either static or dynamic analysis. The strain vector in Eq. ( 11 1 iJy
IH
•• Rc•r. 2·' und for generally an-
' ( ., 11,,111111/r· 11111lllrit1I N, Nl'l'
ot,opic (m·1/1111r1111/1•) i1111ll' t 1n1N, St e
~ l'nr o, thol(OHlllly u11lM
l {(I I , "

110
J12
Two- and Three-Dimensional Continua I oc. 7.2 Stresses and Strains i n Conti nua 313

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
yx a 0
ax
a (10)
d= 0
ay

·-J dy

dx
1111d the vector u is
a a
ay ax
y
(11)

L,
' u = {u, 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:
0 = 'Yzx = 0 (12)
(a)
<J', = 'Ty, = 1'zx =0 € , =I=- Yyz

111 particular, note that the strain normal to the plate is nonzero. Using these
nmditions in Eqs. (3. 2-5), we find that

If-- ~
l~ : ~ 1
dy
i)V ,w
V I ily dy
(13)
E = -----=E
1 - V2 1- V
0 0 -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
V,V 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
i!u
= =0 '=z = ')'yz = Yzx =0 (15)
U I dx <J', =I=- 0 1'yz T,x
i!x
(ll) 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
(11) Stresses und (b) struins In two dimensions. nqs. (3.2-5), we obtain
(16)

1111d
 314 315
Two- and Three-Dimensional Continua Sec. 7.2 Stresses and Strains in Continua

>rvl
y
y

E - (I + v)~I - 2v{ : v I I
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 X
X

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' = [~: (J 8) z
Tzx z'
(b)
In which complementary shearing stresses are included. Then the rotation-of (a)
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' y
I X J rotation matrix R in Eq. (19) has the form y
y'
A1 1
Yp
R = A2 1
(20)
[
,\31 x'

111 the lultcr matrix the terms / 1, m 1 , and so on are slightly more efficient symbolN
i'
Im the direction cosines A11 , A12 , and so on, that were explained in Sec. 3.fi ~
0
X
Nolu lhul lhe rows of R contain the x, y, and z components of the unit veclor11 j~ X

I' , .I', und k' shown in Fig. 7.2(a). Id)

(c)
Similarly, the strain vector in Eq. (2) may be restated as the 3 x 3 sy,11
111t1l1 k matrix
. . dimensions: (a) inclined; (b) principal-two
Figure 7.2 0 rthogona1 aXes-three
dimensions: (c) inclined; (d) pnnc1pal. .
€x 'Yxy 'Yxz]
E = ')'yx €y 'Yyz (21 l . 1matrix (or spectral matrix)
111 this equation the symbol represents a diagona

l]
[ O'p
'Yzx ')'zy €,

which in<.:ludcs dependent shearing strains . Again, the rotation operation is nl principfil normal stressesu:lr ~' (24)
(2.11
111 whld1 the 3 X 3 matrix e' is similar to E in Eq . (21), but for inclined 11xt•11
Wt• cun obtain principal normal streS.\'<'.I' und I heir clircclions us the solut i1111 .1 fl (23) denotes the rotation matrix for the
111 1111 ,•ll{1·n1 ul11,· prohl<•111 141. Using thut nmthod )'iws
1
111 nddilion, the symb~>l ~,, ; ~~) This mntrix is the transpose of the nor-
1,,/m·liwl rm•s shown 111 hg. ·,2 , · ,
er,, 1{1, er H /, (.! I) ,111111/t· d ,·IR1'111•,•,·1or 111,11ril c1,N I hnl IK,
 .1111
Two- 011d 1111011 Dlrnonslonal Continua Chap, 1 Sec. 7.2 Stresses and Strains in Continua 317

Rp = <t>X (2') Therefore, the stress transformation matrix '(,. is proven to be the transposed
lnwhichthc rowsof R,(orc l . f<I>) inverse of the strain transformation matrix T,. The inversion implied by Eq. (33)
Simi larly . .' l o umns o. N are normalized to have unit lengths
' prznczpa . normal strains may be calculated as is not actually necessary, because expansion ofEq. (19) shows that the parts of
T" bear the following relationships to the parts of T,:
. Ep = RpE R] (2(11
In tlu s expression the symbol Ep stands for a di
l'1ll'rcsponcfing to the stresses m
· <Tp.
.
agonal matnx of normal strain-
TO' = ['t.11 TO'l2] = [IT,,, 2T,12J (34)
. T0'21 To22 2 T,21 Tm
In subsequent work it will become nece As before, the subscripts 1 and 2 on the submatrices of TO' refer to normal and
IIIHtrix E from one set of coordinates to ano~~d to trans~orm the stress-stnriu
shearing stresses, respectively.
purpose, we rewrite the expanded results of Eq. (~~/~~atton of axes. For th,~
Now the transformation of E' to E can be accomplished by first writing
e' = T,e stress-strain relationships in the primed coordinates as
(27)
l lcrc the strains e and e' are in the form ofE ( 2). d u' = E' e' (35)
.l'lm/11 tronsformation matrix T, in Eq. (27)qis mstea ofEq. (21). The 6 X ,,
Next, we substitute Eqs. (27) and (32) into Eq. (35) , producing
/7 TO'<T = E 'T,E (36)
Ii
/ .I2
Then premultiply Eq. (36) with T;;', and use Eq. (33) to find that
·r: • 2/1 /2 CT = TJ E 'T.:E (37)

2/2/3 Thus, we see that

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'' = OET<T

substitute the transposed incremental form f Eq

( 10) sin (} J (41)
' l'h1111
ohf11l11 ° · (27) into Eq. (30) to cos (}
111 which the rows contain the x and y components of vectors i' and j'. Further-
1110rc, the symmetric 2 x 2 matrix of strains becomes
( 11,
I l11111't'. we conclude that
E= [Ex Yxy] (42)
er' = 1;,o 'Yyx Ey
wh1•11•
whklr <.:ur1 h1; applkd in Eq. (22) to find strains E' for the inclined axes.
'(;, • 'J' ;- I
By solving II Nl'l'Olld Ol'(kf' t't/•l·nv11llll' prnhlcm , we cun also obtain prin-
1 lp11l 11011111d 1-1111•:,i,t•1, 1111d Nl111i111, in two di11111111,lo111,. Thus, tht• l'igt·nv11h1t• prob-
 318 Two- and Three-Dimensional Continua 319
Sec. 7.3 Natural Coordinates

lem associated with the 2 x 2 stress matrix in Eq. ( 40) is

(O' - ,\;l)<I>; = 0 (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
ll = 0

o-y T~ A; I= 0 (45) 11 =
Expanding this determinant yields 1
y i; = - 1 i; = -
2
+ + =0
L.
AT - (o-x 0-y)A; 0-xO'y - T;y (46)
ll = - 1
From this quadratic equation, we obtain the roots
(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

~~=~+
i; = 1

T. = [---:: - - - - :: I - - - - --- -
- --4-
2l,l2 2m 1m2 / l,m2 l2m,
J (48)

2
Also, the stress transformation matrix T" becomes

If2 mf2 I
I
2/, m, J y
s= - 1
T" =[ l2 m2
------- -4- - - --- - - -
11/2 m,m2 I l,m2 + l2m,
which is related to T, in accordance with Eq. (34).
I 2/2m2 (49)
L. ll = - 1

(b)
s = --12

Figure 7.3 Natural Coordinates for quadrilaterals: (a) straight; (b) curved.
7.3 NATURAL COORDINATES
pl une are

111 lllis Sl't'ti on we de fine natural coordinates for q1111drilatcrals and hcxahedrn in xH = Hx1 + X2 + X3 + X4) Ys = HY1 + Y2 + y3 + y4) . (1)
prnp11rnt io11 for· dl'Yl'lopment of isopurnnwt, k 1•lt·1111•11t 11 having these shapes . Wt· . ( ) arc the x and y coordinates of po mt 1,
111 these averaging expressions, Xi, Yi .. , .. ·1 the same as the
li1·1il11 with 11 <J11<ulr llt1t1•r,il that has st11d ~ht 1•dg1•11, 11~ ill11stratl1d in !lig. 7..l(i',) N t ti I tl w gl•n1111.llrk t'l'llll' I' is no t ncccss,m Y .
'I'll!' poi nt l11lwl1•d ,-: is tilt' Mr'o1111•trl1 1·,·1111•1. !111 \\ hi, II lh1• 1•rnudi1111t1•s in lhl' 1 1• 1111d 1-10 on. o l' Ill Al I i I th . llHllll' un· till' di11w11.1·im1f<•,1·.1·
l l' llll Old 111 th1• q1111dtil11h•11il t\ NII S HlW II I l
 321
1:10
lwn "'"' ll11nn 111111,1111111111111 Cu 11 t1 111111 <:hni, '1111:. / ,'J N11t11111t Cow dl1111lot1

11,111m1/ 1·11<1rdl,u111•.1· t and T/

Ill poi111 /.:, it is i111portant l<~
Alli
ru·
I
i','.H'.~ 1' f111 >j1 '1 ls1•w rn.onlinutcs huvc !hell' rn 1)!111
11or this arrangcmcnl, terms in the coefficient matrix are easily obtained by
differentiating Eqs. (2). This array is called the Jacobian matrix J, which
pol111 o11 lhc clement. We see t~' 1("~
I ml Ihr Ii dlt l'l'fions arc different ut l'Vl'I V l'Ontains derivatives of the global coordinates with respect to the local coordi-
' I, T/ . I ulong edge 3-4, an~
pol11!1011 111 both the 1: and d" .
J_ / "
1011
~ 7dgc 1-2, ~ I _along l'd1,tr
,dong cdge 4-1 · With linear i111r 1
s nates. Thus, we have
!> r, 1rect1ons form la . f 1 .
unywhcrc on the quadrilateral become ' u s or ocatmg a generic po1111 J= [111 112] = [x,g Y,g] (8)
121 122 X,') Y,ri
x= I4

1=1
f;x;
4
y = 2, f;y;
. I
( ,, Terms in this 2 x 2 matrix are
4
4
111 which the geometric interpolation functions '~e 111 = x,g = L f;,gx; 112 = Y,g = L f;,gY;
i= I (9)
!1 = Hl - g)(l - r,) f2 = !(I + [)(1 _ r,) i=l
4
4

, , • • / j = !(1 + g)(l + r,) f4 = HI - [)(1 + r,) ( I) f 21 = X ,T/ = 'v +.

~Jl,T/X I·
122 = Y.') = L AriYi
i= l
i=I
I Ill Sl' I 1111ct1ons give the global coord· t
1
f .
l!~1111s of lhc natural coordinat H ma es o a pomt on the quadrilaternl 111 Casting these expressions into matrix form yields
hypl11·holic), the local coordina~es~ gao~ever, because Eqs. (2) are bilinear (rn (10)
ol Iill' global coordinates x and n 71 cannot be expressed directly in te111111
J = DLCN
W ·11 y. The array DL in this equation contains derivatives with respect to local coordi-
c w1 need derivatives of the function "
dl11plun•111cnt relationships The h . s J ! through f4 to use in stra in nates, as follows:
u•spl'l'I lo x and y gives . 1 for differentiation off([, r,) with
c am rue

af = af ag + af a71
DL = [ !1,g fi.g /3,g !4,{ J
!1.ri Ari Ari Ari
ax ag ax
af = af ag +
ar, ax
af ar,
(ti)
=4
[-(1 -
1 TJ)
-(1 - g)
(1 - r,)
-(1 + g)
(1 + r,)
(1 + g)
-(1 +
(1 - g)
r,)J (11)

ay ag ay ar, ay and the matrix CN consists of nodal coordinates in the arrangement

01'

wh1•n• the s m .·
[J.x] = [S.x 'Y/,x] [J, g]
!,y. S,y 'Y/.y !.ri ( )
5 CN = f:: ::1
X3 Y3
(12)

,, . · Y bol f. x means differentiation off ·th X4 Y4

I 1•1111s 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' By comparing Eqs. (5) and (7), we can see that the coefficient matrix in
1 ior !> an r, m terms f d
'""'' I1ll' opposite approach and differenti . o x an y. However, if w,• the former expression is the inverse of the Jacobian matrix. Using the formal
111lt• produces ate/with respect to sand r,, the chain
definition of the inverse, we can obtain
1
from J as r
-y,g]
af = af ax +
ag ax ag
af ay
ay ag
((l)
-112]
f1 1 = m1 [
-x, 71
Y.ri
X,g
(13)

af = af ax + af ay where Ja denotes the adjoint matrix of J and IJI is its determinant. The latter
a,, ax ar, ay ar, quantity is calculated by
1)1
\JI = l,1122 - Ji1l12 = X,gY.ri - X,riY,g (14)

f,1
f = [.x., Y.cl [J., l
IJ.,, x,,, Y.,, /1,.I ('/)
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
Sec. 7.3 Natural Coordinates 323

fi.x] =
[ fi.y J-1[fi·fi.7]s] (i = 1, 2, 3, 4) (15)
numerical integration for both straight-sided and curved quadrilaterals, as ex-

Altogether, we have plained in Sec

Now . 7.4.
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
(16)
1 8 1 8 1 8
The matrix DG given by this expression consists of derivatives of Ji with respect Xg = 8L X; Yg =- L Yi Zg = 8 L Z; (19)
to the global coordinates. That is, 1=) 8 i=I i= l

DG = [fi.x fz.x .h.x f4.x] (1 )

T)

ii .y fz.y h.y Ay 7
3
Evaluating terms in DG, we find
]
Dc11 = 4 / J / [ -(1 - YJ)l22 + (1 - [)1 12] y

Dc12

I
1
= 4 / J / [(1

Dc13 = 4 / J / [(1
- YJ)l22

+ YJ)l22 -
+ (1 +

(I +
[)1 12]

{)112] z
)-. 5

I
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 + YJ)l21 + (I + [)Ji i]
1
Dc24 = 4 / J / [(1 + YJ)l21 + (1 - [)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 z 5
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 ~ =1
11011s, Wl' 11suully cannot integrate terms explicitly to obtain stiffnesses, consis lb)
lli11t ou,sscs , und equivalent nodal loaclli. lnstcnd , ii hl•comes necessury lo 11s1•
Fl11111·1• 7.4 Nuturul coordinurcs for hcxnhcdrn : (11) straight; (b) curved .
 il-(
I w11 nrn I 1111 n.- I>l111111111lur111I Cu11tlr11u1 Chnp Ill ,.:1 Nntw 111 (.;oor dltrntuu 325

wht'll' ,,, I''.' "."d i, 1~1·c the 0 1rt<:siu11 l'11111dl1111h•11 of poi nts J thrnugh 8. 'l'lw fi ,,,., (I - 77)(1 - () (1 + 77)(1 - ()
11lN11 Hhows t/1111,•11.,·101,le.1·.1· natural coor,lln,,r,•,,· (.. '>'J und r Nol<: lhal 1: I.ti
(I ~)(I - I,')
fm•t• l l 7 ( 11 + (1 + {)(1 - ()
• >: 77 . J on face 3-4-8-7, nnd so 0 11 .' With ' linear 11
b· b I - {)(1 - () -(1 g)(l - ()
interpolation i 11 th
~~ 'I• und ( d1rcct1 ons, the location of any point in the hexahedron may be wrlflrn - ( I - {)( l - 77) - (1 + g)(l - 77) -(1 + g)(l + TJ)

s s 8 - (1 + 17)(1 - () -(1 - T))(l + t) (1 - 77)(1 + ()

X = 2_ fix; y = 2, f;y; z = 2, f;z; ( .'ll I
i=l
(1 - g)(l - () -(1 - g)(l + {) -(1 + g)(l + ()
i= I
111 whk h the geometric interpolation fu,nctions are - (1 - g)(l + 17) (1 - g)(l - 17) (1 + g)(l - TJ)

/, ff{ I - {)(1 - T))(l - t)

./'i HI + g)( I + 77)(1 - {)
!2 = ko + t)(1 - 11)(1 - c) (1 + 17)(1 + t) -(1 + ~)(l + I,')]
Ii = k(I - {)(1 + 77)(1 - C) (1 + {)(1 + () (1 - g)(l + () (24)
.f'.. HI - g)( I - 77)(1 + ?) 16 = ko + t)(1 - TJ)(1 + c) ('II (1 + g)(l + 17) (1 - g)(l + 17)
.Ii ff{ I I s)(I + TJ)( l + {) f s = k(I - f)(l + 17)(1 + () \ 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
1111d , 1111rnot be expressed rn terms of the global coordinates
I ' II . ' d. .
d
x' y' an z.
' 1
1
X1 Yi Z1 1 (25)
' ucc. 11ncns1ons the chain rule for differentiation with res ect to lhr
11111\rntl rnordrnatcs for a hexahedron leads to the foll
CN = -~2. Y2 .~2.
· 3 P
111 ,,r,'II : owmg X 3 l aco/,/, 111
r Xs Ys zs
The inverse of the Jacobian matrix may be expressed as
1 = [;~: ;~: ;~:] = [:: r·: :::] () l,
h, 1 32 ] 33 x,, YJ z., (26)
1
'f't 1111s in thi s matrix are found by the differentiations indicated. Thus,
S 8
where the symbol Jarepresents the adjoint matrix of J, and IJI is its determinant.
1 12 = 2, f;,€Yi 11 3 == 2, f;,€z; l'o find the derivatives of all of the functions with respect to global coordinates,
I I i= l i= I
8 S s Wl' use Eq. (16) . In this case, the matrix DG consists of the following terms:
.!2, - ~ fi.r1X1 122 = 2, f;,.,,y; 123 = 2, J;,.,,z;
I I i~ I i= I
(2 1)
_ [/1,x A x Ax /4,x fs ,x Ax / 1,x fs.x]
(27)
8 8 8 DG - !1.y !i,y /3,y !4.y fs ,y f6 s !1.y fs.y
.I 11 L !t.,x, 1 32 == 2, f; ,,Y; ] 33 = 2, J;,,z; f i,z f2.z Az /4,z fs ,z /6.z h,z f s.z
I I i• I i• l

AN hl•IOi l',. lh.csc calculations may be arranged in the matrix format given b p 1 A hexahedron with curved edges appears in Fig. 7.4(b). Geometric inter-
( I !I) In lh1s 111slance, the n:iatrix Dl is the following 3 x 8 array of derivt~i~1 ! polation functions for this solid may be quadratic, cubic, and so on. The natural 'I

will, H·s1wc1 to local coordmates: " 1•oordinates 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,

u,
.Ii., .Ii., .li., A, 1~., /6,, !1.,
.Ii ,1/ .Ii.,, /1.1, 14.,, 1~.r, A,, Ari ls.ri t.,] 11llllough the functions to be differentiated are of higher order. As for a quadri-
l1111•rt1l with straight or curved edges, it becomes necessary to use numerical
l11tl·~ration to evaluate stiffnesses, consistent masses, and equivalent nodal
.Ii., .Ii., /i,, ./~., 1.~., Ji1,( !1., .IH.t
l1111ds.
 328 327
lw11 111111 11111111 l>l1111m11lo1111I Cc111tl111111 111 /JI N11111n1 lo11l l111u11111tlo11

7.4 NUMERICAL INTEGRATION

The process of computing the valuc ol II dt•flnitc integral I see Fig . 7 .5(11ll

l
'l
Ix = /(x) d x t IJ
x,
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. Wht•11 1111
plied to the integration of a function of a single variable, the method is rcl1•1111I
to as mechanical quadrature.
If interpolation formulas for numerical integration are polynomial~ 111
sufficiently high order relative to those assumed for displacement (or 0111111 (a)
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
x = H(l - t)x, + (1 + t)x2] ( 'I
Substitution of Eq. (2) into the function in Eq. (1) gives
J(x) = cp(g) ( ,, (b)
Also,

( II
Then substituting Eqs. (3) and (4) into Eq. (1) and changing the limits 111
integration yields
3

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 Figure 7.5 Gaussian quadrature.

In this cx.pression ~ is (he location of the integration point} relative to the t·t·111!·1,
N1 is 11 1w·i~liti111< fa ctor f'or point j, and n is the number of point.1· ul whic:h c/,( ( I
is to lw t·nku luted. Yult1es of these pururnt•(l\rs nn: listed in Table 7. I.
 329
1111 ,II N1111111tl1111t 1i1111u1111lt111
l\,\111 111111 tl111111 Ul11111111 lo1111I cw111!11111 < 11111 Ml /

2
Nu111L·1h;ul intcgrntinn hy, ,, 111111
dl'gl'l'C 2 11 I 'fl' t '.
,
,111 ,111,1rlm1t11·1· IS CXIICI for poly11011111tl I ,, ~ 1?1c/,(~1) (1)(3 ~O t ( 1)(3 - gD
' . ,H IS , 011f y 11111' i11(r)tl111i . • .
111tcgration of a linear functi , I oo point rs required f'or tlw ,·,11 j I

und so on. For example, con~;~ ~.. ~:1. p11111t ''.' \nccd_ed for a cubi c poly1111111111I = (2)(2.666
. .) = 5.333 .. .
ing in Fig. 7.5(c). First let n = ·tron <!>W
I ~t111111,111r11 , 11,11,~ uncT
mm
3 - g2upp1·111
able 7 .1
=
that
, hlrh is exact.
' t lit Now let us apply Gaussian quadrature to quadrilaterals in Cartesian coor-
g, = 0 /? 1 =2 1ll1111tcs. The type of integration to be performed is
Y2 l X2 f(x , y) dx dy (7)
TABLE 7.1 Coefficients for
Gaussian Quadrature
I=
lYI X\

llnwcver, this integral is more easily evaluated if it is first transformed to the

n ±(; R; 1111t11ral coordinates for a quadrilateral. We accomplish this by expressing the
l 0.0 2.0
llull'I ion f in terms of g and 1/ and using the limits - 1 to 1 for each of the
l11tl·grals. In addition, the infinitesimal area dA = dx dy must be replaced by an
2 0.5773502692 1.0 11ppropriate expression in terms of d(; and dr,. For this purpose, Fig. 7.6 shows
0.7745966692 1111 infinitesimal area dA in the natural coordinates . Vector r locates a generic
3 0.5555555556
0.0 0. 8888888889 pnlnt in the Cartesian coordinates x and y, as follows:
r =x+ y = xi + yj (8)
4 0.8611363116 0.3478548451
0.33998 10436 0.6521451549
lhc rate of change of r with respect to g is
0. 9061798459 0.236926885 1
5 0.538469310 I 0.4786286705 ar ax. ay .
- = - 1 +-J (9)
0.0 0.5688888889 ag ag ag
0.9324695142 0.1713244924
6 0.6612093865 0.3607615730 iJr
- dTj
0.2386191861 0.4679139346 ax ilTJ
- dT) dA
dT)
0.9491079123 0.1294849662 y
7 0.7415311856 0. 2797053915
0.4058451514 0.3818300505 iiy \ i!r
0.0 0.4179591837 - d-ri I - d~
dT) ~ i)~

0.9602898565 0.1012285363
iJr
0. 7966664774 0.2223810345 r + - dTJ
8
0.5255324099 0.3137066459 dT)
0.1834346425 0.3626837834
y
'l'hl'll from Eq. (6) we have

1E = R, </>((;,) = (2)(3) = 6
which is approximate. Next ' let n = 2 and o btam
. f rom Table 7. J:

=- ~=
X

(;, = - (;2 -0.5773 . R, = R2 = I

z
I knee, we lind from Eq. (6) li'IKllr(! 7.6 Infinitesimal area in natural coordinates.
:no I Wt1 1111il I 1111111 l>lt11,1111l111111I <.1111tl1111n
( I'/)
I' d I v,I I ~k
AINo , tltl' rntl' ol chungc ol I' With 1111,111·1 I lo t/ IN
1111 t ,tll'S ot' drnnge of,. wit h respect to , , T/ , uml 'arc
/Ir d , I I iJ y I
~ ax ay az
o'YJ ll tJ ii 'Y/' or _ - i + - j + - k
ag - ag ag ag
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 a ay az (18)
ar = ~i + -j + -k
111 cu may be detennined from the following vector triple product a'Y/ a11 a11 a11
a +-
ar _ 2-i ayj + -azk
dA == ( :~ d g x :~ d 11) · k I I 11 ~ - ai ai ai
Substitution of Eqs. (9) and (10) into Eq. (11) produces t t l the symbols a, b , and c denote
ar ar (19)
dA (ax~_ ax ay) dg d 11 a dg C = -d~
== (I - _! b = - d11 ai
ag a11 a11 ag a - ag a11
. . as the edges of the infinitesimal pai::al-
The expression in the parentheses of Eq. (12) may be written as a 2 • 78
lfo.isc vectors are shown m_ Fig. . be determined from the following
dctcrminant. That is, !11lcpiped of volume dV. This volume may

I;,,,,,s
Vl'Clor triple product:
dA = Y.s I dg d11 == IJI dg d11 (I I)
Y.,,,
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 4

( I •II

Two successive applications of Gaussian quadrature result in

n n
I == L L RjRd(g, TJd IJ(g, 1/k) I ( ISl
(a)
(b)

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 4
4

I ==
iz2 JY2Lx2 f(x, y,
' I YI Xi
z) dx dy dz (16)

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
(d)
must replace the infinitesimal volume dV == dx dy dz by an expression involving (c)
tlg, d'f/, and d{ Toward that end, Fig. 7.8 shows an infinitesimal volume dV in . -l·(b) n = 2;(c)n =3;
Figure 7 .7 Integration points for quadnlateral: (a) n - '
the natural coordinates. Also depicted is a vector r, which locates a generic point
(d) n = 4 (each way).
ln the space. Thus,
 lw11 n111f 1111 1111 Olt1111111h1111,1 Co11t l111111
111 I II ltmp11111111111tln Ou11drll11t11111I• Im 1'1 11 1111 Strose end Pl ane Strain 333
V

1, ISOPARAMETRIC QUADRILATERALS FOR PLANE STRESS

II )( b 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-
111111Ik as well as displacement compatibility conditions. If the geometric inter-
p11l11tion 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 iso-

p11111mctric 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 quad-
1 //,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 coordi-
111tll'S gand 7J are orthogonal throughout the element. For either figure the generic
- j
I /z
_ _ _ _ _ _ _J.,/
I / dl~placements at a typical point are
u = {u, v} (1)
X

z N,,dal displacements indicated in Fig. 7.9 consist of x and y translations at each

11rn lc, as follows :
Figure 7.8 Infinitesimal volume in natural coordinates. (2)
dV = (a x b) . c = / J / d( d7J d( 11,11· both elements we assume the displacement shape functions
ax ay az 4 4
ag ag ag u = ""ifiu; V = Lj;V; (3)
= ax ay az i= l i=I

aTJ aTJ aTJ d( dTJ dt <1111 111 these expressions the functionsJ;,!2 , Jj , andf4 are the same as those in the
ax ay az v.1•0111etric interpolation formulas given by Eqs. (7.3-3). Therefore, the Q4
ac ac ac 1 lt•ment is isoparametric, and previous statements regarding differentiation and

l11h•µration of functions hold true. Equations (3) may also be written in the matrix
in which J is the 3 x 3 Jacobian matrix d /J / . . . I, 11111
revised form of the integral in Eq. (16) 'b:~omes IS Its determmant. Hence, lhl1
U; = f;q; (i = l,2,3,4) (4)

I= fl fl f1!((,TJ,t) /JI d(dTJd( (.! I)

111 which

Three successive applications of Gaussian quadrature yield

n n n
f.r 1 OJ
= [O 1 Jri (5)
1
= .Z: .Z: .Z: RjRkRtf(Sj, T/k, ?,) 1J(Q, T/k , ?,) 1
l = J k= I j= I (.' 'I I h1• !'Cncric displacements u1 in Eq. (4) represent translations at any point due
Integration points for n = 1, 2, 3, and 4 each way number I 8 27 IcI h1lht• displacements q, al node i. As n furth t:r effi ciency of notation, we can write
respectively. , , , unc ,. , 1111• l'unction .fi us

./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
Vnlucs of g1 and 'Y/, for this element are listed in Table 7.2

T~ --+---:_-u~J-,.
4
Tdt
TABLE 7 .2 Nodal Coordinates for Element Q4

g,
I

-I
2
1
3
1 -1
4

2b ----- 1/i -1 -1 I I
91

L,1~ ~tq-,
Similarly , the strain-displacement relationships for element Q4 may be
11111cisely expressed as
----Lli_____:J2tq~ q3 ~; = B;q1 (i = 1, 2, 3, 4) (8)
whore
i-+--~~~~~2a~~~~~-/

(a)
B1 = d f1 = 0
a
ax
0
a
ay
f1 =
[f;, /,]
0 (9a)
TJ f;,y fi .x
a a
ay ax
lfrfcrring to Eqs. (7.3-17) and (7 .3-18) , we see that the submatrix B1 can also
h written
q,_
B1 =
Dc11
0
0
Dc21
J (9b)
[
DG21 Dc11
Next, we express the stiffness matrix for the Q4 element (with constant
-+ lhlckness h) in Cartesian coordinates as
+
G1
G2
K = h L BT(x,y) E B(x,y) dx dy (10a)
y
Ilowcver, in natural coordinates this formula becomes

L. (b )
1111d with two applications of Gaussian quadrature we have
(10b)

l•'lgurc 7.9
Element Q4: (a) rectangular parent; (b) isoparamctric countc1p111t.
n II

K '= h L L R1RkBJ.kE Bj.k IJj.k I (10c)

k I J I
 338
l•oi, 11111111111,t, 1111nd1llnl111nl• 1111 1'1111111 t.11 111111 nrul l'ln1111 h1111l11 337
In this expression for nun11:r 1rnl 11111•pr11111111, llw 1u1ttrix n1,4 und the <ktl•1111111
lhrn•tw e, liqs . ( I Oh) 1111d (III,) 111111plll y to
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 (x, y) f(.x, y) d.x dy

K """ rr I 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]
i (15)

I h,· resulting 8 x 8 matrices are given in Ref. 2 and need not b~ repeated here.
or 11,u,lly, the equivalent nodal loads in Eq. (12b) also take the simpler form

M = ph L
II

k E I j=I
IJ

L R1RdJ.k ~ .k J Jj,k j Pb(t) = abh f,f, fT(g, 'Y])b(g, 'YJ, t) dg d'Y] (16)

I ,111nple 7.1 . .
in which p and h are assumed to be constant. 111' 11 vc numerically the consistent mass term M24 for the isoparametric Q4 elemen~1~
Fig·
In addition, equivalent nodal loads due to body forces may be statl·d 11 / 'l(h), using Gaussian integration with n = 2 each way. Assume that p an 6 aa;~
follows: , 1t1INlunts and that the coordinates of nodes 1, 2, 3 , and 4 are (3, I), (8, 2), (6, ),

Pb(t) =h L fT(x ,y)b(x,y, t) dxdy

1 ',~). respectively .
T
. . · ( G
I Eq ( 11 c) we must set up the network of four mtegrat1on pomts or auss
Y G3
o appG2,
l'•illlls) GI, · Iar, ~or n -- 2 we have
· , and' G4 indicated in Fig. 7.9(b). In part1cu
or 11, Rk = 1 (from Table 7 .1), so that
2 2 )
M24 = ph L L (fit.k(/it.k IJj.k I (a
k=l / • I

or I lw l'unctions/1 andfi are needed for evaluating the term M2 4 becau_se of the arrangement
111111ntrix f shown in Eq. (5). Substituting/1 andfi from Eq. (6) mto Eq. (a) produces
n n
Pb(t) = h L L R1RkfJ.kb(t)j,k IJJ,k J
M24 = f!._
h 2
L L [(1
2
- g2)(1 - 17)2] j,k IJj,k I (b)
k=I jE I
}6 k - 1 j=I
where both Pb and b are functions of time. 111 Implement this formula, we first calculate the Jacobian matrix using Eq, (7 .3-1 0) , as
Except in special cases, the integrals in Eqs. (IO), (11) , and (12) must h l11llows:
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 (1 - 11) (1 + 11) -(1 + 17)]
38 21]
be handled by explicit line integrations. Of course, if the body forces consist of -(l +t) (1 + t) (1 - ~) 6 6
point loads, no integration is required at all.
2 5
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 =4
l[ 9 - 17
- 3-~ ~] (c)

J= [ ~ ~] JJJ = ab ( 11)
1111'11 the determinant of J is
IJ I H39 I I; 417) (d)
in which a and bare half the width and height of the rectangle in Fig. 7. 9(11)
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
339
1
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• r1•Nt1II N yl11ldN

Mi,, I . I .I<) pit

which can be finalized using numerical values for p and h.
4 7 3
Now we shall consider a higher-order quadrilateral element that is b111.i 1l V

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 2b
t
•-u
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
y

LX ---
8 ----
gl
I
6

the nodal displacement vector is

I
I
q = {qi, qz, ... , q16} = {ui , V1, . . . , vs} (1 7)
I
which contains x and y translations at each node. We also assume the fo!Jowi11 j1 q1- 15 2
quadratic displacement shape functions: 2a
r
q2
8 (a)
8
u = 2, f;u; V = 2, f;v;
i=J
i= l
( I H)
where

.ft = i (1 + fo)(l + 'Y/o)( - I + (o + 'Y/o)

(i = 1, 2, 3, 4)
.ft = f (I - ( )(1 + 'Y/o)
2

f; = HI + (o)(l - r, 2)
(i = 5, 7) (, ,,,
(i = 6, 8)
'l'hll values
i11 Table of 9 and 'Y/; required in these formulas [see also Eqs. (7)) are glvrn
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 elc111c•111
hc•co,nc quadratic curves [7]. Thus, we locate any point on the quadrilatrn,I
(including point g) by the formulas
8
8
X = 2, .ftx; y = 2,f;y;
i= l (.!II)
1111el l'lllment Q8 is seen to be isoparametric . Formulations of stiffnesses, co1111i~ q 1 - . __ _ _ _ _ _ __
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~
t
q2
(Ii)
lh1• s111m· pattern as before, even though the lornl coordim1tcs arc curwd .
Fl11111·r 7, IO Elunu.inl Qll: (n) i 1:ct11111111l111 p111 \1111 : (h) iN01111111111tltric cou111crp11rt.
 J40
tw 111111 11111111 I lh1111111 l1111nl < 111111111111 111, / ,II 1'11111111111 l>YNAl'li 1111 1'1101111,trnH nnd 1'111110 S1111ln 341

'l'Alll ,1!. 7, I
- Nmlitl t 111111ll11nl1•• tor lr.11•111N1I QH TAIILE 7.5
-
Slrudur11I Dutu for Program DYNAPS

- (·, - I
2 ~

()
6 7
0
Type of Datu
No. of
Lines
Items on Data Lines

,,, - I - I - I 0 hlem identification l Descriptive title

1ctural parameters 1 NN , NE,NRN,IPS,NEN,E,PR,RHO,H

l'ln ne stress (strain) data

TABLE 7.4 Shape Functions and Derivatives for Element Q8 n) Nodal coordinates NN J, X(J), Y(J)
b) Element information• NE I, IN(I, 1), IN(I, 2), ... , lN(I, NEN)

- I

I (I
Ji fi.t Ji.,, c) Nodal restraints NRN J, NRL(2J-I), NRL(2J)

- 2
~( I - 71)(-{ - 71 - 1)/ 4
( I I- ~( I - 71)({ - 71 - 1)/4
(2{ + 71)(1 - 71)/ 4 ( I - m211 • Jlor sequences of node numbers, see Figs. 7.9(b) and 7. IO(b).

-3 (2{ - 71)(1 - 71)/4 (I+ m2r1 TABLE 7.6 Dynamic Load Data for Program DYNAPS
(I + m 1 + 71)({ + 71 - 1)/ 4 (2{ + 71)( 1 + 71)/ 4 o + m211 No. of
4 Type of Data Items on Data Lines
-5 ( I - ~( I + 71)( - { + 71 - 1)/4 (2{ - 71)(! + 71)/4 o - m271 Lines

(I - §2)(1 - TJ)/2 - g(I - 71) - (I - C1 Dynamic parameters 1 !SOLVE, NTS, DT, DAMPR
(l + ~( I - 71 2)/2
"
7 (I - f)(J + 71)/2
(I - 712)/2

-[(I + 71)
-( I I ' )11
(I - €2} /1.
Initial conditions
(a) Condition parameters
(b) Displacements
I
NNID
NNID, NNIV
J, D0(2J-1), D0(2J)
(c) Velocities NNIV J , V0(2J-l), V0(2J)
H ( I - ~( I - 71 2)/2 -(I - 71 2)/2 - (I - ~ )11
Applied actions
(a) Load parameters I NLN,NEL,NEV
(b) Nodal loads NLN J, AS(2J-l), AS(2J)
(c) Line loads• NEL J, K, BL!, BL2, BL3 , BL4
7 .6 PROGRAM DYNAPS FOR PLANE STRESS AND PLANE STRAIN
(d) Volume loads NEV I, BVI, BV2

Now we shall discuss Program DYNAPS for dynamic analysis of thin plnll·~ Ill ( lround accelerations
plunc stress or prismatic solids in plane strain. For this purpose it is assumed tha (a) Acceleration parameter l IGA
H ttivcn continuum has been discretized using either Q4 or Q8 elemc111,, a (b) Acceleration factors l GAX,GAY
dt•st·ribcd in Sec. 7.5. All such quadrilaterals in the analytical model huw th
Forcing function
1111111c thickness h, and the material is taken to be homogeneous and iso1111pl (a) Function parameter l NFO
Table 7.5 shows preparation of structural data for Program DYNAPS . I hi (b) Function ordinates NFO K, T(K), FO(K)
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 "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).
1•k1s . 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 quadril11lt•111I ,111d NEV in Table 7.6 must be explained. If NEL * 0, at least one line load
Wht•rc NEN = 4 for Q4 elements and NEN = 8 for Q8 elements. The otht•r 1w1 t lmt·c per unit length) exists on the edgejk of an element. The types of line loads
sl rurturul parameters included are Poisson's ratio PR and the thicknt·~, II 1111 quadrilateral elements are illustrated in Fig. 7 . 11. For the Q4 element a
hrr lhcrmore, clement information now consists of the element number I 111111 ll111•11rly varying line load is specified by the force intensities BLl through BL4
ll(l(k numbers one through four for element Q4 and one through cip.hl fur
t·lt•111t·nt Q8. 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,
Prcpuration of dynamic load data appenn, in Tublc 7.6, which is si 1111l111 111 ,h llrll'd by the force intensities BLI 1h1011gh BL6 (sec the footnote below the
Tuhk 4. 1 f'or plane trusses (sec Sec. 4. 10). l lowt·YC.1r, the loud p111·11111t·ll•1s NI I 111hll 1
).
 341
I W11 1111d lh rnn lll111111111lc11111I (;01111111111
111 I II 1•11, 11 111111 l>YNAI' 1111 l'ln11n 1,111111 nnd f'luno Strain 343

· st,·esst•s o., 0 1, ·111., 1111t I <r,, tl1e, last

Ihm• v11ry1ng , of which is nonzero for£the 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 y

L. b,,

la) 5
8
b,s
/2
4
t
·- 11
b,, 3
/1 i

L.
j 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 f«H l
13 V2 in the y direction.
In Program DYNAPS we take n = 2 each way Lo locale point s 1111 n,
numerical integration. At each of the four points lhe computer evul1111t t•s 1111
IC'lic1u·1• 7.12 l'lnm• Hlcnlc1 with ()H ,,1,111tc.llllN.
344
Two- and Three-Dimensional Co ntinua Chap. 7 Sec. 7.7 lsoparametric Hexahedra for General Solids 345

E = 207 X 106 kN/m 2 v = 0.30 p = 7.85 Mg/ m3 7.7 ISOPARAMETRIC HEXAHEDRA FOR GENERAL SOLIDS
R2 = 2Ri = 0.02 m h=Im bx= I kN/m
Figure 7 .14(a) illustrates the parent rectangular solid RS8 of the isoparametric
for which the material is steel and the units are SI.
hexahedron H8, shown in Fig. 7.14(b). In the former element, the natural
We ran this data with Program DYNAPS, using IPS = 1, NEN = 8, DAMPR _
coordinates g, 1/, and ( are orthogonal everywhere. For either element the
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 generic displacements at a typical point are
SY, an? SZ near node 29 are plotted in Fig. 7.13(b). Maximum values of the nodai u = {u, v, w} (1)
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 Nodal displacements consist of x , y, and z translations at each comer node,
the normal stresses near node 29 is less than a twentieth of SX.) y ~ which fill the vector:
DJ l q = {qi , qz, q3, ... ' q24} = {u1, Vi , W1, . . . ' Wg} (2)
E
,._I:; 0, 06 Trilinear displacement shape functions may be expressed as
8 8 8
I
lf)

w u = J;u; v = Ifiv; w = Ifiw; (3)

Cl i= I i=I i=l
~ 0.04
I-
where
<
(/)
f; = k(1 + to)(l + T/o)(l + (o) (4)
z
;'. 0. 02 nnd
I-
<
-'
(/)
T/o = TJ;TJ (5)
z
< 0. 00
o:: The formulas represented by Eq. (4) are the same as those given previously in
I- o.os o. l 0 0 . 15 0.20 llqs. (7 .3-21). Values oft;, 1J;, and (; required for Eqs. (5) appear in Table 7 .7.
(ms)
For the parent rectangular solid RS8, explicit integrations are feasible, and
(a) Hliffnesses for an orthotropic material were presented by Melosh [8]. Fur-
lhcrmore, consistent masses for this simplified element are easy to derive and
were given in Ref. 2.
~ l 20 SY Turning now to the more general isoparametric H8 element [9] , we take
a.."' lhc geometric interpolation functions to be those that were given earlier as Eqs.
:i::
~ l 00
( I .3-20). Because the geometric and displacement shape functions are the same,
(})
N lhc H8 element proves to be isoparametric. Equations (3) can also be stated as
w an lhc matrix expression:
Cl
0
z 60 z U; = f;Q; (i = 1, 2, ... , 8) (6)
er
< 111 which
~ 40

~
(/)

(/'/
w
20 f; = [~
0
~
0
~]/;
1
(7)
Ct".
·-
CJ) 0
Ai, lwl'ore, the generic displacements u1 in Eq. (6) denote translations at any
p111nt due lo the displ11cc mcnts q, at node i.
( lil l1111ddilio11 , Hlnlin displut.'l'llll'lll n•l11tionships 11111y be written efficiently as:
1"1111111• 7, 11 lfr~llllllNl N Il l pJ11111 Nflllill I ~11111pl!1 fll Nll1p 1111111
t I ll1«f1 (/ I , 2. • 'H) (8)
 I w11 n11d 1111 nu l>lmonsionai Continua
Sec. 7.7 lsoparametric Hexahedra for General So lids 347

TABLE 7.7 Nodal Coordinates for Element HS

,, g;
i 7/i C,
1 -1 -1 -1
2 1 -I -1
3 1 1 - I
4 -1 I -1
y 5 -I -I 1

)-, v,
where
6
7
8
1
1
-1
-I
I
I
1
1
1

.f fi .x 0 0 0 0
I• - • ti DG!i
/ 0 f;,y 0 0 D G2i 0
0 0 f;,z 0 0 D G3;
(a)
B; = d f; = (9)
f; ,y Ji,x 0 DG2; D Gli 0
0 J;,z f;,y 0 D G3; DG2i

f; ,z 0 f; ,x D G3; 0 DGli .

Terms appearing in the submatrix B; were discussed in Sec. 7. 3.

Next, we express the stiffness matrix for element H8 in Cartesian coordi-
IU\tes to be

K = fv BT(x,y,z) E B(x,y,z) dx dy dz (10a)

In natural coordinates, this equation becomes

I v, und three applications of Gaussian quadrature give

5
i r n 11 rt

• - • II K =II I R j RkR1B[k,1E Bj,k,l IJj,k,/ 1 (10c)

6 /
w,
l=l k= l j= l

' l'his formula for numerical integration implies that the matrix Bj.kJ and the
(b) determinant IJ j,k,1 I are evaluated at each integration point, where the coordinates
1•'1,curl! 7.14 lll'l' (~, 'Y/k, (1).
111111 .
Element 118: (11) parent rectangular sol id; (b) isoparamctric countllf Similarly, the consistent mass matrix for element H8 is

M p LfT(x,y,z) f(x, y,z) dx dy dz (1 la)

M ,,f1F, (r'u. ,,.l ll'(/ , 1/,tl l ,IU , r1,[)l,ll,/,,t1l (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 " ft

M =PI I I Rj RkR,ffk.l lj,(., j,JJ,k,/ 1

/= I k= I j = I (a)
in which p is assumed to be constant.
wlll'rc
Furthermore, equivalent nodal loads due to body forces have the f'cu 11
1
J; = kO - TJ)O - ()
Ph(t) = L fT (x , y , z) b(x, y,z,t) dx dy dz
( I 111) 1
,11hslilution of Eq. (b) into Eq. (a) produces
g)(l - (b)

or
M,, = a:~p LL L (1 - g)2 (1 - T/)2(1 - ()2 dg drJ d(

= Sabcp (c)
27
or
11 n n
1'1 occcding in a similar manner, we can find all of the terms in the first column of matrix
M, us follows:
Pb(t) =L L L RjRkR1ffk,1b(t)j,k,I IJj,k,d (I 1
1 J
l = I k= I j = I pV
where both Ph and bare functions of time . (M)col.1 = {8, 0, 0, 4, 0 , 0, 2, 0, 0, 4, 0, 0 , 4, 0, 0, 2, 0, 0 , 1, 0 , 0, 2, 0, O}
216
. F?r the rectangular solid RS8 , the Jacoblan matrix and its deterniirumt (d)
specialize to

J = r~ ~ iJ
0 0 C
JJj = abc

I(ere _the constants a, b, and c are half the dimensions in the g TJ , 1 ~

( Ill
111 which the volume V = 8abc.

Now let us examine a higher order hexahedral element that is formulated

11Ning quadratic geometric and displacement shape functions. The parent rectan-
»11lur 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
1111
10ns, as shown in Fig. 7. 14(a). Thus, Eqs. (10b), (lib), and (l,2h)
d_1rcct_
111 1111tural coordinates g, TJ , and ( are orthogonal, and nodes 9 through 20 are
simplified, as follows:
lol'nted at midlengths of the straight edges. Both elements have the nodal
1ll11placement vector
( II) Q = {q,, q2, q3, , , . , q60} = {u,, V1, Wt , . . . , W20} (17)
lllld
1h111 contains x, y, and z translations at each of the 20 nodes. Quadratic displace-
111t•nl shape functions for these elements are
20 20 20

Ullcl
u= I J;u; v = Ifi v; w=IJ;w; (18)
i= l i=l i= I

P,,(t)
11:x111u11fo 7.3
= abc r, r. f I fT(g, TJ, ()b(t, TJ, ( , t) dt dTJ d( ( lr1)
wh~·rc

/, H I + so)O + 7Jo)(l + (o)(to + T/O + (o - 2) (i = 1, 2, ... , 8)

I, ,t( I - g2)( I + TJo)(l + (o) (i = 9, 11, 17, 19)
Ass1 u11i11g .1hu1 P _is constant, let us derive the terms in the first column of the l'OIININII I, Ht TJ 2)( 1 + (o)( I + so)
r1111Ns mutn x M for the rectangular solid clement RS8 shown in Fig 7 J '1(u) JI II 111 (i = 10, 12, 18, 20)
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, '
111
· 1 ,.,, l/1, _11111 L•
i" /1 .t ( 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 Solids 351

1'AU1.11; 7 .H Nodnl Coordinates for Element H20

I t, T/1 {, i t, TJ; t,
I - I -1 - 1 11 0 1 -1
2 1 - 1 - 1 12 - 1 0 -1
V 10 -1 -1 -1 0
1 1 1 13

I
)-, 20
-----, 2
v,
4
5
6
7
8
-1
-1
1
1
- 1
-1

1
1

- 1
1
-1
1
1
1
1
14
15
16
17
18
1
1
- 1
0
1
- 1

- 1
1
1

0
0
0
0
1
1
-1 -1 19 1 1
. er
I
9 0
0 - 1 20 -1
0
0 1
• II 10 1
5--------e----..JI
17 2a 6 /
w,
111 this instance, the faces and edges of the element are quadratic surfaces and
(al 1 111 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
3 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
d«•vclopment of element H20 are easily obtained and need not be tabulated. For
1•K11mple,
V
!1.{ = kO + 2g + T/ + ?)(1 - 71)(1 - ?)

l
)- X

5
2
v,
I >I' course, numerical integration is required for this element.

1,8 PROGRAM DYNASO FOR GENERAL SOLIDS

i
t
• -11
/ 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
(b) .tHN umed to be homogeneous and isotropic.
The manner of preparing structural data for Program DYNASO is shown
ll'll(Ut't' 7.15 Element H20: (a) parent rec tangular solid ; (b) isoparamclrk·
l'Ollll(l\lJ)Url .
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 hexa-
lu•dru are H8 elements; and if NEN = 20, they are H20 elements . Therefore,
Y11 '.111·s ,0 1' §,, 'Y/1 , un~I (, for these formulas arc listed in Table 7.8. Ex 1,111 11
r ll'lllent information consists of the element number I and node numbers 1
111111,1111,i1011s a.re poss ible for the subparametric parent clement.
I 120 clement 1101 in p·,g · 7· I5(b) , we use gco1111•11 I1 011ough 8 for element H8 and 1 through 20 for element H20.
I I • hir I ·lhc 1sopurumctric
· . Table 7. IO gives preparation of dynamic load data for Program DYNASO.
111 ipo 1tl1<'.1~ lunc11ons th al arc the same as the displacement shape f'unctionN 111
1<c111 ( 11) ). llrns, l•11 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
.1(1 20 20 p111m11ch.:rs NEI., NEA , und NHV ll' q11111• Home explanation . If NEL :/= 0, at
•\'" ,/'I 11 Y ~ .l,v, z }: .fi z1 ( 1111 h•111111u1t• li111.• loud (lon'l' pl' I' uni I h-n~tlt) 1•111 ~tH011 lhc l'dgl' jk of' an c lement. For
I I I I I I
, 1«•1111•111 I IH II lt11l'1til y v11 1yi ,11• 11111• Ii 111d ii, d, f1111•d hy tlw 1'011·1• i111rnsitks Ill. I
 Jb2
TABLE 7. 9
Type of Data
Problem identification
Structural parameters
General solid data
(a) Nodal coordinates
(b) Element information•
(c) Nodal restraints
a .._
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
Dynamic parameters
Initial conditions
(a) Condition parameters
(b) Displacements
(c) Velocities
l\ppl icd actions
(11) Load parameters
(b) Nodal loads
(c) Line loads"
(d) Arca loadsb
(c) Volume loads
( lround accelerations
(11) Acceleration parameter
(h) Acceleration factors
Fmt'i ng function
(11) Function parameter
(h) Function ordinates
= define the geometry of this simple network, we need only state that the x-coordinates of
"'I t•11111nh only ro clc_mcnt HS (NEN = 8).
nodr numbe1h ond nine force intensities .
"Fm L' k1111.ml 1120 use eight node numbers and 24 force intensities.
Iwo 1111«1 I '111111 l>l11111111lcm11I c·0111'111111 '111c. / .8 1•,011111111 DYNASO for General Solids 353
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
Structurul D11tt1 for Program DYNASO
llrst foo tnote) . The firs t six force intensities have the same meanings as in Fig.
No. of I . I I (b) for element QS , and the last three are for the z direction.
Items on Data Lines
Lines 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
1 Descriptive title
I
loads on face jklm are defined by 12 numbers, of which the first four (BA l
NN,NE,NRN, NEN, E, PR, RHO
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
NN J, X(J), Y(J), Z(J) lour (BA9 through BA12) apply to the z direction. It is assumed that each
NE
I, IN(I, 1), IN(I, 2), ... , IN(I , NEN) component of area loading has a bilinear variation over the surface jklm. On the
NRN J, NRL(3J-2), NRL(3J-l ), NRL(3J) 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
No. of
Lines Items on Data Lines directions. Also note that GAZ is included in the list of acceleration factors , as
for any three-dimensional analysis.
I In a manner similar to Program DYNAPS, we use n = 2 in each of three
!SOLVE, NTS, DT, DAMPR
ways to locate points for numerical integration. Thus, there are eight such
l
points, at which the time-varying stresses CTx, CTy , CT,, 'Txy , Ty,, and Tzx are deter-
NNID, NNIV
NNID mined in Program DYNASO. Again, we had to use n = 3 each way for terms
J , D0(3J-2), D0(3J-l), D0(3J)
NNIV in the consistent mass matrix of element H20 to retain sufficient accuracy .
J, V0(3J-2), V0(3J-l ), V0(3J)
11:xample 7 .4
l NLN,NEL,NEA,NEV
NLN J, AS(3J-2), AS(3J- l), AS(3J)
'l'hc tapered cantilever beam in Fig. 7. l 7(a) is doubly symmetric, and the parabolas with
NEL 11pcxcs at the support determine the rate of taper. This beam is made of reinforced
J, K, BL!, BL2, ... , BL6
NEA l'Oncrcte and has a rectangular impulse of magnitude Pi and duration t 1 applied in they
J, K, L, M, BAJ , BA2, . . . , BAl 2
NEV I, BVJ, BV2, BV3 direction at its free end. Physical parameters are given as follows:
E = 3.6 x 103 k/ in. 2 11 = 0.15 p = 2.25 x 10- 7 k-s 2 / in.4
J IGA
l
L = 120 in. Pi = 400 k t1 = 5 ms
GAX , GAY , GAZ
where the units are seen to be US.
Figure 7 . I 7(b) shows discretization of a quarter of the beam into two H20 ele-
I NFO 111cnts, with nodal restraints imposed for symmetric and antisymmetric deformations. To
NFO K , T(K), FO(K)
=
For clement H20 (NEN = 20) wc need rhrl'l' 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

y
0.1L

z
)-,
0.3L

y
(a)
(a)

y
j2

Sym.
t
•· - i,
25 21 /
13 j3 I

29 4

12
24 /20
/ 32
Sym. Typ.

(ll)

(h ) Figure 7. 17 (a) Tapered c11111ilcvcr beam; (b) 1-120 clements.

li'l1111r1• 7. 16 t\rcu loud~ for hcxuhcdrn: {u) t•lcnll'nt IIH; (h) clcnw111 1120
 t wu 111d lli11m 1>11 1111 111 1011111 Co11t l111111 ~.Lt lnop111 1111111t1 ii I hHIIMIII N 1111 A,<11y11111111tllo tlolld•
Ill 367

,h pil'IH II qu111 l(:r ol l hl• 11•1•t1111Httl111 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.
."'Q.
10 0 \ 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 .

;! 50 '-
(.0
,0 ISOPARAMETRIC ELEMENTS FOR AXISYMMETRIC
~ 0 I
' ' I
SOLIDS
0 2 4 6
2'
8

<-so l-
\ 11 nx isymmetric solid is defined as a three-dimensional body that may be
o d1•vdoped by rotation of a planar section about an axis. This type of body is
<.
3 - 100 - _,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
(a )
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

C
h11pcs, but we will deal only with isoparametric quadrilateral sections. Al-
~ 0.10 lhough nodes are shown as dots on the cross section of a ring element, they are
(.0 m•1ually nodal circles. If the loads on an axisymmetric solid are also axisym-
111c1ric, we may analyze a representative cross section as if it were a two-
~ o.os
0
7 Z,V

'..t, 0, 00 I
2 6 8
7 I
Cl T s)
;- -o.os
<
-J
U)
~ -o. 10
,0-
I (b)
>-
r
1--- -
-. I , 00

IJ'l
(\J
o.so ---
~ o.oo I
0 2 4 8
7
T Cms)
10 1z
~ -o.so I
Iii
:;,:

~ - 1 .oo -
(c)

!figure 7.18 Tapered beam: (a) loud; (b) displucemcnt ; (c) stress.
 369
Sec. 7.9 lsoparametric Elements for Axisymmetric Solids
358 Two- and Three-Dimensional Continua
Figure 7 .20(a) shows the cross section of element AXQ4, which derives its
dimensional problem. At first, only axisymmetric patterns of loads will h~ characteristics from the quadrilateral element Q4 in Sec . 7. 5. Bilinear displace-
considered, but nonaxisymmetric loads also will be covered later in the section ment shape functions in matrix f are the same as those for element Q4, and the
. For any point on an axisymmetrically loaded ring element, the gencrk strain-displacement submatrix Bi becomes
displacements are
u={u,v} (I) .h,r 0

Tr~nslat~ons u an? v occur in the rand z directions, as indicated in Fig. 7.1'>

0 fi.,
(i = 1, 2,3 , 4) (7)
With ax1symmetnc loads, the translation w in the () direction is zero and tlu Bi= L 0
shearing strains 'Yr& and y,8 are also zero. However, the figure shows f;ur typl'~ r
of strains that are nonzero, as follows: fi., .h,r
(" ) which is obtained by using the operator d in Eq. (4) on submatrix ( from
Relationships between these strains and the generic displacements in Eq. ( 1) 111 r Eq. (7 .5-5). The radius r in Eq. (7) is found as
seen to be 4
_ au 21r(r + u) - 21rr u (8)
r = "'Zfir;
Er - - €9 = =-
ar 21rr r i= l
( I)
av au av In addition, the derivatives/;,r = Deli , and so on, are given by Eqs. (7.3-18) ,
E, = az 'Yrz = dZ + ar except that r and z replace the coordinates x and y .
The stiffness matrix for element AXQ4 may be formulated in natural
These relationships are embodied in the differential operator coordinates as
a
ar
0
a
K = ff 1
Jo27T BTE B \ J \r d(J dg d17

L, rl
1
0
az <•I) = 21T BTE B \ J \r dg d17
(9)
d=
1
0
r Similarly, the consistent-mass matrix is
a
az
a
ar
M = p rl L, L27T fTf \ J \r d8 dg d17

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
= 21Tp rl rl fTf \ J \r dg d17
(10)

/\lso, equivalent nodal loads due to body forces are

depicted in Fig. 7. 19 are
(J' = { CTr, er,, (T9, Tr, }

For an isotropic material , the stress-strain matrix is

( S)
Pb(t) = L, rl r1T fT b(t) \ J \r d8 dg d17

E = E 1J
1J

1J
Sym .
,,
( )
= 21T rl rl fTb(t) \J\rdgd17
(11)

(I + 11)(1 - 2v) II 1J 1J Numerical integration is required to evaluate Eqs. (9), (10), and (11).
I - 2 11
0 0 0 The cross section of element AXQ8 appears in Fig. 7.20(b) . Its properties
2 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 H1·c the sam e as for clement Q8.
'l'ht M•I X •I 111111 y i1; H01tll111 tu till' I X I 111utii x 1111 pl1111,• Nlnii11 In l\q ('/ ,). 11)
360 Two- and Three-Dimensional Continua Chap. 7 Sec. 7.9 lsoparametric Elements for Axisymmetric Solids 361

Expressions for B;, r , K, M, and ph(t) are similar to those for element AXQ4
z given by Eqs. (7) through (11), except that i = I , 2, ... , 8.
\
\ 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 2
M 35 = 21rp L L (ht.k(J3)j,k IJ,.k lrJ,k (a)
k=I j= I

In this case the bilinear functions h and h are needed to evaluate the term M35 because
• ---.... u .
of the arrangement of matrix fin Eq. (7. 5-5). Substituting Ji and.fi from Eq. (7 .5-6) into

ks
i '
Eq. (a) yields

(b)
r
I To apply this formula , we first calculate the Jacobian matrix from Eq. (7. 3-10), as
follows:
(a)

(1 - 'l'/) (1 + 'l'/) -(1 + .,.,)] 14

10 42~1
-(I + g) (1 + g) (1 - g) [ 15
z 11
' '\
I (c)
3

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
M35 = 21.061rp = 66. l5p (e)

I 5 2 •-u.'
i
which can be finalized using a numerical value for p.

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
k==ser----..1__0- - r 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
I tl/'composition [12] of the symmetric loads form harmonic terms produces
(b)
Ill Ill

Figure 7.20 Quadrilateral sections: (a) element AXQ4; (b) element AXQ8 . b,. = I b,j cos JO b, =I b,j cos JO
J• O J• O
(12)
Ill

ho L IJ111 sin JO
I o
 Sec. 7.9 lsoparametric Elements for Axisymmetric Solids 363
362 Two- and Th ree-Dimensional Cont inu a Chap. 7

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
e
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:
z
E = { Er, Ez, Ee , Yrz, Yzo, Y,o} (14)
Strain-displacement relationships developed by Love [13] are
au av u 1 aw
Er = ar E, = az Ee = -; + -; a8
au+av
--
av 1 aw
"1
rrz - az -
ar r ae 'Yo= - - + -
z az ( 15)

1 au aw w
-va=-- + - - -
rr ae ar r (bl

I lcrc we see that the radius r appears in the denominators of several expressions z
From Eqs. (15) we can form the operator d as
a 0 0
ar
0
a 0
az
0
1 a
r r ae
d= a a 0
( I hi
az ar
1 a a
0 Figure 7.21 Nonaxisymmetric loads: (a) b,1 cos 0; (b) b,1 cos O; (c) bo1 sin 0.
r ae az
1 a a Stress-strain relationships are easily extended to cover six types of stresses and
r ae 0
or ,. 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 slrcss vector for nonaxisymmctric loads musl contain r, 11 and 7', 11 , 11 '1 Tlw response of Hll nxisymmclric solid to II series of symmetric , harmonic
lollows: lo11ds rnusists of 11 1w1 il'S of symnwt, it·, h111111011k, pentric displ11ccmcnts that
er !11,, 11,, 11 11 , 11 , 1 ,,, 1111 ) ( 1/) 11111y Ill' t'X PH'HIH'd 111'
 364
lwo- 1111d l 1111111 l>l111nr11l111ml Lo11tl11u11
n1 / , 10 1'1011111111 l>YAXSO for Axleymmotrlc Solids 365
II/

=L
Ill

u Uj cos J(:) V =L VJ cos JO IINl11g llqs . (2 1) in Eq. (22) , we find that

j• O

Lf
j• O
Ill
M = k1rp fT f IJ I r dg dYJ (24)
w = L w1 sin}(:) I
j=O
1liiN l'ormula is the same for j = 1, 2, ... , m, and it does not change for
Again, i~ the l?a~s were antisymmetric with respect to the plane of syn11111·1, v ,1 111tsy111111etric displacements. Thus, to determine frequencies and mode shapes
~he funct10ns sm J() and cos JO would be interchanged. Applying the opcr111111 d 1111 nny value of j, we use the (variable) stiffness matrix K1 from Eq. (20) and
m Eq. (16) to Eqs. (18) expressed in terms off, we find a typical partilio11 uf 1111• (constant) mass matrix M from Eq. (24). However, in the latter equation note
the strain-displacement matrix to be
11!111 k = 2 for the case of axisymmetric vibrations, where j = 0.
Ji., cos}() 0 0 Equivalent nodal loads for each harmonic set of symmetric body forces
111k1• the form
0 Ji.z COS J() 0
Ji- cos}'()
r
0 }.Ji
- cos}'()
r
p,,(t)1 =f, f, f 77

fTcJc1b(t)j IJ I rdOdgdYJ
(B;)1 =
= f f
Ji., cos}() ( I!))
Ji., cos J() 0
hr fTb(t)1 !JI r dg dYJ (j = 0, 1, 2, .. . , m) (25)
0 .Ji Slllj
-1- . '() Ji.z sin j()
1 I

r whore
.Ji
-1-
r
. '()
Slllj 0 (ii fi) . '()
i,r - ; ~Ill}
b(t)j = {b,j, bzj, be1} (26)
Finally, the stresses for each harmonic response are
where i = 1, 2, ... , nen and j = 0, 1, 2, ... , m.
An element stiffness matrix for each harmonic set of symmetric displ111 (j = 0, 1, 2, . .. , m) (27)
ments may be written in natural coordinates as
t >I course, such stresses, as well as nodal displacements, must be added at the

~= f, {, f" BJE B1 JJI rd() dg dYJ

, 11tl of the analysis.

= k1r f, L BJE B1 IJ I r dgdYJ (j = 0, 1, 2, . .. , m)

( 111)

/ ,10 PROGRAM DVAXSO FOR AXISVMMETRIC SOLIDS

where k = 2 for j = 0, and k = 1 for j = 1, 2, . .. , m. The latter co1111111111 I , t us now consider Program DYAXSO for dynamic analysis of axisymmetric
(k = I) appears as a consequence of
~ollds with axisymmetric loads. We assume that such a solid has been discretized
2,r [21T 11110 ring elements AXQ4 or AXQ8, which were decribed in the preceding
1
0
2
cos J() d() = Jo sin 2 J() d(J = 1r ( 'II Nt•1·tion. As before, the material of the solid. is taken to be homogeneous and
IHOll'Opic.
Similarly, the consistent mass matrix for each harmonic set of symn11•1111 With very few modifications , Program DYNAPS in Sec. 7.6 can be con-
di splacements becomes vm 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
( "l 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
i11 which
, vnluatc not only BJ.k and IJ1. k I, but also ,i. k at each numerical integration point.
rco;/° 0 'll 111ilur comments also apply to the generation of consistent masses and equiv-
CJ
()
cos JO
() ,}J (j 0, I , 2, . . . , 111 ) (
1
11
11lt111t nodal lontls.
St rncturnl dnt n lor l'rngrum DYNAPS (s1·1• '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:1.1•tl is 1111 nxisynrnwtric solid .
 367
Sec. 7.10 Program DYAXSO for Axisymmetric Solids
366 Two- and Three-Dimensional Continua Chap. 1

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 z
and the acceleration factors GAX and GAY must be replaced by GAZ. I

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~ I
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 elemt•nl I I
AXQ8. \ ~Stem

Example 7.6
- iI L
l---~~~~ 6L ~~~~~~·1
An axisymmetric titanium valve head is discretized using seven AXQ8 elements, 11-
1 I 2' , I 3 _ _ _ _ _ _ _ _ _ _ _ _ -,
shown 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 I
edge of the finite-element network is divided into equal lengths between the nodes on thnt I
edge. For this problem the physical parameters are I
I
E = 1.7 x 104 k/ in. 2 v = 0.33 p = 4.20 x 10- 1 k-s2 / in.4 I
L = 0.25 in. (p, )max = 1.53 k/ in. 2 I
I
and the units are US. I
For this example we need to calculate equivalent nodal loads due to pressur,• I' I
on a surface of an axisymmetric solid element AXQ4 or AXQ8. Considering any of tl11 - I L
9
four surfaces of such an element, we can find the components of p, in the direction~ ol " I
r and z, as follows: "\ I Restra int
29
b, = p,e~· (Ill

ti
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 2L

~~2~
7
by first calculating a unit row vector in the direction of g' as
1 k::::'.'.~ -/~ / ~30~
\ __,____ r
e,, = -[r,f Z,f ] (!})
C

where
c = Y(r.d + (z.,,)2
4L I ///
1
~r::;i;es~:face

l~//
(d
Second, from known orthogonality relationships, we can write the unit normal vcctrn In
the form
L
eT), = - [ -z.,. r .f]
C

Then the equivalent nodal loads on surface nodes become Figure 7.22 Valve head with AXQ8 elements.

(ii

In this lormuln the dctrnnlnnnt of .J' is

 ,11111
IW11 ,11111 I limn Dl1tw1111lo1111I Cot1ll111111 369
1111111111111 ""

"
y
Pi, = 2np,.
f r1' [-z.f]
I

- I 1', f
r df (h)

w 11111N on these equivalent nodal loads are automatically determined by Eq. (h) for
~ I ,00 111, ~sure in the positive sense of the normal direction "f/ '.
V)
V) We processed the foregoing data with Program DYAXSO using NEN = 8,
w
0:: 11/\ MPR = 0.05 , and Subprogram NUMINT for responses. The computer plot in Fig.
a_
I 'l(a) gives the variation of internal pressure acting on the valve head, and Fig. 7 .23(b)
°<! 0, 50 i111ws time histories of the normal stresses SR, SZ, and ST at the integration point near
z 1111dl· 34 . Maximum values of these stresses are 40.04, 34.54, and 50.67 ksi at time
0::
w
.... 80 ms .
z
o.oo
so I 00 200
r (m s l REFERENCES
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,
sr Prentice-Hall, Englewood Cliffs, N.J., 1984.
::. so
"' Lckhnitskii, S. G., Theory of Elasticity of an Anisotropic Body, translation from
-"
Russian by P. Fern, Holden-Day, San Francisco, 1963.
.., 40
Gere, J. M., and Weaver, W., Jr., Matrix Algebra for Engineers, 2nd ed. , Brooks-
I")

w Cole, Monterey, Calif. , 1983.

C1 30
0 ~ Scarborough, J. B., Numerical Mathematical Analysis, 6th ed., Johns Hopkins
z
0::
Press, Baltimore, Md. , 1966.
i:'_j 20 ri . Zienkiewicz, 0. C., The Finite Element Method, 4th ed. , McGraw-Hill, Maiden-
z head, Berkshire, England, 1987.
V)
w I0
V)
Ergatoudis , B., Irons , B. M. , and Zienkiewicz, 0. C., "Curved Isoparametric
V)
w
'Quadrilateral' Elements for Finite Element Analysis," Int. J. Solids Struct., Vol.
0: 4, No. 1, 1968, pp. 31-42.
'
V)
0
200 H. Melosh, R. J., "Structural Analysis of Solids," ASCE J. Struct. Div., Vol. 89, No.
r (m s\
ST4, 1963, pp. 205-223.
( b) 11, Irans, B. M., "Engineering Applications of Numerical Integration in Stiffness Meth-
ods, " A/AA J., Vol. 4, No. 11 , 1966, pp. 2035-2037.
1r1gurc 7.23 Valve head with internal pressure: (a) load; (b) stresses.
I(), 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 .
r.e· ling., London) , 1968, pp. 37-50.
IJ'I = - z.e· If/ 11 Wi lson, E. L. , " Structural Analysis of Axisymmetric Solids," AIAA J., Vol. 3, No.
C C
12, 1965 , pp. 2269-2274.
I' Snkoln ikoff, I. S. , and Redheffer, R. M., Mathematics of Physics and Modern
'I'hls. ''.oh.ll'lninant tr~nsforms arc length instead of area because its second mw iN 11111 il'118i11eering, McGraw-Hill , New York, 1966.
1111111:1.u l lo have unrt length. Expanding the determinant produces
II Love, A. E. H., The Mathematical Theory of Elasticity, 4th ed., Cambridge Univer-
Nily Press , Cambridge, 1927.
IJ' I = :.[(r.d + (z.(') 2'1 = c 111

S11hNfil111i11g liqN. (a) , (d) , 1111d (g) into Eq. M , wt• find 111 111
 371
1111, 11.~ I h11111111t IOI l'lnt111 111 ll11mll1111

8 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 iso-
Plates p111·111netric hexahedron H20 to become a plate-bending quadrilateral called ele-
111,•111 PBQ8. While an H8 hexahedron could also be specialized, the resulting

and Shells 1111ight-sided quadrilateral would not be suitable for conversion to a shell ele-
11111nt later in the chapter.
Figure 8. l(a) shows the original H20 element, which has quadratic inter-
polation 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 t,
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

8 .1 INTRODUCTION

When a plate is subjected to forces applied in the direction normal to itN own
5
plunc, it bends and is said to be in a state of flexure. For this type of prohh 111
z
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 compl11·111 ~
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 flt•x11111I
X
)-,
1111d shearing stresses and strains, but also those associated with membrn11,· 111r
i11 plunc) deformations. 16
Finite clements for dynamic analyses of plates and shells will be h1111r1I 2

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] . 3
. Computer programs in this chapter perform dynamic analyses of pl111,•11 111
bending , general shells, and axisymmetric shells. All structures that we nn11ly1 la)
urc nssumed to be composed of linearly elastic materials with small struim, 111111
Figure 8.1 Spcd11H1111l1111 111 ht• K11hl·dron: (a) element H20; (b) rectangular parent
lllspluccmcnts . Guyan reduction (see Sec. 6. 7) is used in the plate and Nhrll
PQR8 of clllll1l'III l'll<)H 111•11111• l 1t11Hl11il11tH: (c) constrained nodal displacements.
programs to eliminalc the nodal rotations and rl.ltain the translations.

ll/0
 111 , u.7 I 11111111111 1111 1'1111011 111 llondlno 373

(l 3

.t .t j

) 'l
I ------ 2
/
h;
2

;t
3
l t1
~.u
lb)
k

21z 1
/ 2
/
i -3

Figure 8.1 (cont.)

le)

)-, h;
2

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
_l k k
------ 5 k

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
7/t 9
8
4/j
6
of constraints to be introduced are:
(b) (c)
la)
I . Nodes on the same normal to the middle surface have equal translation Ill
the [ direction. Figure 8.2 Nodal displacements: (a) comer of PQR8; (b) midedge of PQR8; (c)
node of PBQ8.
2. Normals to the middle surface remain straight (but no longer non11111)
during deformation. 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
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 hi
0 0
matrix: 2
0 0 0 hi
0 0
2
0 0 0
0 0
1 0 0 Cb,= (3)
hi
hi 0 0
0 0 2
2
hi
hi 0 0
Cai= 0 0 ( ) 2
2
0 0
1 0 0
hi wIuc· h 1s · 6 x 3 · If we were to apply each of these constraint· matrices
· of size 1 in
t
0 0 lour locations , we wou ld be ublc to reduce the number of n~dal d_1sp acemen s
2
I (4)(9) + (4)(6) 60 lo (8)(1) 24. Instead of following th1~ path , how-
h, iom we will pursue 11 11101t• diil'l'I lorn1ulat1
,•vcr, , ·on o f e 1ement PBQ8 m a manner
0 0
2
"imilur to that in Rd. I, . · · 1
() () Figure 8 . I show Nt•h•uu 111 l'IIIJH, ol t 011Nl11nt thickness h , with its neutra

J/'1
 l'l11t111 1111d Hh11ll1 1111. 11.2 I h11111111t Im l'lnl1111 In lh111cll11u 376

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

0 0 /!.2
-,-
)I, V
h
4 f; = 0 0 f; (i = 1, 2, . . . , 8) (6)
8 2
lJ
1 0 0

'- l'o isolate terms in submatrix f; that multiply (h/2, we let

'-
X, U

i t " 3
O O OJ 0 lJ f;
•-e . 6
fAi = 0 0 0 f; -1 0 (7)
/ y,
[1 0 0 0 0
0.,

l'hcn
Figure 8.3 Element PBQ8.
h
N111 fuce Iyin~ in the x-y plane. Its geometry is defined to be the same as thut I1,1 f.I = fA. + ~Y-2 fo·
I I
(8)
t'lll111cnt Q8 m Sec. 7.5. Thus,
8 8 nnd
X = 2_, /;X; y = 2., f;y; (z = 0) (7 .5- 20)
i= I i=I (9)
where
'l'he formula in Eq. (9) will prove useful for obtaining the consistent mass matrix
fi = W + fo)(l + 7Jo)(-1 + {o + 110) (i = 1, 2, 3, 4) nnd equivalent nodal loads for element PBQ8.
f; = f(l - {2)(1 + 110) (i = 5, 7) (7 .5- 1')) The 3 x 3 Jacobian matrix required for this element is
(i = 6, 8) Xg
Y,€ OJ
J = X ,7) Y.7) 0 (10)
Clcncric displacements at any point off the neutral surface are
[
0 0 z,,
u = {u, v, w} ! II
We assume that w, 0.,, and Oy vary quadratically over the element, so thul
where z,, = h/2 and
8 and so on.
u = z{)y = z L f;Oyi i= I
i= I
8 'l'he inverse of J becomes
V = -z()x = -z"' I'()xi
L, Ji {,x 71,x
i= I
J I = J* = {; ~y
(11)
8
[
w = 2.,fiw;
/• I
where (,z = 2/h.
wht•i:t· z. (~,(~· I1~ ll~ese expressions the displacement shape functions.ti IUl' 1h, We need cc1·tnl11 dl 1ivutlv1•N with respect to local coordinates, which are
N11111t us I hose 111 f!qs, (7 .5- 19) . Note thnl tht· rotations o, und ()1, urt· l'lmNrii mllccted into the l'ollnwlup I I 1111111 ix:
 J/6
I 1l111un ond Sholle 377
'luc. 8.2 Elomo11t l or Plates In Bonding

h
U,(: v,, w,, '2,fi,l)y; - (1/i,/Jxl f i.,w, As with the submatrix f 1, we can isolate terms in B1 that multiply (h/2, as
lollows:
8
h
u,,,, v ,11 w,r, =I ( 2fi.110y;
h
-( f; . 11 0x1 f;,r,W/ 0 0 0 0 0 a;
i=l 2
0 0 0 0 -b; 0
h h
u., v., w., 2f;Oy; - J;ex, 0 BAi = 0 0 0 Bs; = 0 -a; b; (17)
2
b; -f; 0 0 0 0
Tra~sformation of these derivatives to global coordinates is accomplished usin~
the 10verse of the Jacobian matrix, as follows: a; 0 f; 0 0 0
'l'hcn

[
: :: ::: : ::] = J-1[:.·: :.·: :::] BI = BA +
l
h
(-BB'
2 I (18)
U,z V,z W,z U,( V,( W,C
und
h h
( a10y1 -( a10x1 a;W; h
2 2 B = BA + '2BB (19)
8
=I ( 2h b10y; -( 2h b10x1 b;W; ( I 11 E4uation (19) will be convenient when finding the stiffness matrix for element
i= I
l'BQ8.
f; Oy1 -J; ()xi 0 Stresses corresponding to the strains in Eq. (15) are
i11 which
(20)
a;= Jf,f;., + lf2fi,11 b; = lf1fi., + lf2J;, 11 (l•ll
Then the stress-strain matrix for an isotropic material becomes
The five types of nonzero strains to be considered for element PBQ8 11111 1- JI Sym.
Ex U,x JI 1 - JI
Ey V,y I - 2v
E= E 0 0 (21)
E = 'Yxy = U,y + V,x ( I ~l (1 + v)(l - 211) 2
'Yyz V,z + W,y I - 2v
0 0 0
'Yzx W,x + U,z 2(1.2)
1 - 2v
lly inspection ~f the second version of this strain vector, we can assemble 11 11 0 0 0 0
/lh purl of matnx B from terms in Eq. (13) as 2(1.2)

h l'his matrix is similar to that in Eq. (3.2-10) , but the third row and column
0 0 (2a; (1•orresponding 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
h Nlrcsses produce too little strain energy [4].
0 - (-b · 0
2 I We may write the stiffness matrix for element PBQ8 as
B,=
0

b,
h
- (2a,

- J,
h
(2b;

0
(i = 1, 2, ... , 8)
K = J' 1 J' 1 r I
1
H 11: HI.JI di; d11 d(
1
(/1 () ,r, - J',J', /',(11, 1 c''.11,,) 1•:(n,, ,~ nu)I Jld{dTJdt (22)
 3/8 379
l1lnl111 1111d ShollN Ch•P 111 11.~I 1'11111111111l>YN/\1'111111 l'l11h1• 111 llm1cll11u

In this expression the malrices B11 and n11 . .1 hns 1111ll y 111 pllll'l' of 1, /2; and b, in Eq. (27) now has the
array contains onJ t h urc hoth 01 .~lzc 5 X 24, but tht• lull 111 this cxprcssion
Y erms t at are multiplied t rJ /2 1 . . 1111•11ning of forcc pcr unit 11n•11 . Note that this body force causes no equivalent
through the thickness of the element gives )Y .,, , . ntegrat10n of Ilq (
1111dnl moments.

K = f , f/2BJE B11 + !2
BJE Be)JJJ dg dTJ
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)
which must be evaluated nume . II I .
multiplied by h/2 from the thir~~;:~f n this process the factors~ and hi(<• II I •or best accuracy, these stresses should be calculated at the numerical integra-
hand h 3 /12. Thus th fi t f IJ_J [see.Eq . (10)], producmg the lm·1o
, e rs part o matnx K m Eq (23) · d 111111 points (5].
shearing deformations, whereas the second .. 1~ ue t~ transvrr
deformations. part is associated with Jlc1111,1I 1,1rnmple 8.1
1'111' the rectangular parent of element PBQ8 (after constraints are imposed), find the
The consistent mass matrix for element PBQ8 becomes 111111;istent mass terms M11 and M22 , In the first case, Eq. (25) specializes to

M =pf, f, f, fTfJJ I dg dTJ d( Mll = pabh LL /T d{ dr,

(a)

=Pf 1 f 1 f i(r + Ci r + cirs) I

In this equation the matrices f and f are b th
11

.
fs r( 11 JI dg d,,, d(
111hstituting the shape function/1 from Eqs. (7.5-19) into Eq. (a) yields

M1 , = pabh
16
f1 f
-I
i
- 1
(1 - t)2(1 - r,)2(-1 - g- r,)2 d{ dr, (b)
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 1'111forming the integrations indicated in Eq, (b) results in
M11 = ipabh
M - P _ _ fl JI ( 1
_
1
2fif11 h2 fJfe)
+6 JJ / dg dTJ which is simply a fraction of the total mass. Similarly, in the second case we have
(c)

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 M22 = fl fl /T
pab h3 _
12 1
_
1
d{ dr,
rotational (or rotary) inertias.
1
a wna mertias, and the second part givt1
1
Equivalent nodal loads due to bod fo = -pabh 3 (d)
using only matrix f as follow , y rces on element PBQ8 are calcu l11ll'd 96
A, S.
which has units of mass moment of inertia.
Pb(t) =f 1 f I f l fJb(t) I JI dg dTJ d(

= 2 f1 f rJb(t)JJI dgd,,,
11,3 PROGRAM DYNAPB FOR PLATES IN BENDING

Wc shall now describe a computer program named DYNAPB for the dynamic
in which 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
. b(t) = {O, 0, b,} ( 171 p111t11ining to element Q8 (see Sec. 7 .6), because the geometric and displacement
und b: is force per unit volume in the z dir . , 11h11pe functions are the same for both. However, in program DYNAPB the
lhc factor h/2 from IJI and rewn·t E (e2c6t)1~n . Alternatively' we may cxll m I
e q. m the form 11111trices K, M , and p,lt) must he handled according to the expressions devel-
11pt·d in Sec. 8.2.
P&(t) = {, {, fJb(t)/JI dgd,,, ( 'HI
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 1
l'htht• n11el Hhnll• 1111, 11.:1 1110111 11111 DYNAl'II 1111 l'lnt n1 111 lln111ll1111 381

TAUi.iC H. 1 NII 111'1 111111 1111111 for 1'1•0141·11111 l>YNAPII PIH, 7. I6(b) J. Finully, nolr th11l ll1l' llt'l'l' lrrution factor GAZ is for the z direction
1111ly,
Type of Data No. nl As in Program DYNAPS, we lake n = 2 each way to locate points for
LillCN Items on Datu Lines
1111111crical integration. With a plate-bending element, the computer evaluates the
Problem identification l Descriptive title rl111c-varying stresses a:,, O'y , Txy , ry,, and T,x at each of the four integration points.
Structural parameters 1 lll•cause the displacement shape functions are quadratic, we found it necessary
NN,NE , NRN,E,PR,RHO, H
Plate bending data 111 use n = 3 each way for terms in the consistent-mass matrix of element PBQ8.
(a) Nodal coordinates NN J, X(J), Y(J) 11.xumple 8.2
(b) Element information• NE I, IN(I, 1), IN(I, 2), .. . , IN(! , 8)
(c) Nodal restraints NRN fllgu rc 8 .4 shows half of a square, symmetric, simply supported plate that is divided into
J, NRL(3J-2), NRL(3J-l), NRL(3J)
a ,,lght PBQ8 elements. Also indicated in the figure is a moving load P that travels in the
For sequence of node numbers, see Fig. 8.3. 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].
!PS andd NfEN are omitted. In addition, there are three possible nodal restnunt
mstea o two per node. · This problem is analogous to Example 6.5 for a moving load on a simply-
Nllpported beam divided into four flexural elements. However, the displacement shape
Dynamic load data for Program DYNAPB is given in Table 8 2 1-1, .• l11nctions for the PBQ8 plate element are quadratic instead of cubic. Therefore, the
?ave three possible initial displacements, initial velocities and ~o-datt ~ Junctions given by Eqs. (7.5-19) must be used in Eq. (a) of Example 6.5, and y,(t)
mstedad of/wo per node (as in Table 7.6). Line loads act in th~ z direction u~:1,•,: I 11Jplaces x,(t) in both Eqs. (a) and (b). Of course , we need to extend Program DYNAPB
an e ge o an element and ma~ vary_ ~uadratically. Therefore, we require lh1 t1 to handle one or more moving loads, as described previously for Program DYNACB.
;~de numbers and three for~e mtens1hes (per unit length) to describe them IN'1 Physical parameters in this example are
g ..7. l !(b)J. Also, ~uadratically varying area loads (force per unit area · rl E = 69 X 106 kN/ m 2 V = 0.33 p = 2.62 Mg/ m 3
111
z direction) need eight force intensities for a complete de~criptio~ N~
1 a = 0.1 m h = 0.025 m P = 20 kN

TABLE 8.2 Dynamic Load Data for Program DYNAPB lor which the material is aluminum and the units are SI. We ran this data on the extended

j3
Type of Data No. of
Lines Items on Data Lines
z t
I • - j2
Dynamic parameters
ISOLVE, NTS, DT, DAMPR /
Initial conditions j1
(a) Condition parameters I NNID, NNIV
(b) Displacements
(c) Velocities
NNID
NNIV
J, D0(3J-2), D0(3J-l), D0(3J) r------------7---v
J, V0(3J-2), V0(3J-1) , V0(3J) / p /
Applied actions Sym . I, // y(t)--4- Typ. //
(a) Load parameters
(b) Nodal loads
1
NLN
NLN,NEL,NEA
J, AS(3J-2) , AS(3J- l) , AS(3J)
_L~ I/ 8 13 I 21 j_ /
(c) Line loads NEL
(d) Area loads J , K, L, BL!, BL2, BL3
NEA I , BA! , BA2, ... , BAB
Ground accelerations
(a) Acceleration parameter
IGA
(b) Acceleration factor
GAZ
Forcing fun ction
Typ.
(a) Function parameter I NFO
_
(b) Function ordinates NFO
_ _ _ __ _ _ ___J__ _ __J__ K, T(K), FO(K)
-------=- 11'111111·11 tl.4 Simply-supported plate with moving load .
 382
Platos and Shells Chap. e Sec. 8.4 Element for General Shells 383
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 1 0 0 0 0
(VOP = 153.7 mis), the plot shows a maximum translation of 2.320 mm; and for 0 1 0 0 0
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
0 0 1 0 0
applied gradually at node 21) are 1.576 and l.382, respectively. As for the beam, lh1· hi
1 0 0 0
values of VOP and AOP used in this example both give travel times equal to llw 2
fundamental period of the plate, which is 5.206 ms.
h;
Ca;= 0 0 0 (1)
YOP CON Sr. 2
0 0 1 0 0
AOP CONST. h;
0 0 0
2,00 2
h;
0 1 0 0
2
E
E
0 0 0 0
~ I .SO
- When this 9 x 5 matrix is compared with Eq. (8.2-2), it is seen that columns
( \J
I and 2 have been added. Similarly, the constraint matrix Chi for a midedge node
w
Cl of the rectangular parent becomes
0
z
.._ I . 00 h;
< 1 0 0 0
2
-,
Cl h;
0 0 0
2
o.so
0 0 1 0 0
Cb;= h; (2)
0 0 0
2
o.oo h;
3 4 5 0 1 0 0
T <ms J 2

Figure 8.5 Translational responses at center of plate. 0 0 0 0

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
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

L !, IX;l
Pl3Q8 (sec Sec. 8.2) for plate bending. However, the constraint conditions 1111 3
h• [ l
Y1 + L fd-:j_
g 8 ;]
111odilied because two additional translations, u1 and v1, occur at each node of 1111• m3; (3)
Nlwll eh:menl. Thus, the constraint matrix C111 for a corner node of 1·h e rectn11g111i11 I I %1 I I n 3;
fllll'l'III clL·111cn1 Isc-e Fig . R. I (b) l has two more 1:olumns thun before, us f'olfowi,
'l'ht· i11lll1pol11lio11 '1111Plt111111 / 1 11p111•111i11g In Eq . (3) nrc given by Eqs. (7.5-19).
 3114 ' 110, 11.4 I ltt1111rnt lor lln11n111I hnll• 3811

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)

t1oints j and k in the figure are at the surfaces of the shell. In a computer program
z',{; 1•lther the coordinates of points j and k or the direction cosines for V3; must be

/,W 2
\i' ,,/
, ./
y
Hivcn as data.
Generic displacements at any point in the shell element are taken to be in
the directions of global axes. Thus,
/~v,
)- y,v
u = {u, v, w} (5)

(a) u, C<,~ 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,
,.,,, (6)
q; = {u;, v;, W;, a;, f3;} (i = 1, 2, . .. , 8)
z',~
<lcneric displacements in terms of nodal displacements are

UJ = I8f; [U;J + I8f; 'ih fJ.; [a·] (7)

[ V

W
t=l
V;

W;
, =I
I

/3,
v'
In this formula the symbol µ ; denotes the following matrix:

fJ.; =
-/2;
-m2; mli
Iii J (8)
[
-n2; nli

( '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 ,~,. 11•M111·1 tl Vl' ly Tlwlr vulucs arc
 386
Plates and Shells Sec. 8.4 Element for General Shells 387

h•
v' = -(~a; and so on.
Contributions of these terms to the generic displacements at any point are given The inverse of J becomes
by the second summation in Eq. (7).
As for the plate element, the displacement shape functions in Eq. (7) may t,x 'T/,x {x]
be cast into the matrix form
r 1
= J* = f y TJ.y
[ f , 'T/,z (.,
t.y (17)
h
1 0 0 -(-;j_l2; /!!z
2 11
We need certain derivatives of the generic displacements [see Eq. (7)] with
h; h; respect to local coordinates. These derivatives are listed in a column vector of
f; = 0 1 0 -(2 m2; ( m1; f; (i = 1, 2, ... , 8)
2 (12) nine terms, as follows:

0 0 1
h;
-(2 n2; (2h; nli u., fi.s 0 0 - C/;,gl2; Cfi.iu
U;
u,,, Ji.,, 0 0 - (f;,,,l2; CJ;,,,l1;
In order to isolate terms in submatrix f; multiplying (, we let u., 0 0 0 -f;l2; f;lu V;

0 fi., 0 -(f;,gm2; (f;,,mli

~ G! f f f}
V,g
8 W;
v,,, =I 0 fi .,, 0 -(f;J,f1 m2·I Cfi.11mu
h;
(18)
f,, (13a) v., i= l
0 0 0 -f;m2; fim1; -a·
2 I
11nd w., 0 0 Ji., -Cfi.,n2; (Ji.,n1;
w,11 0 0 Ji.1) -(fi.,,n2; (fi. ,,n1;

C,; ~G
0 0
0 0
-l2;
1, l h•
-m2; mli -;j_J;
J w.c O O O -}in2; Jinli
~{3;

~e:f
(13b)
Transformation of these derivatives to global coordinates requires that the in-
0 0 -n2; nu
Then
verse of the Jacobian mla:~x be aprp;~ed ore. :·' I
f; = fAi + (fs; (14)
und U, y O J* 0 ,'1 (19)
f = fA + (fs

The last of these formulas will later be used to derive the consistent mass matrix
( 15)
. . .
W ,z
0 0 J*
11 ...
W ,(

und equivalent nodal loads for element SHQ8. Multiplying the terms in this equation, we obtain
The 3 x 3 Jacobian matrix required in this formulation is U ,x a; 0 0 . -d;l2; d;lu
x,s Y.t z,s] U,y b; 0 0 -e;l2; e;l1;
J= x,,, Y.,, z,,, 0 0 -g;l2; g;lli
[x, Y.c z. 1
( I<, l U,z

V,x
C;

0 a; 0 -d;m2; d;m1;
U;

V;
1 8

We find the derivatives in matrix J from Eq. (3), as follows: V,y =I

l• l
0 b; 0 - e;m2; e;m1; W; (20)
v .• 0 c, 0 -g;m2; g;mt; a;
8 8 h,
x., = I I I
fi.,x, + I !i.d2'"
I I
W,1 () 0 (11 - d1n21 d1n1; {3;
IV,1, () () /11 ('1 1121 <11n11
\,,, ~
H
Ji,,, 1, ~'
... -4
,,,
I, ,, (. J I ,, II', , () () t'1 R1 ll i1 {<11111
I I I I
111 whldl
 388
Plates and Shells Sec. 8.4 Element for General Shells 389

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).
(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' = T.-E (7.2-27)
However, the third row of matrix T, must be deleted, because E,· is not to be
For element SHQ8 we consider six types of nonzero strains, as follows:
included in vector E'. For the purpose of evaluating T, at an integration point,
Ex
we need the direction cosines for vectors Vi, V2 , and V:i at the point. This may
U,x
be done with the following sequence of calculations:
€y V,y
€, w,, (28)
E = = (22)
'Yxy U,y + V,x In these expressions the vector (J 1)norm. denotes the first row of the Jacobian
'Yyz V,z + W ,y matrix normalized to unit length, and so on.
'Yzx W ,x + U,z
When calculating stresses in local directions, it is also useful to have
(29)
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 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
a, 0 0 -d,121 d;lli using matrix B', as follows:
0 b, 0 -e1m 2;
B,=
0 0 c, -g,n21
e1mli
g,nli
(2))
K = L, f f I I (B')TE B' df dY] d(
b, a, 0 -e,L2, - d,m2; e,lli + d,m1,
0 c, b, - g1m2, - e1n21 g,m1; + e,nli = f f f,I
I
(B~ + tBs?E(B~ + tBe) 1JI df dY] dt C30)
C; 0 a, -d,n2, - g,12; d,n1, + g,Li,
I Iere the matrices B ~ and Be are both of size 5 x 40, but the latter array contains
(i = l, 2, . '8) only terms that are to be multiplied by (. Integration of Eq. (30) through the
.Similar to the plate element, we can isolate terms in submatrix B, that multiply rhickness* of the element leads to
( lo find
K={, f 1
[2(B~?E B~ + HBe?E BeJ IJI df dY] (31 )
B; = BA; + (Ba; (24)
The remaining integrals in Eq . (31) must be evaluated numerically, using two
Submatriccs B111 and Bn; are composed from Eqs. (21) and (23), but the actuul integration points in each of the gand YJ directions [5] . In this process the factors
dc111ils are omitted. Altogether, we have
) 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
(2~)
purls of matrix K.
which will be useful when determining the stiffness matrix for the shell ch.lllll' III The consistent mass matrix for element SHQ8 is
The following nonzero stresses in the directions of primed axes will ho
l'ons idcrcd:

1111d llw r01 1·csponding slruins nrc (32)

lt1,•, t\•, Y,•,·, Y, , Yl','I • 111-11111•ll l y 1111q1111ll11111l11111111h 1hr lllll ~111 "'· tr1111~ 11111111111, J 1111111111111111, 111 111·11ln 1rd
 JOO
I lt,11 a 1111d Sholl H •1t11i. U.b 1'1 011111111 DYNASH f<>t Clo111111tl llhollu 391

Rcc:ull lhut the matrices f11 and fJJ arc bolh of r,1111 I •10, but the scc:on<.l hus 11111 TABLE 8.3 Strncturul l)utu for Program DYNASH
lcnns to be multiplied by (. Integrating Eq . ( ll) th rnugh the thickness y1t•ld
No. of
M =pfIfI[2flf11 + HEruJ /JI dt d71 ( I ~)
T ype of Data
Lines
Items on Data Lines

1iroblcm identification I Descriptive title

Because the factors 2 and j effectively become h; and hr /12, the first and sc1:111ul Structural parameters I NN,NE,NRN,E,PR, RHO,H
parts of matrix M contain translational and rotational inertias, respectively
General shell data
Equivalent nodal loads due to body forces on element SHQ8 may be fc,111111 (a) Nodal coordinates; h1 NN J, X(J), Y(J), Z(l) , H(J)
using only matrix f 11 , as follows: (b) Nonna] vectors NN J, V3X(1) , V3Y(J), V3Z(J)
(c) Element infonnation 3 NE I, IN(I , I), IN(l, 2), ... , IN(J, 8)
Pb(t) = fff 1 I I
rib (t) /J / dt d71 d( (d) Nodal restraints NRN J, NRL(SJ-4) , NRL(5J-3), ... , NRL(5J)

"For sequence of node numbers, see Fig. 8.6(a).

= 2 ff 1 1
rib(t) /JI dtd71 ( II) 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).
In this expression, the load vector b(t) is assumed to contain components of 101, Table 8.4 lists the dynamic load data required for Program DYNASH,
(per unit volume) that are uniform through the thickness of the shell. ThuN, which is similar to that in Table 7 .10 for Program DYNASO. However, now we
b(t) = {bx, by, b,} ( I'\ )
h11ve five possible initial displacements, initial velocities, and nodal loads in-
M lcad of three per node. As for element H20, we may have a quadratically
As for the plate element, these body forces do not cause any equiyalent 111ul11I vnrying line load on an edge of the shell element, with components in _the x, y,
111omcnts.
11nd z directions. For this purpose, we need three node numbers and nme force
After the time-varying nodal displacements in the vector q(t) have lwrn lnlensities (per unit length). Also, the area loads require 24 force intensities (per
obtuincd, accompanying stresses in the element may be calculated for Im nl
(primed) directions . That is, TABLE 8.4 Dynamic Load Data for Program DYNASH
0'
1
(t) =E T,B q(t) = E B'q(t) ( ,,,, Type of Data
No. of
Items on Data Lines
~uch stresses should be determined at the sampling points for numerical intc~111 Lines
11011.
Dynamic parameters 1 !SOLVE, NTS, DT, DAMPR

Initial conditions
(a) Condition parameters 1 NNID, NNIV
8.6 PROGRAM DVNASH FOR GENERAL SHELLS
(b) Displacements NNID 1, D0(51-4) , D0(5J-3) , .. . , D0(51)
(c) Velocities NNIV 1, V0(5J-4) , VO(SJ-3) , ... , V0(5J)
111 Sec . 8. 4 we developed element SHQ8 for the dynamic analysis of ge1w111I
Applied actions
shells. Now we present a computer program called DYNASH, which is b11N1 ii (a) Load parameters I NLN,NEL, NEA,NEV
011 lhu~ clement. The easiest way to compose this program is to modify sta111 (b) Nodal loads NLN J, AS(SJ-4), AS(51-3), ... , AS(5J)
IIH.·nts in Program DYNASO relating to element H20 (see Sec. 7.8) , from whh h (c) Line loads NEL J, K, L, BL1 , BL2, .. . , BL9
ulcment SHQ8 is derived. But in Program DYNASH, we must use the forn111l11- (d) Area loads NEA I , BAI , BA2, ... , BA24
given in Sec. 8.4 to construct the matrices K , M , and Pb(t). (e) Volume loads NEV I, BVI , BV2, BV3
Preparation of structural data for Program DYNASH appears in Tabfo H I Ground accelerations
C'ompuring this data with that in Table 7. 9, we observe thatNEN is omiUcd 1111111 (a) Acceleration parameter I IGA
(b) Acceleration factors 1 GAX, GAY ,GAZ
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 Forcing function
Is included with the nodal coordinates for 11 case where the thickness vuries ov, 1 (u) Jiu netIon p111111111•t111 I NFO
1 1
llll l lt•111l nl. The lint' of normal vectors in tlw f11hll t'onf nins u node 1111111lwr .I 1111d
1
(b) F11m·tl1111 rn.11111111•• Nl10 K , T(K), FO(K)
'= ~
 393
Sec. 8.5 Program DYNASH for General Shells
392 fllutoe end Shell s

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 .
Example 8.3
A quarter of a doubly symmetric cylindrical roof shell appears in Fig. 8. 7. This port11111
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 . Fig-
ure 8 .8(a) shows a computer plot of the ground acceleration, and the res_ultmg trans-
lntions 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
~
'"·
C
50

of the shell is divided into four SHQ8 elements of constant thickness. Note that the shl•II
0 1000 1200
is symmetric with respect to the x-z and y-z planes. Consequently, nodal restraints 11111•1 T (ms)
(.)
be used to prevent translations across those planes and rotations in the planes, 11• <
-50
indicated at nodes 4 and 14. On the other hand, simple supports at the ends of the ~hl'II
0::
prevent translations in the x and z directions, as at node 20. A rigid-body grouml c..,
acceleration D8 J(t) occurs in the z direction, and we wish to find the response ol tht ,!_, -100

structure due to this influence.

(a)
For this problem the physical parameters are
E = 3.6 x 103 k/in. 2 v = 0.15 p = 2.25 x 10- 1 k-s 2/ in. 4 OJ3

L = 100 in. R = 3L h = 3 in. <I>= 40°

C

(D8 J )max = 115.9 in./s 2 :::; 1 .oo

If)
z
w
Cl
0 o.oo
z

/
Sym . 1.5L~::.L/ ,-
<
<fl - 1 . 00

1 T 6 s/ 14u 17/
t----=
z
<
0::
I-

( b)
Sym. sx .
G) I
I _:; 0. 60
I - 0-40

IVp /I ---71 w
Cl O. 20
// I // I 0
z
/ 20 / I o.oo
0
:/ 0 I
l ____,s_ _ _2_ 1, /
... ._./_ _s____
r.k" Typ.
I
I
0::
<
W -0.20
z
~ -0.40
(/)

~
I / t ', .~ o·
0,60 -

', I 1 R'--('''' I ....

(/) (r l
',
'-J
I '
~
I
1,·111111,1, H.H ( 'ylli111t l111I lll'll 111111 (11) p11111111l 11n rk1111h111, (h) dlNpl11n•1m•111s, (,•)
tk,111111 ~, 111~1 ~
 394
Plates and Shells Chap. 8 Sec. 8 _6 Element for Axisymmetric Shells 395

the upper surface) for the integration point near node 1. Their minimum values are SX' ::. z
- 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.

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
i • ---+- U ;
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. I 2
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 ~l~______i___e - - r

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' (a)
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 v', 'Tl
displacements indicated at point i in that figure are '

(i = 1, 2, 3)
(I)
where a a small positive rotation about an axis normal to the t-11 plane. Figurl'
1 is

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:

I. Nodes on the same normal to the middle surface have equal translations in
the 1J direction. x', s
(b) (c)
2. Normals to the middle surface remain straight (but no longer normul)
during deformation.
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 1
With these criteria we can relate the six nodal translations in Fig. 8. IO(a) to 1hr displacements.
lhrce nodal displacements in Fig. 8. lO(c) by the following 6 x 3 constraint
mutri x: 8 imilarly, the four nodal translations in Fig. 8: IO(b) a~e related to the three nodal
1 0 displacements in Fig. 8. IO(c) by the constramt matnx
0
0 0 h;
0
h1 2
0
2 0 0
C,,1 = ( ))
h;
(3)
0 0
0
2

()
0
"'
)

()
() ()

which is ol si:11· •I \, 11 wt· w1•11• 111 11pply Pq , ( >) 111 tht• 1•11d~ 1111<1 I\(( , (1) tit
 .100
l'h,11111 n11d Uhnlln
• 1111, 11.e l 1011111111 for Axl11y1tm1011 lo Shull• 397
4 ')

i
1

3-r i - , 1, V
I
------ ---......
""-
,\

,1~1
h, r,
-
2 ------

. 1 ..
I
y'
• I

3
/
h,

.,,~ '
2

k
_J_ 5 k -3
z' x 1

k
t
6 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 S1•1 •
8.2 and 8 .4 for plates and general shell elements. h;
T/ -c, .
y',ri,v'
2 ,
Figure 8.11 shows element AXSH3, for which the coordinates of any po1111
may be stated as

[r]
z
= ± [r;] + ±
fiTJ~ [ l2;
2 m 2;
i= l
f;
zi i=I
J t I)
In this equation the direction cosines for the normal vector Vi are
h; .
l2; = cos 'Yi m 2; = sin Yi 2 sin Y;
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
g
!, = --(] - t) h = I -g2
2
z' x',1;,u'
'l'hercf'orc, the th ickness h may vary quadratically in the g direction.
Generic displacements at any point in the element are (b )

u = {u , v} (h I 11111111·•• R.11 (a) Element AXSH3; (b) nodal rotation.

11ssu111ing that the loads arc axisymmetric. These di1,placements cnn he exp1·1·ss1•cl
In tl•1111s of the nodal displaccmcnls 111, i,1, 1111<1 us fo llowi.;: ,r,,
 308
Plates and Shells Sec. 8.6 Element for Axisymmetric Shells 399

[ U]
V
= ±
i=I
j; [Ui]
V;
+ ±
i=l
fiYJ!!!.
2
r- sin ,Y;J
COS 'Yi
ai
(7)
These derivatives are transformed to global coordinates by the operation

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)

n
mlo the matnx format
h,

fi = [1 0 ""i
- YJ sin 'Yi] 0
-d,s'.ny,]
0 1
h-
YJi cos 'Y;
f; (i = 1, 2, 3) (H)
-± r::i=l 0
0
ai
- ei Sill
d; cos 'Yi
'Yi
Vi
ai
(15)

0 bi ei cos ,'i
To isolate terms in submatrix fi that multiply YJ, let us define
where
fAi = [oJ 01 00]/; fsi = [00 00 -sin 'Yi]!!!./;
cos 'Yi 2
(I})

Then (16)

( I ti)
und We consider four types of nonzero strains for element AXSH3. They are

( 11)
€, u.,
€, v,,
The _formula in_ Eq. (11) will be used later to determine the consistent nw11N
u1ulrix and equivalent nodal loads for element AXSH3. e= = u (17)
€9
The Jacobian matrix required for this element is r
'Yrz u,, + v,,
J = [;,·: :::] ( I lj 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 -di sin 'Yi
0 bi ei cos 'Yi
and so on.

The inverse of J becomes

Bi= I! 0 - drf; T/hi. sin y; (18)
r
b; ai d; cos 'Yi - ei sin 'Yi
1- 1 = J* = (, YJ.,]
[ g_, YJ. ,
( I I) (i = 1, 2, 3)
We w_ill need derivatives of the generic displacements with respecl to 1h Sl111ilar to the general shell element, we isolate terms in submatrix Bi multiplying
lorn! coordmates, as follows: 1/, which gives
ut !,,, 0 -Ji.1;'YJ sin y, B; = BAI + 'l7B0; (19)
u,,, 3 0 0 -Ji sin y1 llor the whole matrix, we have
L ( I I)

r "·'
11, ,,
I I [ 0
()
Ji.,'Y] cos y,
.fi CON ,'1
1lw 111st 10111111111
n nA I r,n,,
wlll lu•lp IIN to d1•11v1· tht· htlll11nHM 11111td x 101 1•1l 111t•11t
(20)
AXSI I I,
 400
1•1111111, " '"' m1011,. ,11 11.0 I 101110111 lur A><l•y11111111tr lu 8hulle
401
Stresses in till· d itl'l'flo11 111 ,,I p1 l1111•d 11x1 i, (rnused by ux isy111111etrlc loud'9 1
11 1111 ly te rms lo be mullipli(.)d by TJ. Integration of Eq. (26) through the thickness*
er' Iti I.
I r, I' T, ')•'} 111 1hr clement g ives
and the corresponding strains fll'l'

e' = {e.,., e,, , 'Y.,•y,} r

K = 21r f pcB,:.?E B,:. + i(Bli?E Be] /J /r dg (27)

The stress <ry· and th~ strain. Ey• normal to the middle surface have been 1111 I 1lw remaining integral must be evaluated numerically, using two integration
Thus, the stress-strain matrix for an isotropic material has the form
0111 p11l11ts in the gdirection [5].
The consiste nt mass matrix for element AXSH3 has the form
II II 0
E=
(1 +
E
v)(l - 2v) II V 0
M = p _ r r frr
1
_,
0
fTf/J/rdOd{d77

0 I - 2v
0 = 21rp {, { / fA + T]fa?( fA + T]fa) /J /rd( d77 (28)
2(1 .2)
111 thi s case the matrices fA and f 8 are both of size 2 x 9 , but the second has only
In or~e~ to relate local strains in the vector e' to global strains in the v1·, 1t h I ms that are to be multiplied by 71. Integrating Eq. (28) through the thickness
E, we specialize the 6 x 6 strain transformation matrix T. in Eq. (7.2- 28) 1 t 1, Nlllts in
11

T =
[
IT
0 0
0
I
(4
M = 21rp fyn fA + HHa)/ J / rd( (29)
e 21, 12 2m1m2 0 .\N before, the first and second parts of M consist of translational and rotational
l1111rtias .
which is of size 3 X 4 . Then the strain relationship becomes
We find equivalent nodal loads caused by body forces using only matrix
11 Thus,
E' = T€e (7. 2- J I l
'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
done as follows: 11111
yI Pb(t) = f 1
{, frr fJb(t) / J / r dO d{ d71

e, = (J,)norm = {l1, m,} e2 = {l2, m2} = {- m1, / }

1
( ' ~) = 41r f I
fJb(t)/ J /rd{ (30)
As for the general shell element, it is also convenient to have
1111d the components of force per unit volume (constant through the thickness) are
B' = T.B (8.4 2HI b (t) = {b,, b,} (31)
fo_r fhe purpose of calculating stresses in local directions . In this case, mar, I, n I hcsc body forces do not produce any equivalent nodal moments.
wi ll h;~ve only three rows, due to the form of T. in Eq . (24).
After solving for time-varying nodal displacements in the vector q(t} , we
J he clement stiffness matrix may now be formulated using matrix D 1 • 'f'hu- 1 101 determine stresses for local directions in the element using

K= f f frr(B '?EB'/ J /
I I
r dO dg d71 u' (t) = E T. B q (t) = E B 'q (t) (8.4-36)
l'h1•sc 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:

111 this equation the matrices B,; a nd Bn arc both of size 3 x 9, buf flw 1 11 ,
11 1 1
fin ~ ~To Ni mpllfy lntl~f.1111lio11 thro1111h thl' lhil•k,wss, 11,1111s In r 1111(1 rn11trl x .I contuining 1/ urc
111•11h·l'l1·d ,
 402
Plates and Shells Chap, I Sec. 8.6 Element for Axisymmetric Shells 403
m m
br = L brj COS j() hz = I b,j cos J() and
j=O j=O o o o -sin y,.
m
(7.9-12) = 0 0 0 cosy,.
bo = L b 8j sin J()
fBi
[0 0 0 0
(35b)
j=O
If the loa_ds were anti.symmetric with respect to the r-z plane, the functions sin jtl Thus, Eqs. (10) and (11) apply equally well to any case of nonaxisymmetric
un~ cos 18 woul~ be mterchanged. The response of an axisymmetric shell to lhi~ loads, but matrix f is now of size 3 x 15.
s~nes of hannomc loads consists of a series of hannonic generic displacemenlN For this analysis we must determine the derivatives of the generic displace-
l•or element AXSH3 these displacements are expressed as follows: ments in Eq . (32) with respect to local coordinates. They are

[u]V = t [cos J()

O COS
0
j() 0
0 ·J u,s
u,11
f;,{Cj
0
0
0
0
0
-f;,s17S;Cj
-.,t;s,.cj
0
0 U;
1 0
w - 0 0 sin }8 U,e -J.,t;sj 0 0 jf;77s;Sj 0 V;
(.l'I v ,s 0 f;,sCj 0 0
X
3
~.,t;
[u;]
V; +
3
Li77j
h[-sin Y;
cosy,. - v,11 =I
3
0 0 0
f;, s17C;Cj
j;C;Cj 0
W;
(36)
{ i= l
1- l W; j 1= 1 0 V,8 0 -J.,t;sj 0 -jj;77C;Sj 0 h;
w,s 0 0 f; ' sS·J 0
2a,.
wh.crc t~e angle /3,. is a small rotation about the x' axis [see Fig. 8. ll(a) 1. A- f;, g17Sj
bcforc'. 1f_the loads were antisymmetric with respect to the r-z plane, the furn w,11 0 0
lions sm 18 and cos JO would be interchanged. j 0 0
0 0 f;sj %/3;
W,e jf;cj 0 jf;77cj
Equation (32) may be stated more efficiently as
In the coefficient matrix of this expression, the following abbreviations are used
m m 3
u = L cl qj
j=O
= L cj L f;Qu
j=O i= I
(11, s,. = sin y,. sj = sin JO
(37)
in which
C; =COSY; Cj = COS j()
'l'he required Jacobian matrix is
cos J() 0
rg
Cj = 0 cos J() (7.9- 31)
[ J= r ,1) (38)
0 0 [ 0
Also, we put the displacement shape functions into the matrix form
In this case the inverse of J is seen to be
0 0
h;
-77 sin y,. 0
2
h; fr 17,·r OJ
f, = 0 0 772 cos y, 0 .,t; (i = 1, 2, 3)
r' = J* = [ to,z 77,z 0 (39)
( II) 0 1
0 0 0
h,
77- l )Hing this inverse matrix , we can transform the derivatives in Eq. (36) to global
2 ,•oordinates, as follows:
ANlwf'orc , lcl us is<>lale lcrms in submalrix f1 that arc multiplied by r,. I k 1111 U ,r u, ,

1~11 f/) ~ ; ~ ;:1./i

() () () ()
(l''\11)
u,,
II, 11

11', II /
fJ '
()

0
0
.I "'
0 ~1
., ...
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,, a;Cj0 0 - d;S; Cj
where i = 1, 2, 3, and} = 0, 1, 2, ... , m. As before, we can decompose (B; )j
0 Into
U, z b;Cj0 0 -e;S;Cj 0 (B;)i = (BA;)i + TJ(Bs;t (44)
U, IJ -JJ;sj 0 0 ! (Jf;'r]s;sA) 0 11ncl state the total matrix in the form
V,r 3 0 a;Cj 0 d;C;Cj 0
V,z =2 0 b;Cj 0 e;C;Cj 0
B1 = BAJ + 'r]Ba1 (45)
i= I
V,o 0 -JJ;si 0 -! (jJ;71c;sA) 0 which will be used to find stiffnesses.
W ,r 0 0 a;Sj 0 For the present analysis, nonzero stresses in the directions of primed axes
d;si lll'C
W ,z 0 0 b;Sj 0 e;s1
w , 11 i 0 0 jJ;ci 0 ! }firJcih; (46)
where the constants a;, b;, d;, and e; are given by Eqs. (16). lo which the following strains correspond:

For no?axisymmetric loads on element AXSH3, we consider six typt ~

1
uf E = {Ex•, Ez• , 'Yx' y' , 'Yy' z', 'Yz' .,,.} (47)
nonzero strains . Thus,
In this case the normal stress <Ty· and the strain Ey• are omitted. Also, the
Er U,r
HIrcss-strain matrix E for an isotropic material is similar to that for element
E, PBQ8 in Eq. (8.2-21), except that £ 33 instead of E 55 is divided by 1. 2. Addi-
V,z
llonally, the 6 x 6 strain transformation matrix T. serves to relate E to E , as in 1

1 l'.q. (7.2-27), except that the second row is deleted. Another useful matrix is
E9 -r (u + w 11 )
.
= (48)
Ej 'Yrz = U ,z + V,r (I )
which is of size 5 x 15.
'YzlJ W
,z
+ V,11 We write the stiffness matrix for each harmonic loading as
r
'YrlJ
j
-U,r IJ + w ,r
w
r
Ki= f f, f"
1
(BJ)TEBJ JJlrdOdgd'rJ

The strain-clisplac~ment relationships shown in the second form of this vccto = f br

1
[2(B,11)TE B,;i + HB~JTE B~J JJ Jr df (49)
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 where k = 2 for j = 0 and k = I for j = I, 2, ... , m. Next, the consistent
11 muss matrix becomes
a19 0 O
0
0 b; CJ Q
0
I I .
- J,c; 0 I I .
r ; rfic1 - 2rfi'rJS;cA
(H,)1 a:
2r }/, 'f'JC/11 (50)
b,c1 a1c1 0 (d;c1 - e1s1)c1 0
I 1 which is the same for j = 1, 2, . .. , m (see Sec. 7.9). Then equivalent nodal
0 - ; jfis; b,s1 - 2,)fi'rJC1s/11 e1.11 londs for each harmonic set of body forces are

J', J', f" f'XcJc1b(t)1 IJlrdOdgd 'rJ

I
- r J!,.11 0
("
1
-1). 1
1
I .
2r )/,rJ.1·,.11h1 (r1, Ji·q!,')
2,. .,,
p,,(t)1 =

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 407
in which only matrix fA is used. Components of force per unit volume (constant
lhrough the thickness) consist of TABLE 8.6 Dynamic Load Data for Program DYAXSH

b(t)j = {b,j , b,j, b0J (52) Type of Data

No. of
Items on Data Lines
Lines
Finally, the stresses for each harmonic response may be written as
Dynamic parameters 1 !SOLVE, NTS, DT, DAMPR
u'(t)j = E BJ q(t)j (j = 0, 1, 2, . . . , m) (5.l) Initial conditions
which are in the directions of local axes. (a) Condition parameters 1 NNID, NNIV
(b) Displacements NNID J, D0(3J-2), D0(3J-l), D0(3J)
(c) Velocities NNIV J, V0(3J-2), V0(3J-l), VO(JJ)
8. 7 PROGRAM DYAXSH FOR AXISYMMETRIC SHELLS Applied actions
(a) Load parameters 1 NLN,NEA,NEV
Let us now consider a computer program named DYAXSH for the dynanik (b) Nodal loads NLN J, AS(3J-2), AS(3J- l), AS(3J)
(c) Area loads NEA I, BAI, BA2, . . . , BA6
analysis of axisymmetric shells with axisymmetric loads. This program USl'~ (d) Volume loads NEV I, BVl, BV2
clement AXSH3 from the preceding section. It is generated by modifying thl•
Ground accelerations
part of Program DYAXSO dealing with element AXQ8 (see Sec. 7 .10), 011
(a) Acceleration parameter 1 IGA
which element AXSH3 is based. However, in Program DYAXSH we mw,t (b) Acceleration factor 1 GAZ
formulate K, M, and Ph (t) in accordance with the equations developed 111
SL'C. 8.6. Forcing function
(a) Function parameter I NFO
In Table 8.5 we see how structural data is to be prepared for Progru111 (b) Function ordinates NFO K, T(K), FO(K)
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' With element AXSH3 we use n = 2 in the f direction [see Fig. 8.1 l(a)]
thickness varies over an element. That line of data also contains the compone11t11 to locate points for numerical integration. At each of the~e two p.oint~ the
Y2R(J) and V2Z(J) of the normal vector Vi for node J [see Eq. (8.6-4)]. Finally, computer evaluates the time-varying stresses <rx·, <r,·, and Tx'y' m local direction~ .
three types of nodal restraints are indicated in line (c). Again, we had to use n = 3 when evaluating terms in the consistent mass matnx
for element AXSH3.
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 Example 8.4
instead of two per node. Area loads for element AXSH3 are similar to those Jor The vaporous gas storage tank in Fig. 8.12 has a hemispherical top and a circular
AXQ8, but only the element number I (not the node numbers J, K, and L) lll'l'd
z
be given. Also, the acceleration factors GAX and GAY are replaced by GAZ I111 I
the uxisymmetric shell.

TABLE 8.5 Structural Data for Program DYAXSH

Type of Data No. of

Lines Items on Data 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'
Descriptive title
NN, NE, NRN, E, PR, RHO, II
NN J, R(J) , Z(J), V2R(J), V2Z(J), ll(IJ
NE I, IN(T, I), IN(l ,2), IN(l ,3)
NRN J, NRL(JJ-2), NRL(3J- I), NIU ,( \f J
1111
Fl11 t! I 1(11) ,
 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.
j2 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
t
J•-j1
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
j3
llcxural stress at that location.
16

® ..
3 . 00

I 0.4H
a.
I:

-+
_l_
0.4H
~ 2.00
:::,
(/)
(/)
w
0::::
a.
_J
<
I. 0 0
z
0::::
w
1-
z
- o. oo
2 4 6 8
(b)
T Cms l
Figure 8.12 (cont.) (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 MEMBRANE
this inllucncc. Physical parameters in this example are
a."' 200
E = 207 X 106 k:N/m 2 v = 0.30 p = 7.85 Mg/m 3
I:

R = H = 2m h = 0.025 m (p.,.)mnx = 3 MPa

"-
I")

w
when: the material is steel and the units are SI. O 0
0 8
As in Example 7 .6 , we need to find equivalent nodal loads due to press111·1• I' z T Cm s l
which we assume acts in the direction normal to the middle surface of a shell ek•11w11t 0::::
<
(lhl· t / direction). These equivalent loads arc w
z

f r-z J
-200
X
p1, = 27Tp,,. l fr• ,.:; rd( 1111
(/)
1

'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 11uto11111t11·11 lh
d11t1•111il11l·d hy liq. (n) l'or pressure t11ken to be positive in the positive sonsl• of1h1• 1111111111l ( b)
tliH'l'Iio11 r/.
lt'l1111111 II II VIIJ11UUII N )lllN Nllli ll)ll' 11111k: (11) Jond ; (h) Nll\'NNCS,
 lnt11• nrut Hhnll1 Ch

REF R NC 8

I. M1ndl11.1, R. 1)., "l111lucncc of Rotutwy Mo11111111111I Shl'llt on Flcxurul Mot11u11

9
Isotropic, Elastic Plates," J . Appl. Mt•ch ., 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
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
Within
New York, 1981. '
6. Yoshida,. D. M.'. and Weaver, W. , Jr. , "Finite-Element Analysis of BeaniH und Flexible Structures
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 instL·ud of 6111•

411
 412 lll11lcl tloclh11 within Flexible Structures Chip Sec. 9.2 Rigid Bodies in Framed Structures 413

2
In this chapter we examine the effects of including rigid bodies within
t_, framed structures and discretized continua. It is assumed that the bodies are
/ A connected and supported by linearly elastic, flexible materials and that their
3 displacements are small. We also assume that the bodies are not rigidly con-
nected to each other or to the supports, which would require additional con-
straints .
y

)-, (a)
9.2 RIGID BODIES IN FRAMED STRUCTURES

z Figure 9.2 shows x, y, and z axes and six indexes p I, p2, ... , p6 for
3 components of actions or displacements at a reference point p on a three-
t_2 dimensional 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
1
/ 6 y

f Rig id body
z /-5
~y
X

(b)

)-, /
X
/
/
/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

11shell 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>l·lcd I through 6 in the n~ure. Othcrwi1-K·, ii' the shell Wl'11.• uxl11v111
111l't11c. lhl· 1111111hl·r of iigid body motions would lw ll'Wl'l • 1>'111111·1• 1>.2 Rigid hody consti uint~ .
 415
414 Rigid Bodies within Flexible Structures Sec. 9.2 Rigid Bodies in Framed Structures

~ix indexes j I, j2, ... , j6 for actions or displacements at point}, as indkul where
111 thc figure. An offset vector rpj is directed from pointp to point} and has Sllll Zpj -ypj]
l'0111~onents XpJ, YpJ, and zpJ· We may calculate the statically equivalent al'flmt (10)
0
~ ,~ point p due to actions at point j using the concept of translation of axes I >, I I
Xpj

I hus, -Xpj 0

II 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
In Ihis equation the symbol Aj denotes a column vector of six action comprn11 nl
to that in Eq. (4.5-1) for small rigid-body ground displacements.
111 point}, as follows:
A similar transformation matrix TpJ can be derived for each of the other five
A1 = {A11, A12, · · · , A16} 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
wh~rc the first three are forces, and the last three are moments. Also, the w11
first five matrices in the table can be found by deleting appropriate rows and
A,, 111 Eq.
(1) contains six statically equivalent action components at poi111 I'
columns from the sixth. For example, point p for a rigid lamina in a plane truss
Aµ = {Aµ1 , Aµ2 , ... , Ap6} would have three displacements, which are translations in the x and y directions
l<'inully, the transformation matrix TpJ in Eq. (1) has the form
TABLE 9.1 Transformation Matrices Tp;

Tpj = [:;j :J for Framed Structures

1. Beams (x-y plane of bending)

i11 which l, is an identity matrix of order 3 and
Tpj = [L ~]
~
-zpJ 2. Plane trusses (structure in x-y plane)
CpJ =[ 0
-y~ T·=[
Pl
b ~]
- ypj Xpj

Thi s skew-symmetric submatrix contains positive and negative values ot th 3. Plane frames (structure in x-y plane)
n>mponcnts of the offset vector rµJ· 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}. Tp;=[ b ~ ~]
S11nilurly, the kmematically equivalent displacements at point j 11111y I -ypj Xpj I
rnkulntcd from those at point p with the relationship 4. Grids (structure in x-y plane)

= T;o,, 1ft I O
D1
b -t
YPi]
T p; = [~
111•11· th~ ~ymbol DP represents a vector of six small displacement co1111>11t1l'III
5. Space trusses (transposed to save space)
ol lhl' 11gtd body at the reference point p. That is,
D,, = {D,,1, D,,2, . .. , D,,6} T;; = [b ~ ~ -~; t -~]
0 0 1 YPi
2

-Xpj 0
whl'H' thl.' first lhrcc arc translations , and the last three arc rotations. In 11dd111111
6. Space frames
tlt1· Vl'l'tor 1>1 rnnsists of six kincmatically equivalent displacements at point
I 0 0 0 0
D, {D11 , l>p, . .. , D,c,) (N 0 I 0 0 0

l
0 0 0 0
'1'111• 11p1•111l111 '1' 1'.1 i11 nq (h) is T,,, () z,,, Y,,1 I 0
:,, () ,,,, () I
() () ()
1,,1
 416 Rigid Bodies within Flexible Structures Chap I Sec. 9.2 Rigid Bodies in Framed Structures 417

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 ] (I I)
O Tqk Ak
'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
I I
l I
T; = [;j ~J ( 11)
I --- ..__
--- ..__
--- --- -- -- --
I
I
I
'l'ht• trunsf'onnation matrix T1 is a combined operator that converts the m:ti1111" Ill --- --- ..J
AM, (al lhc ends of the member) to the statically equivalent actions in A111 (111 lh
Figure 9.3 Space frame member with rigid bodies at both ends.
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'.
The displacement vectors in this eq~ation corres~ond to the action vectors in
Eqs. ( 13). Equation (15) may be wntten more bnefly as
DiHplHl'Cllll'nls at joints j and k wi ll also be expressed in terms ol' lh11111 ti
I' 1111d ,1 hy 1111 t·xtcndcd fmm of Eq . (6), as follows: (16)

I I
u,I
l)A
T,~,
O
o ,, n,. ,
'l',/1 l>,1
Whl'll'

( I 'I)
 418 Rigid Bodies in Framed Structures 419
Rigid Bodies within Flexible Structures Sec. 9.2

and 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
T; = [; ;JJ m 0 0 0 0 0
. In addition, we can transform the member stiffness matrix Ki from joinl• 0 m 0 0 0 0
J and k to the reference points p and q. For this purpose, consider lh 0 0 m 0 0 0
Mc= (26)
action-displacement relationships 0 0 0 f;u -lxy -fxz

AM;= K;DM; 0 0 0 -fyx fyy -fyz

Substitution of Eq. (16) for DM; gives 0 0 0 -Jzx -lzy fzz

AM;= K;T;D8 ; (20) In this symmetric 6 x 6 array, the symbol m denotes the mass of the body,
Use of this expression in Eq. (12) produces which is computed from

Asi = T;K;T;D8 ; (l I) m = fvpdV (27)

I lence, the matrix relating ABi to D8 ; is

The mass moment of inertia lxx is obtained as
K8 ; = T;K;T; (" )
The symbol_ Ka! denotes the member stiffness matrix for actions at points p urul fxx = fv p(y~ + z~) dV (28)
f/ due lo unit displacements at those points.
Similarly, the consistent-mass matrix M; for the member may be trn1111 in which Ye and zc are they- and z-distances of a typical point in the body from
lonncd from joints j and k to points p and q. Reasoning as above, but w11h the center of mass. The mass product of inertia lxy is
uccclerations instead of displacements, we can derive the fonnula
fxy = L
PXcYc dV (29)

I, Iere the sy~bol Mi_ re.presents the consistent-mass matrix for member i al jn1111• where x is the x -distance from point c. Other moments and products of inertia
J nnd k, while MBi 1s its counterpart for the reference points p and q. in matri~ M have similar definitions. The negative signs on mass products of
Equations of motion for all of the possible rigid bodies in a framed sll 111 inertia in Eq~ (26) result from Euler's equations for small motions of rigid bodies
lure may be written as 14]. Table 9.2 contains various forms of matrix Mc required for all types of
MaaDa + SaaDa = A 8 (t) ( '4) framed structures.
The matrix M for a rigid body may be transfonned to a working point p
To form the matrices in this equation, we assemble contributions from individ1111I hy an operation si;ilar to that for consistent masses of membe_r i ~iven in Eq.
lllcmbers by the direct stiffness method, as follows: (23). However, only the points p and care involved, as shown m Fig. 9.2. The
,,, required congruence multiplication is
S,111 = L KB/ Mao= L"• Mo; Ao (t)
"e
= L As (t); ( 1)
Mp = TpcMc TJc (30)
This assembly process is similar to that described previously in Sec. 3.5, hut 1h 1n this equation the transformation matrix Tpc is of the sam_e _form as ?1a~ix TpJ
111otions arc at working points instead of joints. If a rigid body docs nol l'Xllll 111 discussed earlier, except that c replaces j. Note that for a ng1d body_ m either a
11 particular joint, all of the offset vectors for members framing into that joinl 111 plnne truss or a plane frame, the matrix Tpc must be the same (and of size 3 x 3).
t11kl·1t to hove zero lengths . Also, the rotational displacements arc 0111illt·d lor Also, for a three-dimensional body in either a space truss or a space frame, Tpc
11011t·xislt•111 rigid bodies i11 plane and space trusses. 111 11gain the snml' (1111d of si7,l' 6 x 6) . . .
As yet, lht• muss 111111,ix M1111 in Eqs. (24) and (25) is tkvoid ol 11111 As the st·~·o11d 1111•p 111 11HNl'11thling cquutrons of motion , we must add the
l11h11tio1'.s 110111 lht· rigid hodit·s lht·111st·lws, whidt 11111y lw lht• 1110s! 1111p,;111111t 11111 ssi•s 1111 d 11111s1, 1111111,rrl! M111 1111·11111 lt11 tht• ri~id bodies to the matrix M1111 in Eq.
11•1111~ li.11d1 111-t1<I hody 111 lltl• 111111ly11nrl 111od1•l lt11s Its 1•1•1111•1 11/111t1.1·,1· lornh·d 141 ()•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)

Mc= [~ I~J
2. Plane trusses and plane frames (structures in x-y plane) i..l,- - -; ; - - - ~ -- - - ~ - - -- -·+---;;---
j~
Mc= [ ~
0
3. Grids (structure in x-y plane)
~z ~]
0 I,,
-
Mc =
f xx
- f yx
-fxy
f yy
OJ
0
....,
~

[
0 0 m
4. Space trusses and space frames
,,;
0 0 0 0
g"'

J]
m 0 0 0 ..,c::
0 m 0 0
M, -[} 0 0 f,x -fxy "'
-a.
.s..,"'
0 0 - f yx ! yy -ly,
0 0 -fzx -1,y I,,
"'c::
'§
...., .$
"b 00

M1s =Mes+ L Mpk ( II)

-0
'@i
k=l 62
wkrc 111, is the number of bodies. ...,.
Q\
After solving the eigenvalue problem for the augmented form of Eq. ( 1 11 ...
QI

wt• cun add modal damping to our analytical model. Initial conditions 111111 So
~
g1ound motions may also be included, along with applied actions, if dcsii11I
St·t·Iion 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
1wglt·clcd), 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 convl•11lr11t
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 diH•t111111
ol pri11l'ip11I body axes are not known in advance, they may be found by solvl11
1111 l•igl'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' i111plit·d rigid bodies do not exist, we could also eliminate the displ11n•111111t
111 11111ssk ss 11odt•s by stuti,· r<'d11ctio11 , as discussed in Sec. 6.7 .
K1rn111111t, •>.1
1•1µ1111• 11 •I Hli11wH II p1111i11u 11111 pl111111 1111NN wilh lwo dgid 1l'l'l11111;11l111 l11111ii1111• 1·11111111 l
liy II pilN11111t11 1111·111lw1 r 111 1111111~ / 1111d A 11,1 11~ dl'Vl'l11p 1•1111l11h111l1111N111 lhl' 1111111h
1111d 1hr 111111111111 l111111d11111p1•tl 1•q11111i1111N 111111111111111111 p111t1l ~ /I 11111l 1/ llt1 ll11, lipid 11111(1

.....
 422 Rigid Bodies within Flexible Structures 423
1UC. 9.2 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 2 Sym.
Statically equivalent actions at the working points p and q may be obtai111•d II II 0 2
transformation matrices for plane trusses from Table 9. I , as follows: 5pAL -8L 6L 50£2 J
·-------- - ----i-- (g)

~ ~J
3 1 0 -4L I 2
I
TP, = [ O 1 3l l0 2
-4L 3L 4L - 4L -28£2 I 8L -8L 64L2

Substituting these arrays into Eq. (14), we form the operator T; and use it in h1 1I
Thus, which is found using Eq. (23). .
Turning now to the rigid laminae, we take the followmg mass and mass-moment-
ABi = T;AM; = {Ap, Aq} 111 -inertia matrices from Table 9. 2 for plane trusses:

= {A11, A,2, -L(4A,1 - 3A12),: Aki, Ak2 , 4L(Ak1 - Ak2)}

mi
12 0 0 ]
Second, we set up the 4 x 4 stiffness matrix for member i, using Eq. ( I , Mei= - 0 12 0 (h)
12 [
0 0 25l 2
0.64 Sym.1
K; =
K.ii K,k] EA 0.48 0.36 J

1'1nnsformation operators required here are the 3 x 3 arrays

[ Ki\i i - IOL -0.64 -0.48 10.64
r
Kkk

- 0.48 -0.36 ! 0.48 0.36

Then we transform matrix K; to the working points p and q with Eq. (22), p1111h11 I
,:, n l-:L ; :1 (i)

0.64
0.48 0.36
Sym .
0 ~] = [ ~ 0 ~]
Xqcz 1 2£ -2l 1
= -
EA - I.l2L - 0.84L 1.96L 2 i
lOL ------------------1---- I hey are applied to convert matrices Mc I and Mc 2 to the reference points p and q,
- 0.64 - 0.48 l . l2L l 0.64
111"rnrding to Eq. (30). Thus,
- 0.48 -0.36 0.84L I 0.48 0.36
6
- 0.64L - 0.48l l.l2l 2 J 0.64L 0.48l 0.641.' T _ m1
Mp= Tpc1 Mc, Tpc, - 6 0
Third , we write the cons istent-mass matrix for member i from Eq. ( I ~ \ [ -12L
Thul is,
(j)

M, 5pAI,
1
r~ I
2
0
Sym.

2
M,,
, T
1,,, i M,. 2T ,,,·2 - 3
_ m2

[
3
0
61,
0
3
61,
~~L]
321}
0 I () ')
1lw 11 11, 1 o r thl'Nl' 111 rnys 1111n1111·11ts s 11h1111111 Ix M,,1,1 in Eq . (Hl, 1111d the Nl'l' ond is nddcd 10
'1'11111sl11111111111111 nl 11111ld x M, lo llll' wm·~11111 points 11 111ul ,, yi1•lcls M,,.,, Thr N Htr11 t111till H th1• 11h11•1 II Vl'N Ht11h•d Ill lh1· '11•11 11111111)\ 111 th1• t'Xlltllpk
 Rigid Laminae in Multistory Buildings 425
424 Rigid Bodies w ithin Flexible Structures ftoc. 9.4

y
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 L- ---.1.--

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 L
key to easily extending the programs for framed structures to incl ud1· 11 0 0
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
j2
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
t
·-i1 4L
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 t•q111I
j3/ i
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 M;
(9.2-31). After the equations for rigid-body motions have been solved , w, - - -- ---x
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.
Example 9.2
Figure 9.5 Rigid body in a plane frame.
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 1•11,1. 9.6, the first of which depicts the impulsive x-g~oun.d ac~eleration due to the
to ground acceleration D 8 1(t), caused by an underground blast. underground blast. The plot in Fig. 9.6(b) gives the resultmg_t1me history of x_-translat10n
If all members of the frame have WlO x 45 cross sections, we can Ninh 111 node 3 (the center of mass of the rigid body). Also, Fig. 9.6(~) contains y~ots of
physical properties as lli1tc•varying member end-moments AM6 for element.s 2 and 4. Max~mum (or rrumm~m)
E, = 3 .0 X 104 k/in. 2 p, = 7 .35 X 10- 7 k-s 2 /in. 4 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.
L = 36 in . A = 13 .2 in . 2 I, = 249. 0 in . 4
where the subscript s implies steel. Al so, the relevant properties of the d gid l1111l y
7 lt,4 RIGID LAMINAE IN MULTISTORY BUILDINGS
t>, 2.25 x 10 k-s 2/ in .'1 Ill {), [ , I 1.050 X 10 2
ks
1
/ i11

I Afl, I.' 2.268 k s 2-in. M1tllistory bui ldings arc usually constructed in tiers of one, two, or three .stories
for whk h lltt' s11hst·1lpl ,. dt•11otcs rnut·1t•l1•. , 111 11 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· cilk·s. Tlw stl·c l skl'l\.\lons of these lull buil~mgs
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 111•111111 p1111h11·1·~ 1111• 1•11111p111l I plnl 1 l!IININI nl ll jllll'l' 11111111·11 111111 IIIIIHI l11•q111•nll y 1111w tht•ir llll' itllll'l'I, lllrttl\At'<l 111 llll
 Rigid Laminae in Multistory Buildings 427
426 Rigid Bodies within Flexible Structures Chlll Sec. 9.4

400 Slab or deck

N ., y-beam
'- 200
C
x-beam

0 60 80
20
(.) T (ms ) P,(t)

T---J--.----+-----
<
a:: - 2 00
{!)
I
X
- 400
Y FP
(a)
C
·.:; 0 , 0 4
I")

~ 0 . 02
0
ll'---.1 F

z i - - - - -- X , c - - -- - + I

< o. oo
z
0
20
T
80
(m a )
- - -- -X,p-- - --,.... P,(t) t
;: -0. 0 2 k--- - - -- --X,; - -- - -- - --1
<
..J
(/)
Figure 9.7 Typical floor plan of multistory (tier) building .
~ -o. 0 4
a::
I-
I
( b) rnlumns that are parallel to the z axis (not shown in Fig. 9. 7). An origin of
)(
ELEM,2
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
100 ,. represents the center of mass that is tributary to the floor, and point j is a typical
C
joint at this framing level.
I Because each horizontal lamina is taken to be rigid in its own plane, all
~o 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
lO
r
< - I 00
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
( c) 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
li'IKurc 9 .6 Rigid body in a plane frame: (a) ground acceleratio n; (b) displut'l'1111111 . 11t her hand, the x and y rotations and the z translation at a typical joint j remain
{c) 1'1HI lllOIIIClltS.
11s unconstrained displacements.
111 Ihogonnl ( or l'L'<.: langu lar) pnllcrn . Figure 9.7 shows a lypical floor (111 11 For simplicity in the analysis , we assume that the geometric layout of the
pl1111 of II lit•r building, (.'()Jlsisling o f x hcnms, Y· beams, und a sluh 111 ,It, k ti li 11ming is lhe same at ull levels, that there is only one distinct lamina at each
wr shull l'llllN1dt•r lo tw I igid i11 its own plnnL'. This l11111i1111 is fk x1hh 111 I ll•vcl, and thut lhL:rt nn• oo shear walls or lateral bracing in the building.
dtrt'\'111111 11<1111 1111 to ii ~ plt11a• 1111d t·ntdd Ill' 1li s1•11•ti:,,t•d wilh plult' lw111lt11ai I Mc.·111bc1s of llw spnl'<' t 1111m• nrc 111kcn to he prisnmt ic and of a lincorly elastic
1111·111 , A 1111,lt-1 l11rl 111111r 1111111111111l y !Hird upp1011d1 IN t11111cl11d1· 11 tr 111111111 v 1111111•1inl Wt• ulNo IIHNIIIIII" 111111 joi11tNIn llw r, 111111• IIH' I i1tiil 1111d thllt displ11L'l'll1Cl1I S
ol lhl' Ninh 111 tlw 1,11,.'I .,r, 111111 111 1· 111 Ii lw11111 Thl''1' hr11111 ~ 11111111• 111111 v rt 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 "' I"- "~. ( :s
stiffness matrix Ki for such a beam is the same as that for the grid mcn1h1 1 II
Sec. 6.3. Thus,
,.,•- - - - -
'.zr -
o '~.s
------=--------'< -
·-. S:l
-~ 2
VJ
·,;:
oj
N

~~
0.
GlxL 2
C:
Sym . E
::,
0 4ElyL 2 lI 0u

0 -6EJ L I 12Ely ~
- ------- - --L- ~- ------ -----2
-GfxL2 0 l O GlxL
(I
E
-in'~
'< ;,...
0 2ElyL i -6ElyL
2
0 2
a:l
0 6ElyL l - I2Ely 0 12Efy =;;i
~
0.
As hcf'orc, the symbol Ix represents the torsion constant, and ly is the 1w11111 E
oj
1110111ent of the cross-sectional area with respect to the Ym axis. ..8
The second type of beam to be considered has its axis parallel to thl' I' 11 I ein'
as indicated in Fig. 9.8(b). In this case Ym is chosen to be the member uxlN, 1111 ·,;:
oj
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 2
E
.;
111ember stiffness matrix for the y-beam is
~
'<

K, = [~j
B
Kik
Kkk
J 0 ~
0.

Q)
Kki i !:Q
~

4EI.,, L 2
QC>

0 GlyL 2 I
I
Syni.
,.,• - - - -
Q\
..
::,
I .!:/1
6ElxL 0 12Elx I ....
. ------------1 ------
I..' 2ii;i} 0 6ElxL I 4ElxL 2
,,l

0 - GI y L 2 0 I 0 GlyL 2
6El,L 0 - 12£/, 1-6£1,L 0 12W,
Noh• thul for lhis type of me mber the cross-sectiona l properties h11 v1 11
Ny 111hols lwt·nusc of lhc orientation of the memhcr uxis . T hut is, / 1, is 1111w th
1t11 slo11 rn11s1u11l , whlk / 1 is the st•rnnd 111011wnt ol' the cmss Sl'<:lio,wl 1111·11 wit
11'Nlll'l'I lo till' 1111 uxis
 430 Rigid Bodies within Flexible Structures 431
oc. 9 .4 Rigid Laminae in Multistory Buildings

Figure 9.8(c) depicts a typical column with its member axis z,,, p11111II For this member the cross-sectional properties are defined as follow.s:hA, = art:·
the z reference axis. Its principal bending axes x,,, and y111 are assu,m·il I . t t and I x and I y = second moments of area wit respec o
/ = torsion cons an ,
1
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 lhe x,,,;::t~~:ex;:~sence of the rigid laminae, certain geometric tr~nsforfmationhs
terms from the member stiffness matrix to joint and floor stiffness matrkr~ 11 . . t . and c to those at pomt F or eac
ure required to relate the motions at p~m s J . t . and their
is , indexes j I, }2, and }3 at the top of the column correspond to joinl dr~pl lraming level Figure 9.9(a) shows actionsAj4, Ajs, and Ai6 at pom J .
ments that are not directly associated with the rigid-body motions of tlu n .
Nlat1cal . ·1ents AFI, An, and A F3 at point F. Values of the latter actions are
eqmva
above. These displacements are the same types as those discussed previou~I n1lculated from the former' as follows:
the beams. Similarly, indexes kl, k2, and k3 at the bottom of the colun111 p ti (5)
lo joint displacements not directly associated with the rigid-body motio1111111t AF = TFjAj
floor below. However, the indexes j 4, j 5, and j 6 at the top and k 4, k 5, 111111 !'his equation is similar to Eq. (9. 2-1)' but point F replaces point p. In this case
ul the bottom correspond to displacements that are directly dependt·111 11 1he vector Ai contains (6)
rigid-body motions of the floors above and below. They consist of Lru11~l11II Ai = {Aj4, Ajs, Aj6}
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-- Sym.
[Ki, I I
K31
I
K32 I K 33
! K42
I
Kt3 ~
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 (a)
1111t·o11strained displacements at level j from those at level k, as well as I111111 I
t·onstrnined displacements at both levels. The submatrices in Eq. (3) 1111
4fx 0
K11 = K 22 = f [
i 4~

2/., 0
K2, - I [
_E
i 2;;

~l
- ly
_ 6E
K11 K,i • K,11 = K42 - /} I, ro 0
() ()

I JW,, () 0 (IJ)
KJ, K,, I ()
K11
f I
I JFI, ()
li'IMIII ,, 11 ,11 I h 111111 II II ll llll •" l11111111IIIIII N (II) 111111111 11•l11l l1111MhlpN, (t,) d lNpl111•1111111nl
(I () <;/ I '
1rl11lh111ijhlp~
 433
Sec. 9.4 Rigid Laminae in Multistory Buildings
432 Rigid Bodies within Flexible Structures

l
which has the same meaning as Eq. (9.2-22), except that the subscript F replaces
Vector AF is 8. The results of Eq. (14) in expanded form become
AF= {AFI, AF2, An}
_!_(~~-~---- - - Sym.
and the transformation operator TFJ has the form
K Fi = _ K 21 -+-- K22 - ~-- - - - - - - (15)
[ TFJ K31 l TFJ K 32 I TFJK33 TJ:1

T,1 - [ _L ,!, ~] TFk K41

I
I TFk K42 ! TFk ~ 3 T TFj
I
TFk ~

Note that the transformed stiffness matrix KFi is still partitioned in the manner
TT
Fk

indicated for K, in Eq. (3) . When the submatrices in Eq. (15) are evaluated, we
~u~ing next to .displacements, we see in Fig . 9.9(b) the effect at polnl
find that the equalities among them are the same as those given in Eqs. (4),
of a n ~1d-body rotation DF 3 at point F This rotation causes a negativl'
provided that reference points at all levels lie on the z axis.
lran.slation ~nd a positive y-translation at point j. Hence, the kinemuti1 1111
In lieu of more detailed information, we assume that the mass of the
equivalent displacements at j due to rigid-body motions of point F are
building tributary to a given framing level is uniformly distributed over the area
Dj = TJ:jDF 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
Again , this equation is similar to Eq. (9.2-6), but with point F replacing poh
lo point c. Thus, the mass and mass-moment-of-inertia matrix with respect to
p . Here the vector DF contains the terms
e
point c for the lamina at level may be written as
t Ill
1 0
und lhe vector Di is (16)
Mee= me O 1
(
DJ = {D,4, DJ 5 , DJ6} t 11 0 0
Also, the transformation operator becomes where me is the mass at that level. Furthermore, the mass-inertia matrix with

~
respect to point Fis obtained as
TJ:, = [~ tI 0
0 0 1 (17)
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 opi·r ntlt This transformation is of the same type as that in Eq. (9.2-30). The symbol rF
n.:qu'.rcs a combmed action-displacement transformation that makes usl' 111 ti In Eq. (17) denotes the radius of gyration of the floor with respect to point F. It
11111tnx Is related to that with respect to point c by
(18)
Sym.l
which is a familiar expression for translation of axes from elementary dynam-
tI
ics [4).
T,.'k 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
'l'h1· tr:u11,1'0 1111ation matri x 'I; conlains f'our 3 X 3 suh111atric1:s on till' d111~1111 I point F, and the stalically equivalent actions at that point are
of whrch thl' first 1111d Nl't·ond 11r1: id1:ntit y mntriccs. The third nnd fo11r11l ul
111atri1·1·s lltl' ol till' typt• giVl'n by Eq . (8), 11nd tht•ir suhs<:tipts dt noh' j11lt11 I A,:(t)r T,..,,A,, (t) r (19)
th1• 11001 11h11Vl' 1111d joint ~ in tilt• flow lwlow INt't' Fii:. <) H(c)I , 'l'lll' ·,h ~11
In whkh
l11111 11 lll111111f11111 nl 1hr n1l111111111frll111•10111111f11x tH'
(10)
A,,(/l1 • (/',(I) , /\(()\,
 434 Rigid Bodies within Flexible Structures Sec. 9.5 Rigid Bodies in Finite-Element Networks 435

and

When the forcing influences are horizontal ground accelerations, the resul11111
inertial forces at level are e
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. y

9.5 RIGID BODIES IN FINITE-ELEMENT NETWORKS

I
)-,
In this section we shall consider rigid bodies that are embedded wllhll Figure 9.10 Rigid laminae connected by QB elements.
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 l'igid-body motions, which consist of ~ransl~tions in ~hex an~ y directions and
structures (see Sec. 9.2) will be abandoned in favor of a body-oriented 111,·11,,, 11 rotation in the z sense. At node J the mdexes J 1 and J2 repr~sent ~nly
Figure 9.10 illustrates an analytical model consisting of rigid 111111111 lranslations in the x and y directions (for a Q8 element). Correspondmg actl~ns
connected by Q8 elements for plane stress or plane strain. By omitting till' 11111 nt node j are forces in the x and y directions, and their statical equivalents at pomt
bodies temporarily and excluding support displacements, we can write equ11111111 ,.k become
of undamped motion for the nodal displacements in the finite-element 1w1w111~ (2)
as fo llows:
which is similar to Eq. (9.2-1), but point ck replaces pointp. Actions in vectors
MM MAF] [SM
[ Mfoi\ MFF Op + SFA
[DA] SAF]
SFF DF
[DA] = [AA(t)]
Ap(t)
II Aj and Ak are
(3)
In this equation the subscript A refers to nodes attached to rigid bocli1•11, 111
subscript F denotes free nodes (see Fig. 9.10), at which the displact•1111•111 ~ 11 In the latter vector we have forces in the x and y directions and a moment in the
indcpcndcnl. For an arbitrary sequence of numbering nodes , the lcr 11111 11111 sense. The transformation operator in Eq. (2) is
l1ihuti11g lo Eq. (I) must be rearranged to put type A first and typcF st.K·orul N11
ll't us hri ng in the rigid bodies and recogni ze that nodes of type A arc <:0111111,1111 (4)
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. which hus tlw sa111t· frn 111 11N that for plane trusses in Table 9. I , with k replacing
llip1111• 1J Io 111tl1c1111•H rh,· 11111N.~ ,·1·111,•1 , ·1 of hody /.. 1111d 11 1yp11·11 I 111111
I'
wh1•11• l'11·1111•111, 111 1111111l11•d l1uk111·11 /..I ,/.. ' , 1111d /.. I al p111111 , ·1 d1•1111h I Hq 1111111 11 1 ( 1) ll ll I ic 1111 I y Io ti w / 111 11odt• 011 I ht• /.. 111 11 p1d hod y, 11111 I II II II 11odt·~
 437
436 Rig id Bodies w ithin Flexib le Structures Sec. 9.5 Rigid Bodies in Finite-Element Networks

attached to the body are considered, the equation becomes

Ank = TkAAk
DA1l
DA2 = [TI1l
Th D (12)

The expanded form of Eq. (5) is [~:.:' ~i, "

When all rigid bodies are taken into account, we have
(t\) (13)
DA= TlADB
Here the vector D8 consists of displacements at mass centers for all of the rigid
AAn1 k bodies, and DA
contains the dependent displacements at all of the attached nodes.
Of course, T'};A is the transpose of matrix ToA in Eq. (8) .
where nj is the number of nodes on the body. Each subvector in AAk contains tw
Now let us return to Eq. (1), consisting of the equations of motion for
~crms, as in the first of Eqs. (3); and each submatrix in Tk is a 3 x 2 arniy , 11
attached and free nodes, devoid of contributions from rigid bodies. To account
111 Eq. (4).
for the presence of the rigid bodies, we form a transformation operator T , as
Considering now all of the rigid bodies in the system, we express the nrt h1n
follows:
trunsformation relationships by
An= TnAAA
l II th is equation the vector AA contains actions for all of the attached nodes, 1111 I
T = [T~A :J (14)

A,, consists of their statical equivalents at mass centers for all of the r igid hod I" in which IF is an identity matrix of order equal to the number of free nodal
'l'ht transformation matrix T8 A in Eq. (7) is a large, sparse array contuinln displacements . Th~n we have
1nibmatrices of type Tk in diagonal positions, as follows:
(15a)
T1 Sym.
0 T2 Also,
TBA=
0 0 Tk
'" Therefore , premultiplication of Eq. (1) by T and use of the relationships in Eqs.
(15b)

0 0 0 Tnb
( 15) produces
where 11,, is the number of bodies in the system.
Similarly, the kinematically equivalent displacements at node j in 11,
M 8 n MnF J [Dn] [Snn SnFJ [Do] _ [Ao(t)J
DF + SFB SFi DF - AF(t)
(16)
[ MFB MFF
'). 10 may be calculated from those at point ck by the formula
Terms with the subscript B in this equation refer to motions of the rigid bodies
DJ = TiDk (II
in the system. Block stiffness and mass submatrices in Eq. (16) are
whh.: h is of the same form as Eq. (9.2-6). In this case the displacement Vl'l 1111 (17a)
llll'

t 111 and
(17b)
11 1111 nodes on lhc kth ri gid body arc included , the tru11sf'orrnatio11 in h( 1U
ht'l'Ol lll'S llccnusc the nrntri x 1'1111 has the form shown in Eq . (8), we can express Eqs. (17)
IIIIH l' ('Xplil'itly llH

'I'PS IU , 1'l' ft S1111 S/m ' liS~, 1 ( I 811)

1111 whi1 h 1111' 1·,i p11rnh-d 1111111 i11 S
• 11111 I
438 Sec. 9.6 Program DYRBPB for Rigid Bodies in Plate-Bending Continua 439
Rigid Bodies within Flexible Structures

and 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
(18h) introduce information about the rigid bodies that enables the computer to dis-
fork = 1, 2, .. . , nb and C = 1, 2, .. , nb, Matrices Snnk. e and SMk, e in Eq., tinguish attached nodes of type A from free nodes of type F . Then the desired
( J 8a) are submatrices of S88 and SM that contain terms coupling body k wi lh matrices may be generated in the partitioned form shown by Eq. (9.5-1).
body C, and so on. As mentioned previously, the sequence for numbering displacements at a
To complete the equations of rigid-body motion , we need only add 1hr node of the PBQ8 element is z-translation, x-rotation, and y-rotation. We also
mass-inertia matrices for the bodies to matrix M 88 , as follows: 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 corre-
nb
sponding actions at point ck due to unit actions at node j becomes
M;n = Mnn + I Mck ( 111)

~
k= I

where Mck is drawn from Table 9.2. For a typical two-dimensional body k in Fill T kj = [ :~ ~] (1)
9. 10, the matrix Mq is the same as that for plane trusses and plane frames giwn
-xkj O 1
in the table.
We shall now discuss briefly other types of rigid bodies in other types 111 This array is taken from the third, fourth, and fifth rows and columns of the last
discretized continua. If we have three-dimensional rigid bodies embedded in 11 matrix in Table 9 .1, with p replaced by k.
network of H8 or H20 elements, the transformation matrix Tkj becomes the s1111111 Similarly, the mass-inertia matrix for rigid body kin a discretized plate is

[I ~;~ ~!~l
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 M,, - (2)
t•quations of undamped motion remains the same as that described abow 1111
two-dimensional continua, but numbers of actions and displacements art• 111 which is drawn from the third, fourth , and fifth rows and columns of the last
ncased.
matrix in Table 9.2.
Rigid bodies supported by discretized plates in bending are analyzed 111 11 Supplementary rigid-body data required for Program DYRBPB appear in
11111nncr analogous to that for grids, except that the theory is body-oril•111111 Table 9.3, which conveys the number of bodies NB in part (a). Then in part (b)
instead of member-oriented. Also, the sequence of nodal displacements for th we have NB lines of body-node data containing the body number K, the number
plate-bending element is not the same as for the grid element. As shown i11 hj of nodes NJ(K) attached to the body, and the body-node numbers JB(K, 1)
(8.2 I), the z-translation W; is taken before the x- and y-rotations Ox; and IJ,, through JB(K, NJ(K)) . Properties of the rigid bodies are given in part (c), wh~ch
Three-dimensional rigid bodies connected by networks of general 1,lu II indicates NB lines with the body number K, followed by values of the followmg
ckmcnts have characteristics similar to their counterparts in space f'r111111 11 six terms:
1lowcvcr, the rotations in the x, y, and z senses at an attached node are ac11111II
XC(K) = x-coordinate of point ck
dl•pcndcnt components of the independent tangential rotations a; and {31 (s1.•t• I ·11i1
8 .() ). YC(K) = y-coordinate of point ck
Finally, rigid rings in axisymmetric solids or shells with axisy1111111'lll BM(K) = mass of body k
lo11ds would be restrained to translate only in the z direction. If loud'I 111 XXI(K) = mass moment of inertia l:o:k
11111111xisymmctric, such rings can also translate and rotate in the r and () dlll'l
lions und rotate in the z sense .
XYI(K) = mass product of inertia lxyk

YYI(K) = mass moment of inertia l yyk

9.8 PROGRAM DYRBPB FOR RIGID BODIES IN PLATE-BENDING
CONTINUA
'1'11 illusll'lllt• tlw 1111lun· or prog rumming for Ila· hody rn it•nll'd Hppmud1 , w,.,1111
dl Nl'IISN Prn~ tlllll l>YRI\PII 1111' dynun1il' 11111il ys 1~ 1)1 li 1ii d hodi,·N s11pp1111i·d I
pl1111•11 Ill lll'111ll11p 1"01 lhl N p11tpUHI', till' Nltlldll111l 1111d d y 1111111ic loud d11111 1111
We also need supplementary dynamic-load data for Program DYRBPB , as
shown in Table 9.4. Information contained in this table consists of initial condi-
lions 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'llit•H NBIV. In part (b) arc NA()) lines with the body
1111111h1·1 K 1111d lh11·,· po10,1hll' 11111i11l displ11l•1•111,•11ls 1)1\0( 'K 2), DB0(1K I ). und
440 Rigid Bodies within Flexible Structures Chap

TABLE 9.3 Rigid-Body Data for Program DYRBPB•

No. of
Type of Data Items on Data Lines
Lines

Rigid-body data
(u) Number of bodies I NB
(b) Body nodes NB K, NJ(K), JB(K, I), JB(K, 2), ... , JB(K, NJ(K))
(c) Body properties NB 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
Type of Data Items on Data Lines
Lines

lnitiul conditions
(u) Condition parameters I NBID, NBIV
(b) Displacements NBID K, DB0(3K-2), DB0(3K-1), DBO(.IK I
(c) Velocities NBIV K, VB0(3K-2), VB0(3K-l), VBO( IKJ

Applied actions
(u) Load parameter I NLB
(b) Rigid-body loads NLB K, AB(3K-2), AB(3K-1), AB(3K)

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
t·o111putcd 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
t•li11il11atud by Guyan reduction . After solving the equations of motion 1111 rt.aid
hodil·s und f'rcl.l nodes, we can find time histories of displacements for 1111111 h I
11odt·s with Eq . (9.5 13). Then the displacements in vectors 0 11 and D, 11111y I
11Nl·d to obtain ti111u vmying stresses in the finite clements.

1,:,ui111pfo '> ..,

l•l1•t11r 'I 11 d1•ph 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 d1vid1·d 111111 two I'll
 442 Chap. 9. References 443
Rigid Bodies with in Flexible Structures

elements that are fixed along edge 3-5-8 and symmetric with respect to a plane co111111n ~
E
0.20
ing edge 8- 10-13. A step force of magnitude P, = 0.1 kN acts in the z direction 111 th E

mass center of the first rigid body, which is labeled point c 1 in the figure. Phy~1111I 0.1s
properties of the plate are
w
a o. 10
E = 103 x 106 kN/m 2 v = 0 .34 p = 8 .66 Mg/m 3 0
z
L = 0.05 m h = 0.002 m °< 0 . 05

and both the plate and the solid are made of brass. The rigid body is attached to 110,I -, 0. 00
a
I , 4 , 6, 9, and 11, and its properties are as follows:

L
R, = l = 0.05 m h1 =5 = 0.01 m

1rRf .,
~ 0
m, = ph,- = 1.700 X 10- 4 Mg Q..
:c
4

I.u =IYY =m, R, - [ 2(1

16)
4 -97T
-2 +-
12
hf] =3112X
. 10 -8 Mg·m 2
"' -10
w
a
0

m,-Rf(!
z

1r -
2 - -
16) -2 0
I xy = 91r = -8.913 X 10- 9 Mg·m 2 a::
<
w
z
whe re the symbols R, , h 1 , and m, denote the radius, thickness, and mass of the qu111 t X -30
disk . U)

We used this data in Program DYRBPB with DAMPR = 0.05 and direct 1111111 r ( C)
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 Figure 9.12 (cont.)
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 REFERENCES
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 W W Jr "Dynamics of Elastically-Connected Rigid Bodies," in Devel-
hi stories o f the flexural stress SX at the integration point near node 3. Of course, lhl IW !. eaver, . , ., . d w A Shaw Pergamon
opments in Theoretical and Applied Mechanics, Vo1. 3 , e · · · ,
1·111 vcs in each of the plots must be added to obtain total values for the responsck 111v II Press, New York, 1967, PP· 543-56Z.
2. Weaver, W ., Jr., and Gere , J . M ., Matrix Analysis
. of Framed Structures, 2nd. ed. ,
0 .2 0
Van Nostrand Reinhold, New York, 1980. .
o. 15 3. Weaver , W . , Jr. , and Johnston , p · R · , Finite Elements for Structural Analysis,
fl Prentice-Hall, Englewood Cliffs, N.J. , 1984. . .
Q
- 0 , 10 4 . Beer, F. P., and Johnston, E . R. , Jr. , Vector Mechanics for Engineers: Dynamics,
ID
4th ed., McGraw-Hill, New York , 1984. . . .
0 0,05
5. Weaver, W. , Jr., and Ne1son , M . F . , "Three-Dimensional
6 Analysis
385-404of Tier Bu!ld-
ings " ASCE J . Struct. Div., Vol. 92, No. ST6 , 196 ' pp. . . . .
0 00
6. Wea,ver, W . , Jr., Ne lson , M . ~-, and Manning , ~ - A. , "Dynam1ct4~~~:~7:_u1ld-
.1ngs,
. " A<'( " , t , F,"'
o} ' /• , Ml'C"h. Dtv. Vol. 94, No. EM6, 1968, PP·
("i • '

"''""l'I' '), ll lt1111d hmly 111 11 pl1111•· (11) hrnly fl1111Nllllh1IIN, (h) 11nd11l 111111Nl111 11111N, 11)
111·~111111 ~l11·on
 10 Sec. 10.2 Guyan Reduction Methods 445

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 dividl•d 1111
substructures and analyzed by one of the methods in this chapter.
For analysis by substructures, we must distinguish techniques thul 111 Figure 10.1 Radio telescope antenna structure (pointing to zenith).
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
t•q1111tions during a frontal reduction procedure [1, 21. However, in dy1111111I
10.2 GUYAN REDUCTION METHODS
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 Tridiagonal Method for Substructures in Series
111011011 cnn be written. Three approaches that work well for dynamks 1111 th
1i1di11gonul 11wlhod (with modificutions), tht• parnllcl elimination llll·lhcul , 1111 I Pigurc 10.2 shows n two-dimensional discretized continuum divi?ed into
lht• t·o1111HH1t·111 111mk ll't'hnique . Euch ol' these suhslnic1111·c 111c1hods will I s1thstruclurt•s 1h111111l· l'lllllll't'lcd in series. Suhstructurc numbers appear in ~oxes
d('Sl'iilwd in till' l'IISlllll~ Sl'l'liOIIS or thi s d111pl1•1·.
lwlnw till' llg111t• 111,11• 1111• My111hol (1 dl'IHlll'S II lypil'ul s11hstruct11re, 1111d 11, 1s lh(:
 446 Sec. 10.2 Guyan Reduction Methods 447
Substructure Methods Chap, 11)

B, 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 displace-
ments 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,
(5b)
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 / e- Equations (5b) are of the same form as Eq. (6. 7-7b). The acceleration re-
e
while Be+1 represents those joining substructures and + 1. Let us form 11 11 e lationship similar to Eq. (5a) is
column vector of nodal displacements as .. .. ..
DAe = TAeBeDBe + TAeB e+1 Dne+, (5c)

D = {on I' DA p Da2, DA2, . , DAn,' Dsn,+} ( 11 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 0 Then we have
S82A1 SB2B2 S82A2
S= SA2B2 SA2A2 ( 'I (7a)

0
SAnsA"s SA"sBns+I
and
Sn11s+ IA11s Sn,,s+1B,1sf I

m~t1:ix Sis a tridiagonal array of submatrices, this approach is rdr 11111

lll•1..·11usc De= (7b)
lo ns. lhc tridiagonal method. Without writing it, we can also observe tlllll 11 11
t·1111s1stcnt mass matrix has the same tridiagonal form.
Substitution of Eqs. (7) into Eq. (3) and premultiplication of the result by TJ
Suhstructurc .e contributes terms to the equations of motion as follow
produces
MeDe + SeDe = Ae(t) ( 11
111 l'Xpaudcd form, thi s equation is (8)

Mu,n, M11 1A1 In this equation the reduced matrices are

M, ,11, M,1,1, S
' "'

Ist,,
11 rll r
St 'I'~ Sr'I} (9a)
0 M11 111 11 111,
 448 Substructure Methods Chap. 10
Sec. 10.2 Guyan Reduction Methods 449

(9h) Example 10.1

The unrestrained beam in Fig. 10.3 is divided into two substructures, each consist!ng of

A i (t) = TI A e(t) = [ ~;e

ABe+1
J (9tl
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 ( 1011)
MrieBe = M BeBe + T Jen eMA eBe + 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 Figure 10.3 Beam divided into substructures.

Substructures contribute to the matrices in Eq. (11) by the usual direct assenihl v For substructure 1, the stiffness matrix in Eqs. (3) and (4) becomes
pr•occss. The reduced stiffness matrix is still a tridiagonal array of submatrit·,,11
ns follows: 6
3L 2L2 i
-------t---
0 - 6 - 3L i 12
(a)
3L l O 4L
L2
- --- - --T------- 1---
2 1

( I I) 0 0 I - 6 -3L I 6
I I
0 0 l 3L L 2
! - 3L 2L2
0

Subrnatrices of S!n having single asterisks receive contributions from onl y 11111 Terms in this array are drawn from Eq. (3 .4-24) . Similarly , the consistent-mass matrix
.~ubstructu re, but those with double asterisks have two contributing 1rnh for substructure I is
Nltuctures. The assembled mass matrix M% 8 in Eq. (11) has the same tridia111111,1l 156
u11·11 11gcmcnt of submatrices as the assembled stiffness matrix in Eq. (12) . S1111 22L 4L2 i
ll 11rl y, the assembled load vector is --- - - - -- - -l- - -- -
54 13L l 312 . (b)
A*()-
lJ t - {A*
BI' A**
8 2 , A**
83 , · · · , A***
Dn,, An,,,+, } ( l II
- 13L - 3L2 l O 8L i
- - - - - - - - - ~ - - - - - - - - -~-- - -
2

0 0 l 54 13L i 156
wli il'h aga in has single and double contributions. For convenience in co111p11h 1 0 0 - 13L -3L l - 22L
J
2
4L2
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:t·1111•11t
lk cuusc the assembled stiffness and mass matrices both have It idl11v11111il in which the terms arc obtained from Eq. (3 .4-26).
form~. vibn11 ionnl and dynamic analyses can be more efficient 131than 101 111111 1 Next, we use Eqs. (Sb) to fo rm TA tne and T AcD e+ i for substructure 1, as follows:
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 st111t·t111t·. 'I\ 1/11
1
S ,1 1fl I S ,1 1/1 1
"'·
I I 2/,
l
L21
/,
(c)
 450 Sec. 10.2 Guyan Reduction Methods 451
Substructure Methods Chap. IO
and
2 .------~-------------
TA 1B2 = -S;iA I SA 1B 2 = __!_[ 2 L -L ] / 78
I [I]
i
1, Sym.
4L 3 -L 2
1
I
22L 8L
r------ --,----------1
J

With these submatrices, the transformation operator in Eq. (6) may be written as I 27 13L \ 156 I \
M* _ pAL (j)
4L lo O O
BB - 105 II______ ____ TI _ _O____
-13L -6L 2
__ _ I
16L2
I II] II

T I --
In 1 0 J _ = 4L
l 2L
I
_9___~-L~___Q_.
L2
B,

l 2L
1
- L2
A
0
0
0
0
\ 27 13L
i- 13L _ -6L 2 _ -
78 \
22L __ 8L 2 \
TA1B1 TA1B2 (i) B,
[ ,:-3 -L 1 3 -L '
0 In 2 - -- -- -i---- - - - - -
0 0 J 4L O which may now be used in Eq . ( 11).
2
0 0 / 0 4L B
Modified Tridiagonal Method

Then the reduced stiffness matrix becomes Now we consider a modification of the tridiagonal method for sub-
structures in series that abandons the idea of insulating boundary nodes. Instead,
3 there are three definitions for types of displacements, as follows: (1) Subscript
*- T
SI - TI S1 Tl = -
EI 3L 4L2 I
---------f---- Sym.]:' Ae implies dependent displacements in substructure C; (2) subscript Be represents
2L3 -3 -3L I 3 ( fl dependent displacements at boundary nodes joining substructures C and C+ 1;

1 3L 2L2
I
/ -3L 4L2 2
and (3) subscript Ce denotes independent displacements associated with sub-
structures 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 indepen-
dent 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
< \,11sl•quc11!1y, assembling the reduced stiffness and mass matrices f O
Ml= Mi'
b
(hi
Men
:::J[;:J [t
Mee De e
+
ScA
(15)

111111 J yields ' r m su st ruct1111 • I Let the displacements of types Ae be dependent on those of types Be and Ce, as
,- 3 ------------ follows:
I/ 3L OJ
4L2 I Sym. fJ I (16a)

s,~"
m
2/,'
1-1/,3
2//
I3L
6
0
8/~-i--- [II IJ 2 11)
where
TA1llr = s;;,1 ,St1 r1J1 Ti1,cr =- s;~A eSAtCt (16b)
() () I/
() () I I/ ,I ' I/ ,I/ ' II I
Also, Wl' huvl' llw lll'l'l'il'l'Hlion rclnlionship
If

ll1, 'l\1111>11,
II

j 'I\,, 11),
U

r ( I Cw)
fl, fl /11
 452 Substructure Methods 453
Chap. 10 Sec. 10.2 Guyan Reduction Methods

Then the transformation operator becomes C, the mass matrix is diagonal, as in the next example. For such a cas~,
transformation of the associated eigenvalue problem to standard, symmetnc
Tc=[~: T~cJ 0 le c
( If I
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).
With this matrix , the displacement and acceleration transformations are 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
( l Hnl 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
und 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.
De=
[2:J =
..
De e
Te [~BJ
De e
( IHhl y

Substituting Eqs. (18) into Eq. (14) and premultiplication of the equation hy Tl
yidds CD 2
-oci

Me* [Dn]
..
c
De
+ Sc* [Dn] = Ae*(t)
c
'l'hc l'educed matrices in this equation have the definitions
De
( IV)

QJ
T7 L ®
/
DA,

0
st = TISe'fe = [s!B s:c] (.!OnI
Sen Ste e
--- -- -- © - oc,
Mi TIMe'fe = [M! M%c]
= 8
(.ltlhl / 3
/4
McB Mtc e Da, Da,

At(t) = TIAe(t) = [A!] Ac e [I] L

® ®
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
1111•11(s of type Be become a subset within those of type Ae+1. Also, for lht· In 1
L / 5
X
/ 6

suhslrncture, displacements of type B,,, are usually omitted. Thus, tht• 1111111
1•q1111tions of' motion become
* Dc
[M,c .. + Sfc Dc = A7,(t)J,,, l 'II
z
/ I 2L .1
Figure 10.4 Plane frame divided into substructures .
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 Fm suhNhul'1 1111• I , w1• Sl'I up the 6 x 6 stil'f1wss 111ntrix in Eqs. {14) und ( 15), as
I lt1Wl'Vl'I , ii l1111qwtl lllllNSt'H 1111' II Mt•tl ill ('lllljlllll'llo11 wllh lll'l't' lt•mliot1N 111 l\' 11 lollows:
 4114
li111J1 1r1u.h1111 M11thrnh1
1
,110. 10.2 (11 1y1111 llnd111.. tlot1 Molhoda 455
,I/
Syn1. A1 In !his instance, (hem arc no displacements of type B2 because the bases of columns are
I. ' It '
llxcd. Also, displacements of type C2 are the same as those of type C1 in the first
WI J;' () ,I '
S, -u 0 I;' () 21.1 I II,
ijilbstructurc. Terms in matrix Si from Eq. (o) may be added to those in matrix S2 from
I !q. (p) to produce
3L 3L I 3/~ 3L II 12 116L2
I (' ,
- 3L -3L 1 - 3L -3L ! - 12 12 3IL2 116L2 i
A * + S2 = 2£1
(q)
A, B1 C1
S2 = SI --
15£3 -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 9L 9L !' - 126 306
IIH' IINl'd lo form T A e11e and TA ece for substructure I. Hence,

TA 101 -
- -
5 A1A1
-1 S 1[-4 -~J
A1 8 1 =
15 l 'J'his superposition of terms represents an overlay technique that will prove useful for
1·omputer programming in Sec. 10.3.
In preparation for elimination of type A2 displacements, we determine TA 2c 2as
T -
A 1C:1 - -
5 A- 1A
11 S A 1C 1 = 3[-1 11]
SL - 1 II
TAC 2 -_ _ 8, A_21A 2 s'A 2 C 2 _ 1 [0.2449 0.06122]
- - - (r)
2 L 0.2449 0.06122
I JNin,-: lhcs1.: suhmatrices, we compose the transformation operator in Eq. (I '/) 11111 111
Then the abbreviated transformation matrix T2 becomes
-4L L / -9 9
I A1
L - 4L I -9 9 0.2449 0.061221
---------j-------
1 15L O / 0 0
B1
Ti = [TA2C2] = _ _!_ ~}~~- ~:S)~.!_~- (s)
ISL O I ISL
--------~-------
O 0 l c2 L[ -L 0

0 0 / ISL 0 0 -L
0 0 / 0 ISL C , Now we can determine the final reduced stiffness matrix to be

'l'h111t•fon: , the reduced stiffness matrix Si is

S*
[ cc
J2
= S*2 = rs T = L3
2 2 2
El [ 14.45
- 17.39
-17.39]
40.65
(t)

1-~l~--~~~l----
This matrix contains actions of type C due to unit displacements of type C, with those
of types A and B eliminated (or released).
2
St 'f'/' S I T1 = l 5L3
EI
36L 36L / 126
Sym.l :'
Parallel Elimination Method
- 36L -36L !- 126 126 I
A more general manner of substructuring does not produce coefficient
IIN HIVl' IIhy Hq . (20H) . matrices composed of submatrices in a tridiagonal pattern. For example, the
1'1ol'l'l'di11g l'rnm su~structurc I to substructure 2 (sec Fig. 10.4), Wl' 11 ch 1111 two-dimensional discretized continuum in Fig. 10.5 is divided into four sub-
iltNpl11c1·m1•111s_ ol lype lJ in the former to become type A in the htllur. Adcl111 1111 I structures with interior nodes of type Ae and insulating boundary nodes of type
11111111h1111011Nfrn111 suhslruc1url' 2 to the stiffness matrix arc: Be having no special arrangement. In this approach the meaning of subscript Be
is that it includes independent nodal displacements for substructures 1 through
e. I
Si
WI
I,
I
I;;::
()
6L
()
2

0 11 I
As in the tridiagonal method, the technique is to eliminate dependent displace-
ments of type A 1 from each substructure and to retain a growing number of
independent displarcmcnts of type Be.
l/,
A,
"· () 17
(' ,
For each st1•p in the reduction process , substructure
following tc11111, In 1111' l'q1111tions of mo tion:
contributes the e I
<n )
 457
Sec. 10.3 Modified Tridiagonal Method for Multistory Buildings
456 Substructure Methods Chnp

and
..
[1?.AJ = TeDnc (26b)
B, De=
Dn e
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
B, B, 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
A3 A. B would both become "slaves" to the "master" displacements of type C.

10.3 MODIFIED TRIDIAGONAL METHOD FOR MULTISTORY

BUILDINGS

We now apply the modified tridiagonal method described in Sec. 10.2 to two-
and three-dimensional multistory building frames. Most planar building frames
Figure 10.S Substructures in parallel. can be handled in the high-speed core storage of a large-capacity digital com-
puter without dividing them into substructures. However, we wish to use the
This. cx~ressio n again contains the same symbols as Eq. (3) but its ex 1 1 same technique for both plane and space frames in order to take advantage of
Vlll'H10 0 IS ' I 11111
inherent similarities while explaining these two applications.
For a multistory rectangular plane frame, we take_the analytical model
[MM MAB] [~A] + [SM SAnJ [DA] _ [AA] illustrated in Fig. 10.6. The frame is assumed to have linearly elastic prismatic
MnA Mnn Dn e SnA S88 D8 e - A8 e ( \I
members, rigid joints , and fixed bases. We also assume that there are no shear
which is diffcrentfrom Eq . (4). ThedependenceoftypeA d' l
type JJ,. is written as e isp acement" upon walls, diagonal braces, or setbacks in the building. Each substructure in the e
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
DAc = TA elJ eD11r I' II level , and those of type B are at the lower level. From the joint displacements
I ll'rl' .the n~atri ~ TA.i•!J,e has the s~me formula as the first of Eqs. (Sh), hut lh labeled in Fig. 10.6, we see that axial strains in the beams are to be omitted;
111t•11111ng of IJ1 1s different. T he required transformation operator is whereas, those in columns are to be retained. The reason for keeping the latter

T1 = f ~:'l. 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,
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' d11q1l11t•t•111t•11t 111111 lll't't•ll'rn1io11 tn111sl'o1111Hlions urt· 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
n,
1
1) I
u,,
I t
'I', I >11 I ( 1t111)
l'OIIIIIIIIS , AINO, till' Nlli>M'ripl ~ llll lhmll' 111 typl' F (lllp lo botto m) Hl'l' I , 2, .. .•
 458 Substructure Methods 459
Sec. 10.3 Modified Tridiagonal Method for Multistory Buildings

t t Ym

/ /
t n (j) 1~
t t -/),
Level A
-~- '-------=------"7 ----
i2/ i k2/ k Xm

/ /
y /
(a)

t ~ ·,
B __ loA, _ Leve l A t0A3 DA2nc-•

~~~....C....~~~~~-4~~~--; _,, z Ym

/
DA4
/
DA2nc

Level 8
t 83 t DB2nc- 1 Level A .
t
11
0
_,, "-.
- ~ -
I
--13
.

/ / j2/
084 082n c

(j)
t t _,,
/ /
Level 8

y ~!_ - k 3 - -- - Xm

}--,
---x

/ k2/t
/ /

-~

Figure 10.6 Analytical model for Program DYMSPF. z Zm

(b)

n,, where ns is the number of stories. Finally, this analytical model has tr ih11t111 Figure 10.7 (a) Beam parallel to x axis; (b) column parallel toy axis.
masses m 1, m2 , •• • , mn, lumped at the framing levels, which includt· 11111
tributions from beams and columns.
· 1 ·2 k L and k2 imply that the 4 x 4 member stiffness matrix is the same as
As for a tier building, every member in Fig. I0.6 is parallel to om· 111 th .I 'J ' '
that in Eq . (3.4-24) for a flexural element. . . .
structural axes; so no rotation-of-axes transformations arc rcquirl•d . 1•11:1 111 On tht• otht·r h1111d , Fig . I0.7(b) shows a column with its member axis 1m
10.7(a) depicts a beam with its member uxis I',,, pnrallel to tht· 1 11xiN 111111 It purnllcl to 1lw v 11xt'l 1111tl ti s prinripnl bending axis z,,, ugain par~llel to the z axis.
principal bending axis z,,, purnlkl lo lht· ., uxis . Aclio1111nd displ11c1•1111;111 1111h 11 Action urnl dt ~plm i· 11 w 11 1111d1•x1•s 1111•111ls.J 1111d k ml' 1111111lwn·d 111 a s1•q11c11ct thut
 480
i,11h11t1111 turn Mnll10d1 line. 10.3 Mudl!lnd I I lrlln11111111I Molhod lor Mulllstory Buildings 461

IN l'<'.IHhll'.i;l' fo~· l1'1111Sll' lll11p 11111111111 ~llll111•~Nl'S lo joint 1111d Horn 11tl the next substrucllltl' 11re added to the residual arrays from the preceding sub-
111ut11ct·s. I hus, indcxcs1 I 1111111 'Ill lhl' lop ol lhc column arc lhl' .~1111 11• 11 t structure. On the other hand, the number of rows and columns of type F
Im II lw11111 ul lcvcl A , whllL· A I 1111d A.> 111 lht• bottom coincidc with tho~ h Increases from 2 for the first substructure tons for the last substructure. For each
hl•11n1 at level B. However, lhl' indl·xcs j I al the top and k3 ul lfll' hull 11cw story (except the last), we pick up one more lateral displacement, so that
crn IL'spo'.'~. to the x-~ranslations or tloors al levels A and B. Wu muy w, 11 I matrices s;F and SJF keep expanding in size. This overlay technique requires
'1 "< <> s~1flness matnx for a column as a partitioned array consistinp ol , 1·omputer core storage for only one substructure and n, framing levels.
s11h111utnces, as follows:
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
r,
Sym.
JI of undamped motion
0 2 _J __ _
________ 4L i }2 A (5)
K,-- -r,
_ El , Ir 0 1 kl
In which MFF is a diagonal matrix of lumped masses and AF(t) is the vector of
L3 0 I O 4L2 I
2L2 II II
--------~--------t----- k2 !literal forces. Damped or undamped story displacements may then be found
0 6L I O 6L I 12 j3 using either the normal-mode method or direct numerical integration from Chap-
0 -6L I O -6L I -12 12 k3 ter 4 or 5.
.i I }2 kI k2 j3 After time histories of story displacements DFhave been determined, other
k3
Items of interest may be calculated in a backward-substitution procedure. Start-
A B F ing at the lowest level and working upward , we compute joint displacements DA
from
whid1 .is a rearranged version of Eq. (6.2-1). Note that subscripts in Eq , ( t)
l11l..1·11 111 the sequence A , B, and F (6)
Wh~n stiffn~ss~s for beams and columns in substructure arc m.~r 111 hl
lhl· resultrng matrix 1s
e 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
Sym .J become displacements D 8 for the level e-
l above. Of course, the vector D0
Is null for the lowest substructure.
Sn- e Time histories of member end-actions for beams may be obtained by
Bl•11111s cont~ibute ter~s only to submatrix SM, but columns contribull' 11, nll 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;
till' su~n~utnces. Matnx Se augments stiffnesses from previous subst 111r111,
11 prcl~n1111~ 1ry lo reduction . From the modified tridiagonal method , Ntrlln yields:
ll'lh1L't1011 formulas for each substructure are (7)

S1t1 = Sm1 + s,IA TAB = Sou - SIJA S,J SAIi in which the end-action vector AM; contains a shearing force and a bending
I1 moment at each end. For columns, time histories of end-actions are found in a
s;,. = s,JF + s,1ATAF = s,JF - s,1As;;-J sA,.. I \I similar manner, but vector DM; contains two terms each from DA, Do, and DF.
s:,. Sn: + s,.i\ TM' = Sn: - Sri\ SA) SA/o' n 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
Whl'll'
hcams and columns could be added to any static actions existing at time t = 0.
T,1/1 - SA) SAIJ TA/o' = - SA) SA/o' (~ Turning now to three-dimensional multistory tier buildings, we must ac-
rnunt for the rigid lamina existing at each framing level. Figure 10.8 shows
A~ llw .~L·rics elimination proceeds from one substructure to the nt·xt, thi· 1111111 1 ~ubstructure e, consisting of a rigid lamina, the beams at level A , and the
nl 111ws 111.1d <.'ol1111111s of _types A and B remains equal to 211 ThL•H•lorr , 1
,
l'Olumns below . Recall that stiffness matrices for x-beams, y-beams, and z-
11•d1·'.111~·. d1spl11L'l''.1,'L'nls of .type 11 i.11 subs1r'.1t·turc (' to hL'L'mnc tyrw A 111 ~111
1 l olumns were giwn in Eqs. (9.4-1), (9.4-2), and (9.4-15) and that the last is
Hlllll lt11l ( I I . I his 1l'qtlll'l'S pl11c1ng 11111(11L'l'Nof' typl' JI 11110 posillonic nl IVI trnnsl'ormed lo I lit id hndy rnordinates . In addition, formulas for the transformed
, \ 11111•1 hJs, ( \) llll' itpplll'd IO l'lll'h SllhStllll'lllll' f'lll'II lll'W l 'Olllllhllllllll ll frt1I 11111ss inert in 1111d ~1111 y 11111d 11111trices appear in Eqs. (9 .4 17) and (9 .4- 19) .
 462 Substructure Methods
/ I
/ In this expression , TT is the transpose of the matrix in Eq. (~.4-13~,. and DF,.
/
/ I
/ I
Level A
x-beam --------- z-column
/ /
/ /
/ /
/ /
Level 8 / /
/ /
- ----,
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 11·1 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 rel1•11•m
point F to joint j at the top and joint k at the bottom. Therefore, Wl' haw
Sec. 10.4 Programs DYMSPF and DYMSTB 463
contains story displacements from levels A and B i~ the last s~x P?s1t1ons.
I/ / 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
1-------t
I
1?.
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
I beams and we can visualize a one-dimensional rigid body at each framing level.
I 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.
10.4 PROGRAMS DYMSPF AND DYMSTB
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
1----
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
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' ll•vi-l M
)
 464 Sec. 10.4 Programs DYMSPF and DYMSTB 465
Substructure Methods

d. Ground acceleration (in x direction) Preparation of dynamic load data for multistory plane frames appears in
e. Forcing function (piecewise-linear) Table 10.2. The dynamic parameters are the same as before, and the initial
5. Calculate story displacements conditions involve NS story translations and velocities at time t = 0. Under
a. Set up modal damping matrix applied actions we see the load parameter IAF, indicating whether story loads
b. Cal_culate story displacements for each time step exist or not. As shown in Fig. 10.6, each story load is a force An in the x
c. Wnte and/or plot story displacement-time histories e.
direction at framing level The ground-acceleration and forcing-function data
d. Find and write maximum/minimum story displacements in the table have been discussed before in conjunction with other types of
6. Determine. member end-actions (by substructures) structures.
a. Retrieve story information from auxiliary storage TABLE 10.2 Dynamic Load Data for Multistory Plane Frames
b. Calculate joint displacements DA
c. Cal_culate member end-actions for each time step Type of Data
No. of
Items on Data Lines
d. Wnte and/or plot member end-action time histories Lines
e. Fi~d and write maximum/minimum member end-actions Dynamic parameters I !SOLVE, NTS, DT, DAMPR
f. Shift elements of DA into D8
g. Repeat steps a through f for each story (bottom-to-top) Initial conditions
(a) Condition parameters 1 IND, INV
a
·1:ublc JO . I shows preparation of structural data for multistory plane frnr11 (b) Story displacements 00(1), 00(2), ... , DO(NS)
a
(c) Story velocities VO(l), V0(2), .. . , VO(NS)
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 Applied actions
DYMSJ .1 requ1r~s mm1mal mformation. Each line of that data gives the nurnh (a) Load parameter I IAF
a
N~JM of sequential _occurrences, followed by the repeated dimension of tlw hll (b) Story loads AF(!), AF(2), ... , AF(NS)
w1d1h BW (left-to-right) or the story height SH (top-to-bottom). Each of thr N
Ground accelerations
blocks of substructure data contains the superimposed story mass SMA
(a) Acceleration parameter 1 !GA
(t,o p lo-bottom) and member information (left-to-right). For the latter wt· 11r J (b) Acceleration factor I GAX
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 respct·t to th Forcing function
, ,,, axis. (a) Function parameter 1 NFO
(b) Function ordinates NFO K, T(K), FO(K)

TABLE IO.I Structural Data for Multistory Plane Frames • As required .

Type of Data No. of Example 10.3

Lines Items on Data Li
Figure 10.9 shows a two-bay, ten-story plane building frame having a rectangular layout.
l'mb 1cm identification I Beams in this structure are all steel rolled sections of size W 21 x 55, but the steel
Descriptive title
St,uc·turnl purumctcrs I columns vary, as follows: (Cl) W 8 x 31, (C2) W 10 x qO, (C3) W 12 x 85, and (C4)
NB, NS,E, RII ()
D11nc•n,,ionN
W 12 X 106. The length Lis 144 in., and the mass superimposed at each framing level
(11) Buy widths n
(b ) Story heights
NUM, BW is 0.06 k-s2 /in.
ft
NUM, SIi For this frame we have two dynamic loading conditions. The first is an atmospheric
Suh, tI Ul' llll C d11tu11 blast that causes the triangular force An in Fig. 10. lO(a) at each level , except t~~ top
(11) Stoiy lllllSN (where it is half as much). The second consists of rigid-body ground acceleration D81 in
I SMASS
(h) Ml•111lw1 info,mution the x direction that has the sawtooth shape in Fig. 10. IO(b). We analyzed the frame for
I llt'IIII I~ • NlJM, AX, /I
these two load ing conditions using Program DYMSPF, with a damping ratio of 0.10.
) ('nh111111N • NlJM, AY, t'I 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
• A~ 11•q11i11•d 11111xi n111111 1·x1111~H111 lrn lontl 1·nsc (h) is 4 \'i7 111. nt tinw t 2 . 16 s. After those
"N'I hl111~~ IIIIIXIIIIII, 1hr 1np1111\1·, il1111llll\h h1•1·11111w ot d11111p111p
41111 i;111>atr uutur u Muthocl11 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

C2 -0.1
10@ l (b)

t (s)
011 (in.) 0

- 1

-2
ly
7,7,:-- X

'}l
~~ - - ~I
Dy,

2L
~~
-3

4
(c)

Fl1t11n• I0. > Ten

1 Nl ory cxumplc 1hr Program l)YMS1'J1 Figure 1().10 Results for ten-story example: (a) atmospheric blast; (b) ground
m;cclcrntion ; (c) 1r11nslutinn 1),. 1 due to (a) und (h).
 Sec. 10.5 Component-Mode Method 469
488 8ul111tr ucturn M11thncl1

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 ccn11·1 111 111
(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 displaL·1·111r111~ 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 pict·cw1,,· 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 limitL·d 1111111
of vibrational mode shapes and a finite number of nodal displacement, 111 Ill Figure 10.11 Substructure e for component-mode method.
luting boundaries. The component-mode theory produces mass (or dy1111111I
coupling between the modal and nodal accelerations in a typical suhsl111d\l 2, ... , m" where m, is the number of retained modes. By normalizing the
Figure 10.11 shows a substructure C that is arbitrarily located within modal vectors with respect to the mass submatrix MAAl, we can state the
discretized continuum. Nodes of type A are indicated at interior position, . whll relationship
those of type Bare at boundary locations. Undamped equations of mot11111 l111 lh
(4)
substructure may be written as
In this expression, the symbol DNe represents a vector of m, normal coordinates,
Me De + Se De = Ae(t) (I
and «l>Ne is the normalized modal matrix.
whil'h is the same as Eq. (10.2-14). However, the expanded form is 1111w For nonzero boundary displacements (occurring statically), we have

[MM MAB] [DA] [SAA SABJ [DA] [AA I (5)

MBB Ds ( + SsA s,JB
M11A = A,, D11 I I
where
With boundary nodes restrained and zero loads, we can set up 111111 "'Iv~ lh TAB e = -S;leSABt (6)
1·1~l·nv1llue problem
Now let us define a transformation matrix Te as

111 wl11l'11 (1)11 is the ungulur frt•qut'ncy of motk I und <1>11 is tlw n111,·,p1111,ll11 T( = [<l>N T,111] (7)
0 I,, ,
111111k sl1111w. Frn this ll11111l·d vihrnt101111l 111111lys1s, tht• 111od11I indl'Xl'' 1111· 1
 470
Substructure Methods
Sec. 10.6 Component-Mode Method for Trusses 471

This operator relates the displacements in vector De to generalized displacemt•nl - ns - - n, -

111 a vector De, as follows:
M = I Me A(t) = L Ae(t) (J 5)

De = [~:l T{~:l
= = TeDe (Hu)
t• J (=I

Eqs. (15) must be expanded with zeros to become conformable for addition.
(=I

where n, is the number of substructures . Of course, the substructure matrices in

~:;~~ !:t !~:on~;,;~~:o~e~;ti~:::i~s DNe in the first part and D e in the scrnnd 8
Now the equations of motion in Eq. (14) can be solved for the displace-
ments in vector D by either the normal-mode method or direct numerical integra-
tion. Then displacements in vector DA e for each substructure are found using Eq.
De= Te.De (Hh) (5) as a back-substitution expression. In the next section we shall apply the
.Substitution of Eqs (8) int0 Eq (1) d · · · component-mode method to trusses, for which some additional complications
yields · · an premult1phcat10n of the result by 'I J
arise.
MeDe + SeDe = Ae(t) (1))
In expanded form, this equation is 10.6 COMPONENT-MODE METHOD FOR TRUSSES

[ ~,,
MoN
1\1NB] [~NJ +
Moo Do e
[ro;,,O _o
S88
J[DN]
D e
8
= [AN]
A e (10)
In earlier chapters of this book, we considered only axial deformations for
dynamic analyses of trusses. However, inertial and body forces also occur along
8
The mass submatrices in this equation (without the subscript e) are the members in transverse directions, which cause flexural deformations as well.
These influences are most significant in trusses composed of only a few mem-
MMN = ct>iMM cf>N = I m, ( I I ft I bers. We have found that the best approach for analyzing such structures is the
MNB = ct>iMAA TAB + ct>XMAB = MIN component-mode method [ 11, 12]. By this technique, a member with both axial
( 11 h) and flexural deformations constitutes a substructure. Treating the member as a
Moo = Moo + T1 MAB + MBA TAB + T1 MAA TAB (lld simply supported beam, we include a limited number of its exact vibrational
Also , the stiffness submatrices become mode shapes as flexural displacement shape functions.
Figure 10.12 shows a prismatic plane truss member i with local (primed)
SNN = «1>isM cf>N = ro m,2 ( I '11) 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
sNo = ct>XsAA TAo + «1>isAB = o = siN ( I h) shape functions in member i are taken to be the vibrational mode shapes of a
Son = Sno + T Xn SAB ( I l) simply supported beam [13]. These mode shapes will be superimposed on linear
und the applied-action subvectors are 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
AN= «1>rAA (I In) written in terms of modal and nodal displacements as
Ao = Ao + TXoAA ( I lh) u: = f Df (1)
:;• El(l, ( l )I"!, the symbol Im, den~tes an identity matrix of order m,. Also, .., in or
<I 12a 1s the spectral matnx for the substruct r · · "'•
1111g11l11r frequencies w2 2 2 . . u. ~· contammg sq11111(•11 ur (2)
Nllh11111trix c.lefincd ·ab~~:r~ .M. '.
Wm, m diagonal P?S1t1ons. The most i111prn l11111
11fi11R terms between accel~rati;~toEqi;n~ DJJ,), which represents dy11a1111, '""
T N JJ , In this equation the symbols DA; and D8; represent the displacement vectors
o HSSl'lll~le e~1uutions of motion for all substructures we simpl udd th
11111 1
1 in·s 11 PPl'm111g 111 Eq . (9), u~.ing the direct stiffness 111~ll10d. Thl'~l·~ull I DA, Ds,I
In lhi~ t•q1111111111
M I) I S
llw ll~~l·111hll·d 11111ldn·~
I)

Jill
A(t}
( Ill 1>At -=

I
0A2
D111

• • • I
D,11
Ds2
DJo
1)/,,1 I
(3)
 472 Substructure Methods 473
Sec. 10.6 Component-Mode Method for Trusses

y'

I v'

D~,
D~, __...
tr-----+.::...._....:.....;_
t- _ _0_ _ _ _ t
u'
-<J - - -- - x
,
j k ~

~ -- - -- - L~-----~

(a)

1~

V (b) (e)

(f)

Figure 10.12 Plane truss member with local and global axes. (c)

'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
(d)
s11h111ntrix is
Figure 10.13 Displacement shape functions for plane truss member.
f = <I> = [ 0 0 0 .. ·]
A sin 1r{ sin 21r{ sin 31r{ ... 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, tL·d prismatic beam. To keep them dimensionless, these mode shapcs 1111· 11111
.o
y' a2
1 (7)
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 L 2 ag2 ;
The first submatrix in B; has the resulting form
0 g OJ
.. .l
t ~, 0
I ~ 0 g I 7T2 [ 0 0
8 Ai = d;f" = Y L2 sin 1rf 4 sin 21rf (8)
9 sin 31r(
'lht•sl' li11l'11 r shupl· !'unctions result froin unit displuccments of' 1>;11 th ro11ph /J 111
IINl11dic111t•d i11 Figs. 10. Uk) nnd (!') . und the second is
S1111i11 diNplllL'L'llll' llt n•l11tio11ships f'tu· lhi s L'lt•111t•11t 11111y lw st11tl•d nil
II, cl, I' 111 , II,, I , I'll/ cl, r,, ,
I
I o
llllllll,
o I ('))
 474 Substructure Methods Chap. 10 Sec. 10.6 Component-Mode Method for Trusses 475

The stiffness matrix Sf in local directions for member i has the known frn 11 1 2
2 Sym.l
s: = L BrE B; dV = oJ
[soM snB i
(tm
,
M,m -
_
Iv ,,r,r, dY -
1
_ pAL 0
6
[
~ 0 2
(17)

1 0 2 ;
In this equation the stress-strain matrix E is
which is the same as matrix M' in Eq. (3.5-32).
(I f l
Also, equivalent modal and nodal loads in local directions are calculated
where 12 is an identity matrix of order 2. The first submatrix in Sf becontl'H as

-sMi -- f BT
v
7T4EJ,
Ai E BA; dV = - -
2L3 [~ 16
0 81
Sy1 A; Ct) = r r1 b'Ct) dL
Ji
= [A~J
Aa ;
Here the vector b' (t) contains body forces (per unit length) in the x' and y'
(18)

directions, as follows:

_ (pAL
- -2-w"',
2 )
; (I 'I
b'(t) = [by'bx'] (19)

and the second submatrix is found to be The first subvector in Af (t) has the form

,
SBBi -
_
r BB;E Ba; dV = -L
)v
1 EA

[
1
0 0
-1 0
0 0
Sym.]
1
0 0 ;
(II) A,; - i ~Tb'(t) dL - 1[:~f::]~, dL
(20)

which is the same as matrix K' in Eq. (3.5-25) . and the second is
. Next, the consistent mass matrix Mf (in local directions) for memhl't 1 111

l [g ~b;i::1
obtamed from
= l f];b'(t) dL = (21)
Mf = ( pf f dV =
Jv
1
[M~ M~J
MBA Mos ;
( Ill
An;

thy' i
dL

The first submatrix in Mf is For local (or member) axes , the undamped equations of motion for small
displacements of the plane truss member are
Mf Df + Sf Df = Af (t) (22)
As before, the3mbol I,,,, represents an identity matrix of order 111,. < >11 111 By rotation of axes for the parts of type B, Eq. (22) may be transformed to global
submatrices in Mf are (or structural) axes to become
M;D; + S;D; = A;(t) (23)

~
I 0 l

T\it'ill/I - f v p<l>'I' f 11 r/V - pAL [

7T' ()
1/2
I/ l
()
()
1/2
1/J
(If,)
The displacemcnls D/ and.accelerations fi; in Eq. (22)_are related to the corre-
sponding vectors 1)1 ond 1)1 in Eq . (23) t,y the expressions

'
n; ft,1>1 f>; ft,n, (24)
1111d l111h1•1111 11·l11ti1111 -.hlpH tlu· 1011111111111111,mM 111111Hl11111111lHlll 11111111 ,i 111
 476 Sec. 10.7 Programs COMOPT and COMOST 477
Substructure Methods Chap, 10

in which
R; = ( 1~) m m
s = LS; A(t) = L A;(t) (35)
Note from the form of matrix R; that only the nodal displacements D i=I i=I
transformed fr~m structural to member directions. (Displacements D 81 th;.:11;":, where m is the number of members. The equations of motion for the structure
'f?n4
lS
. · E q. ( ~,)
are shown m the x and y directions in Fig. 10.12.) Submatn·x R'811n can now be solved by the normal-mode method or by direct numerical integra-
tion. Either approach may be preceded by a vibrational analysis, from which a
damping matrix can be established for the whole structure, as before.
Rn;= [R 0 R;
OJ ( '" I
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
where 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
R; = [ _ Cx Cy] are still the sine functions sin 7Tg, sin 27Tg, sin 37T{, and so on. Definition of a
Cy Cx i (.! II
principal plane in space may be aided by using a third point p in addition to
Dire~tio~ cosines in the rotation matrix R; were defined in E s (3 5-24) 11 points j and k, as explained in Sec. 6.4.
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~
- 'T- ' 10.7 PROGRAMS COMOPT AND COMOST
S; = R; Sf R; ( 'HI
M l
= RTMiR'.
l l l
( '')I
Now we briefly discuss Programs COMOPT and COMOST for dynamic anal-
yses of plane and space trusses by the component-mode method of the preceding
A;(t) = RT A; (t) ( 1111 section. These programs calculate responses to initial conditions and piecewise-
The results of these operations are as follows: linear forcing functions that may be either applied actions or translational ground

s:]
accelerations. Using the normal-mode method, we determine axial forces and
c; bending moments in the members, as well as time histories of modal and nodal
c2y
Son,= R.];,Snn,R.n, = EA CxCy responses.
[ -c;
L c2X ( 111
Example 10.4
-cxCy 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
-::12]
Cx/3
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
• . • I 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
(I - f)(c,.bx· - cyby,) I
A - R" TA'
Ill - /Ji
-1
01 -
1.
(I - g)(cyb,,
•
+ c,by•)
g(c,bx· - Cy by,)
' dl
( I\)
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
{(cyb, • + C\ bv,)
. I 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
Noll· tlu.11 submutrix. M11111 M f1111 is invariant with rotution axes or dimensionless time , ,.,. is time r divided by the period of the first mode for the truss
I Afl~r tlw rotat,011 or nxcs lntnsf'ornrntion f'or cuch nw111h<.1r is ~·0111plrt I solution without lk•x11n•. The curves in H11, 10. l<i 111l' lnhck•d with encircled numbers 0 ,
I I(' l'ljlllll IOIIS of Ill()( ion for ( ht• whok• H( 1111.'ltll'l' 11u1y lw USSl'lllhlt•:, (O t11il11l1'.' I, ... , 'i t11 lntlil'11ll' how 1111111y lkx11rnl 111odl'S 1m1 1111•111h1·1 llll' i11d11tl1·d i11 thl· nnnlytic-nl
Mi> I Sh A(() 111od1•I Wr 11111 Hr1• 1h111 1111• Hol11tu111 with 1111 lh•K111r I~ q111h• 1lllh•11•111 1111111 lhos1• wi th
( 11)
478
Substructure Methods Chap. 10 479
Sec. 10.7 Programs COMOPT and COMOST
y

\
\
\
\
\
\
\
\
\
3L

(a) (b)

2
----x

~I.-3L-.I
Figure 10.14 Plane truss with two members.

TABLE 10.3 Dimensionless Frequencies (c) (d)

for Two-Member Truss

Figure IO.IS Mode shapes for two-member plane truss: (a) mode 1; (b) mode 2;
Mode w lf< (c) mode 3; (d) mode 4.
Mode w lf<

I 0 .204 7 3.034
2 0.117 8 3. 163
J 0.7% 9 1.628
11 I , I 8() 10 ~. 146
~ I .MH 11 t ~·, ,
h J II I l :l H )ti/
 400
•;11lin11111ill1111 MulhocfN
Sue. 10, / 1110111111111 C:O MOI' I 1111d COMOS1'
481
y

2.5
I p

4
,
D;t-~-t 3 5
2.0

I TL

1.5
l 6--------....:o----...,__,..;u-----s
4 6

1.0 '
I
'/.
Figure 10.17 Plane truss with np panels.
I
0.5
I
I
I

0
1.0 1.5 ;,.o R
Figure 10.16 Displacement time histories for D t.
1.6

1.5
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 1.4
Example 10.5
1.3
Figure I0. I7 illustrates a truss with an indefinite number of members, having u Nh•p 11111 1
P applied in they direction at the center of member 1. By increasing the numbL·r ol 1111111 I
1.2
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 compml·d 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 1- 0.9
incrcascs. Thus, the flexural deformations in the members have little cffcl'I 011 1hr 1111111
rusponscs when the number of mem bers becomes large.
l•'l uun• J0. 18 Displacement ratio versus number of panels.
 482
!h1h11t1uut11111 M11tlu1d1

R R NC S

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, Act1dl•111h 1111
London, 1977 .
\ , Wcuvcr, W. , Jr., and Yoshida, D. M., "The Eigenvalue Problem for B1111d1•d 1\11111 I
ccs," J . Comp. Struct., Vol. 1, No. 4, 1971, pp. 651-664.
'1 . Weaver, W., Jr. , and Nelson, M. F., "Three-DimensionaJ Analysis of Ti1•1 ll11tltl Notation
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. 1. MATRICES AND VECTORS
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. Symbol Definition
11 , Wl·uvcr, W. , Jr. , and Loh, C. L., "Dynamics of Trusses by Componc111 Mn1h 0 Null matrix
Mclhod," ASCE J. Struct. Eng., Vol. 111, No. 12, 1985, pp. 2526- 2575. A Action vector (also coefficient matrix)
12. Loh , C . L., "Dynamics of Trusses by Component-Mode Method," Ph .Jl ,/11 B Strain-displacement matrix
.1·1•rratio11, Department of Civil Engineering, Stanford University, May I 98'1 C Strain-stress matrix (also damping
11 , Timoshenko, S. P., Young, D. H., and Weaver, W., Jr., Vibration Proli/1•1111 111 matrix and constraint matrix)
l:11t,1i11t•t•rl11g, 4th cd., Wiley, New York, 1974.
D Displacement vector
E Stress-strain matrix
F Flexibility matrix
G Constraint matrix
H Characteristic matrix (also Householder
matrix)
I Identity matrix
J Jacobian matrix
K Element stiffness matrix
L Lower triangular matrix
M Mass matrix (also concentrated
moments)
Concentrated force vector

483
484 Notation Notation 485

Symbol Definition Symbol Definition

Q Factor in QR algorithm b Body
R Rotation matrix (also factor in QR d Damped
algorithm) Element
e
s Stiffness matrix Forced
T
f
Transformation matrix (also tridiagonal g Ground
matrix) Index
u Upper triangular matrix j Index
V Eigenvector matrix
k Index
X Vector of unknowns
y C Index
Vector of unknowns Number
m
z Vector of unknowns
n Number
A Spectral matrix
p Working (or reference) point
<I> Eigenvector matrix
q Working (or reference) point
b Body force vector for element
I' Radial direction
d Linear differential operator for strain
s Structure
displacement relationships
X x direction
e Unit vector
y y direction
f Interpolation function matrix
z z direction
Unit vector
j Unit vector
k Unit vector 3. SIMPLE VARIABLES
p Nodal load vector for element
Symbol Definition
q Nodal displacement vector for elc11w111 A Area
u Generic displacement vector for elc111rnt Constant
B
C Constant
2. SUBSCRIPTS FOR MATRICES AND VECTORS D Displacement
Symbol Definition E Young's modulus of elasticity
A Nodal displacements eliminated (<1/.w1 G Shearing modulus of elasticity
attached) I Moment of inertia (second moment
8 Nodal displacements retained (al.1·0 of area)
body and boundary) J Polar moment of inertia
F Free (also floor or fram ing leve l) L Length
/, Lumped M Moment
M Member p Force
N Nornrn l rnordinntes R Radius
I' J>ri11l'ip11I rnordin111t•s 'f' Period
N l{~•Nll llllll'd (} Strnin l'llorgy
II l111tlul \' l 1111r11t 111l l'lll'I HY
 486
Notation Notation 487

Symbol Definition Symbol Definition

w Work a Rotation or angle (also Hilber constant)
X Generalized action Magnification factor (also Newmark
a /3
Constant (also acceleration) constant)
b Constant Shearing strain (also damping ratio and
)'
C Newmark constant)
Constant (also damping constant)
d Constant (also displacement) 5 Increment
e Base of natural logarithm E Normal strain
f Interpolation function (also frequency , ( Dimensionless coordinate
cycles/ sec) rJ Dimensionless coordinate
h Thickness () Rotation or angle (also Wilson constant)
Index for ... (also v=T) A Direction cosine (also eigenvalue)
j Index for . . . Frequency coefficient
µ,
k Index for ... (also spring constanl) V Poisson's ratio
e Index for ... (also length) t Dimensionless coordinate
m Number of ... (also mass) 'TT 3.1416 ....
n Number of degrees of freedom (also p Mass density
damping parameter) a Normal stress
p Action at element node T Shearing stress
q Displacement of element node
</> Curvature
r Radius (also cylindrical coordinate) !/; Twist (dOx/ dx)
s Segment length w Angular frequency (also angular
t Time velocity)
u Translation in x direction
V
Translation in y direction
w 5. PROGRAM NOTATION
Translation in z direction
X
Cartesian coordinate
Symbol Definition
y Cartesian coordinate AO( ) Initial accelerations of nodes
z Cartesian coordinate (also cylindric11l AOP Initial acceleration of moving load
coordinate) Actions applied to bodies
AB( )
AE( ) Actions at element nodes
AF( ) Actions at free nodes (also actions at
4 . GREEK LETTERS floors or framing levels)
ALPHA Hilber parameter a
Symbol Definition AM( ) Actions at ends of members
~
...,.
Increment AN( ) Actions in normal coordinates
Summation AR( ) Support reactions
<I>
FU1K·lio11 or ,nod<.· AS( ) Actions at structural nodes
u A11~11l111 I 11•q111•111 y AX( Cross Sl'l'lionnl nn·as A,
 488
Notutltm 489

Symbol Ocfi11ltlo11 Definition

BETA Symbol
Nl•w111111•k puramclcr f3 ID( ) Displacement indexes
BI( ) Body i11crtit1s IBO( ) Element numbers for output of stresses
BL1 , BL2,. Jntcnsilics of line loads !G A Indicator for ground accelerations
BM( ) Body masses IML Indicator for moving load
BS1,BS2, .. . Intensities of surface loads IN( ), JN( ) Indexes for nodes of elements
BV1,BV2, . . . Intensities of volume loads IPL Indicator for plotting
BW Bay width IPS Indicator for plane stress or plane strain
CME( , ) Consistent mass matrix for clc111t'11I IR,IC Row and column indexes
CMS ( , ) Consistent mass matrix for strnt·l111, IRO Indicator for eliminating rotations
CV( ) Characteristic values (eigenvalul'N ) !SOLVE Indicator for method of solution
CX,CY, CZ Direction cosines c.,, cy, and c, IWR Indicator for writing
DO( ) Initial displacements of nodes JB( ' ) Body-node numbers
DAMPR Damping ratio JNO( ) Node numbers for output of
DBO( ) Initial displacements of bodies displacements
DE( ) Displacements of element nodes Loading number
LN
DF( ) Displacements of free nodes (also Number of bodies (also number of
NB
displacements at floors o r 1'1·1111111111 bays)
levels) NBID Number of bodies with initial
DM( ) Displacements at ends of membt•rN displacements
DN( ) Displacements in normal coorcli,11111·11 Number of bodies with initial velocities
NBIV
DR( ) Displacements of restraints Number of columns
NC
DS( ) Displacements of structural nodes Number of degrees of freedom
NDF
DT Duration of time step /:l.t Number of elements
NE
E Elasticity modulus Number of elements with line loads
NEL
E l ,E2, . Elasticity constants Number of element nodes
NEN
EL( ) Element lengths Number of elements for output
NEO
F( ) Frequencies (cps) Number of elements with surface loads
NES
FO( ) Function ordinates Number of elements with volume loads
NEV
G Shearing modulus Number of function ordinates
NFO
GAMMA Newmark parameter y Number of joints (or nodes) on a body
NJ
G AX, GAY, G AZ Ground acceleration faclors for 1, I' , Number of loaded bodies
NLB
and z directions NLN Number of loaded nodes
II Thickness NLS Number of loading systems
l,J, K, L Indexes NMODES Number of modes
IA('
Indicalor for imposing axial co11s1111l111 ~ NN Number of nodes
IAF lndicalo r for actions al lloors NNA Number of nodes of type A
490

Symbol l)()llnlllon
NNI> N11111lw1 of nodal displucc11K•111s
NNF N111uhcr of nodes of typu F
NNII > Number of nodes with initial
di spl acements
NNIV Number of nodes with initial wlrn llh
NNO Number of nodes for output
NNR Number of nodal restraints General
NRL( Nodal restraint list
NRN Number of restrained nodes References
NS Number of stories
N'l'S Number of time steps
NUM Number of repetitions
OM EGA( Angular frequencies w
p Moving load
Pl II ( ) Eigenvectors ct> (mode shapes)
PR Poisson's ratio
R( ) Rotation matrix
TEXTBOOKS ON STRUCTURAL DYNAMICS
I{( '( ) Radii of gyration of rigid bodies with (CHRONOLOGICAL ORDER)
respect to centers of mass
RI 10 Mass density p 1. Rogers, G. L. , Dynamics of Framed Structures, Wiley, New York, 1959.
SH( Element stiffness matrix 2. Norris, C. H. , et al. , Structural Design for Dynamic loads, McGraw-Hill, New
SIi Story height York, 1959.
SS( Structural stiffness matrix 3. Biggs, 1. M. ,Introduction to Structural Dynamics, McGraw-Hill, New York , 1964.
SX ,SY, . Stresses 4 . Hurty, W. C., and Rubinstein , M. F., Dynamics of Structures, Prentice-Hall, En-
glewood Cliffs, N. J., 1964.
T( ) Times
5. Lin, Y. K., Probabilistic Theory ofStructural Dynamics, McGraw-Hill, New York,
TIME Time 1967.
VO( ) Initial velocities of nodes 6. Przemieniecki, J. S., Theory of Matrix Structural Analysis, McGraw-Hill, New
VOP Initial velocity of moving load York, 1968.
VHO( ) Initial velocities of bodies 7. Rubinstein, M. F., Structural Systems-Statics, Dynamics, and Stability, Prentice-
VN( ) Velocities in normal coordinates Hall, Englewood Cliffs, N. J., 1970. ·
VS( ) Velocities of structural nodes 8. Fryba, L., Vibration of Solids and Structures under Moving loads, Noordhoff,
Groningen , The Netherlands, 1972.
X( ),Y( ), Z( ) Nodal coordinates
9. Fertis, D. G. , Dynamics and Vibrations of Structures, Wiley, New York, 1973 .
XC'( ),YC( ), ZC( ) Coordinates of point c 10. Clough, R. W. , and Penzien, J. , Dynamics ofStructures, McGraw-Hill, New York ,
XC'.I( ),YCJ( ), ZCJ( ) Components of offset vectors 1975.
XC'K( ),Y CK( ), ZCK( ) Components of offset vectors 11. Belytschko, T., Osias , J. R ., and Marca!, P. V., Finite Element Analysis of Transient
X I( ) Torsion c:onslanls /, of cross scc:lions Nonlinear S1r1U'/11ml ll<•ltGvior , ASME, AMD , Vol. 14, 1975.
YI( ). ZI( ) Sl•t·wul 111011wnts of arcu /1, 1111d I of 12. Bathe, K . J,, 1111d WIINon, E. L ., Numerical Methods in Finite Element Analysis,

l ' I OHM ~1·1·1io11s

Prcnlicc-llnll , I111plnv1111cl C'lilTN, N. J., 1976 .

491
 492
{1111111r11l llnftunm 111 493
13. Blevins, R. I)., Floll' l11rl11, ,•rl I l/1111tlr111.,·, V1111 Nostrnnd lfrlnliold , NI'\\ , 1
1977. TEXTBOOKS ON FINITE ELEMENTS
14. Simu, E., and Scanlan, R. II ., Wind HJ)i•ct.1· 011 Strt1ct11r,•s, Wiley, New Y111 ~, I •J (CHRONOLOGICAL ORDER)
15. Meirovitch , L., Computational Methods in Structural Dy11a111ic·.1·, Sljlliufl
Noordhoff, Alphen aan den Rijn, The Netherlands, 1980. 1. Przemieniecki, J. S. , Theory of Matrix Structural Analysis, McGraw-Hill, New
16. Craig, R. R ., Structural Dynamics, Wiley, New York, 198 1. York, 1968.
17. Paz, M. , Structural Dynamics, 2nd ed., Van Nostrand Reinhold , New Ylll ~. l'Jk 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,
TEXTBOOKS ON VIBRATIONS 1972.
(CHRONOLOGICAL ORDER) 4. Martin, H. C. , and Carey, G . F., Introduction to Finite Element Analysis, McGraw-
HiU, New York, 1973.
1. Rayleigh, J. W. S., The Theory of Sound, Dover, New York, 1945. 5. Norrie, D. H. , and de Vries, G., The Finite Element Method, Academic Press, New
York, 1973.
2. Den Hartog, J.P., Mechanical Vibrations, 4th ed. , McGraw-Hill, New YOI~, l•J~h
6. Strang, G. , and Fix, G. J., An Analysis of the Finite Element Method , Prentice-Hall,
3. Myklestad , N. 0., Fundamentals of Vibration Analysis, McGraw-Hill, Nm• , 111h
1956. Englewood Cliffs, N . J., 1973. .
7. Gallagher, R. H., Finite Element Analysis Fundamentals, Prentice-Hall, Englewood
4 . Jacobsen, L. S. , and Ayre, R. S., Engineering Vibrations, McGraw-lltll , N
York, 1958. Cliffs, N. J. , 1975.
8. Bathe, K. J., and Wilson, E. L., Numerical Methods in Finite Element Analysis,
5. Bishop, R . E. D. , and Johnson, D. C., The Mechanics of Vibration, C11111l11hl11
University Press, London, 1960. Prentice-Hall, Englewood Cliffs, N. J. , 1976.
6. Tong, K. N., Theory of Mechanical Vibration, Wiley, New York, 1960 . 9. Hinton, E. , and Owen, D.R. J. , Finite Element Programming, Academic Press,
London, I 977.
7. Church, A.H. , Mechanical Vibrations, 2nd ed. , Wiley, New York, 19(,l
10. Desai, C. S . , Elementary Finite Element Method, Prentice-Hall, Englewood Cliffs,
8. Crandall , S. H., and Mark, W. D., Random Vibration in Mechanical S1•1t, 111 N. J., 1979.
Academic Press, New York, 1963.
11. Cheung, Y. K., and Yeo, M. F. , A Practicallntroduction to Finite Element Analysis,
9. Bishop, R. E. D., Gladwell, G. M. L. , and Michaelson, S., The Matri, A1111/\ 11 Pitman, London, 1979.
of Vibra,ion , Cambridge University Press, London, 1965.
12. Hinton, E., and Owen, D.R. J. , An Introduction to Finite Element Computations,
10. Chen, Y., Vibrations: Theoretical Methods, Addison-Wesley, Reading, ~111 ~ Pineridge Press, Swansea, Wales (United Kingdom), 1979.
1966.
13. Owen, D. R. J. , and Hinton, E., Finite Elements in Plasticity, Pineridge Press,
11. Anderson, R. A., Fundamentals of Vibrations, Macmillan, New York, I% I
Swansea Wales (United Kingdom), 1980.
12. Vernon, J.B ., Linear Vibration Theory, Wiley, New York, 1967.
14. Cook, R.' D., Concepts and Applications ofFinite Element Analysis, 2nd ed. , Wiley,
11. Vierck, R. K., Vibration Analysis, International Textbook, Scranton, Pu. , J 11(, I New York, 1981.
14 . llaberman, C. M. , Vibration Analysis, Charles E . Merrill, Columbus, Ohru , l!JriN 15. Becker, E. B., et al., Finite Elements (five volumes) , Prentice-Hall, Englewood
15. Thomson, W. T., Theory of Vibration with Applications, Prentice-Hall , Englrn 11, ut Cliffs, N. J., 1981-1984.
Cliffs, N. J., 1972.
16. Bathe, K. J., Finite Element Procedures in .Engineering Analysis, Prentice-Hall,
l <i. Timoshenko, S . P., Young, D. H ., and Weaver, W. , Jr. , Vibration Prohl,•1111 /11 Englewood Cliffs, N. J. , 1982. .
H11,:i11eering, 4th ed., Wiley, New York, 1974. 17. Huebner, K. H. , The Finite Element Method for Engineers, 2nd ed., Wiley, New
17. Meirovich, L. , Elements of Vibration Analysis, McGraw-Hi ll , New York, Jilt, York, 1983.
IH. Newland, D. E., An Introduction to Random Vibrations and Spectral 1\11,1/\ 111 18. Weaver, w., Jr., and Johnston, P. R. , Finite Elements for Structural Analysis,
Longmans, London, 1975. Prentice-Hall, Englewood Cliffs, N. J., 1984.
I9. Tse , F S., Morse, I. E., and Hinkle, R. T. , Mechanical Vibrations 'J'h,·1111• ,111 I 19. Segerlind , L. J., Applied Finite Element Analysis, 2nd ed., Wiley, New ~ork, ~985.
Applications , 2nd ed., Allyn and Bacon, Boston, 1978. 20. Zienkiewicz, O. C . , The Finite Element Method, 4th ed., McGraw-Hill , Maiden-
head, 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, 1•11
glewood Cliffs, N. J. , 1966.
4. Hall, A. S., and Woodhead, R. W., Frame Analysis, 2nd ed. , Wiley, New York
Appendix A
1967.
5. Willems, N. , and Lucas , W. M. , Jr., Matrix Analysis for Structural Engin1'1'11
Prentice-Hall, Englewood Cliffs , N. J. , 1968.
Systems of Units
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
and Material Properties
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. A.1 SYSTEMS OF UNITS
11. Vanderbilt, M. D. , Matrix Structural Analysis, Quantum, New York, 1974 ,
12. McGuire, W., and Gallagher, R.H., Matrix Structural Analysis, Wiley, New Y111k The two most commonly used systems of units are the International System (SJ
1979. units) and the United States Customary (US units). The first of these is called
13 . Weaver, W., Jr., and Gere, J. M ., Matrix Analysis of Framed Structures, 2nd Id an absolute system because the fundamental quantity of mass is independent of
Van Nostrand Reinhold, New York, 1980. where it is measured . On the other hand, the US system has force as a funda-
14. Meyers, V. J., Matrix Analysis of Structures, Harper and Row, New York , 1'1"1 mental quantity. It is referred to as a gravitational system because the unit of
15. Holzer, S. H., Computer Analysis of Structures, Elsevier, New York, 19H~ 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 dy-
namics 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'11t•t·,

lhlll t'Olllt'N I 111111 lltt 111111111111 1111111N lcH('l'/ncn•krnlion ,

496 Systems of Units and Material Properties Sec. A.2 Material Properties 49:

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
(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.

TABLE A.I Conversion of US Units to SI Units

A.2 MATERIAL PROPERTIES
Quantity US Units x Factor = SI Units
To analyze solids and structures composed of various materials, we need to
Length inch (in.) 2.540 X 10- 2 meter (m) know certain physical properties. For structural dynamics, the essential material
Force kilopound (kip or k) 4.448 kilonewton (kN)
Moment kip-inch (k-in.) 1.130 X 10- 1
properties are modulus of elasticity E, Poisson' s ratio v, and mass density p.
kilonewton · meter (k.N · ru)
Stress kip/inch2 (k/in. 2 or ksi) 6.895 X 103 kilopascal (k.Pa) Table A.3 gives these properties in both US and SI units for some commonly
Mass kip-sec2/inch (k-s2/ in.) 1.751 X 102 megagram (Mg) 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 TABLE A.3 Properties of Materials•
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 Modulus of Elasticity E
Poisson's
Mass Density p
units for force, length, and time appear in Table A.2. For instance, in SI( I ) wr Material
k/in. 2 GPa Ratio 11 k-s2 /in. 4 Mg/m3
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 Aluminum 1.0 X 104 69 0.33 2.45 X I0- 7 2.62
acceleration ii in millimeters per second squared (mm/s2 ), and so on. Brass 1.5 X 104 103 0.34 8. 10 X I0- 7 8.66
Bronze 1.5 X 104 103 0.34 7.80 X I0- 7 8.34
TABLE A.2 Consistent Systems of Units Cast iron 1.4 X 104 97 0.25 6.90 X 10- 7 7.37
Concrete 3.6 X 103 25 0. 15 2.25 X 10- 7 2.40
System Force Length Time Magnesium 6.5 X 103 45 0.35 1.71 X 10- 7 1.83
Nickel 3.0 X 104 207 0.31 8.25 X 10- 7 8.82
(I) newton millimeter second Steel 3.0 X 104 207 0.30 7.35 X 10- 7 7.85
SI (2) kilonewton meter second Titanium 1.7 X 104 117 0.33 4.20 X 10- 7 4.49
(3) meganewton kilometer second Tungsten 5.5 X 104 379 0.20 1.80 X 10- 6 19.2

(!) pound foot second "Numbers in this table are taken from J. M. Gere and S. P. Timoshenko, Mechanics of
us (2) kilopound inch second Materials, 2nd ed. , Brooks/Cole, Monterey,. Calif., 1984.
(3) megapound yard 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 gig111H·wtons 1wr squ1111• 1111 I
 Sec. B.1 Inverse Iteration 499

on the left-hand side, we find that the reverse iteration converges to the funda-
mental mode. To simplify notation, let
X; = <I>; (4)
Then Eq. (3) beco mes
(5)

Appendix B 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
Eigenvalues of ones. Thus,
(X 1) 1 = {l , I , I , ... , 1} (6)
and Eigenvectors 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 = (A 1).(X 1) 1 (8)
This expression is only an approximation, unless the estimated mode shape
8 .1 INVERSE ITERATION 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), .
All eigenvalue solution routines are iterative because we seek the roots ot tho That is,
<'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 (A) = (}j), (1 :s j :s n) (9)
I I (XJ,
problem to be solved. Thus, we have
S <I>; = WTM <I>; (i = 1, 2, . . . , n) (I) 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 for11•111tl)
ifl'f'otion of Eq . (1) would involve substitution of trial vectors for <I>, 011 lh 1
(X,h = (A,), (Y1)1 (10)
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 in which the normalization constant is (11. 1) 1 itself. This procedure is repeated
wt·lor <1>11 • J lowever, to extract the smallest angular frequency first , we muNI 1111 until the eigenvalue A1 and its associated eigenvector X, are determined to some
" '"''I',\'£' (or inverse ) iteration. For this second approach, let A; = I / "'i 1111~
specified accuracy.
fnclor the stiffness matrix S by the modified Cholesky method [2 , 3], as folloWM In the kth iteration, the recurrence equations for the steps described above
s = UTD u () are:

In this lype o f factorizatio n, D is a diagonal matrix; and U is upper trinn itlllM I. Calculate vector (B 1)k as
with vulues of unity in diagonal positions. Substituting Eq . (2) and ,viI/ A
11110 Eq . ( I ) produces (11 )
Ur I) U A1<l>1 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 Inverse Iteration 501

3. Find the new trial vector (X1)k+ 1 to be 2 .72

M = m 1 3.28 (b)
(I J) [
0 0
for which m = pAL/6.
where (A 1h is the largest term in vector (Y1)t. To check convergence of thl' In accordance with Eq. (2), we factor matrix S to obtain
eigenvalue, we use the expression

~
60

( 14)
D = s[l.3 0 0.2°647 ] (c)
0 0 1.000

_[l
111 which nd is the number of significant digits of accuracy desired. and
Bathe [4] recommends calculation of (wr)k = 1/(A1h from the Raylcixh
t//10tie11t - 0.2647 -0.3529]
U = 0 1 1.333 (d)
( 2) _ (XThS(X1h
( I~ ) 0 0 I
WI k - (XT)k M(X1h
As the starting eigenvector, assume
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 (X 1) 1 = {I , 1, 1} (e)
l'l~l·t1vcctor. However, improvement of eigenvector convergence can be altat11l·d In the first iteration, we evaluate (B 1) 1 from the last of Eqs. (4), as follows:
wtlh spectral shifting, as described later in this appendix.
After the fundamental mode has been determined, it is usually elimi1111h·d (B1)1 = M(X1)1 = m{3 .72, 4.28, 3.28} (f)
I111111 the eigenvalue equations by the process of deflation [l , 4]. For this ll'l'h Then solve for (Y1) 1 from Eq. (5):
11iq11l'. we express the orthogonality of the eigenvectors <1> 1 and <l>i with n·~pr, I
lo lhl' mass matrix M, as follows: m
(Y, ), = - {8.000, 23.12, -2 .426} (g)
s
<l>TM <l>i = 0 (i = 2, 3, ... , n) ( It!)
Normalize this vector with respect to its largest term to find the new trial vector
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 1
1111llk becomes dominant and also may be calculated by inverse iteration I hi (X,)z = - () (Y1 ) 1 = {0.3460, 1.000, -0.1049} (h)
A1 I
dt•l1111ion iteration sequence is repeated to extract as many modes as dl'"tt ~d
In this iteration the first approximation to the eigenvalue is
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 m (1')
(A 1) I = 23.12-
<,'111111 Schmidt orthogonalization, as shown in Ref. 4. s
which is the largest term in vector (Y1)1.
E111111111lt• H. I
Results of successive iterations are listed in Table B.l. We see that convergence
N11w w1· sh11II apply inverse iteratio n to extract the fundamental mode from lh1• 1•l1t II (to fo ur significant digits) for the reciprocal A1 of the eigenvalue wf occurs in six cycles
vu ltH· pt 11hll·m I'm the plane truss in Fig. 3. I I (a) (sec Sec. 3.5). For thi s truss 1h1· NI ti In of iteration. However, the eigenvector X1 requires another cycle to attain the same
111111t Ix wns found to he nccuracy. Their final values are
m 2 1 s
0.36 A, = 22.22- w, = - = 0.04501 - (j)
0.481 s A, m
0. 1(1 0.48 t ,1 1111d
0.48 I .M X, {0.2314, 1.000, 0.2472} (k)
w ltr1r ,1 / •/\ / / Al1.o, tlt1• 1•011Mis h•111 11111 ~11 11111t1i x IN wl11l'11 1111• 1111• ~111111· 11~ 1lt11~1· 111111111 hy Pro~rnm Vllll'T 111 Ex11111pll' l •I
 Sec. B.1 Inverse Iteration 503

Spectral Shifting

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
"<I" N close (but not equal) to WT, which is the eigenvalue to be calculated. Then
r-
~
~
- 0 r-
<">0"<1"
NON
N
N
subtract aiM ct>i from both sides of Eq. (1), and divide by wr -
ai to obtain
8: c:;;....;c:;; c--i
(17)
I N
S* A; «l>i = M «l>i
where
"<I" N
~ ;;; 8~ N
'D
8: NON
o....; o
N
c--i
N
s* = S - aiM A;=---
wT- ai
(18)
I

By this manipulation , the dominance of A; in Eq. ( 17) is greater than that of Ai

~
'D
~8~ °' in Eq. (3), because in the second of Eqs. (18) we see that A;--" oo as ai--" wr.
~
NON
N
N After the spectral shift, inverse iteration proceeds as before, except that the
V)
8: 0 ....;c:;;
I
c--i
N factorization of s* is expressed as
s* = u*To* u* (19)
..
~
"""
<"l
<"l
N
§ 0
V)
"""
N
<"l
N
Also, the eigenvalue wr becomes
""" 8: o....; o !"i
2 1
I N
W· =-*+a· (20)
I Ai I

r-
°' § °' which need only be computed once (after convergence).
~
~
'D
"<I"N r- Spectral shifting may be used to aid the following tasks:
N N N
<"l
8: c:;; ....;c:;;
I
!"i
N

1. Extraction of modes near an expected resonance (ai--" n), where !l is the

angular frequency of a periodic forcing function.
0

N
N
~

8:
~§s°'
o....; o """
V)

!"i
2. Extraction of modes corresponding to repeated roots (ai--" w;,) of multi-
plicity m, including rigid-body modes (ai--" O)
I N
3. Extraction of sequential modes by using (ai--" wT) after deflation

If the stiffness matrix Sin Eq. (1) is semidefinite, one or more rigid-body modes
- ~

8: -- - ~
..,.;
N 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
Q.)
u A
..
9 A
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
~
C 8: ~ "'IE: is required when calculating nonzero repeated eigenvalues.
--'-- - - 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 I Sec. B.2 Transformation Methods 505

w1ll ,.11ll11w, cxtrm:tion

I i11st o , 1·,qs. ( 18, ) or the second eigenvalue and eigenvector w1'thout dcflat·1011. I·111111 TABLE B.2 Iteration of Second Mode
111• the modified stiffness matrix becomes

S"' - S - a2M = s -0.72

[
0.3808 -0.72
- 0.8208
-0.48
0.48
J (1111
Cycle

(X2h
1

(X2)1
2

(X2h
3

(X2h
4

(X2)•
5

(X2)s

- 0.48 0.48 0.4592 I 0.8904 0.8677 0.8673 0.8673

Vector 1 -0.1400 -0. 1701 -0.1715 - 0. 1715
1111l•tmi1.11tion or S"' as in Eq. (19) yields
1 1.000 1.000 1.000 1.000
0. 3808 0

111111
D* = s
[
0
0
- 2.182
0 -oLJ (11)
.!... (Ai)t
m
-92.68 -84.2 1 -83.27 -83.25 -83.25

B.2 TRANSFORMATION METHODS

_ [l - 1.891 - 1. 261]
U* = 0 I 0.1~59 When most or all of the eigenvalues and eigenvectors are desired, trans-
Io J
0 0 formation methods [5] prove to be more efficient than inverse iteration. In this
As lw lon:, we ussumc the starting eigenvector section we describe the Jacobi, Givens, and Householder transformation pro-
cedures, as well as QR iteration. To confirm ideas, a numerical example follows
(X2)1 = {I, I , J} q,1 the discussion of each approach.
111 tlw first iteration, the vector (B2 ) 1 is computed to be As a preliminary matter, we assume that the expanded eigenvalue problem
has been converted to the standard, symmetric form:
(82), = M(X2), = m{3.72, 4.28, 3.28} Jql
AV=VA ( l)
wll1d1 rnnt uins the same values as Eq. (f). Solution of Eq. (17) for vector (Y2 ), p, rnhll
as described in Secs . 3.6 and 4 .2. In Eq. (1) the symbol V denotes an orthogonal
m
(Y2)1 = ;{- 82.52, 12.97, -92.68} (I) 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
N111 11111 li z11lion or this vector with respect to its largest term gives of the spectral matrix A.
I The basic process in all transformation methods is to diagonalize matrix A,
(X2h = (A;), (Y2)1 = {0.8904, - 0.1400, 1.000} as follows:
Ill wh1l'11 VTA V = A (2)
When this is accomplished, we have found not only the spectral matrix A but
(At), = - 92.68~ UI also the normalized modal matrix V.
.!'

l'uhk 13.2 ~h.ows .the resu lts of successive iterations. In this case conver~Mlt (hi Jacobi Method
l,u11 At and the second eigenvector X 2 occurs in fi w I y 1I
Nignilk-111_11 d1g_11s) for both
l-111111 vnlm·s for llu s mode arc
Sinee its development in the nineteenth century [6] , the Jacobi method has
111 2 I 1· enjoyed extensive usage by mathematicians, scientists, and engineers through-
At - 83.25
,\'
W2 = 7*' + a2 = 0.3480 • (11) out the world. The essential idea in this approach is to zero a selected off-
"2 Ill
1111d
diagonal term A,,., = Aqp of the coefficient matrix A in Eq. (1) . This is accom-
plished by opl·rnting upon that matrix with a generalized rotation matrix R.kin
X, j0.8<17.l, - 0. 17 15, 1,000} step k, us follow~·
wh11 h 1111· 11i 11111 1•1111111 lo thos1•
1
111 Ex11111pk• l.4 ,
 1508 I lunrwnl11n1 1111d t l11n11vncto11
Soc. B.:.> 1rorrnformotlon Methods 507

The form of the II X II rotutiou 11111111 11 Is

and Eq. (8) becomes
1
(10)
which is an approximation of Eq. (2).
sin fh-------- · row I'
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.
-sin ()k Each sweep of this kind requires approximately 2n 3 multiplications. Equations
cos 8c - - ----- · row q
I I (6) give the formulas to be used for altering either the upper or lower triangular
I I
I I part of matrix A.
I I
I I During a particular sweep, terms that are zeroed do not necessarily remain
I I
col. p zeros. However, the square root of the mean of the squares (RMS) for off-
col. q
diagonal terms is reduced after each annihilation (and even more dramatically
In this rather sparse array, we define the terms after each sweep).
(Rpp)k = (Rqqh = cos fJk (;") Example B.3
(Rpq )k = - (Rqp h = sin fJk 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
(Ruh = 1 (i t= p or q) (~d
0.64 -0.48 0 ]
(Rij h =0 (i or j t= p or q) (111) A = -0.48 1.44 -0.48 (a)
'l'ht! multiplication in Eq. (3) alters only terms in rows and columns p and , 111 [
1 0 - 0.48 1.92
mutri x A , which become
In the first transformation of matrix A, we shall make the term A,2 = A21 equal to zero.
(App)k+r = (APP cos 2 fJ + 2Apq cos () sin fJ + Aqq sin 2 fJ)k (fut) Therefore, p = 1 and q = 2, so Eq. (7) becomes

(Aqq)k+ , = (App sin2 () - 2Apq cos () sin () + Aqq cos2 fJ)k ((1h) = 2A12 = (2)( - 0.48) = I
2 (b)
281
tan A11 - A22 0. 64 - 1.44 ·
(Aµq)k+, = [(Aqq - App) cos ()sin()+ Apq(cos2 fJ - sin2 fJ)]k
From this expression we determine the angle 8, and its sine and cosine as
= (Aqp h+ I ( (1\:)
81 = 0.4380 sin 81 = 0.4242 cos 81 = 0.9056 (c)
(A,1,)H r = (A;p cos () + A;q sin fJ)k = (Ap;)k+, (<111)
Then the first generalized rotation matrix in Eq. (4) takes the form
(Aiq)k+1 = (-AiP sin fJ + Aiq cos 8)k = (Aq)k+ , ((10)
0.9056 0.4242 0~ ]
'l'lw unglc 8k is chosen so that the term (Aµq)k+ , becomes zero. Thus, fro111 I 11 It = - 0.4242 0.9056 (d)
(C,l') we have [
0 0
tan 28k = ( 2Ae9 ) and the first transformation in Eq. (3) produces
Aµp - Aqq k
0.4152 0
whl•rc Io~ I -s 'Tr/ 4. After nr rotations, Eq. (3) yields ,
A2 = R , A1R1 =
, T
0 1.665
- 0.2036]
- 0.4347 (e)
R,,, ... R2R,A 1 R"T""
A ....

1 R2
T A
..•
"'T
R,,r
"" A [
- 0.2036 -0.4347 1.920
Tlwrcforc, the matrix of orthonormal eigenvectors is As a conNl'< t111,m·t•, lht• fll'sl rotation has reduced the RMS of off-diagonal terms in matrix
• ""'
K I'
V -= ""T A from OAH 111 II I 11l•I
Kr R"T
2 •• 11,
h11 1111· N1·111ud 1111us twm11ti o n, WL' ,unl..1• A,, • A II t·q1111l 10 11•111, so 1h11t
 608 Sec. B.2 Transform ation Methods 509
Eigenva lues and Eigenvectors App, a
" _ 2A13 (2)(-0.2036) 0.8927 -0.4 135 0.1523]
tan 2 u2 - ---
A11 - A33 0.4 152 - 1.92 = 0 ·2706 (1) v = RTRI RI = 0.4205 0.7005 -0.5767 (o)
[
Thus, we have 0.1317 0.5817 0 .8026

82 = 0.1321 sin (Ji= 0.1317 Modal vectors appearing column-wise in this matrix are automatically normalized to
cos 82 = 0 .9913 have unit lengths.
and the second rotation matrix is
Givens Method
, [ 0.9913
R2 = 0 (h) Instead of annihilating the p, q term (as in the Jacobi method), Givens [7]
- 0.1317 proposed annihilating the p - 1, q term instead. Then a zeroed term will remain
Using R2 in Eq. (3) yields zero during a forward sweep. For this purpose, we set Eq. (6e) equal to zero to
obtain
0.388 1 - 0.05727 O ]
A3 = R2 A2 RI = -0.0~727 1.665 - 0.4309 (ll tan Ok = (Ap-i ,q) (11)
[ Ap- 1,p k
-0.4309 1.947
for which the RMS of off-diagonal terms is 0.3074. I I
where Ok :s 7r/2 . For the first row of matrix A, rotation in the 2-3 plane gives
The third transformation involves zero1·ng A 23 -- A 32, wh'ch ·
1 gives
tan 283 = 2A23 = (2)(-0.4309) _
A22 - A33 1.665 - 1. 947 - 3.o53 ( j)
Rotation in the 2-4 plane yields
I knee,

(h = 0.6272
sin 83 = 0.5868 cos 83 = 0.8097 (k) (1::)2
Sil the third rotation matrix becomes
and so on. After we clear the first row (and column), matrix A has the form
0
R,- G 0.8097
-0.5868
0.5~68]
0.8097
( I)
X

X
X

X
0
X
0
X
0
X
0
X

0 X X X X X
Suhstituting It into Eq. (3) results in 0 (12)
An-2 = X X X X X

0.3881 -0.04637 0.03361] 0 X X X X X

A,1 = RJ A3 RT = - 0.04637 1.353 0 (Ill)

[
0.03361 0 2.259 0 X X X X X

wlt1•11• tlw RMS of off-diagonal terms is O 0405 1 Note that each off d'
t I• . . · · - 1agona I te1111 IN 1111w T his technique produces a tridiagonal matrix T in a finite number of steps,
n 111st 1111 order ol magnitude smaller than at the beginning of th
At ti . . I r I . . e sweep . consisting of n, = (n - 2)(n - 1)/2 rotations. The entire sweep requires ap-

l
. le ent <~ 11e sweep, the diagonal terms in matri x A4 represent ,m,d 111110 - proximately 4n 3 /3 multiplications.
l11111t1011s lllr the eigenvalues of the original matrix . That is, g 11
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
0 .30881 0 O

111
A
r 0
1.353
0
0
2 2 'ii)
nddllhlll, IIJlJll llXi lllillt' l'i)tl'II Vl'l't\ll ,~ Ill\' l'llk 11) 11(1•d IIN
1111
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
Sec. B.2 Transformation Methods

I 0.48 1,0()
5
A=
- 0.48
U<,
- 1.00
0
1.64
1.64
0
0 0.48
0
1.36 oq (p I
Substituting this matrix into Eq. (3) yields

A A -
1.360
I.109
-l.109
1.412
0
-0.4463
We start by annihilating
Eq. (11) results in the term A, 3 with a rotation in the 2-3 plane. For this p11IJH1~1. A3 = R2A2RI =
[ 0
0
-0.4

0
463 1.538
- 0.1953
-o.iL,]
1.689
-1 - 1.00
01 = tan - I A1 3
- = tan - = 1.123 which is now in tridiagonal form.
A12 -0.48 (q)
Then we have
Householder Method
sin 01 = 0.9015 cos 01 = 0.4327
and the first rotation matrix R. 1 becomes (1)
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
0 itself. The operator has the form
0

ri
A 0 0.4327 0.9015
P =I - 2e eT (13)
R, -

Using this operator in Eq. (3) gives

-0.9015 0.4327
0 0 f] in which e is a col umn vector of unit length. The matrix Pis both symmetric and
orthogonal, so that

p = pT = p - 1
(14)
-1.109 0 Hence, P is equal to its own inverse.
A

A,-R,A,R, -
AT
[ -1.109
13~
i l.412
-0.1092
-o. 1092
l.588
04~27]
0.2077 (II
Consider a vector a to be converted to b = {±s, 0, 0, .
is the length of a. That is, , O}, wheres

0.4327 1.640 0.2077 Pa=b

(15)
In this case only one transformation is required to clear the first row (and column) 011f~hh We form the matrix P by first creating e in Eq. (13) from the vector a , as follows:
thethe
of triangular
operation part. The term A 14 = A4 1 was zero initially and remained unaffct·t,•cl h\
above. fl

The second step in this example is to annihilate element A by a rotation l11 tl11
s2 = L aJ (16a)
3-4 plane. Thus, Eq. (11) produces 24 J=l
c = {a1 ± s, a2, a3, . .. , a,.}
- I A24 -I 0.4327 (16b)
82 = tan - = tan = - I.324 C
A23 - 0.1092 (Ill e=--
From thi s we get . Vc1c (16c)

sin 82 = - 0.9696 In Eq. (16b) the sign of s is taken to be the same as that of a 1 , so that there is
cos 82 = 0.2447 no possibility of getting zero for the first term in vector c. The unit vector e in
Then fhc second rotation matrix R.2 is Eq . (13) is the result of normalizing vector c to unit length, as indicated in Eq.
(16c).
0
Now let us express the product Pa in Eq. (15) in terms of the vector c.

0.2447
0 0
0
0.9696
l (\\ I
Using Eqs. (13) and (16c), we obtain

fi c CT) 2c(c Ta)

0.9696 0.2447 Pa = \1 - 2cTc a = a - (cTc) (17)
Rcsull s ol flll' l1111!'l' pr·oducls (in pi11·e11Jlll'st•s) from llw st•rnnd 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),
n except that k replaced 1 and the lower limit on j is k + 1.
era= (a 1 ± s)a 1 + 2 aJ = s 2 ± a 1s After n - 2 Householder transformations, we evolve the tridiagonal
j =2
matrix
II

c Tc = (a1 ± s)
2
+ 2 aJ = s 2 ± 2a 1s + s 2 = 2(s 2 ± a 1s) (22)
j =2
where the Householder matrix is
Substitution of these expressions into Eq. ( 17) produces
2 ff = Pf PI, .. PI-2 (23)
p a =a _ 2c(s ±
2
s)a1 =a_ c
3
2(s ± a 1s) Equations (22) and (23) require approximately 2n /3 multiplications.
= {±s, 0, 0, ... , O} = b Example B.5
( 18)
as desired. 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
In a similar manner, the last n - 2 terms in the first column of th1• outside the tridiagonal region, as shown by Eqs. (19) and (21).
coefficient matrix A may be zeroed . With this objective, we let
" sf = L" AJ, 1 = (-0 .48)2 + (-1.00)2 = 1.230 (y)
ST= 2 AJ.1 ( 1911)
j =2

j=2 SI = ±1.109 (z)

C1 = {O, A21 ± S1, A 31, A41, . . . , A,.1} ( 191,) C1 = {0, A21 ± s1, A 31, A41} = {0, -1.589, -1.000, 0} (a')
C1 C1
e, = -C1- - e, = \/c"fCi = l. = {O, -0. 8464, -0.5326, O} (b')
( 19\') 878
VcTCi

Then
P, = I - 2e 1 ef

X X
( 19d )
P1 -_ I - 2e, e1T - -r~ 0
- 0.~327
- 0.9015
-0.9~ 15 0 ]
0.4327 0
° (c')

l
X X X 0 0 0 1
S1 X X X X
1.109 0
P,A1
0
0
X X X X
(2()) T
[1300
1.109 1.412 0.1092
X X X X A2 = Pi A1 p I = ~ -04~27 (d')
0.1092 1.588 0.2077
0 X X
. -0.4327 0.2077 1.640
X X
Matrix A2 is the same as that found before in Eq. (t) by the Givens method, except for
Also, the last n - 2 terms in the first row of matrix A may be zeroed as wi•ll the signs on several off-diagonal terms.
by
Similarly, for the second row and column we have
X s, 0 0 0 II

s, X X
s~ = L A].2 = (0.1092) 2 + (-0.4327)2 = 0.1992 (e')
X X j- 3

0 X X = ± 0.4463
A2 = P1A1PT = X X
(.'I l
S2 (f')
0 X X X X C2 = {O, 0, A 32 ± S2, Ad = {O, 0, 0. 5555 , - 0.4327} (g')
C2 CJ
() t• , {O, 0 , 0 /HKI/ , 0 .61 tl~f (h')
,\' ,\'
Vt· !t'J 0.704 f•I
 514
Eigenvalues and Eigenvectors App, II Sec. 8.2 Transformation Methods 515

0
than T1. Recursively, we have
2e2eI = [ 0
P2 = I -
0 I
0
0
0
0
0
- 0.2447
0.9696 0.2447
oL] (I') T k+I = QITkQk = RkQk-A
which iterates to the spectral matrix A. The modal matrix Vis calculated from
(28)

Eq. (25b), in which

1.109
T
['1.109
360 1.412
0
-0.4463
W = Q1Q2,, · Q,!f (29)
A,~ P, A,P, ~ ~
-0.4463
0
1.538
0.1953
oiLl
1.689
(J') 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
Again, matrix A3 is the s~e as t~a.t in Eq. (x), except for the signs on off-diagonal tc 1 IIIN we see that the definition for Rk is
Thus, we see that the desired tndiagonal matrix is
Rk = QITk (30)
(k 'l This formula gives us the means for determining both of the desired matrices,
where
as will be explained next.
0 0 At the operational level for this method, we wish to annihilate a lower-
triangular term (1j;h of the tridiagonal matrix Ti,, where j = i + l. For this
= Pf PI=
-0.4327 0.2206
-08~411 purpose, we set equal to zero the product of row j of a Jacobi rotation matrix and
[i
H ( (' ' )
- 0.9015 - 0.1059 0.4196 column i of matrix Ti,, as follows:
0 0.9696 0.2447
-sin O;(T;;)k + cos 8;(1j;)k = 0 (31 )
QR Algorithm From this expression, we have

Assume that Householder transformations have converted Eq. (I) to tan e; = (T;;T,)
-1!.
k
(32)
TW=WA (.?•I)
where Consequently,
. (} (1j;h (T;;)k
(}~ul sm · = - - cos (} . = -C;- (33)
or ' C; I

where
V=HW
Now let us factor T = T1 into the form (34)
Then the Jacobi rotation matrix (R;)k premultiplies matrix Tk (for i = I, 2, ... ,
T1 = Q1R1
n - l) to produce
which is known as Givens factorization. The symbol Q1 in Eq. (26) dcnoll'II ,111
(35)
o~thogonal matrix obtained by n - 1 Jacobi rotations, and R 1 is an upp1•1
triangular array. The QR algorithm l9] derives its name from the factors in h1 After the first operation in this sequence, matrix Tk is no longer tridiagonal; and
(26). ' after n - 1 operations it becomes the upper triangular matrix Rk in Eq. (28) .
. Prcniultiplication of Eq. (26) by Q'/' and postmultiplication by Q 11~1·11
1
Comparing Eq. (35) with Eq. (30), we see that
gives
(36a)
( I/)
or
l'lw 1ww 1111111 ix 'I'.! is unotlwr 11 idingo1111l 111 rny huving s11mllt•1off di11g111111I 11·1111 111
( Hih)
 516
Eigenvalues and Eigenvectors App. B App. B References 517

Therefore, matrix Qk is defined as the product of the transposes of n - I Jacobi TABLE B.3 First Cycle of QR Iteration
rotation matrices.
(R1)1 T1
Example B.6
[ o.m, ["ffi
~]
0.6341 0 1.109 0
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
-0.6341
0
0
0.7776 0
0
0
1
0
1.109
0
0
1.412
- 0.4463
0
-0.4463
1.538
0.1953
oL]
1.689

C1 = YTf1 + n1 = Y(l .360)2 + (1.109)2= 1.755 (m') (R2)1 (R1)1T1

~] J,,]
. T:i.1 1.109 0 0 -0.2821
1.752
sm 81 = - = - - = 0.6320 (n') 0.6613 -0.7501
C1
Ti.1
1.755
1.360
[! 0.7501
0
0.6613
0
[Lr 0.3935
-0.4463
0
- 0.3459
1.538
0.1953 1.689
cos 81 =- = -- = 0.7749 (0 1)
C1 1.755
(R3), (R2R1)1T1
Then the first 4 x 4 Jacobi rotation matrix becomes
0 0 1.752 -0.282 1
0
ot] [17550 0.5950 -1.383

l
0.7749 0.6320 I

• -0 .6320 0.7749 00 01
0 [! 0
0
0.9684
-0.2495 0.9684
0 0 0
0
0.7578
0.1953
-00.1291
'~"]
1.689
(R1)1 = (p')
0 0 1 0
Qi = (RTRIRD1 R1 = (R3R2R1)1T1
0 0 0 I
Tnhle B.3 shows this matrix in the upper left-hand position. After multiplying it with [0.7749
0.6320
- 0.4180 - 0.4591
0.5152 0.5629 - 0. 1450 0ll83] [1755
0
1.752
0.5950
-0.2821
-1.383 -01~65]
mu1ri x T = T1 (in the upper right-hand position) , we list the product (R. 1) 1 T1 in tla• 0 -0.7501 0.6404 -0.1650 0 0 0.7826 0.5466
positi on below T1. 0 0 0.2495 0.9684 0 0 0 1.604
For the second rotation (in the 3-2 plane) , we have

C2 = YTi2 + T52 = Y(0.3935)

2 + (-0.4463) 2 = 0.5950 (q') Qi = (RfRillr)1 (x')
• (J
sm 2 = C

COS 82 = -
T32

C2
2

T-i.2
=- -
- 0.4463
= 0. 5950 = -0.7501
0.3935
0.5950
= 0.6613
'l'uhle 8 .3 contains the second Jacobi rotation matrix (R.2)1 below the first, and the l'l'Ntrll
(r')

(N')
is listed in the lower left-hand position.
Using Eq. (27), we now compute a second tridiagonal matrix, as follows:

T2 = Ri Qi =
[
2.468
0 .3760
o
0. 3760
1.342
- 0.5870
0 .0000 0.0000
-0.5870 0.0000
0.6375 0.4001
l (y')

of multipl ying it with (R1)1T1 is shown below the latter matrix. 0 .0 0.4001 1.553
The third rotation (in the 4-3 plane) involves
Note that computed zeros must appear in the upper triangular part of matrix T2 to satisfy
c3 = Vr ~J + T~3 = Y(0.7578) 2 + (0.1953)2 = o.7826 (t ' ) 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
. 43 0. 1953 matrix T 2). Spectral shifting can be used in successive cycles to accelerate the rate of
srn 03 = - = - - = 0 .2495 (11 1 )
C3 0 .7826 convergence, as explained in Sec. B. l.
7;, 0.7590
cos 0, - Ci == 0. 7826 = 0.9684 (v' l REFERENCES
'l'lw third rot11tlo11 mulrix (R,) 1 ulso uppeurs in Tublc B.3, ulong with the prodlll'I l . Timoshonko , S. P., Young, D. 11. , and Weaver, W., Jr., Vibra1ion Problems in
R1 (I{ift,((,)1'1'1 (w ' l l~11J(l,11•1•rl11}( , 4th L'd., Wiley, New Yor·k, J1)74,
l. W1•11v,,1, W , Ji , 1111d Ci1•11•, .I . M , Motr/1 /\1111!1w/,1· 1!/' l•111111NIS1r111•111r1•,1·, 2nd 1•d.,
l1111ddil in11 1 th,• 11111t11 ~
Vn11 Nrn1t111111l lfr l11h11ld , N1•w Yrnk, l11HO
618 I l111111vnh1111 nnd I l11111w1u:tor • At>t> I

3. Weaver, W. , Jr. , and Johnston, I'. R., Vi11it1• J:'l,•1111•111.v for S1m, ·111m/ A111,/n11 1
Prentice-Hall , Englewood Cliffs, N.J ., 1984.
4. Bathe, K. 1. , Finite Element Procedures i11 Engineering Analysis, P1·c11t1n• 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 d1•1 Sh,
ularstorungen vorkommenden Gleichungen numerisch aufzuloscn," Crl'l/1•' 1 I ,
Vol. 30, 1846 , pp. 51-94.
Appendix C
7. Givens, W. , "Numerical Computations of the Characteristic Values of a Rl·11I !·i\ 111
metric Matrix," Report No. ORNL-1574 , Oak Ridge National Laborut01 y, I >11t. Flowchart
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
for Program DYNAPT
9. Parlett, B. N. , and Kahan, W., "On the Convergence of a Practical QR Algw 1111111
Proc. IFIP Cong., 1968.

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

Read and write struc1ural

NN. NE. NRN, E. RIIO
parameters.

NN . NE. NRN, E. RHO

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

r
I
I J. X(J), Y(J) Read and wri1c 11od11 I
coordinates.

J
J, X(JJ, Y(J)
I
e. Nodal Restraints
I
L_
CONTINUE
NND = 2*NN Calculate number of
nodal displacement,
possible.
d. Element Information
r-
1
I
Heading: ELEMENT INFORMATION
Subhead: ELEM. J K AX EL CX CY I NRL(J) = 0 Clem· nodal rcs1111i111
I list.
3
I
L_
r-----
1
I Heading: NODAL RESTRA INTS
I I, JN(f), KN(!), AX{!) Read infonnmion ltu
Element I.
Subhead: NODE NR! NR2
I
I XCL = X(KN(JJ) - X(JN(l)) Compute x and y ~·01111~,
I YCL = Y(KN{I)) - Y(JN(J)J ne111s of clement lc11ptl1
I EL(!) = DSQRT(XCL *XCL + YCL*YCL)
CX(J) = XCUEL(I); CY(!) = YCUEL(l)
XCL and YCL. tlw Jt,1111111 I I
and lhc direction l'Cl\1111·~
+ CX and CY.
I
I I, JN(l), KN{!), AX(!). EL{!), CX(I). CY{!) Write infon11111io11 lrn
I Elemc n1 I.
I 2
L ____ _ CONTINUE
22 I l11w,1h 11 lnr 1 1 11111111111 I VNAl'l App I h1w1.h 11t lo1 1'1ot:J1 mi OYNAPT 523

t
,--
I NRL(JJ > 0 ID(JJ = NDF+NI
4 J I
1 I
I K. NRL(2*K-1). NRL(2*K)
Rend 11ml W1lh 11111111 I 6 Otheiwise, put rD(J)
,cstruinl I M L __ _ among supp rt re ·train ts.
I UE
I
I K. NRL(2*K-1), NRL(2*K)
RETUR
I 99 TOP
4
I
L-- CONTINUE END

lni1ializc 1111111h 1 111 11101 I

rc,1rui111, hi 11•111 2. SUDPROGRAM STIFPT FOR SUBPROGRAM VIBPT
a. Clear Structure Stiffness Matrix

,--
1
I
I
NNR = NNR+NRL(J)
Calculalc 111111111< 1 111
nodal rc�11ul111, NNII
I
,- I J = I. NDF

I 5 f--- I K = J. NDF
L-- I
I
SS(J.K) = 0.0
Cal ulah.: 1111111111•1 111 I
NDF = NND - NNR degrees ol l1,·1·1h1111 NI •I I
L_ CONTINUE
NUMBER OF DEGREES OF FREEDOM (NDF) Write NDF 11nd NNII
NUMBER OF NODAL RESTRAINTS (NNR)
b. Calculate Element Stiffness Matrix

I'. Displacement Indexes

r----- 5

Initialize N l 11111111 1
I Compute stiffness
SCE = AX(l)*E/EL(I) constants and ti II
I SCEXX = SCE*CX(l)*CX([)
SCEXY = SCE*CX(l)*CY(I) upper triangular

r-- 6 J = I, NND I SCEYY = SCE*CY(l)*CY{I) part of SE, as shown

•I
by Eq. (3.5-26).
I SE(l.1) = SCEXX; SE(l.2) = SCEXY
1 SE(l ,3) = - SCEXX: SE( 1.4) = -SCEXY
SE(2,2) = SCEYY; SE(2.3) = -SCEXY
I NI = NJ+ NRL(J) By clcf'uull, p111 lllt l1 SE(2,4) = - SCEYY; SE(3,3) = SCEXX
lllllll11g 1'11:1· 1h�11li11 I 1111 111
I ID(J) = J-NI
I SE(3,4) = SCEXY; SE(4.4) = SCEYY

I I
624 f lowi:hnrt for l'rournrn l>YNAl'I App C
App .C f luwohurl for Progrom DYNAPT 525

t c. Calculate Element Displucemcnt Indexes

e. Fill Lower Triangle of Structure Stiffness Matrix

IDE( l) = ID(2*JN(l) - l); IDE(2) = ID(2•JN(I )) Eq,. (J . 5 30)

IDE(3) = ID(2*KN(I) - I); IDE(4) = ID(2*KN(l))

d. Transfer to Structure Stiffness Matrix

JMI = J -1

r--- r
I
Check for row restrnint I Place an upper triangular tenn
SS(J,K) = SS(K,J)
I into a lower triangular position.

r--
I 6

L CONTINUE
I
I 7
Check for column rc,11,11111
I L __
CONTINUE
I
I Detenn ine row and col
IR = IDE(J ): IC = IDE(K) umn indexes for structur ~·
I stiffness matrix.
RETURN
+ I Check whether row intk·~ i,
I less than or equal to coh111111 END

.. + index.
I I If not. interchange row ,111d 3. SUBPROGRAM CMASPT FOR SUBPROGRAM VIBPT
ITEM = IR; IR = IC; IC = ITEM
I I column indexes . This subprogram is similar to STIFPT. but the consistent
mass matrix is generated instead of the stiffness matrix.
I I 2
I I SS(IR,IC) = SS(IR, IC)+SE(J, K ) Transfer a term in mat1 1x
SE to matrix SS, u, i111 4. SUBPROGRAM STASYM FOR SUBPROGRAM VIB
I I plied by the fi rst of Eq~ a. Decompose Stiffness Matrix and Copy Mass Matrix to CMU
3
I IL __ (3.5- 14).
I CONTINUE
I J I , NDF
I r-
I 4 1
L ____ I
CONTINUlo
r
5 I
L _____ CONTINUE L 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---

Dccompo~c Mil Im•" 111,1111

1
DECOMP(U, NDF, & 3)
inro the factorl'd lt11111 I 11 I
as given by Eq. (.I " •11 r-
1 5
L .-----''----,
STIFFNESS MATRIX DECOMPOSED
CMU(J.K) = SS(J,K) Copy stiffness matrix to CMU.

r--- 2 J = I. NDF
1

,-
l

1 2
6

STIFFNESS MATRIX /\ND MASS MA-

TRIX NOT POSITIVE-DEFIN ITE

L 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
Call Subprogram INVERU to
INVERU(U, NDF)
3 obtain the inverse transpose of U.

,- 4 J = I. NDF
1

,-
I

l 4
d. Transform to Standa rd, Symmetric Form
(I) Prcmultiply by u-T

L U(J,K) = CMS(J,K)
lhtnsfcr mu,~ I'
11111111\ 111
r-----
1
I I = NDF-11+1 Set up decreasing index I.
I
DECOMP(U, NDr, & 6) Dcco111pm1• 11111" 11111111,
11110 the iill'lorl'd 1111111 I 11 • ,----
I 9 JI I , NDF

I
Slll+N l·.SS M/\TRIX NOT 1•os 1l"IVI,
1) 1'.l •IN I J'I :, M/\SS M/\TRI X Dl :C'OMPOSED
I I
I I
' J NDF JI f I
Tl•.MP 00
Set up decreasing index J.
1111<.l ini1iali1c TEMP to zero.
I t
I I
 I lowt hnr I for 1'1ou1nm llYNAl 1 1 App C: 1lowr.hurt fur 1'109run1 DYNAPT 828
I I
!
I
I •I ,---
I
X K l ,J
I
I
IL __ _ _
I I
•I I I TEMP = T EMP + U(J.K)•CMU(I.K)
P1 c111u lt1ply C'MlJ hy II 1
indicmcd 111 tht.· 111 , 1 ,,1 I 11
I •IL _____ _
I I (3.6 11 ).
I I IL ___ 8
I I
I I RETURN
I
I •I CMU(l,J) = TEMP Put result back into ('~ ll I
END

•I IL ____ 9 5. SUBPROGRAM EIGEN2 FOR SUBPROGRAM VIB

I
I ______
L 10 Calculate eigenvalues and
eigenvectors to double pre-
cision.

(2) Postmultiply by u- 1

6. SUBPROGRAM TRAVEC FOR SUBPROGRAM VIII

a. Calculate Angular Frequencies

r----- 13 II = l,NDF

I TOL = 0.000001
1
, 'l'I IOl1•1tl lll'I' JIii /l' III

I I = NDF - 11 +I Set up decreasing i ndex I

~ l)ll.'IIVllilll'\

I
I lf mass matrix wa, de
composed, go to 2.
I r-- -- 12 J = l ,NDF

I I
•I I I
TEMP = 0.0 Initialize TEMP to zero
r---
1
I I L Calculate angular fre-

I I r---
CV(I) = 1.0/ DSQRT(CV(I)) quencies from the first
of Eqs. (3.6- 12).
I I I
I I I
I
I •I II TEMP = TEMP

II
+ CMU(K.J)*U(I .K)
Postmultiply CMU by U i, II\
in the first of Eqs. (3 .6 11 l 2

I I L ___ r-
I I 1
I I I Check si ze of eigenval ue .
I
•I
I
I
CMU(l ,J) = TEMP Put result back into CM lJ
•I
I
1130
I l11w1 hnr t fur l'rour nr11 l>YNAl' I App, c: I lowohurt lnr Proumm lJYNAPI 531

I C V(!) = DSQRT(CV(I)) Calculait: 1111g11J11r lh· c. Normalize Eigenvectors with Respect to Largest Values
+ qucncics from rlw 111NI
ofEqs. (3.6 19),
I
I ,-----
I 3
I
I I
I CV(!) = 0. 0 Set small cigcnvullll'
I
equal to zero. Initialize BIG to zero.

•L __ 4 I
I r---- Search for largest value
I 1 in vector I.
I I
b. Back-Transform Eigenvectors
I I CI = DABS(PHl(J,I)) Set up constants C I
I I C2 = DABS(BTG) and C2.
r---- I I
1 I I
I
L ___ _
•I
I
I
C I > C2 B IG = C l
If C J exceeds C'2, n:-
placc BIG wit h (' I .

I I I 8
I Initialize SUM to ~crn
I L __ CONTINUE
I
•II r----
I
I r----
1 I 1
I
I I I I
+ Back-transform cigcnv1'l'l111 ,,
I I S UM = SUM U(K,J)*PHl(K ,I)
as in the second of Eq.~. ( I !I I r, I I P Hl(J, I) = PHl(J,l)/BIG Nom1alize vector I with
respect to B IG.
I I 6 I I 9
I L __ _ + L __
I
•I I
L ___ _ 10
I PHl(J,I) = SUM
Put SUM into PIii.
I
L ____ _ 7

RETURN

EN D
..
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
I I
lDCMP = I I I
If lite lllllSS 11 11111 iX Wu, If K corresponds to a
decomposed. µo lo I I + free displacement, go to 4.
I I
NDP2 = NDF/ 2 Divide rhe n11111lwr 111
degrees of f recd om hy 1
.. I
I EE(J) = 0.0 For a restrained displace-
ment, put a zero into the
I expanded eigenvector.
I I
r---
, I
I
I
I
I
4
I J = NDF - I + I ; TEMP = CV(!) I
I CV(!) = CV(J); CV(J) = TEMP Interchange sy111111c111t·11lly
I t EE(J) = PHI(K ,1) For a free displacement
transfer a nonzero term.
I placed angular frcquc11l'il''
I I 5
I I L_
I I
I I I
I I TEMP = PHI(K ,l) I MODE (!)
Write headings for modal
I I PHI(K, IJ = PHJ(K,J) Interchange sy111111ctric11lly
I ANGULAR FREQUENCY (CV(!))
outpu1.

•I placed eigenvectors. NODE DJI DJ2

I I
l I
I PHI (K,JJ = TEMP
I
I
L __ _ 2 I
I 6
I J,EE(2*J - I).EE(2*J) Write cigc11vec1or for
3
+ mode I.
I
L __ _ 7

CONTIN UE
b. Write Angular Frequencies and Expanded Eigenvectors

c. Normalize Eigenvectors with Respect to the Mass Matrix

r---
1
I r------
I
I ,r- 1
I ,----
I I I
I I Set index K equal 10 ll)(J ). t 1
I I I t Initialize SUM to zero.
I I
I I
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 r--- 8 K = l,NDF Heading: LOADING NO. (L N) OF (NLS)
Subheads: DYNAMI C PARAMETERS
I I !SOLVE NTS OT DAM PR
I I 8
I L Prcmultiply ihl.! 1111M 11111
I SUM = SUM + PHl(K ,l)•CMS(K ,J)
I trix by the trnm,po,l·cl
eigenvector:
I I 9 Read and write dyna-

•I L ____ EE(J) = SUM Put SUM into Iii: (1111

temporary Monigc).
ISOLVE,NTS,DT,DAMPR
mic parameters.

I ISOLVE,NTS.DT,DAMPR
I SUM= 0.0 Reinitiali1:e SUM to 1<•111
I
I
I
I
r--- 10 J = I.NDF b. lnitial Conditions

I 10
I L_ SUM = SUM + EE(J)*PHl(J,I) Postmultiply by thl.! l'lp r- - I J = I ,NND
I envector.
I
I I Clear initial displacement
•I SUM = DSQRT (SUM ) Calculate nonnalil11t1n11
constant.
I
I
DO(J) = 0.0; VO(J) = 0.0 and velocity vectors.

I L_
I r--- II J = 1,NDF

I I II
I L_ Divide eigenvector hy 1h1 Heading: I NITIAL CONDITIONS
PHl(J, I) = PHl(J. 1)/ SUM
I normalinuion con~t,1111 Subhead: NNID NNIV
I 12
L _ ______
Read and write initial
NNID,NNIV
condition parameters.
RETURN

END
536 App.C Flowchart for Program DYNAPT 537
Flowchart for Program DYNAPT App. C

(I) Initial Displacements

Heading: INITIAL VELOCITIES
Subhead: NODE VOi V02

If there are no initial

displacements, go to 4.

r-- 5 I = 1,NNlV

Heading: INITIAL DISPLACEMENTS I

Subhead: NODE DO! DOZ Read and write initial
I J, V0(2* J - 1), V0(2*J) velocities.
I
I
r- - I J,V0(2*J - 1),V0(2*J)

1
I
I J,D0(2*J - l),D0(2*J)
Read and write initial I 5
displacements. L __ CONTINUE
I
I
I J,DO(Z*J - l),DO(Z*J)
r-- 6 J = l ,NND
I
I 2 I
L __ I K = ID(J) Set index K equal to ID(J).
CONTINUE
I
I If K corresponds to a
r-- I K .;; NDF VO(K) = VO(J) free displacement, put

1
I VO(J) into VO(K).

I I
I
Set index K equal to 11)( I)
I 6
l __
I CONTINUE
lf K corresponds to ;i
I K .;; NDF DO(K) = DO(J) free displacement, p 111
I DO(J) into DO(K).
c. Applied Actions
I
I 3 7
L __ CONTINU E r-- 8

1 8
(2) l11i1ial Velocities L__ AS(J) = 0.0 Clear action vector.

If' 1hcrc un.: no ini1i11I

vcloci1ic>. go 10 I Heading: APPLIED ACTIONS
Subhead: NLN NEL
ftJff
Flowchart for Program DYNAPT App.C App. C Flowchart for Program DYNAPT 539

NLN,NEL Read and write load

parameters.
r---- II J = l,NEL

1
I l ,BLI ,BL2,BL3,BL4
Read and write scale
factors for line loads.
I
I
I l,BLI ,BL2 .BL3 ,BL4
( I) Nodal Loads
I
I
I PBS(I) = (2.0* BLI + BL3)*EL(l)l 6.0
Calculate equivalent
If there are no loaded nodal loads, as in Eqs.
nodes, go to 10. I PBS(2) = (2.0*BL2 + BL4) *EL(l)16.0 (4.10-4).
PBS(3) = (BLI + 2.0*BL3)* EL(l}l 6.0
I
Heading: NODAL LOADS
Subhead: NODE AJ I AJ2
•I 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(!)
I AS(J2) = AS(J2) + PBS(2)
I AS(KI) = AS(KI) + PBS(3)
AS(K2) = AS(K2) + PBS(4)
I
r-- 9 l = l,NLN
I II
1
L___ _ CONTI NUE
I J,AS(2* J - 1),AS(2* J) Read and write scale
I factors for nodal loads.
I d. Ground Accelerations
I J.AS(2*J - l), AS(2*J)
12
I
Headi ng: GROUND ACCELERATIONS
I 9 Subhead: !GA
L __
Read and write acceler-
ation 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 ll•ndin)l: U NE LOADS
S11hlw11d : El.EM . IJLI ... BI A
Heading: ACCELERATION FACTORS
Subhead: GAX GAY
1140 Flowchart for Program DYNAPT App. C Flowchart for Program DYNAPT 541
App.C

•I
I r---
GAX,GAY Read and write acccll•1ti I
tion factors.
+ L1 14
Multiply mass and acceler-
I SUM = SUM + CMS(J,K)*GA(K)
ation terms.
GAX,G AY
I
I Augment AS(I) with the
AS(l) "'. AS(!) - SUM
I negative of SUM.
( J) Fill Vecto r GA with Ground Accelerations
I 15
L_____ _ CONTINUE

r-- 13 I = 1,NN 16
1 r- 17
I 1
I JI = !0(2*1 - I); J2 = ID(2*1) Calculate displacernl'III
indexes JI and J2 . I Set the displacement
I I index K equal to ID(]).
I JI ,a; NDF GA(J I ) = GAX If JI is free, put GAX I
I into GA(J I).
I
I K ,a; NDF AS(K) = AS(J)
I
I I lf K is free, move the
corresponding action up
J2 ,a; NDF GA(J2) = GAY If 12 is free , pul GAy I 17 the list.
I mto GA(J2).
L_ CONTINUE
I
I 13
L__ CONTINUE e. Read Forcing Function

(l) Multiply /\ccclcrntions by Mass Matrix Add the parameters

N = NLN + NEL + IGA;NFO = 0
requiring a forcing function.

r ---- If N = 0, skip reading a

forcing function .

'
I
I
Sci lhc displucc11wn1
index J cquul 10 11 )(I)
Heading: FORCING FUNCTION
Subhead: NFO

I
t If' J is l'CSlrai11cd, N~lp
the 111uliiplit·111it111.
Read and write number
I of function ordinates.
I
I l11i1i11lin· SlJM 11 , it•iu
I
 llnwil11111 tor l'1u111iu11 l>YNAl'I
App.C I lowoh111t lot l11011rt1111 DYNAfJI 643

i
lk11d 11111, l •l lN( IION tll<lll NA ,-~
S11hhc11d; K 'l 'IMI • l•A( 1()1( ~
I
I
I
•
,-- 21 I = l ,NTSS

r---..- 18 J 1,NFO
I I
I I
' Increase J by J.

1 I I Detennine step and ramp

I Rcud and write sulN·11pl ,
I I STEP(J) = FO(K) + (I - l)* DST EP for lime increment wilhin
K,T(K),FO(K) a piecewise-linear segment.
I lime, a11d f'w1c1iou 01<11111111 t I 21
I 1s I L __
L- RAMP(J) = DSTEP
r - - -K-.'_r(_
K..L)
,F-· 0-(K-)--~ I
I 22
L____ _

RETURN
END
I'. ( 'nknlntc Step und Ramp for Each Time Increment

r--- 20 J 1.NTS
1 9. SUBPROGRAM TRANOR FOR SUUl'ROGAM NOUMOD
I a. Read and Write Number of Modes
STl iP(J ) = 0.0 Clear slcp and r,1111p
I 0.0; RAMP(J)
vectors.
I 20
L __ CONTINUE
NMODES

NORMAL-MODE SOLUTION
N l'OMI NFO I: J 0 lniti,liizc NFOM I 1111t1 .I NMODES = (NMODES)

,---
1
22 K l ,NFOM I

C.1icula1c number ol 1i111r

b. Calculate Transformation Operator <I?;;'

I ~ ·s·_s_·_ <
_.,_.<K
_ 1_1J__·_rc_K_lJ_11_)1_'__,
I
I
slcp~ for a picccwisl'
I incur M
.lgnicnl. r---- 2 l = l ,NMODES

If' NTSS 0, skip 1111• ~11•11

I
I
anti nunp cakul111ion,.
'~---- 2 J = l ,NDF

•I llST/\JI ( IU(K I I) l'O(K)J/ NTSS

Culcullllc 1hc nuc ol
chnngc ol nnli11u1l·, wilh
1
I
I -i- 1c~pcc1 lo linlL' ,
•I
lnilialize SUM to zero.
1144
Flowchart for Program DYNAPT App.C App.C Flowchart for Program DYNAPT 545

I •
I r---
I r-- - I 1 5

•I L' + PHl(K ,l)*CMS(K,J)

I
I
L SUM = SUM + PHI(J,l)*AF(J) Premultiply vector At
with <Pi. as in Eq. (4.4-4).
I SUM = SUM Calculate <I>; 1
= <I>IM. I 6
I 2
L _____ _ AN([) = SUM Put SUM into vector AN.
L - - - -,L__ TR_o_r_u_.J..,..
l _= _s_u_M_-..J
Put SUM into the trans-
format ion operator TROP.

c. 'lhmsfonn Initial Conditions

l RETURN
END
-r AF equivalent to AS in
COMMON.

,r---- 4 I = l ,NMODES

IO. SUBPROGRAM TIHIST FOR SUBPROGRAM NORMOD

I a. Calculate Response for Each Mode
DSUM = 0.0; VSUM = 0.0 Initialize DSUM and VS UM
I to zero.
I
I
I
r--- r------
1
4 I = I.NMODES

I
•I I' I DSUM = DSUM
VSUM = VSUM
+ 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) Put CV(!) into OMEGA.

I
I I 3 I ( I ) Response Co nstant s

I L __ _ I
I •I For a rigid-body mode ,
•I l)NO(I) DS UM; YNO(l) = VSUM Put DSUM and VSUM i1110
ONO an VNO.
I
I
go to I.

I 4 For a vibrational mode,

l __ __ _ I OM EGA2 = OMEGA*OM EGA
OMEG AD = OMEGA* DSQRT(l.0 - DAMPR* DAMPR)
determine the constant's
I CN = DAMPR* OMEGA: EX = DEXP( - CN* DT) CI through C8 given by
Eqs. (4 . 10-3).
I C = DCOS(OMEGAD*DT); S = DSIN(OMEGAD*DT)
C l = EX*(C + CN *Si OMEGAD)
I C2 = EX* S/OMEGAD; C3 = ( 1.0 - C l) /OMEGA2
cl. 'lh 111Nfor111 A1>pllccl Aclfons
I C4 = (DT - C2 - 2.0* CN* C3) /(0MEGA2* DT)
CS = - OMEGA2 *C2; C6 = Cl - 2.0*CN*C2; C7 = C2: CS = C3/DT
I
r- --- (1 I 1,NMODES I
1
I
•I
I
•I l11 i1i11litl· SlJM 10 11•1, 1,
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).

I
546 Flowchart for Program DYNAPT App, C App.C Flowchart fo r Program DYNAPT 54'
(2) S1cp-hy-S1cp Rc,pon,c

•I 2

I DNJ = DNO(I); YNJ = VNO(l) lni1ializc DNJ and VN.1 12. SUBPROGRAM NUMINT FOR PROGRAM DYNA
I a. Read and Write Integration Parameters

I
I r---- ALPHA.BETA,GAMMA
I 1
I I DN(l,J) = Cl *DNJ + C2*VNJ + C3•AN(I)
Calculate modal n:-

•I I
I
*STEP(J) + C4*AN{l)*RAMP(J)
VNJ = C5*DNJ + C6*VNJ + C7*AN(I)
*STEP(J) + C8*AN(J)*RAMP(J)
sponscs using Eqs. (4. IO I l
and (4. 10-2).
DIRECT NUMERICAL INTEGRATION
ALPHA = (ALPHA) BETA = (BETA)
GAMMA = (GAMMA)
I
I
I
•L___ _
I
DNJ = DN(l,J)

3
CONTINUE b. Calculate Structure Damping Matrix
I
I 4
L _____ _ (I) Delcrminc Factor SA

r-----
RETURN
END
,
I Cl = DSQRT(2.0*DAMPR*CV(I)) Compute Cl = C)iJ .
II. SUBPROGRAM TRABAC FOR SUBPROGRAM NORMOD I
•t - - - - -
r----- 2 J = 1.NTS 1
1 I Initialize SUM to zero.
I I
r----- I
1 I
I
r---
I Initialize SUM to , cro . 1
I I
I I L SUM = SUM + CMS(J,K)*PHl(K.I ) Multiply M by <t>N.

I
I
r---
1
I I = I.NMODES
•
I
SA(J.1) = SUM*CI
Complete factor SA =
M<I>New needed for Eq .
•L
I SUM = SUM + PH l(K,l )*DN(l,J)
Back-1ransfonn cfo,phtn•
mcnt, with Eq . (4 I ~)
I
I
L____ _
2
(4.6- 17).

I 2
L_ _ _ _ _ Place SUM intu thl' ll\'t'
DF(K,J) - SUM
di~phtCCllll' lll lllllll IX I)( •

IWl'l JRN
l•N I>
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---- 1
1
I l
r---- I
I
I
I
' I
I
r-----
I I 1 6

I I L SUM = SUM - SS(J ,J )*DO(J) - CS(l,J)*VO(J)

I
r--- I
l Set up right-hand vec-
•I L' 3

SUM = SUM + SA(K, J)*SA(J,J)

+
I
AA(!) = SUM + AF(l)*STEP(J )
l tor in Eq. (5 . 1- 4).

I Apply Eq. (4.6- 17).

l ______ _ 7
I
I CS(K ,J) = SUM
I 4
L___ _ [ SOLYER(SA,AA.AO,NDF) Solve Eq. (5.1 - 4) for ini-
rial accelerarions.

c. Calculate Initial Accelerations

r- 9 I = 1,NDF
r- 1
I
'
I- -
I I
I AO(]) = 0.0 If rhe mass matrix is
singular, use zero ini-
lial accelerations.

I l_
I SA(l,J ) "' CMS(I.J)
Copy mass 111a1rix lo SA.
I 5
L_
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 = 1.0 + ALPHA; C2 = 1.0/( BETA*DT*DT) Detennine consrants
C3 = GAMMA/( BETA*DT) for Eqs. (5.5- 15) and (5.5-33).
680 I l11woh1111 ror l'r our IHn llYNI\I' I I\JJp, C I lowol11111 for Program DYNAPT
551

l
,---~
I
I. NI> ~
I
t
L ____ II J
1,NDF I AA([) = SUM + AF(!)*RAMP(I) Complete the calculation
of terms in Eq. (5 .5- 16).
I I 14
I L______ _
Calcul111c l!qulv11ll1 111 CONTINUE
I SA(],J) = Cl*SS(l,J) + C2*CMS(l,J)
stiffness n1u11 ix hy
+ C3*CS(!,J)
+ &1s.(5.5 15)111111 (\, 111
I II Solve Eq. (5.5- 14) for the
L __ _ SOLVER(SA,AA ,DD,NDF)
incremental displacements.

r---
Factor cquiv11lc111 still
FACTOR(SA ,NDF,& 101) ncss matrix . If 11 I, 1
s ing ula r, go to IOI , I
DF(l, I) = DO(!) + DD( !)
I AO(!) = AO(!) + C5*DD(I) - Q(I)
Apply Eqs. (5.4- 16), (5.4- 17),
c. Calculate Response for First Time Step (5.4- 18), (5.5- 12), 1111d (5 .5 13) 10
I VO(]) = VO() ) + C6*DD(I) - R(I)
find lhc 10111! rcsponsl!s.
I 15
CI = 1.0/ (BETA*DT) ; C 2 = 1.0 / (2.0*BETA) Dete rmi ne cons1u111s 1111 L __ _
C3 = GAMMA / BETA; C4 = (GAMMA*C2-l.O)*DT Eqs. (5.5- 10) 1h101111h
C5 = Cl/OT; C6 = GAMMA*CI (5.5- 13).

f. Calculate Responses for Subsequent Time Steps

r--- 12 I = l,NDF

I
I Obtain Qand 1 Vl'l'lui.
r-------
,---
Q(J) = C I* VO(J) + C2*AO( l)
R( I) = C3*VO(I ) + C4*A0(1) from Eqs. (5 .5 10)
I
I 12
and (5 . 5- 1 I ). I
I
'
L__ _ I
I
I I Q(l) =
C l *VO( I) + C2* AO(!) Use Eqs. (5.5-)0) an9 (5.5-J I) to
I R(I ) = C3*VO(I) + C4*A0(1) obtain vectors Q and R.
r------- I
I I
L __ _
I
I
•I r-----
r --- -- IJ K I. ND!' I 1
1 I +
L{ SlJM .~lJM I C'MS(l ,K)' Q(K) I C'S(I,K)• R(K) Add lllllSS lllld d11111pl11p
I I
tc 1111, 111 llq . (~ , 111) I I
t
552 Flowchart for Program DYNAPT App.C App. C Flowchart for Prog ram DYNAPT 553

13. SUBPROGRAM RES2PT* FOR PROGRAM DYNAPT

a. Read and Write Output Selections

tI tI
r----~ IWR ,IPL,NNO,NEO
Read and write output

17 r~~~~~~~~-'--~~~~~--,
parameters.
I I
I I I
I I L SUM = SUM + CMS([ ,K)*Q(K) + CS(l ,K)
*R(K) + ALPHA*SS(l ,K)*DD(K)
Heading: OUTPUT SELECTIONS 6
Subhead: !WR IPL NNO NEO
I I Apply Eqs. (5.5- 16) and (5.5-34) to
I I de termine the incremental
load vector.
I I AA(l) = SUM + AF(l)*(STEP(J)
I + RAM P(J) - STEP(J - I) - RAMP(J - I))
I IWR. IPL, NNO,NEO 6
I I 18
t L _____ _ CONTINUE
I (JNO(J), J = l ,NNO)
Read and write nodes
selected.
I
I SOLVER(SA,AA,DD,NDF) Solve Eq . (5.5- 14) for the
incremental displacements.
I NODES: (JNO(J), J = l ,NNO
__) __ 6
I
I
I ,r--- (IEO( l), I = l,NEO) Read and write elements
I I
selected.
DF(l,J) = DF(l,J - I) + DD(l) Use Eqs. (5.4- 16), (5.4- 17),
t I AO(I ) = AO(!) + C5*DD(1) - Q(l) (5.4- 18), (5. 5- 12), and (5.5- 13)
I VO([) = VO(l) + C6*DD(I) - R(I) to o btain the total responses. 6
I ELEMENTS: (IEO(l), I = 1.NEO)
I IL __ _ 19
I
CONTINUE
I 7
I 20
l ______ _
I , write parameters
to plotter.

RETURN b. Transfer Displacements from DF to DS

IOI
FACTORIZATION OF EQUIVALENT
STIPrESS MATRIX f'AILED
If equivalent sti ffness
matrix is singular, write
this message; and stop cal
cu lat ions.
,----
1
4 I = l ,NND

t Set K equal to ID(l).

STOP I
END I
•For convcnicm·c, 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 App.C App. C Fl owchart for Program DYNAPT 555

•I If K corresponds to a IPL = If IPL = I, write K to plotte[

I free displacement, go to 2.
I
I Otherwise, put zero into

•I
DSO(l) = 0.0 DSO (for a restrained dis-
placement).
J = O; TIME = 0.0 Initialize J and TIME to zero.

I r-
I !WR = I
1 J,TIME,(DSO(KO + K). K = 1.2) 6
I L_
I DS(I.J) = 0.0 Also, put zeros into OS. If IWR = I , write initial

•I IPL = I
displacements.

TIME,(DSO(KO + K), K = 1,2) 7

I
I 2
I For a free displacement,

,---
DSO(I) = DO(K) Similarly for IPL = I.
put DO into DSO.
I 5 K = 1,2
I
I ,- I
•I 1
L_ DS(l,J) = DF(K,J) I
I
I
DMAX(K) = DSO(KO + K)
DMIN(K) = DSO(KO + K)
TMAX( K) = 0.0; TMIN(K) = 0.0
Initialize MAX and MIN
terms.
I Also, put OF into OS. I
I IL ___
I 4
I
5
L _ __ _
I

l". Write Displacements

I
I
I
I
,---
I
8 J = 1,NTS

,----
1
I
I
I
I
TIME = DT*J Compute TIME al step J.

The node number K is fo1111d , I I

I K JNO(KK); KO = 2*(K - I) parameter KO is comp111cd, I I r--
I J NRL(KO + I) + NRL(KO + 2)

•I
and the number of r'Cst111i111.,
is obtained.

If J = 2, there arc 110

I
I •I I
I DS(KO + K. J) < DMAX(K ) If displacement DS
go to 6.
< DMAX ,
2 I I I
I displacements al node
I I
I K; so 11011c arc writhlll. I
I I Otherwise, replace DMAX
I DMAX(K) = DS(KO + K. J)
I llc11di111:1: l)ISPLAC'EMENT TIME I I I TMAX(K) = TIME and TMAX with current
values.
I I II S'J ORY FOR NODE (K) 6
I I I 6

i
Suhh,•ud : S'l'lll' 'l'IME DJ I D.12

I
I
I •
I
L __

I I
111111
1 lowuhut I lo t l't or,1111t1 l>YNAl' I Ar,11 (
App. C Flowchart for Program DYNAPT 557

I
I
t
,r--
7 K 1,2 IPL= l = 7
I I
(DMIN(K), K 1,2)

I I Similarly for plotter.

I I DS(KO I K, J) -> l)MIN(K) If' displ11t'l'IIK'1II I >S I ll\llN I
•I I
+
go 10 7.

+
I [WR = I TIME OF MINIMUM
(TMIN(K), K = 1,2)
6

I I DMI N(K)
TM JN(K)
DS(KO+K, J) Otherwise, 1q il11l'1• 111\ II N
11nd 'l'MI N w ill1 1•11111•111 I
'J'JM E
I I value,. I Write times of minima.
9
I I L____ _
I L__ CONTINUE

I I

I I
d. Calculate and Wr ite Axial Forces
t I IWR J,TIME.(DS(KO I K, J), K 1,2) J h
I I
I I r-- - - -
I J- .---------.J If IWR I, w,IJ1• d lN pl111111 11111 14 II = l,NEO
I 1
I I 11'1.
TI ME,(DS(KO + K , J), K 1,2) I
I I I l = IEO( II) Find dc111c111 1w111hc1 I .

I t I
Similarly for 11'1 ,
I
I L __ _ 8
I !WR = I
I CONTINUE t Heading: AXIAL FORCE TIM E
HISTORY FOR ELEMENT (I) 6
I I Subhead: ST EP TIME AM I

I I
IWR
I MAXIMUM
(DMAX(K). K 1,2)
(J I IPL = I 7 Write I to plotter.
I I
I Wri1c 111aximu111 d1 Np h1, 1
I
t 11'1.
rncnt,.
I
I (DMAX(K), K 1,2) 7 I JI= 2*JN(IJ - l ; J2 ~ JI + I
Compute displacement in-
I I Kl = 2*KN(l) + I; K2 = Kl + I
dexes , and so on.
SCM = E*AX(])/ EL(I)
I Sim ilarly l<ir plo111•1 I J = 0: TIME = 0.0
I + AMO = SCM*((DSO(K I) - DSO(Jl))*CX(I) Calculate initial axial
I TIME OF MAXIM UM
(TMAX(K), K 1.2) I + (DSO(K2) - DSO(J2))*CY(I)) force.

I
I
I
' Write 11n1c~ uf 11111, 111111 i
I
IWR = I 6

iI [ I WI{ I ~ I NIMl lM
L ")M IN(K), K l.2)J'1 I
I
If !WR = I , write initial
axial force.

J. J W1111•1111111r1111111cllspl1111 I 1
I 11111111N, t
IPL =_._ _ _~
--T-IM_E_,,AMD : imilarl y for IPL = I.
I 1
I
668 Flowchart for Program DYNAPT App< App. C Flowchart for Program DYNAPT 559

• t
I I !WR = I TIME OF MAXIM UM 6
I AMDMAX = AMO; AMDMIN = AMO Initialize MAX and MIN (TMOMAX)
TMDMAX = 0.0; TMDMIN = 0.0 I
terms.
t I Write time of maximum.
I I
I r---- I ! WR = I M INIMUM(AMDMIN) 6
I 1 I
I I I Write minimum axial force.
I TIME = DT *J; AMO = SCM*((DS(K 1,1)
I - DS(J I ,J))*CX(l) + (DS(K2,J) Calculate TIME and AMI>
I
I I - OS(J2,J))*CY([)) + IPL = I 7
+ t I
If axial force AMO < i\MllM \
I Similarly for plotter.
AMO < AMOMAX
go to 11. I
I !WR = I TIME OF M INIMUM 6

AMOMAX = AMO; TMOMAX - T IME

Otherwise , replace AM l)M/\ \ I (TMOMIN)
and TMDMAX with cu11 ,•111
values. I Write time of minimum.
II I 14
AMO > AMOM IN If axial force AMO · /\M l >t-.ll N L _____ _ CONTINUE
go to 12.

Othe1wise. replace /\MDM IN

AMOM IN = AMO: TMOM IN = T IME and TMDMIN with Clllll' ltl RETURN
values .
12 END

!WR=
14. SUBPROGRAM DECOMP* FOR SUBPROGRAM STASYM
I, write axi11I 111111·

IPL = I 7 r ----- 3 l = l ,NDF Arguments: (A ,NDF, *)

1
t Similarly for IPL I, L ____ _
I 13 I
L __ _ _ I
•I
CONTINUE
t SUM = A(l,J); Kl = I - .t
I
I IWR M/\X IMUM( /\MDM/\ X) 6 I
I
I
I r----
Write 111uxi111u111 11x111I lrn, 1 I 1
I
I
I
IPL (/\Ml)M /\X~ 7 •
I
L_ SUM = SUM - A(K,l)*A(K ,J)

•I --r Si111il11rly !or pli1111•1

I
•Sec Co1111i1111•rl'ro111·11111.1'.f111· S1r111·111n,I /\1/(//1w/,1, hy W, Wcuvcr.
Ii , ll V1111 Nost1 aml , l'1111t·1·to11 , N J , 1%7
11110
I lowoh1111 for flroornm IJYNAPr
App. C Flowchart for Program DYNAPT 561

t t
I I I
I SUM = 0.0: JMI = J - I
I
I
I r---
I 2
I 1
I
A(l,l) = DSQRT(SUM) I L SUM = SUM - A (KJ)*A (K,J)
I +
I
I
L__ _
I 2
A(J,I) = SUM*A(J.J)

I
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
A rguments: (A ,N , *)

15. SUBPROGRAM INVERU* FOR SUBPROGRAM STASYM

r- I I = 1,NDF ArgumcnJs: (A,NDI •)

,-----
1
I I
I
L A(I .I) = 1.0/ A(l .1) I
I r----
NI NDF - I
I 1
+ I SUM = A(l ,J); II = I - I
I I
r----- 3 I l,N I I I
I I I
I I
lI IPJ I I I I
I
I
SUM = SUM - A (K ,l)* A (K ,J)
+
L ___ I I 2
.I J 11'1 ,NI)!'
I L ___ _
+ A(l,J) = SUM
I +
I
•11,111
•Sl'\' M111r/1 ll1111/1w/.1 11{ F ri11111•r/ .\'11111 11111•1, 211d ,·d • by W
W1•i1Vl' I, Jo 1111,i J. M (ll•H·, V1111 Nos111111d 1{1•1 11h11Jd , N1•w Y1 11 i-, l'IKIJ
562 Howchart for Program DYNAPT App c App. C Flowchart for Program DYNAPT 56:i

I
I SUM = A(J,J)
+
I
! r---
I
'II
I
I TEMP = A(K,J)/A(K ,K)
SUM = SUM - TEMP*A(K,J)

I IL__ _ 3

I A(K,J) = TEMP 1 -.- 3 I = I .N

I IL _
3 __.____ _
I X(I ) = X(]) / U(I.I)

I
tL ___ _ 4
A(J,J) = SUM
,----
RETURN
END
'II l = N - 11 +1
SUM = X(I)
K2 = I + I
I
17. SUBPROGRAM SOLVER* FOR SUBPROGRAM NUMINT I
I
t ,r---4
r---- 2 I = l,N A rguments: (U, B,X ,N) I L SUM = SUM - U(I. K)* X(K)
1 I
I I 5
SUM = 8(1); KI = I - I L ___ _
X(I) = SUM
I
I
I r--- RETURN
I 1 END
I L SUM = SU M - U(K, l)*X(K)
I
t 2
L ___ _ X(I) SUM

' Ihid
 Ar1ewo1 11 tu l'r uhlome 565

6 Lf, 88g A
2.3-7. D = \ {j()8 sin flt 2.3-8. D *=
1271T4
G cos ut

2
2.3-9. D = ~; d sin flt 2.3-10. D = - d cos flt
35

2.4-1. u = 13f sin (flt - 8);

1
8 = tan- C)~fln2)

Answers to Problems 2.4-2. u* =

ma .
- f3T sm (flt - 8);
_ _ ( 2nfl )
8 - tan 1 w 2 _ ff

-n,( M fl + Nn . \
2.4-3. u,. = e \N cos wdt + wd sm wdt)
Pl 3 Ml
2.4-4. D = 0.01445 - sin(flt - 8) 2.4-5. 88 = 9.086 ~G cos (flt - 0)
El 7Tr
mal3
2.4-6. D = -3. 185 ~ cos(flt - 8) 2.4-7. /3rcs = 50, 25, . . . , 2.5

2.5-1. P (t) = ;:Pi(sin flt + i sin 3flt + . . )

2.5-2. P (t) = ; P (cos flt - j cos 3flt + .. )
1

CHAPTER 2
2.5-3. P (t) = 3_ Pi (sin flt - ! sin 2flt + ! sin 3flt - ...)
1r 2 3

2.2-1.
1 /3ii 1 /3ii T-- 2 L -Vm
w = L 't;;:;i;; ! = z1rL -v-;;i>
/mL 1r. 2.5-4. P (t) = ~2 - ~1r (sin flt + !sin 2flt + ! sin 3flt + ...)
2 3
4 /3ii 2 /3ii 7Tl /mL a0 ~ a1 cos (i flt - 8;) + b, sin(iflt - 8,)
2.2-2. w = L -V-;;;i; f = 7Tl -V-;;;i ; T = 2 V3Ei 2 5-5 u = - + L.J
• . k 1• 1 k \!(1 - i 2 fl 2 / w2 ) 2 + (2yifl/ w) 2
- l /3EI . - l /3EI. T -- 2 L V(2;i,
2.2-3. w - L Vi;;:;i, . t - z1rL Vi;;:;i, , w 1T. 2.5-6. u = 6.70t sin flt 2.5-7. u = 13.06 k cos flt
Pi

2.2-4. w =
M£.
°V-;;;E ; f
1 ME.
= Z7r °V-;;;E ; T = 21r°VAE.
/mL
2.5-8. u = -6.2 1 Pi . flt
k sm
2,-2 {3;G r2 {3;G 7Tl2 /2pA
2.2-5. w = u V2pA ; f = 7Tl 2V2pA ; T = 7 V~ 2.6-1. u = kPi(·
sm flt fl sm
- -; . wt)\[ _ (D,/
1
1 w)2 J
/3ii 4 /3ii 7Tl /mL
2.2-6. w =
8
L -V-;;;i ; f = 7Tl -V-;;;i ; T = 4 V3Ei 2.6-2. u = t [1 - e- n,(cos wdt + : d sin wdt) J
2.3-1. u = [
1
_ (~ /w) 2 ] d cos flt
1
I
2.3-2. u * = - [ _ (fllw ma
)2 k J t'OH Ht 2.6-3. u = ~ [w 2
t - 2n + e- n,(2n cos wdt - wl - n sin wdt\ ]
2

t, kw Wd }

PL 3 2ML
2.3-3. D '""'
144
W sin flt 2.3-4. 80 = - - -
31rr 4 G
cos n, 2.6-4. u =
P,
k (1 - cos wt) (0 st s ti )

2.3-5. /)
9
5
d sin !lt 2.3-6. 0 "'
ma1.'
. cos
45111
nt
u= t [cos w(t - t,) - cos wrl

11114
 n1111
At111wor11 to Problems
567
I',/ 1 co.~ 1,1 (t CHAPTER 3
k I I

11 - 7Z lcos w(t 1,) cos Ml 3.4-1. pb(t) = {2b 1 + b2, b, + 2b2} L 3.4-2. pb(t) = {1, 3} b2L
6
-i 12
~
[cos w(t - t2) - cos w(t - 11)]
(IJ II
3.4-3. K =EA[-; ~: -~] 3.4-4. M = p1sL[ : -~]
2.6-5. u = -Pi ( l - cos wt - -t + - wt)
sin -
(0 • I
3L 1 -8 7 -! 2
k t, wt, / 1)
3.4-5. pb(t) = {I , 4, 1} bxL 3.4-6. Pb(t) = {5, 3}
mx1L
u = -Pi [ - cos wt sin wt - sin w(t - t ,)]
+- ------- 6 12
wt,
k
P,
2.6-6. u = -(1 - cos wt)
( r1 I)
3.4-7. pb(t) = {L - x, x} :x 3.4-8. pb(t) = {9, 2L, 21, -3L} b2L
60
k (0 • I I ii

u = -P, [ 1 - cos wt - -t - - t, + -sin -w(t-- -

ti)]
(t 1 • I ,,1
k t2-t1 w(t2- ti)
u = P, [ sin w(t - t1) - sin w(t - t2)J
- - cos wt + - - - - - - - - - -
k w(t2 - t1) (I i rl Pb(th = {6x
2
- 6xL, 3x 2L - 4xL2 + L3, - 6x 2
2.6-7. u* =_ a~(!._ _ sin wt)
W V1 Wt1 (0 s I · I i)

u >I< = - a1
- 1
w2
[
+sin-w(t
- -- -t 1)-- -sin
wt,
- wt]
- (t I I)
3.4-11 . Given 3.4-12. Given

2.(,-8. u* = - ao
- (I - cos wt) - -a, -- 2 a- 0 - - -sin- wt)
<.1.>2 w
(t !1 Wl1 (0 /1)
EA EA
3.5-1. S, 11 = 0.64T; S,12 = - 0.48L ; S,13 = O; etc.

u* = - :: [cos w(t - t 1) - cos wt]

M,11 = 2. 72 p~L; M,, 2 = O; M,13 = 0.06 pAL; etc.

EA EA EA
- --
a 1 - ao [ sin w(t - t ,) - sin
2-
wt + COS W
(t - ( >]
I
3.5-2. S,11 = I.64L; S,12 = 0.48T; S,13 = -L; etc.
w wt 1 I)
pAL pAL
2.6-9. u* = - ~2 [ ~ - - 2 -( t - cos 2
wt)] M,11 = 3.28 - -; M,12 = O; M, 13 = 0.64 ; etc.
w d w tr (0 "'!:. I I iJ
6 6
3EA EA EA
3.5-3. S, 11 = L ; S,12 = - L; S, 13 = - L; etc.
u* =- a ~{ w; ,,2[cos wt - cos w(t - t,) + wt, sin w(r -
w /1 )] 2 2
4pAL _ . _ pAL .
M,1 1 = - -; M,12 - 0, M,13 - - - , etc.
+ cos w(t - t,)} I)
3 6
EA EA
3.5-4. S,11 = L; S,12 = O; S.,13 = -L; etc.
l.7- 1. 11 10 = 0 2, 7-2, U1 0 = kPi 2.7-3. Given 2.7-4. Given
pAL pAL
Pi P, M,11 = - -; M, 12 = O; M,13 = 0.64--; etc.
2.7-5. 112() .., • 2.546 k U20 = - 4 -
2.7-6. 2.7-7. U1 0
Pi
0.9 135 k
3 6
k EA EA
Pi 3.5-5. S,1 1 = I.64L ; S,12 = 0.48L; S,13 = O; etc.
2,7-fl. //1 0 0.3 186t 2.7-9. 1110 = 1.379 -( I'1
2.7-JO. 11 10 I. 102 ,..
M, 11 = Zp:L ; M, 12 = O; M,,3 = 0.06 pAL; etc.
 ti68
An11wurn to f'tuhl1t1111
Answers to Problems
EA3EA
3.5-6. S, 11 = U; S,.,2 = - l; Sm == O; etc.
2
4pAL _ . _ pAL. 3.7-3. (a) w1 , J == V04.•
16.64, 95.21
L2
e J ; cf>
{ii. <I>i = [ -0.5868
3
== [ e
16. 86
J
M, , I = -3-; Ms12 - 0, M,13 - 6 ' etc.

3.5-7. S,11 =
4EJ
T; S ,1 2 = T2EJ ; Ssl 3 = O; etc. (b) W 2.4 =
53.28, 139.2 /El. <I> = [
L2 vpA,
I J.
-0.1011 • 2
<1>
4
== [ 1 ]
0.10n
3 0.9258 /E . ·
pAL pAL3 . _ . 3.7-4. (a) w 1 == ~ yp; mode shape 1s unity.
M,
'
1
=-
105·' M,12 = ---,
140 M,13 - 0, etc.
/E mode shape 1s. uruty.
T1 134 yp; .
3.5-8. S, 11 =
8EJ
T; Ss12 = u;
6EJ 2EJ
Ssl3 = T; etc. (b) w 2 =
3
2pAL 13pAL2. __ pAL 3 . 3.7-5. (a) w2,J == 1.485,L 1.732 /E.
yp' <I>2 -[l]·
- o ' <I> == [OJ
1
Ms11 = ~ ; M,12 = - ~ , M , 13 - 140 , etc. 3

3.5-9. s,, I = u;
12£/
S , 12 = u;
6E!
Ssl3 -
_ 6EJ.
u' etc. (b) W1 , 4 ==
0.6650, 2.909
L
/E.
Vp'
<I> _ [ J
I -
1 ] . <I> == [
0.8801 ' 4
l
-0.6091
= = llpAL2. Ms1J = - 1 J
M 13pAL. M i2 13pAL2; etc.
sll 35 , s 210 , 420 3.7-6. (a) W1 ,4 == 0.6581, @.<I>== [ 1.344
L 1.974 Vp' l ]·<I>= [
' -1.092
1 4

3.5-10. S , 11 =
8EJ
T; S , 12 = -u; 6EJ
Ssl3 =
2EJ
T; etc. l.271, 1.927 /E. <I> -[ 1 ]· = [ l ]
3 (b) W2,J = L Vp' 2
- 0.08687 '
<1>
3 -7.365
2pAL 13pAL2 . __ pAL3 .
M
sll -- - - ,· M sl 2 = - 420 ' Ms 13 -
105 140 , etc.
3.7-7. (a) W3,4 ==
1.352, 2.218
L
/E.
Vp' <I>3 _ [
-
I
-0.3242 '
] . <1>
4
== [ 1
2.055
J
3.6-1. Answer not provided.

3.6-2. W1,2 -
_ 0.5412,
L 1.307 Vp'
@. ,I>
'¥1
= r2.414J
1 '. <I> = r-0.4142
1
(b) W1,2 =
0 , 1.225
l
/E ; <f>1 = [J
yp l] . _[ , <f>2 -
1 ]
-0.5
2
1
3.6-3. W1 ,2 = 0.9393~ 3.529 /E ; <f>1 = r-21
yp ; <f>i -
_ r]J
2
1
CHAPTER 4
3.6-4. w,,2 = 0.6049,
l 1.815 /E.
vp' <1>1 -
_ [1].
3 ' <1>2
= r-31
1

5.546 , 21.04 /El. <I>1 _ [ L ] . <I> = [ L ]

43 _ [ (9 sin w1 t)/ Wt + (sin w2t)/ w2 JDoi
3.6-5. W1.2 = L2 vpA., - -0.3947 • 2 - 26.6 1 . -1. D - (3 sin w,t)/ wt - (3 sin w2t)/w2 10

3.6-6. W1 , 2 = 15.14,
L228.98 /El. -[ -II]·,
V04.· <1>1 - <1>
2
= [1]1 43 2 _ [I.961L cos w, t + 0.03923L cos Wzt] PoL2
• • • D - 2. 701 cos W1 t + 0.2990 cos W2t 6EJ

13.32, 34.79 /El . <I> _ [ l ] . <I> = [ I ]

3 _ [ 1.207 COS W1 t - 0.2072 COS Wz( d
4.3- . D - 0.5000 cos Wt t + 0.5000 cos W2f
J
3.6-7. W1,2 == L2 V04. · -0.707 1 1 - ' 2 0.707 1
4 - r c-2 sin W1t)/w1 + (2 sin W2t)/w2]Dot
/El /El 4.3- . D - (sin w, t)/w, + (4 sin w2t)/w2 5
3.7-1. (a) w1 = ~ ypA.;
22.74
(b) w2 = ~ ypA.
81.98
27 cos w, t - 3 cos W2t] Pol
81 .98 /El 4.3-5. D == - [ 81 cos w 1t + cos Wzt 18EA .
3.7-2. (a) w 2 = ~ ypA.; h . .
modes ape 1s unity.
3 (l.015L sin w, r)/w, - (0.0J506L sin w2r)/w2 1 D0,

(b) w, .J ==
53 .28, 139.2
,.2
{ij
v;x; c,, 1
r 0.1m,
, 1.1· ,,) 1 0.1m ,
, . 4. -6. D == [ ( 0.4006 sin <JJ,t)/w, I (0.4006 sin <u2r)/w2 f ,

1 1 4.).7, I) • / : / (} u ~·os 10,1

 670
Answers to Problems Answers to Problems 571
4.3-8. D = [ 2.707 cos Wit+ 1.293 cos w 2 t ]M, L 0
-1.914 COS Wit+ 0.9142 COS Wzt 14£/ 3
_ - [t + 24(r- ~ 2 sin w2t)/w~1___!:_
4.3-9. D = [t + (2 sin Wzt)/wz]D01 4.4-9. D - 3
( 1 ) z l08t1m
t - (sin w2t)/w2 3 t - 12 t - wz sin W 2 t /w2
pAL
4 .4 •1. (a) D = [ 9(1 - cos Wit)/wf + (1 - cos w 2 t)/w~]_!j_ where m = - -
6
3(1 - cos wit)/wt - 3(1 - cos w 2t)/w~ 10m
(b) D = [3(1 - cos w,t)/wt - 3(1 - cos w 2 t)/w'i]_!j_ 4.5-1. D* - [ -9s, - Sz J~
si = (r-
(
~i
1
sin w,r)/wt
)
(1 - cos wit)/wt + 9(1 - cos w 2 t)/w~ 10m -3si + 3s2 lOt, 2
s2 = t - wz sin w2t / w2
3.28 pAL
where m = ; (D2)A, = (D1)A 2
6 _ [ (1.000 - 1.221c, + 0 .2205c2)L] ~ c, = cos w, t
4.5-2. D - - 1.681c, + l.68lc2 L C2 = cos W2t
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 453 _ [ - 1.000 + 1.207c, + 0.207lc2 JLO c1 = cos wit
• • · D- -1.000 + 0.5000c1 + 0.5000c2 ' c2 = cos w2t
(b) D = [(0.02675{3i/ wt + 0.2878{3z/ w~)l] M 2 sin flt
0.03683/31/wr + 2.193{32 /wi mL2 4.5-4. D -
* _ [ 2b,b bi = 13d wt
- 2b2
4b
b _ r.i. / 2
J~ cos .at
pAL - 2 - JJ2 W2
I - 2 5
where m = ; (Dz)A, = (D,)A 2
210 (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
O. 353/t - J_ sin w1 t\ I wt - 0. 353/t - J_ sin w) / w~1 wheres= EAIL ; mi = pAL/6; m = 3.28m1; bi = /3i/wt; b2 = {3z/wi
4.4-3. D =_ \' w, / \ Wz J .!:_
_ [(-1.000 + 1.053c, - 0.0532c2)l] (J c1 = cos w1 t
[ 456
0.1465(t - ~i sin w1~/wt + 0.8536(- ~ sin w 2 ~/w~ t,m • • • D - -1.000 - 0.4156c, + l.415c2 ' c2 = cos W2t
2
where m = pAL
~
4.5-7. D* = - [ J b2 (J, sin flt where b2 = /32/ Wi
4.4-4. D =r-
2(1 - cos w,t)/wt + 2(1 - cos Wzt)/w'i]_!j_
(1 - cos w,t)/wt + 4(1 - cos w 2 t)/w~ 10m 458 o.05776b1 - o. t882b2J so, . fl s = s + 3mfl2
h 2.72 pAL • • • D = [ -0.04084b 1 - 0.133lb2 --;;;- sm t s = 2El/l
were m =
6 b, = {3i/ WT; b2 = {3z/ Wi m = pAL3 /420
4.4-5. 0 =[ /3,/wt + 9{3z/ w~] Pi cos flt where 3.28 pAL
3{3,/wt - 3{3z/w~ 10m m = 6
CHAPTER 5
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 + 0.1816(1 - cos w 2 t)/w~ m 5 .2-1. and 5.2-11. Given
pAL3 Pi Pi
where m = - - 5.2•2, U10 = 0.9819k 5.2-3. Uzo = -3.063k
2 10

4 •4•7. I) rJ I /31Pi L sin flt

- I 88wrm
where m = pAL
420
3
5.2-4. Ut 6 = 0.3622k
A
5.2-5. U20 = -2.595k
A

P, P,
- 0 .05776(1 _!_ sin w,,)!wr+ 0 . 1882(, I .
SIil W1 f ) / (1/11
5.2-6. u20 = - 4.004k 5.2-7. u,o = 0.9182k
4.4-8. I) • ( Wi W2 /\I J P,
= J .387 kPi
r
w lu•11• 111
0.04084 r
/IA/ I
,,'., sin ,,,,,)/M
t I 0, I 1.l 1(, 1
(tJ1
sin 10,,)/1,d 111/I
5.2-8.

5.2- IO.
111 0 ""

II 111
0.3 146k

I , 107
I'1
1.
5.2-9. U1 0

•I '0
 572 Answers to Problems S'l
Answers to Problems

5.2-12. (D1)20 = - 19.94~; (D2ho = -19.85~ 5.4-4. (8) U15 = 0.3676t (b) U15 = 0.3659t
s

5.2-13. (D1)max = -0.9992~; (D2)max

EA
s
= 0.7950~
s 5.4-5. (a) U 20 = -2.513t (b) U20 = -2.541 t
wheres= -
L 5.4-6. (8) U20 = -3.942~
k (b) U20 = -3 .963~
k

For problem sets 5 .3 and 5 .4 ' parts (a) and (b) denote average- and linear-acceleration
methods. 5.4-7. (8) Uio = 0.9087~
k (b) U10 = 0 .9119~
k

5.4-8. (a) U10 = 0.3191~

k (b) U10 = 0 .3176~
k
5.3-1. and 5.3-11. Given

5.3-2. (a) U10 = 1.031 t (b) U10 = l.016t 5.4-9. (a) U10 = l.372t (b) U10 = l.377t

5.3-3. (a) U20 = -3.023~ (b) U20 = -3 038~

5.4-10. (a) U10 = J.094~
k (b) U10 = 1 .098~
k
k . k
5.4-12. (a) (D1ho = -21.22~ (b) (D 1)20 = -21.40~
5.3-4. (a) U1 6 = 0.3676t (b) U16 = 0.3659t s s

5.3-5. (a) Uzo = - 2.513t" (b) U20 = -2.541 t"

5.4-13. (a) (D1)max = - 0.9974~ (b) (D1)max = - 0.9993~
5.3-6. (a) U zo = -3.942t" (b) U20 = -3.963t" s s
(D2)max = 0.7917~ (D2)max = 0.7861~
5.3-7. (a) U10 = 0.9087t (b) U10 = 0.9119t s s
- 4 2 -
5.4-14. S = S + (AtJ1M + tltj C; Aj+I = Aj+I + M Qj + C Rj
5.3-8. (a) U10 = 0.3 191 ~I (b) U10 = 0.3176t
- 4 4, .. _ 2 .

5.3-9. (a) U10 = l.372t (b) Uw = l.37?t where Qi= (Ati)2Di + Mi Di+ Di; Ri = Ati Di + Di
- _ 2 Ati - - _
5.4-15. C - C + -,.-M + -S; Ai+t = A1·+ 1 + M Q·1 +SR1·
5.3-10. (a) uw = l.094t" (b) uio = l.098t" 1.1tj 2
- 2 , .. - flt ..
where Qi= - D1· + D1 • R1 = -D1 - ___!D1·
Atj ' 2
5.3-12. (a) (D1ho = -20.45~ (b) (D1ho = - 20.32~
- _ At1 (A tj)1 -
s
5.4-16. M - M +
2
c + - -s; Aj+I = Aj+I + c -Qj + s _Rj
(D2ho = - 19.51 ~
(D2)zo = -19.47~ 4
s s where -Q · = -i>1 - AtJfr. R 1 = -D1 - MD .. - (At1 )2D..1
1 1 1 1
1 2 ' 4
5.3-13. (a) (D 1)ma. = -0.9976: (b) (D 1)max = -0.9994~
s - At 2 -
5.4-17. C = C +-M + ___!S· AA=
1 AA1· + MQ1 + SR1
At1 2 '
(D2)max = 0.7917~ (D2)max = 0.7860~
s s where Q1 = 2DJ; RJ = -AtJDJ
5.4-1 and 5.4-11 Given

5.4-2. (a) U1 0 = J .03 ] t (b) U10 = l.0J 6t

5.4-18. M = M + Atic + (AiY s· AA1· = AA1 +
2 4'
c -Q·1 + s -R1-

5.4•3. (11) Pi 1 0 lH /'i

11 111 - .1.02.\ k (h) 11 111
. A
 1174

Annwur " to l'rublome

5
5.4-19. S* '"' S I (A:)2 M I ~lt1 C; At, 1 • A1,, I M Q/ I CR/

= (A 6 )2 D1 + -;.-
where Q/ 6 D1· + 2D1;
·· R/ == -;.- . I At D1
3 Di -1 2D1 "
1
u.tj u.lj u.t1
2
5.4-20. C* = C + AtM
2 +3 At1S ; Aj+I
* = A1+1 + M QJ* + S RI *
J

2At1 CHAPTER 6
where Q*
1 2 · 1 + D-·
= -D ··1 R -* = - D. - -
1
D· - -(At1)2 D··
A t1 '
1
3 1 6 1
Lby-{t) pi:k{t) = Lby·(t)
5.4-21. M * AfJC + ~
=M +2 (AtJ)2S ; A*
j+ I = A~+l + CQ* + SRj * j
6.2-1. pi:i,) = OJ l92;
[1IL
78 [OJ
18 l92
-SL
where Q.*
1 = -D· 1 - -2D (At-)2 .•
At ·· · R * = -D - At-D. - - --='---D'J
1 2 1 ;, 1 1 3 54
2 Pbi(t) = 234]Lby-{t) ; Pbk(t)
312 =[ 72 ]Lby-(t)
5.4-22. C* = C + A M + Atj S; AA/ = AA; + M Q/ + S R/ [ 960 960
u.tj 3 55L -25L

5.4-23. M*
where

=M +
Q/ = 2D··1;

~jc + (At
. - T
R/ = -AtiDi
2
(At)2
D ··j

s; AA/= AA1 + C Q/ + s R/
6.2-2. p;/t) = [1~]
3L
P~f); p;k(t) = [ 5~
-9L
J~f)
where Q*
J
= -ArD·
J 1,
R*
J
= -AtD - (AtJ 2
2 1) J J J

PPit) =[ -10
to
J Py,(t)
64\/2 ; p,,k(r) =[ - 54
54
J/~,,(I)
64V2
A 1 y A A A 3V2L - 9V2L
5.5-1. S = S + /3(AtJ 2 M + /3 Ati C; Ai+1 = A1+1 + M Qi + C R
1 M,(t)
A 1 1 ° 1 - 2(3 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 y (1 - "jjy) Di - o (
1- y) At1D• 1 PM(t) = {-12, -6, -VSL, M,(t)
12, 6, -VSL} 4V5L
213 Lb2(t)
6.2-4. pi:(t) = {O, 9, 2L, 0, 21, -3L}6()
5.5-2. C = C + + 1
M + /3 yAt S; A1+1 = A1+1 + M
y u.tj lt + s Rj
1 • 1-y•• pb(t) = {- 36, 27, IOL , -84, 63, -ISL} Lb2(t)
where Q·1 = - - D1 + - D
A

y At1 y 1 300

and R.1 = - D1 - (1 - ~) Mi)1 - G- ~)(At1)2 D

1
Py,(t)
6.2-5. p;(r) = {O, 9, 2L, 0, 9, - 2L}--
9
5.5-3. M= M + y dtjC + /3(Af_j) 2 S; Aj+I = A1+ 1 + C Qj +s~ ' r,:
p,,(t) = {-1 8, -9, 2V5L, -18 , -9, - 2v5L} V5
Py,(t)
where Q1 = -D1 - (1 - y) At1Di 9

and R.1 = -D1 - At1Di - G- /3)(AtYD1

M,(t)
6.2-6. p,&(t) = {O, -18, -L, 0, 18, -L}~

1 M,(t)
PM(t) = {18, -18, -V2L, -18, 18, -V2L} 8V2L
5.5-4. C = C + -y Atj-M + /3 ydtj S; AA.j = dA; + M Qj + S R.1

where Q1 = ;D 1; R.1 = -At1Di - G- ~)(dt1>2 D1

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 577

' , ..I • I , p,,' (I ) ,,o, 5/,, IH , 0, 11/ , /Hf 1,/, (I)

192 6.4-4. p,:(t) = {O, 2, 0, 0, 3, O} Py,(t)
5
I'" (I) (20/,, I5L, 90, 44/,, 331., 390} Lb, (!) PPi (t) = ~PPk(t) = {O, 0.2828, - 0.2828}Py,(t)
960
6.4-5. ph(t) = {O, -1, 0, 0, l, O} M~(t); L = V29
,,.J..z. 1>,\(1) ... {O, 4L , 20, 0, 2L , 7} p;~)
PMi(t) = -pMk(t) = {-0.07773, -0.1184, - 0. 1201}M,,(t)
1>,,(1) { 4L , - 12L, 20VW, 2L, 6L, 7VJO} P, (i) 6.4-6. ph(t) = {O, 0, I, 0, 0, -1} M~(t); L = V29
27VJO
tt.J.J . ,,;,(,) - {O, 3l, 18, 0, -5l, - 18} My-(t) PMi (t) = -pMk(t) = {-0.1332, -0.03805, 0.1327}My-(t)
16L 6.4-7. PMt) = pbk{t) = {- 1.000, 0.4242, O}b(t)
l>M(l) ... { 3L, 3l, 18V2, 5l, - SL, - 18V2} My•(!) Pbi (t) = Pbk (t) = {-0.5992, -0.1141, -0.8988}b(t)
16V2l 6.4-8. pi:)t) = 4Pi:h) = {-0 .3334, 0.6009, O}b2 (t)
t,..1.4, 1,:1(1) {12l, 3L, 18 , 4L, - 5L, - 18} My(t) Pbi(t) = 4Pbk(t) = {-0.3234, 0.3637, -0.4851}bi(t)
16V2L 6.5-1. Pbi(t) = {-1.057 , 2.439, -0.4269, - 1.769, - 0.49 13, 1.572}b,,(t)
P.11(1) :.:: {9l , 15L, 18V2, 9L - l - l8V2}My{t) Pbk (t) = {-1.057, 2.439, -0.4269, 1.769, 0.491 3, - 1.572}/JAI)
' ' 32l 6.5-2. Pbi(t) = {- 1.601, - 0.4447, 1.423, 0.6523, - 1.505, 0.26'.'ltlfh,,,(1)
td-5. 1>J,(1) ""' {12L, -3l, - 18, 4l, 5L, 18} M;)!j_ Pbk(t) = {- 0.3695, - 0.1026, 0.3284 , - 0.2965, 0.6842, 0. 1 l<J7fi,,.,(t)
16 2l 6.5-3. PPj(t) = {-0.3675, 0.4492, 0.6 125, 0.3811, 0. 1429, (), 111, f/ \,•(/)
l>M (l) {1 5l, 9l, - J8V2, - l, 9l, J8V2}Mx(t) PPk (t) = {-0.06805 , 0.08318, 0.1 134, 0. 1270, 0.047M, 0 111 )} /\,•(/)
32£ 6.5-4. ppit) = {0.1132, - 0.04244 , 0.09902, 0.07639, (),()() l.1(1, (), I l'/ \f/ 1 , (/)
,, . .1.(,. J>,;(I) {O - 3L 21 0 2L 9}lbi(t) PPk(t) = {0.6111, - 0.2292, 0.5347 , 0.2292, 0.280 I , 0. IHI C) )/ 1,(/)
' ' ' ' ' 60
_ { My-(t)
1>,,(1) { 9L, - l2l, 105, - 6l, -8L , 45} Lbi(t) 6.5-5. PMj (t) - -4, 0, 0, 0, 0, O}---is-
300
_ { My-(t)
(t,J.7, 1>/,(1) ... {O, JL J6 0 3L 16}?,(t) PMk ()
t - 4, 0, 0, 0, - 4, 3}---is-
' ' ' ' 16
_ { M,,(t)
6•5•6• PMj ()
t - 0, -16, 12, 5, 0, 0} - -
1'1· (t) {9L, I 2l, 80, - 9l , - 12l 80} P, (t) 75
' 80
_ { M,,(t)
(,.J. ff , 1•J,(1) (0 I O O - I O} My•(t) PMk ()
t - 0, 16, -12, 0, 0, 0}75
' 3 , t ' '

6.5-7. Pbi(t) = {1.5, 0, 0, 0, 0.5, 0.5}bx(t)

l>M (I) {2 I O - 2 - I O}M,.-(t) Pbk(t) = {1.5, 0, 0, 0, -0.5, - 0.S}b,.(t)
' . ' . ' 3\/5
6.5-8. Pbi(t) = {- 0.01111 , 0.02222, 1.028, -0 .3000, - 0.1500, O}b 1(t)
,,,4- L 1>,;(1) {O, I, 0, 0, I, O}Lb)"(t) ; L = VJ4 Phk(t) = {0.01111, - 0.02222, 0.4722, 0.2000, 0.1000, O}b 1(t)
4
1>1,1(/) l>11A(I) ""' { 0.3043, 0.7 161 , 0.5192}b,,,(t) 6.7-1. W1 = 15.69 /El,
,I 1
'•""•• 1>,, I '( ) {()
,
()
, I , 0, 0, I - - ; L = v 14
} / ,/),,(/) • r,-; L2 -V;;;.
'
1>,,,(t) 1>114(/)
6
{ 0.3126, 0.2233, 0.49 12}/J,,(,) 6 , 7 • 2, W1 2 = 5.603, 31.19 2l lfil,
- ; <f> = [ 1.000 1.000 J
t,,4.,1. 1>,\(t) {O , 0, 3, 0, 0 , 2) '\(I) ' L pA 0.5435 -0.9364

I>/'/(/) fo,,d1) (- 0.4899, 0.2449, 0.2449}/>.,(t)

6.7-3. W1 = 9.941
l2 -v/El,
pA
1118
An"wnr11 to l't uhln1111

t,.7-4, t,)1,1 . , • 0, 0, 22.47 /}1 Jn pA ; (I>

6.7-5. W 1,2 -
I
1. 648 , 5.540L2
{iii. ; <I> =
Y"iA [0.8844
1.000

6.7-6. W 1.2 = 0.9044, 3.840 L2I Y"iA;

{iii, <I>= [ 01
O
IJ

/G] _ 5.943 /GJ

6.7-7. w, __ 6. 149 Y"iA
u:- 6.1•8 . w, - u:- Y"iA 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 component-
Action vector:
mode method, 470; in tridiago-
applied, 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
u111l11111pt•d l'it•t• vih1·111 ions, l 11, l O'i co11strni111 s 11uuim.t, NO
580 Index Index 581

Axisymmetric loads, 357 dynamic coupling in, 470 two- and three-dimensional, Degrees of freedom, number of, 105
Axisymmetric shells: for plane trusses, 471: Program 310-69 Density, mass, 80, 497
element AXSH3 , 394 COMOPT, 477 Continuous beams, 243 Dependent displacements, 282 , 291
nonaxisymmetric loads , 401 for space trusses , 477: Program loads , 265 Determinant of Jacobian matrix, 330,
Program DYAXSH, 406 COMOST, 477 Program DYNACB, 265 332
Axisymmetric solids, 357 substructure equations of Convergence, rate of, 206 Diagonalization, 140
element AXQ4, 359 motion, 468 Conversion factors, 496 Differential equations of motion, 75
element AXQ8, 359 Computer programs, 9 Conversion of US units to SI units Dimensionless coordinates, 319
nonaxisymmetric loads, 361 Condensation, matrix, 282 (table), 496 Direct iteration, 498
Program DYAXSO, 365 Conditional stability, 208 Coordinates: Direct linear extrapolation, 211
Consistent-mass matrix: natural, 318 average-acceleration method, 212
for axial element, 84 normal , 139 linear-acceleration method, 213
Backward substitution, 461 for axisymmetric shell element relative, 22, 159 Direct numerical integration methods ,
Beams: with nonaxisymmetric loads , Corrector, 203 195-240
continuous, 243 , 405 Critical damping , 30 Direct stiffness method, 98
frequency coefficients (table), 288 for axisymmetric solid element Critical time step, 198, 224 Direction cosines, 96, 101
Blast load, 50 with nonaxisymmetric loads, Curvature, 90 Discretization of structures, 6
Block mass submatrices, 437 364 Displacement equations of motion ,
Block stiffness submatrices , 437 for element, 82 145
Body axes, principal, 420 for element AXQ4, 359 D' Alembert's principle, 3, 13 Displacements:
Body forces, 78 for element AXSH3 , 401 Damped forced vibrations, 30 dependent, 282 , 291
Body-oriented method, 434 for element H8, 347 Damped free vibrations, 27 generic, 75 , 78
Boundary nodes, 446 for element PBQ8, 378 Damped response of MDOF systems: ground, 21: rigid-body, 157
Bracing in tier buildings, 468 for element Q4 , 336 to arbitrary forcing functions , 172 incremental, 38, 212 ,
Buildings, multistory, 425 for element SHQ8, 389 to periodic forcing functions, 168 independent, 282, 291
for flexural element, 91 Damping: initial, 14, 147
for grid member, 251 absolute, 166 kinematically equivalent, 414
Center of mass, 418, 427, 434 for plane frame member, 246 critical, 30 nodal, 75, 79: free, 99; restrained,
Central-difference predictor, 197 for plane truss member, 103: in effects of, 25 99
Characteristic equation, 105, 498 component-mode method, 474 in MDOF systems, 164 relative, 23 , 159
Characteristic matrix, 106 for space frame member, 261 modal , 167 Displacement shape functions , 74 , 79
Characteristic-value matrix, 141 for space truss member, 255 numerical, 218 for axial element, 82
Cholesky method, modified, 498 for structure, 98 proportional, 165 for element AXSH3 , 398
Cholesky square-root method, 106 for torsional element, 87 relative, 166 for element H8, 345
Coefficients: Consistent systems of units (table), Damping coefficients, 165 for element H20 , 349
damping, 165 496 Damping constant, 3, 27 for element PBQ8, 375
ncxibility, 145 Constant-acceleration method, 219 modal, 165 for element Q4, 333
ror Gaussian Quadrature (table) , Constraint, modal, 500 Damping matrix, 164 for element Q8, 338
328 Constraint matrix, 291, 372, 382, 394 Damping ratio, 31 for element SHQ8, 386
frcqullncy, 287 , (t(lb/e), 288 Constraints: modal , 165 for flexural element, 89
lllUSS , 82 against axiul strains, 290 Decomposition: for torsional element, 85
st i ffrH.'SS, 82 rlgid-body, 4 I I , 427 Fourier, 36 1 Dissipative force, 3
<'0111pom·11t rnodl' nwthod , 468 C'ontinuu: spectral , 223 Duhumcl' s integral , 39, 152, 173
IINNl'lllhly of s11h11t1m'llll'l'N, 470 Hll'l'SSl'S and Hll'HillN in , J 10 Dl'flution , 500 Dy11n 111 il' t•q11ilibl'i11111 , I
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
l•w·nvulucs, 106, 498
ll'Pl'Ull'd, I 07
1'.11\l' nvcclor matrix, 140, 315
1 1),tl'llVCClors, 106, 498
1 Flexibilities, I 0
11m111ulization of, 142
l•,lusticity , modulus, 497
1'111st it'ily force, 3
l•lt•111l' II( /\XQ4 , 359
i'.h·111L·11t /\XQ8, 359
l•h•1m·111 /\XSll3 , 394
hlt'llll 1l( /\XSR3, 394
1 Flexural strains, 89
l•l1•111t•nt 118, 345
(<'.lt•Jlll'II( 11 2(), 349
I •,lt•111L•nt PBQ8 , 371
I k1m·11l PQR8, J7 I
l•k1m·11t ()4 , 1 n
1•11·1111•111 ()8, JJ8
I l1•1m•11t R4 , 1 U
l•h•1111•11t RH , \ lH
l•h•1111•111 RS8, '4.'l
l•h•1m•11t RS l O, W>
l•h•1m•11t SI 1()8, IH)
I•I11111111111011.
1111 WIii d, ,1(,1
lnda,c Index
Clauss .Jordun, 291 Fourier series , 35, 169, 361, 401
Elimination method, purolld, 4~':t Framed structures, 6, 241-309
Equations of motion , J , 11, 7':t, 18 Frames:
action, 140 plane, 243 , 244
displacement, 145 space, 244, 259
for finite elements, 78 Free nodal displacements, 99
incremental , 2 12 Free vibrations, 12, 27
structural, 98 Frequency, 14, 28
Equivalent nodal loads: Frequency coefficients, 287, (table).
for element, 82 288
nonproportional, 267
for structure, 98
for surface pressure, 366, 408 Gaussian quadrature, 328
Equivalent viscous damping, 26 coefficients for (table), 328
Euler's extrapolation formu la, 20~ Gauss-Jordan elimination, 291
Explicit formulas for extrapolation , 11) I Generalized acceleration method,
Extrapolation , direct linear, 2 11 217
Generalized rotation matrix, 505
Generalized stresses and strains, 92
Factorization: General shells:
Cholesky, l06, 498 element SHQ8, 382
Givens, 514 Program DYNASH, 390
Factors , conversion, (table), 496 General solids:
Finite-difference formulations, 195 element H8, 345
element H20, 349
Flexibility coefficients, 145 Program DYNASO, 351
Flexible-body motions, l 38 Generic displacements, 75, 78
Flexural deformations, 90 Geometric center, 318
in plate-bending, 378 Geometric interpolation functions ,
Flexural element, 88 320, 324
Flexural rigidity , 94 for element AXSH3, 396
for element HS, 345
Forced vibrations , 19 , 30 for element H20, 350
Forces, body , 78 for element PBQ8, 374
Force systems, types of, 241 for element Q4, 333
forcing functions: for element Q8, 338
arbitrary, 38, 172 for element SHQ8, 383
hurmon ic, 19 Geometric transformations, 431
periodic, 35, 168 Givens factorization, 514
piecl'wise lincur, 46, 175, 182 Givens method, 509
Fonn f1w1or , 177 Global axes, 95
FORTRAN, 9 Sram-Schmidt orthogonalization,
Fm wnrd t•li11ii11ntio11, 4<> I 107, 500
F111w111d lll'llllioo, 41>8 Grnvitalionnl system of units, 49.5
1•1111111·1 nwll1r11·11h, IC, <liid~. 24 I, 249
583
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', I•I'I
Implicit fo11n11lt1Nfrn ill' lllll1111 , IO\
Incremental ncn·kratH111s, 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 co-
ordinates , 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
I111l•1fo1 lllldl'N, 44(1
 ll84
lllll'I polutioo f'1111ctio11s: ct1uivulcnt nodul :
gL·onwtlie, J20, 325 (.1·<<' olso CJ co
1
for elcmenl , 82
111etric inlerpolation functions)
for structure, 98
piecewise-linear, 46, 175, 182
nonaxisymmetric, 36 1
Inverse ilcralion, 498 normal-mode, 152
Inverse of Jacobian matrix , 321 , 325 on grids, 27 1
In verse of normalized modal matrix ,
on hexahedra, 351
143
on plane frames, 267
lsoparamctric elements, 310- 69
on plane trusses, 179
clement AXQ4, 359
on quadrilateral s, 340
elcmenl AXQ8, 359
on space frames , 278
hcxahcdron H8 , 345
on space trusses, 274
hexahedron H20, 349
Local axes, 95
quadrilateral Q4, 333 Lumped mass matrix, 99
quadrilateral Q8, 338
Iso tropic materials, 77 , 311
Iteration (for eigenvalues), 498
Magnification factor, 20, 32, 169
ltcralion with implicit formulas, 203
Main program:
for DYNA (flow chart), 226
for NOMO (flow chart) , 178
,lncobian matrix, 321, 324
for VIB (flow chart), 11 9
determinant of, 321 , 325
Mass
inverse of, 321, 325
center of, 418 , 427, 434
Jacobi method, 505
of a body, 419
Joinls, 241
Mass coefficients, 82
Mass density, 80 , 497
Mass matrix:
Kilogram (kg), 495
for element, 82
Kilopound (kip) , 497
for structure, 98
Kincmalically equivalent displace- lumped, 99
ments, 414
principal , 140
reduced, 283
Mass moments of inertia of a rigid
Lu~ time, 17
body, 41 9
I,inear-acceleration method, 219
principal, 420
direct Iinear extrapolation, 2 13
Mass products of inertia of a rigid
iteration wilh , 206
body , 419
Linear differential operator, 79 Master displacements, 283, I 0.2- 1<)
I ,incur extrapolation , dirccl, 2 11 Material properties (table), 497
I .oud oplll'ltt or, 223
Materials:
Louds:
anisotropic, 311
11xi sy 11111wtric, 357
isotropic, 77, 311
rn1 rn11lin11011s hcums, 265
orthotropi<:, J I I
dy11111n il', I
pmpl'I I il•s of (/o/,/1') , ,11>7
11•1·ip1t w 11l tlwc111·111 , 15,1
Mntd \ l 'Clll<ll' IINIIIIClll , 18 )
h11 hue Index
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 eigenval-
ues) , 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 k•s. 15 I
585
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
N111111•111 11I lll'l'tll ary, .U I
1 18
N 11111e•11111l d11111p111p,
 ~88
N11111l'lll'lll 111tt•g1allo11, I 2C,
N11111t•1 il-111 stabilit y, 22.l
( )uc-dlmcnslonal clements, 82
<)1 tho).\onulily 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
Pnrnllcl elimination method, 455
1'11rc11t rectangle:
t·lcmenl 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
Pt•riod, 4, 14, 28
l't'riod elongation, 225, 234
l't11iodic forcing functions, 35
d11111pcd response of MDOF sys-
tems, 168
l',•1 iodic ground acceleration, 38
Ph1111c angle, 16, 32, 105 , 169
l'IIIINC plane, 14
I'it·t·cw isc-1inear forcing functions ,
46 , 175, 182
l'l1111t• frumcs, 243, 244
11xi11l constraint s, 290: program
DYPFAC, 299
dynnmic loud duta (table), 269
1'10g1·11111 DYNAPF, 267
1lgid bodies: Program DYRAPF,
,1}4
Nlllll'tlll'lil d11t11 (tnl>/t•), 268
l11rla11
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', ' element AXSR3, 394
l11clox
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
587
(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 re-
sponse, 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 analy-
sis, 118
main program (flow chart) , 119
Program VIBPT, 11 8
structural dutu (tr1bll') , I 20
Proportional dH111pinH, I (1 ~
Properties of' 11111tt·1 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:
clement PQR8, 371
 688
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 func-
tion:
l'orced part, 20
free part, 20
Rt·sponse spectra, 5 1
Rt•strai necl nodal displacements, 99
Retai ned modes , 469, 472
l{t•vcrse 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
lndox
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 k·r, JS
s11h~l111l'll11 t•11 111 ,14c,
1 Stability:
of im•11111 Sl•lhncks 111 1ll•t h111!dl11~11. ,l(,H
Index
589
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 eigen-
element 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 Axisymmet- 405
ric solids)
Solids: for axisymmetric solid element
with nonaxisymmetric loads,
axisymmetric: isoparametric ele- 364
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 DYS-
for 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
0
for torsional element, 87
Spectral shifting, 503
for x-beam, 428
Square-root method , Cholcsky, I 06
for y-beam, 428
fill' z l'llhlltlll , ,1W
('Olldi1101111l , ' OH pli11l'1p11l , 1·10
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
Index Index 591
Subprogram FACTOR (flow chart) , normal-mode response, 157 Truncation, modal, 139, 143
561 Support reactions, 99 Trusses:
Subprogram INVERU (flow chart), Support restraints, independent mo- component-mode method, 471
560 tions of, 162 plane, 243, 471
Subprogram NORMOD, 227 Surlace pressure, equivalent nodal space, 243, 253, 477
Subprogram NUMINT, 227, (flow loads for, 366, 408 Twist, 85
chart), 547 Symmetric, standard form (of eigen- Two-dimensional continua, 310-69
Subprogram RESl , 118 value problem), 106, 505 Types of force systems, 241
Subprogram RESlPT (flow chart), Symmetric modes, 112 Types of framed structures, 241
532 Systems of units:
Subprogram RES2, 179, 227 absolute and gravitational, 495
Subprogram RES2PT (flow chart), consistent (table), 496 Unconditional stability, 208
553 Undamped forced vibrations, 19
Subprogram SDAT, 118 Undamped free vibrations, 12
Subprogram SDATPT (flow chart), Three-dimensional continua, Underconstrained frames, 290
519 310-69 Unit-load method, 146
Subprogram SOLVER (flow chart), Tier buildings, 425, 461 Units, systems of (SI and US), 495
562 bracing, setbacks, and shear cores, Unit vectors, 96
Subprogram STASYM, 118, (flow 468
chart), 525 Program DYMSTB, 463
Subprogram STIF, 118 Time step, critical, 198, 224
Subprogram STIFPT (flow chart), Torsional element, 84 Velocities:
523 Torsional rigidity, 93 incremental, 38, 212
Subprogram TIHIST, 177, (flow Torsional strains, 85 initial, 14, 147
chart), 545 Transformation matrices for framed Vibrational analysis, 105
Subprogram TRABAC, 179, (flow structures (table), 415 Program VIB , 118
chart), 546 Transformation methods for eigen- Vibrational motion, 152
Subprogram TRANOR, 177, (flow value problems, 505 Vibrations:
chart), 543 Transformations, geometric, 431 forced, damped, 30
Subprogram TRAVEC, 118, (flow Transient response: forced, undamped, 19
chart), 529 with damping, 34 free, damped, 27
Subprogram VIB, 177, 225 without damping, 25 free, undamped, 12
Substitution, backward, 461 Translational inertias, 91, 378, 390, Virtual strain energy, 80
Substructure equations of motion, 401 density, 316
446 Translation of axes, 157, 414 Virtual work, 80
in component-mode method, 4C,H Trapezoidal rule, 203 Viscous damping, 26, 164
Substructures, 8, 444- 82 Triangular impulse, 45
Substructures in parallel, 455 Tridiagonal matrix, 509
Substrnctures in series, 445 Tridiagonal method, 445 Wilson-8 method, 220
modified tridiagonal method, •I~ I modified, 451 Work, virtual, 80
tridingonal method, 445 Trigonometric series, 35 Working points, 416
Supc1 purnml·lric ckml•11ts, \ \ \
S11pp1111 111111,ons, •I
i111l1•p1•1ul1•111 , I <,J